classifying aerosols based on fuzzy clustering and their...
TRANSCRIPT
Research ArticleClassifying Aerosols Based on Fuzzy Clustering and TheirOptical and Microphysical Properties Study in Beijing China
Wenhao Zhang Hui Xu and Fengjie Zheng
Institute of Remote Sensing and Digital Earth Chinese Academy of Sciences Beijing 100101 China
Correspondence should be addressed to Hui Xu xhradi163com
Received 14 March 2017 Revised 27 May 2017 Accepted 4 June 2017 Published 24 July 2017
Academic Editor Harry D Kambezidis
Copyright copy 2017 Wenhao Zhang et alThis is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
Classification of Beijing aerosol is carried out based on clustering optical properties obtained from three Aerosol Robotic Network(AERONET) sitesThe fuzzy 119888-mean (FCM) clustering algorithm is used to classify fourteen-year (2001ndash2014) observations totallyof 6732 records into six aerosol types They are identified as fine particle nonabsorbing two kinds of fine particle moderatelyabsorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust and desert dust aerosol These aerosol typesexhibit obvious optical characteristics difference While five of them show similarities with aerosol types identified elsewherethe polluted dust aerosol has no comparable prototype Then the membership degree a significant parameter provided by fuzzyclustering is used to analyze internal variation of optical properties of each aerosol type Finally temporal variations of aerosoltypes are investigatedThe dominant aerosol types are polluted dust and desert dust in spring fine particle nonabsorbing aerosol insummer and fine particle highly absorbing aerosol inwinterThefine particlemoderately absorbing aerosol occurs during thewholeyear Optical properties of the six types can also be used for radiative forcing estimation and satellite aerosol retrieval Additionallymethodology of this study can be applied to identify aerosol types on a global scale
1 Introduction
Aerosol is one of the largest sources of uncertainty in theradiative forcing and plays a key role in global climate change[1 2] It has been identified that aerosols directly influence theearthrsquos energy budget and indirectly alter the cloud processes[3 4] Furthermore different types of atmospheric aerosolswill also lead to different radiative effects which depend ontheir optical and microphysical properties [5] For examplethe present of strong absorbing aerosols like black carbonwill lead to a positive radiative forcing (warming) On thecontrary the presence of nonabsorbing aerosols such asfine hygroscopic particles (sulfate for example) will resultin a negative radiative forcing (cooling) [6 7] In additionthis uncertainty arises in regions where aerosol particles arecontributed by complex components and show significantspatial and temporal variability [8] Therefore it is necessaryto understand different types of aerosol properties (especiallyin regions with higher concentration of aerosols such asChina where they are found with a high positive trend ofAOD from 2001 to 2010 [9]) to reduce their uncertainty
in radiative forcing estimates and other related scientificfields
To better depict aerosol properties many studies werecarried out in the last decades [10ndash17] Some of the stud-ies [10ndash13] classify aerosol types by thresholds of aerosolproperties such as aerosol optical depth (AOD) Angstromexponent (AE) fine-mode fraction (FMF) single scatteringalbedo (SSA) or their combinations For example to dis-criminate aerosol types of biomass-urban desert dust cleanmaritime and mixed-type appropriate thresholds for AODand AE are applied [10] In addition by using SSA and FMFobserved byAerosol RoboticNetwork (AERONET) Lee et al(2010) classified global aerosol into four types they are dustnonabsorbing black carbon and mixture [13] However inthesemethods characteristics of aerosol types at one locationor site are represented by mean values of measurementsThe shortcomings are obvious because aerosol propertiesare easily affected by meteorology events such as wind andrain [18] Thus aerosol type can be changed as short as afew hours Besides in regions affected by multiple aerosolsources (such as China which is affected by dust from desert
HindawiAdvances in MeteorologyVolume 2017 Article ID 4197652 18 pageshttpsdoiorg10115520174197652
2 Advances in Meteorology
biomass burning aerosol from agriculture and black carbonfrom industrial emissions [13 19]) the temporal distributionsand characteristics are rather complex Consequently meanvalues of long-term aerosol properties observations are insuf-ficient to represent characteristics of aerosol types [14]
To overcome this shortcoming the others [14ndash17] takeadvantage of clustering to classify aerosol types Clusteringis a statistical tool used for grouping a set of objects intoseveral clusters using predefined variables so that objectsin the same cluster are more similar to each other thanthose in other clusters Omar et al (2005) applied clusteranalysis to global ground-based observations and obtainedsix categories backgroundrural industrial pollution dirtypollution biomass burning desert dust and polluted marine[14] In addition Levy et al (2007) performed a ldquosubjectiverdquocluster analysis for AERONET records which were dividedinto three fine-sized dominated spherical types (typicallythe ldquolowrdquo ldquomediumrdquo and ldquohighrdquo absorbing types) and onecoarse-sized dominated spheroid type [15] Besides Qinand Mitchell (2009) applied Locally Scaled Density BasedClustering (LSDBC) algorithm to classify Australian aerosolinto four classes including aged biomass burning smoke freshsmoke coarse dust and super-absorptive aerosol [16] Theseresults show that cluster analysis is suitable for classifyingaerosol types However the limitation of these classificationsis that the values of cluster center are used to representthe characteristics of corresponding aerosol types As amatter of fact the differences between records at edge ofcluster and those near center are inevitable Particularly formultiparticles mixture aerosol records are usually on theboundaries between several clusters and only representingthem by the center of each cluster is unreasonable [20]To overcome this limitation Wu and Zeng (2014) appliedGustafson-Kessel fuzzy clustering algorithm to identify theoptical properties of pure dust aerosol type [20] Comparedwith aforementioned clustering algorithm not only center ofcluster but also a significant parameter named membershipdegree (between 0 and 1) is provided by fuzzy clustering Themembership degree is used to describe the confidence degreeof one record belonging to a cluster The records near thecenter of cluster will have larger value of membership (higherdegree of confidence) than those at edge Thus membershipdegree will be helpful to analyze the internal variation ofoptical properties within each aerosol type The capability offuzzy cluster analysis to obtain aerosol properties in singletype dominant regions has been documented whereas itscapacity to obtain aerosol properties in regions which arefrequently influenced by various sources of aerosols still hasnot been well reported
Beijing one of the largest megacities in the worldhas been suffering severe issues of aerosols loading fordecades because of rapid development of both economy andpopulation It is famous for its complex aerosol particlesand high proportion of anthropogenic aerosols Due to itsimportance to climate and environment many studies werecarried out to obtain the aerosol properties in this regionThe researchers are interested in (i) characteristics of fine orcoarse particle aerosols [21 22] (ii) components and sources[23 24] (iii) aerosol optical properties [25ndash29] Nevertheless
the dominant aerosol types and their temporal variation stillhave not been well reported But the definite aerosol typesare important to improve radiative forcing estimation andsatellite aerosol retrieval To this end this paper focuses onclassifying identical aerosol types and studying their opticalproperties
In this study long-term (14 years) observations ofAERONET are used to determine dominant aerosol typesand their optical properties over Beijing One of the mostwidely used fuzzy clustering algorithms fuzzy 119888-mean (FCM)algorithm [30] is applied to classify distinct aerosol typesThis is the first comprehensive study investigating aerosoltypes and their temporal distribution in Beijing This workwill provide a primary parameter for the estimation ofaerosol radiative forcing and retrieval of aerosol propertiesfrom satellite aerosol as well as reference for methodologyof identifying aerosol types in other regions The paper isstructured as follows Section 2 briefly describes the datasetand method used in this study a detailed analysis of aerosolsproperties is conducted in Section 3 at last the conclusionsare drawn in Section 4
2 Data and Methodology
21 Description of Aerosol Sites and Data The AERONET isa global network of ground-based CIMEL sun-sky radiome-ter AERONET inversion products include a comprehensivedata set of aerosol optical and microphysical propertiesaerosol optical depth (AOD) at 340 380 440 500 675 870and 1020 nm the single scattering albedo (SSA) complexrefractive indices asymmetry parameter (ASYM) at fourwavelengths (440 676 869 and 1020 nm) and the parametersof particle size distribution [31 32]
In this study three AERONET sites (surrounding BeijingCity) are selected they are Beijing Xianghe and XinglongBy choosing these sites over 14 years (2001ndash2014) a totalof 6732 records are obtained It should be noted that wecollect records of AERONET ldquoAll Pointsrdquo Level 20 inver-sion products which can be obtained from httpaeronetgsfcnasagov Figure 1 and Table 1 show the detailed infor-mation of the selected sites
The following 22 parameters obtained from AERONETinversion products are applied in the cluster analysis
(a) Single scattering albedo (SSA) at 440 676 869 and1020 nm
(b) Real part of refractive index (REFR) at 440 676 869and 1020 nm
(c) Imaginary part of refractive index (REFI) at 440 676869 and 1020 nm
(d) Asymmetry parameter (ASYM) at 440 676 869 and1020 nm
(e) Parameters of particle size distribution finecoarseparticle volume concentration (VolConFVolConC)finecoarse particle volume median radius (VolMedi-anRadFVolMedianRadC) finecoarse particle stan-dard deviation (StdDevFStdDevC)
Advances in Meteorology 3
50∘
45∘
40∘
35∘
30∘
25∘
20∘
50∘
45∘
40∘
35∘
30∘
25∘
20∘
75∘
80∘
85∘
90∘
95∘
100∘
105∘
110∘
115∘
120∘
125∘
130∘
135∘
75∘
80∘
85∘
90∘
95∘
100∘
105∘
110∘
115∘
120∘
125∘
130∘
135∘
(a)
Xianghe
Beijing Xinglong
Hebei ProvinceBohai Sea
TianjinCity
BeijingCity
43∘
42∘
41∘
40∘
39∘
38∘
37∘
36∘
43∘
42∘
41∘
40∘
39∘
38∘
37∘
36∘
114∘
115∘
116∘
117∘
118∘
119∘
114∘
115∘
116∘
117∘
118∘
119∘
(b)
Figure 1 Geographical location of the selected three AERONET sites
Table 1 Information of the selected three AERONET sites
Site name Observing period Number of records Longitude(degree)
Latitude(degree)
Elevation(meter)
Beijing 2001ndash2014 3418 116381 39977 92Xianghe 2001 2004ndash2012 2797 116962 39754 36Xinglong 2006ndash2012 517 117578 40396 970
It should be noted that the aerosol optical depth (AOD)fine fraction by volume (FFV) sphericity parameter (SP) andwater vapor (WV) are also used in later discussions Howeverthese four parameters are not applied in clustering
22 Fuzzy Clustering Cluster analysis is a statistical toolused for grouping the data sets into several clusters based onpredefined variables [14] The basic theory of cluster analysisis records in the same cluster which are more similar to eachother than to those in other clusters In fuzzy clustering everyrecord has a degree of belonging to each cluster rather thanjust completely belonging to one cluster [20] This degreeis represented by a parameter named membership degree(between 0 and 1) which indicates the confidence degreeof one record belonging to a cluster The records on theboundary of cluster will have smaller values than those nearthe center
In this study fuzzy 119888-mean (FCM) algorithm one ofthe most widely used fuzzy clustering algorithms is appliedGiven the data set containing 119899 records119883 = 1199091 1199092 119909119899and supposing there are 119888 clusters with the centers 119881 =V1 V2 119881119888 The FCM aims to minimize the objectivefunction
min 119869 (119906 V) =119899
sum119896=1
119888
sum119894=1
1205831198981198941198961198892119894119896 (1)
where 1198892119894119896 = 119909119896minusV1198942 represents the distance between record
119909119896 and center V119894 And 120583119898119894119896 is the degree to which record 119909119896belongs to cluster V119894 and is defined as
119906119894119896 =1
sum119888119895 (1003817100381710038171003817119909119896 minus V119894
1003817100381710038171003817 10038171003817100381710038171003817119909119896 minus V119895
10038171003817100381710038171003817)2(119898minus1) (2)
For a certain record sum119888119894 119906119894119896 = 1 The fuzzifier 119898 deter-mines the level of cluster fuzziness meaning the overlappingdegree of the cluster and 119898 ge 1 The center of the cluster isdefined as
V119894 =sum119899119896=1 119906
119898119894119896119909119896
sum119899119896=1 119906119898119894119896 (3)
The step of FCM is described as follows Firstly a randomfunction is used to determine the initial center of eachcluster Calculate membership degree and objective functionbased on functions (1) and (2) Meanwhile centers of eachcluster are calculated based on function (3) Repeat untilthe algorithm has converged that is the membership degreechange between two iterations is no more than the givensensitivity threshold It should be noted that before clusteringthe 22 parameters are normalized by the standard deviationto ensure that each parameter makes the same contributionto the distance calculation
3 Results and Discussions
31 Results By applying the FCM clustering a total of 6732records are classified into six clusters The characteristics ofoptical and microphysical properties of each cluster centerare shown in Table 2 Figure 2 displays aerosol optical depthsingle scattering albedo asymmetry parameter fine fractionby volume sphericity parameter and water vapor of eachcluster center Figure 3 shows the particle size distributions
4 Advances in Meteorology
Table 2 Results of the fuzzy cluster analysis
Properties Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
AOD
440 nm 125 156 079 076 073 087676 nm 077 094 045 043 044 064869 nm 054 066 031 031 033 0551020 nm 042 054 026 026 028 051
SSA
440 nm 095 090 090 084 088 089676 nm 096 092 091 087 091 094869 nm 095 091 090 085 090 0951020 nm 095 090 090 084 090 095
REFR
440 nm 144 150 143 148 151 150676 nm 144 151 146 151 153 153869 nm 144 152 147 152 154 1531020 nm 144 151 148 153 154 152
REFI
440 nm 0007 0015 0012 0023 0014 0007676 nm 0005 0009 0008 0014 0008 0003869 nm 0005 0009 0008 0015 0008 00031020 nm 0005 0010 0009 0016 0008 0003
ASYM
440 nm 074 071 071 068 069 072676 nm 070 065 064 062 064 068869 nm 067 062 061 061 063 0681020 nm 065 062 061 061 063 069
VolConF (120583m3120583m2) 016 017 011 009 008 007VolMedianRadF (120583m) 026 021 017 016 016 015
StdDevF 057 054 050 051 048 052VolConC (120583m3120583m2) 009 018 011 011 016 035VolMedianRadC (120583m) 294 260 273 279 271 235
StdDevC 055 057 060 063 065 060FFV 064 049 050 045 033 017SP () 5232 5783 5828 6427 2899 651
WV (gcm2) 218 113 134 064 091 091Number of records 917 1124 1188 1252 1394 857
Percentage to total number () 136 167 176 186 207 127
and Figure 4 represents the relationship between sphericityparameter and coarse fraction by volume
32 Characteristics of Aerosol Types
321 Fine Particle Nonabsorbing Aerosol As shown in Fig-ure 2(d) cluster 1 exhibits the largest value of fine fractionby volume (FFV) that is 064 Obviously cluster 1 is fineparticle dominated aerosol Figure 3 shows the particle sizedistribution of cluster 1 described by a fine radius andstandard deviation of 026120583m and 057 and a coarse radiusand standard deviation of 294 120583m and 055 Moreover thistype of aerosol displays clearly nonabsorbing properties asthe single scattering albedos (SSAs) at four wavelengths(440 676 869 and 1020 nm) are 095 096 095 and 095(Figure 2(b)) respectively As shown in Table 2 the real partsof refractive index (REFRs) of cluster 1 at fourwavelengths are144 which is lower than other clusters According to Brown(1984) lower value of real part of refractive index is possibly
associated with high hygroscopicity of aerosol [33] Whenthere is high relative humidity water is usually attached tothe surface of particle As a result it will enhance the abilityof particle scattering incident waveThis analysis is consistentwith large value of water vapor (218 gcm2) of cluster 1 whichis shown in Figure 2(f) Moreover this type of aerosol isusually observed in summer (Section 34) in which seasonBeijing is less affected by smoke and dust Therefore it islikely that these fine nonabsorbing particles are mainly theemissions of heavy traffic and industrial factories
There are similarities between cluster 1 and polluted con-tinental classified by Omar et al (2005) [14] But the formeris stronger in particle scattering and higher in fine fraction byvolume Consequently we identified cluster 1 as fine particlenonabsorbing aerosol To verify the relationship between fineparticle nonabsorbing aerosol and air pollution of Beijing wecollected the data of air quality index (AQI) over one year(October 2013 to October 2014) of Beijing (because the dataof AQI is not open to public only 2013-2014 is available to
Advances in Meteorology 5
400 500 600 700 800 900 1000 110000
03
06
09
12
15
Aero
sol o
ptic
al d
epth
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(a) AOD
400 500 600 700 800 900 1000 1100
084086088090092094096
Sing
le sc
atte
ring
albe
do
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(b) SSA
400 500 600 700 800 900 1000 1100
060063066069072075
Asy
mm
etry
par
amet
er
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(c) ASYM
Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 601
02
03
04
05
06
07
Fine
frac
tion
by v
olum
e
Cluster
(d) FMF
010203040506070
Sphe
ricity
par
amet
er (
)
ClusterCluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
(e) SP
00
04
08
12
16
20
24
Cluster
Wat
er v
apor
(g=G
2)
Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
(f) WV
Figure 2 The characteristics of six cluster centers (a) aerosol optical depth (b) single scattering albedo (c) asymmetry parameter (d) finefraction by volume (e) sphericity parameter (f) water vapor
us) The AQI is an index that indicates the pollution level ofthe atmosphere while a higher AQI value means a heavieratmospheric pollution The value of AQI level increase fromone to six represents air qualities of excellent good slightpollution moderate pollution heavy pollution and severepollution respectively Before comparison we collocate AQIdata with classified aerosol types and obtain 88 matchupsFigure 5 shows numbers of each aerosol type observed inper-AQI level over 2013-2014 at Beijing site It can be seenfrom Figure 5 that fine particle nonabsorbing aerosol is mostfrequently observed at AQI level 4 (moderate pollution)followed by AQI-levels 5 (heavy pollution) and 6 (severepollution)
322 Fine Particle Moderately Absorbing Aerosol The finefraction by volume of cluster 2 and cluster 3 is 049 and 050respectively Besides the SSAs at four wavelengths (440 676869 and 1020 nm) are 090 092 091 and 090 for cluster2 and 090 091 090 and 090 for cluster 3 Therefore bothclusters are considered as fine particle moderately absorbingaerosols (cluster 2 and cluster 3 named as fine-MA1 and fine-MA2 resp)
The fine-MA1 and fine-MA2 aerosols are mainly dis-tinguished by real parts of the refractive index (REFRs)and fine and coarse particle volume concentrations TheREFRs of fine-MA1 at four wavelengths are 150 151 152and 151 respectively Correspondingly the REFRs of fine-MA2 are much lower with REFRs being 143 146 147
6 Advances in Meteorology
000
005
010
015
020
025
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
Figure 3 The particle size distributions of six cluster centers
and 148 According to Bohren and Huffman (1983) [34]dry particle basically has a higher REFR usually being15ndash16 That is consistent with the fact that the water vaporof fine-MA1 (113 gcm2) is lower than that of fine-MA2(134 gcm2) which can be seen in Figure 2(f) Additionallythe fine particle volume concentration of fine-MA1 is thelargest one in six aerosol types (017 120583m3120583m2) and thecoarse particle volume concentration is also found being highlevel (018 120583m3120583m2) Correspondingly both fine and coarseparticle volume concentration of fine-MA2 are much lower(both are 011 120583m3120583m2) than fine-MA1
Compared with Omar et al (2005) [14] fine-MA1 is likepolluted continental aerosol and dirty pollution aerosol Butthe absorption is lower than the former and higher than thelatter It can be seen from Figure 5 that fine-MA1 is the mostfrequently observed at AQI levels 5 and 6 Besides fine-MA1is just a little lower than fine particle nonabsorbing aerosol atAQI level 4 Furthermore fine-MA1 is hardly observed atAQIlevels 1 (excellent) and 2 (good) On the contrary fine-MA2 ischaracterized by a low optical depth (Figure 2(a)) and a highfrequency of incidence at lowAQI level when the atmosphereis expected to be relatively clean The characteristics aremore likely the backgroundrural aerosol classified by Omaret al (2005) [14] Consequently fine-MA2 is possibly thebackground aerosol of Beijing
323 Fine Particle Highly Absorbing Aerosol Cluster 4 is ofinterest in view of its high absorption The SSAs at fourwavelengths (440 676 869 and 1020 nm) are 084 087 085and 084 (Figure 2(b)) respectively Besides the fine fractionby volume is 045 (Figure 2(d)) The fine particle volumeconcentration is 009120583m3120583m2 and the coarse particle vol-ume concentration is 011120583m3120583m2 In addition Figure 3shows the particle size distribution of cluster 4 describedby a fine radius and standard deviation of 016 120583m and 051and coarse radius and standard deviation of 279 120583m and
063 Therefore we identified cluster 4 as fine particle highlyabsorbing aerosol
Figure 2(e) shows that the sphericity parameter (thehigher value indicates the particle is closer to sphericity)of fine particle highly absorbing aerosol exhibits the largestvalue (6427) Moreover it can be clearly seen from Fig-ure 4(d) that this kind of aerosol has much more recordswith sphericity parameter higher than 90 It demonstratesthat fine particle highly absorbing aerosol is more likelya spherical aerosol These characteristics are like biomassburning aerosol classified by Omar et al (2005) [14] Addi-tionally it can be seen from Table 2 that the real parts ofthe refractive index at a high level are 148 151 152 and153 at four wavelengths According to the study result ofDubovik et al (2002) [11] higher REFRs and lower SSAs arepossibly associated with high concentrations of black carbonFurthermore this kind of aerosol usually occurs in winter(Section 34) in which season coal is frequently used forheating supply in northern China [23 28] Consequently fineparticle highly absorbing aerosol is possibly mainly producedby fossil fuel combustion
324 Polluted Dust Cluster 5 has the largest number ofrecords with 207 of total records being classified inthis type The SSAs at four wavelengths (440 676 869and 1020 nm) are 088 091 090 and 090 (Figure 2(b))respectively Besides the fine fraction by volume is 033(Figure 2(d)) The fine particle volume concentration is008120583m3120583m2 and coarse particle volume concentration is016 120583m3120583m2 Figure 3 shows the particle size distributionof cluster 5 described by a fine radius and standard devi-ation of 016 120583m and 048 and coarse radius and standarddeviation of 271 120583m and 065 As shown in Figure 4(e)most records of cluster 5 fall within sphericity parameterlower than 50 and coarse fraction by volume higher than05 Thus it indicates that cluster 5 is nonspherical coarseparticle dominated aerosol According to Okada et al (2001)nonspherical coarse particle is possibly contributed by dust[35] However there is no significant difference between SSAsat 440 nm and longer wavelengths But previous studies showthat when wavelength changes from 440 nm to longer theSSA will increase largely (Section 325) for dust aerosolBesides this kind of aerosol is frequently observed in spring(Section 34) in which season Beijing is affected by long-range transported dust Thus cluster 5 is possibly the dustmixes with anthropogenic aerosol As shown in Figure 4(e)cluster 5 also has some recordswith high sphericity parameteror low coarse fraction by volume Consequently we identifiedcluster 5 as polluted dust It should be noted that this polluteddust aerosol is the first time classified as a unique type bycluster analysis method The fact that polluted dust can beidentified by FCM demonstrates the practicability of fuzzyclustering in aerosol classification
325 Desert Dust Cluster 6 is significantly characterized bysmallest fine fraction by volume (017) and lowest sphericityparameter (651) Besides it shows the smallest fine particlevolume concentration (008120583m3120583m2) and largest coarseparticle volume concentration (016 120583m3120583m2) It indicates
Advances in Meteorology 7
0 25 50 75 10000
02
04
06
08
10C
oars
e fra
ctio
n by
vol
ume
Spherical parameter
(a) Cluster 1
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(b) Cluster 2
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(c) Cluster 3
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
eSpherical parameter
(d) Cluster 4
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(e) Cluster 5
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(f) Cluster 6
Figure 4 Relationships of sphericity parameter and coarse fraction by volume
that this is a completely coarse particle dominated aerosolIn addition as shown in Figure 4(f) most records of cluster6 concentrate in the range of sphericity parameter lowerthan 15 and coarse fraction by volume higher than 06 Itdemonstrates that cluster 6 is completely nonspherical coarseparticles dominated aerosol Moreover the SSAs of cluster 6exhibit a distinct feature with lower value at 440 nm (089)and a sharp increase to 094 095 and 095 at 660 nm 896 nmand 1020 nm (Figure 2(b)) respectively This is consistentwith the characteristic of desert dust derived by Levy et al(2007) [15] and Qin and Mitchell (2009) [16] Furthermoreaccording to Dubovik et al (2002) desert dust is possiblyabsorbed in the short spectral region due to the containediron oxide hematite [11] The SSA of dust particle might belower at short wavelength regions when it contains the FeO2Consequently these characteristics clearly certify that cluster6 is desert dust aerosol
326 Comparison with Other Results To assess the proposedfuzzy methodology we take a comparative study of our resultwith other published researches in aerosol classification
Considering study area and methodology two publishedresults are chosen global aerosol types by Omar et al (2005)[14] and East Asian aerosol types by Lee and Kim (2010) [17]
(1) Comparison with Global Aerosol TypesOmar et al (2005)applied cluster analysis to global ground-based observationsand obtained six categories backgroundrural industrialpollution dirty pollution biomass burning desert dustand polluted marine [14] Their characteristics are shownin Table 3 Five of the six aerosol types we classifiedshow similarities with Omarrsquos global aerosol types they arefine particle nonabsorbing aerosol to polluted continentalaerosol fine-MA2 to background aerosol fine particle highlyabsorbing aerosol to biomass burning aerosol and polluteddust and desert dust to dust
Figures 6 and 7 demonstrate comparisons of characteris-tics of aerosol types between global and Beijing aerosol typesincluding particle size distributions single scattering albedoand asymmetry parameter Particle size distributions (Fig-ure 6) show that the values of 119889119881119889 ln 119903 varied significantlybetween two researches The particle volume concentration
8 Advances in Meteorology
Table 3 Characteristics of global aerosol types by Omar et al (2005)
Properties (676 nm) Polluted continental Backgroundrural Biomass burning Dirty pollution Dust MarineSSA 092 088 080 072 093 093REFR 14098 14494 15202 14104 14520 13943REFI 00063 00092 00245 00337 00036 00044ASYM 0612 0580 0603 0594 0668 0711FFV 053 038 033 049 022 026
1 2 3 4 5 60
2
4
6
8
10
Num
ber o
f day
s
AQI-level
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 5 Numbers of each cluster observed in per-AQI level from2013 to 2014 at Beijing site
of Beijing is much higher than global aerosol types which isbecause Beijing usually suffers high concentration of aerosol[9] The SSAs of three fine particle aerosols of Beijing are095 091 and 087 which is larger than the value of globalaerosol 092 088 and 08 Interestingly all particle volumeconcentration SSA and asymmetry parameter of global dustare larger than polluted dust and smaller than desert dustwhich indicates that tow dust aerosol may be more ldquopurerdquothan dust in global aerosol types
(2) Comparison with East Asian Aerosol Types Lee and Kim(2010) performed a cluster analysis for AERONET recordsof East Asia which were divided into four fine-mode types(CAT (category) 1 to 4) and two coarse-mode types (CAT5and CAT6) [17] In addition the two coarse-modes werereferred to as dusty aerosols and the others were considereda mixed type of pollution Four of the six aerosol types inBeijing show similarities with Leersquos in East Asia they are fineparticle nonabsorbing aerosol to CAT2 fine particle highlyabsorbing aerosol to CAT3 and polluted dust and desert dustto two coarse-modes Table 4 shows the characteristics ofeach aerosol type identified by Lee et al (2010)
Figure 8 displays comparisons of characteristics of fouraerosol types between East Asia and Beijing Figures 8(a) and8(g) show that fine particle nonabsorbing aerosol of Beijingis stronger in particle scattering and lower in particle volume
concentration But the wavelength dependence of asymmetryparameter demonstrates similarities (Figure 8(d))The CAT3in East Asia and fine particle highly absorbing aerosol inBeijing exhibit similar particle size distributions and asym-metry parameter But the absorption of the latter is muchstronger than the former (Figure 8(b)) owing to the fossilfuel combustion in Beijing (see Section 323) It can beseen from Figures 8(c) 8(f) and 8(i) that the properties ofdesert dust in Beijing and CAT5 in East Asia are exactlysimilar to each other Meanwhile higher SSAs and particlevolume concentration of desert dust compared with CAT6probably due to sites close to dust source regions (like Yulinselected by Lee et al (2010)) were not included in thispaper It is interesting to note the pronounced differencesbetween polluted dust in Beijing and CAT5CAT6 in EastAsia indicating apparently polluted dust a unique aerosoltype
33 Variation of Optical Characteristics with MembershipDegrees The optical properties discussed in Section 32 arerepresented by the center values of each cluster But thedifferences between records at edge and near center ofcluster are inevitable Particularly for multiparticles mixtureaerosol only representing the aerosol by the center of clusterwill produce large errors As described in Section 22 themembership degree indicates the confidence degree of onerecord belonging to clusters which is an optimal parameterto analyze these particles at boundary ofmulticlustersThere-fore characteristics of aerosol (optical and microphysical)with different membership degree intervals are investigatedhere It should be noted thatwe focus on the internal variationof optical properties for each aerosol type
The membership degrees (MD) of six aerosol types aredivided into five intervals 017 le MD lt 027 027 le MD lt037 037 leMD lt 047 047 leMD lt 057 and 057 leMDWechose the value of 017 because when membership degree ofthe record to one type is larger than 16 (as there are six types)it means that this recordmay belong to that type Andwe takethe step of 01 to make sure that there is a similar amount ofrecords within each interval The numbers of records withineach interval are listed in Table 5
Figure 9 shows the AODs of six aerosol types varyingwith the membership degree and their mean values withineach membership degree interval at wavelength of 440 nmFine particle nonabsorbing aerosol (cluster 1 Figure 9(a))and fine-MA1 (cluster 2 Figure 9(b)) show large range ofAOD variation (from 0 to 4) The large range is reasonablebecause both types are frequently observed at polluted days(discussed in Section 32) Nevertheless the other four
Advances in Meteorology 9
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 1Continental
dV
dFH
r(
G3
G2)
(a)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 3Background
dV
dFH
r(
G3
G2)
(b)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 4Biomass
dV
dFH
r(
G3
G2)
(c)
000
005
010
015
020
025
030
Radius (G)
10minus1 100 101
Cluster 5Cluster 6Dust
dV
dFH
r(
G3
G2)
(d)
Figure 6 Size distribution comparison with global aerosol types
aerosol types show relatively smaller variation ranges (from0 to 2) And there are fewer that occurred during polluteddays Furthermore the AODs obviously concentrate in thelow varying range with increasing membership degree for allaerosol types
Figure 10 shows the mean values of AOD within the fivemembership degree intervals at four wavelengths (440 676869 and 1020 nm) Figures 10(a) and 10(b) show that theaverage AODs of fine particle moderately absorbing aerosolfor all wavelengths exhibit relatively flat behaviors Howeverthe other five aerosol types exhibit clearly declination withthe increase of themembership degree intervalMoreover thetrends are nearly the same between different wavelengths foreach aerosol type
The SSAs (440 nm) of six aerosol types varying withmembership degree and their mean values within eachmembership degree interval are plotted in Figure 11 Same
as AODs points of SSA concentrate in the low varyingrange with increasing membership degree for all aerosoltypes It can be seen from Figure 11(a) that the SSAs offine particle nonabsorbing aerosol (cluster 1) fall within therange between 09 and 10 This characteristic is consistentwith the nonabsorbing aerosol (strong scattering) of cluster1 Besides it is interesting to find that there are somestrongly absorbing particles (SSA lt 08) at membershipdegrees between 027 and 047 in Figure 11(d) (cluster 4) Asdiscussed in Section 323 these particles are probably theblack carbon which are frequently observed in Beijing withhigh absorption [23 28]
The mean values of SSA within the five membershipdegree intervals at four wavelengths (440 676 869 and1020 nm) are plotted in Figure 12 Means of SSA showclear descending trend in Figure 12(a) which indicates thatwith the increase of membership degree scattering ability of
10 Advances in Meteorology
Cluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
010203040506070809101112
Sing
le sc
atte
ring
albe
do
Aerosol typeCluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
00
01
02
03
04
05
06
07
08
Asy
mm
etry
fact
or
Aerosol type
This paperOmar et al
This paperOmar et al
Figure 7 SSA and ASYM comparison with global aerosol types
Table 4 Characteristics of East Asian aerosol types by Lee et al (2010)
Properties CAT11 CAT2 CAT3 CAT4 CAT5 CAT6
SSA
440 nm 0915 0927 0904 0908 0893 0900676 nm 0927 0941 0908 0914 0939 0957869 nm 0918 0933 0897 0907 0945 09641020 nm 0912 0928 0889 0903 0948 0966
REFR
440 nm 1468 1478 1441 1454 1508 1549676 nm 1480 1483 1458 1472 1535 1549869 nm 1485 1483 1468 1482 1536 15371020 nm 1481 1476 1468 1481 1528 1525
REFI
440 nm 00119 00099 00127 00112 0007 00049676 nm 00086 00074 001 00088 00037 00024869 nm 00088 00078 00102 0009 00036 000231020 nm 00091 0008 00104 00092 00036 00025
ASYM
440 nm 0721 0729 0716 0713 073 0748676 nm 0664 0685 0654 0655 0693 0714869 nm 0636 0657 0626 0635 0691 07071020 nm 0626 0643 0618 0634 0696 0707
1CAT category
fine particle nonabsorbing aerosol (cluster 1) is enhancingThis characteristic confirms that the higher the membershipdegree the purer the nonabsorbing aerosol Figures 12(b) and12(c) show the mean values of SSA obviously descendingwhen membership degree increases from MDI1 to MDI2However the changes are hardly observed when member-ship degree interval increases from MDI2 to MDI5 Thesecharacteristics demonstrate that records at the edge of cluster2 (fine-M1) and cluster 3 (fine-M2) have stronger scattering(or lower absorbing) than those near center Figure 12(d)exhibits clearly ascending trend of SSA means which is amanifestation of the reasonability to identify cluster 4 ashighly absorbing aerosol The ascending trend also indicatesthat the higher membership degree denotes the purer highlyabsorbing aerosol
Figures 12(e) and 12(f) show some similarities betweenpolluted dust and desert dust at wavelengths of 440 nmFirstly mean values of SSA slightly decrease with the increaseof membership degree interval Furthermore mean values ofSSA at 440 nm are obviously lower than other wavelengthsAs discussed in Sections 324 and 325 these SSA trendsagree well with the relationships between SSA and dustcompositions which certify our identification that bothcluster 5 (polluted dust) and cluster 6 (desert dust) aredust-related aerosols However there are some differencesbetween two clusters SSAs at 440 nm are distinctly smallerthan at other wavelengths Besides Figures 12(e) and 12(f)show different trends at wavelengths longer than 440 nm theformer displays descending trend while the latter shows noclear variation These differences can be attributed to cluster
Advances in Meteorology 11
400
500
600
700
800
900
1000
1100
085
090
095
100Si
ngle
scat
terin
g al
bedo
Wavelengths (nm)
This paper cluster 1Lee et al CAT2
(a)
085
090
095
100
Sing
le sc
atte
ring
albe
do
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(b)
085
090
095
100
Sing
le sc
atte
ring
albe
do
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(c)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 1Lee et al CAT2
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(d)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(e)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(f)
000
005
010
015
020
025
This paper cluster 1Lee et al CAT2
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(g)
000
005
010
015
This paper cluster 4Lee et al CAT3
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(h)
00010203040506070809
Radius (G)
10minus1 100 101
This paper
This papercluster 6
cluster 5Lee et al
Lee et alCAT5
CAT6
dV
dFH
r(
G3
G2)
(i)
Figure 8 Comparison with East Asian aerosol types
Table 5 The record numbers within each membership degree interval
Membership degree intervals (MDI) C11 C2 C3 C4 C5 C6MDI1 017 le MD2 lt 027 156 131 60 117 43 86MDI2 027 leMD lt 037 508 569 298 603 404 308MDI3 037 leMD lt 047 305 326 240 456 413 234MDI4 047 leMD lt 057 115 128 197 181 275 153MDI5 057 leMD 40 34 122 37 117 761C cluster 2MD membership degree
12 Advances in Meteorology
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 9 AODs at 440 nmwith membership degrees asterisk represents mean value of AOD in each interval and the vertical bars representthe standard deviation
5 (polluted dust) containing absorbing anthropogenic aerosol(such as black carbon) while cluster 6 is dominated onlyby dust These characteristics validate that it is reasonable toidentify cluster 5 as polluted dust and cluster 6 as desert dust
Figure 13 is scatterplot of sphericity parameter versusmembership degree for six aerosol types Figures 13(a) 13(b)13(c) and 13(d) show high variation of sphericity parameterswhich indicates that the first four aerosol types show nocorrelation between sphericity parameters and membershipdegree However Figure 13(e) exhibits clearly descendingtrends and all mean values of sphericity parameter are lower
than 40 Moreover records in Figure 13(f) are more con-centrated with 95 of them falling below 20 These indicatethat both cluster 5 and cluster 6 are close to nonsphericityBecause most dust types are nonsphericity particles [35]these analyses again confirm our identification of polluteddust (cluster 5) and desert dust (cluster 6)
The above analyses confirm reasonability of our results ofclustering and identification of aerosol types Moreover theinternal variation of optical properties of aerosol type can bewell investigated with the help ofmembership degree interval(MDI)
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
2 Advances in Meteorology
biomass burning aerosol from agriculture and black carbonfrom industrial emissions [13 19]) the temporal distributionsand characteristics are rather complex Consequently meanvalues of long-term aerosol properties observations are insuf-ficient to represent characteristics of aerosol types [14]
To overcome this shortcoming the others [14ndash17] takeadvantage of clustering to classify aerosol types Clusteringis a statistical tool used for grouping a set of objects intoseveral clusters using predefined variables so that objectsin the same cluster are more similar to each other thanthose in other clusters Omar et al (2005) applied clusteranalysis to global ground-based observations and obtainedsix categories backgroundrural industrial pollution dirtypollution biomass burning desert dust and polluted marine[14] In addition Levy et al (2007) performed a ldquosubjectiverdquocluster analysis for AERONET records which were dividedinto three fine-sized dominated spherical types (typicallythe ldquolowrdquo ldquomediumrdquo and ldquohighrdquo absorbing types) and onecoarse-sized dominated spheroid type [15] Besides Qinand Mitchell (2009) applied Locally Scaled Density BasedClustering (LSDBC) algorithm to classify Australian aerosolinto four classes including aged biomass burning smoke freshsmoke coarse dust and super-absorptive aerosol [16] Theseresults show that cluster analysis is suitable for classifyingaerosol types However the limitation of these classificationsis that the values of cluster center are used to representthe characteristics of corresponding aerosol types As amatter of fact the differences between records at edge ofcluster and those near center are inevitable Particularly formultiparticles mixture aerosol records are usually on theboundaries between several clusters and only representingthem by the center of each cluster is unreasonable [20]To overcome this limitation Wu and Zeng (2014) appliedGustafson-Kessel fuzzy clustering algorithm to identify theoptical properties of pure dust aerosol type [20] Comparedwith aforementioned clustering algorithm not only center ofcluster but also a significant parameter named membershipdegree (between 0 and 1) is provided by fuzzy clustering Themembership degree is used to describe the confidence degreeof one record belonging to a cluster The records near thecenter of cluster will have larger value of membership (higherdegree of confidence) than those at edge Thus membershipdegree will be helpful to analyze the internal variation ofoptical properties within each aerosol type The capability offuzzy cluster analysis to obtain aerosol properties in singletype dominant regions has been documented whereas itscapacity to obtain aerosol properties in regions which arefrequently influenced by various sources of aerosols still hasnot been well reported
Beijing one of the largest megacities in the worldhas been suffering severe issues of aerosols loading fordecades because of rapid development of both economy andpopulation It is famous for its complex aerosol particlesand high proportion of anthropogenic aerosols Due to itsimportance to climate and environment many studies werecarried out to obtain the aerosol properties in this regionThe researchers are interested in (i) characteristics of fine orcoarse particle aerosols [21 22] (ii) components and sources[23 24] (iii) aerosol optical properties [25ndash29] Nevertheless
the dominant aerosol types and their temporal variation stillhave not been well reported But the definite aerosol typesare important to improve radiative forcing estimation andsatellite aerosol retrieval To this end this paper focuses onclassifying identical aerosol types and studying their opticalproperties
In this study long-term (14 years) observations ofAERONET are used to determine dominant aerosol typesand their optical properties over Beijing One of the mostwidely used fuzzy clustering algorithms fuzzy 119888-mean (FCM)algorithm [30] is applied to classify distinct aerosol typesThis is the first comprehensive study investigating aerosoltypes and their temporal distribution in Beijing This workwill provide a primary parameter for the estimation ofaerosol radiative forcing and retrieval of aerosol propertiesfrom satellite aerosol as well as reference for methodologyof identifying aerosol types in other regions The paper isstructured as follows Section 2 briefly describes the datasetand method used in this study a detailed analysis of aerosolsproperties is conducted in Section 3 at last the conclusionsare drawn in Section 4
2 Data and Methodology
21 Description of Aerosol Sites and Data The AERONET isa global network of ground-based CIMEL sun-sky radiome-ter AERONET inversion products include a comprehensivedata set of aerosol optical and microphysical propertiesaerosol optical depth (AOD) at 340 380 440 500 675 870and 1020 nm the single scattering albedo (SSA) complexrefractive indices asymmetry parameter (ASYM) at fourwavelengths (440 676 869 and 1020 nm) and the parametersof particle size distribution [31 32]
In this study three AERONET sites (surrounding BeijingCity) are selected they are Beijing Xianghe and XinglongBy choosing these sites over 14 years (2001ndash2014) a totalof 6732 records are obtained It should be noted that wecollect records of AERONET ldquoAll Pointsrdquo Level 20 inver-sion products which can be obtained from httpaeronetgsfcnasagov Figure 1 and Table 1 show the detailed infor-mation of the selected sites
The following 22 parameters obtained from AERONETinversion products are applied in the cluster analysis
(a) Single scattering albedo (SSA) at 440 676 869 and1020 nm
(b) Real part of refractive index (REFR) at 440 676 869and 1020 nm
(c) Imaginary part of refractive index (REFI) at 440 676869 and 1020 nm
(d) Asymmetry parameter (ASYM) at 440 676 869 and1020 nm
(e) Parameters of particle size distribution finecoarseparticle volume concentration (VolConFVolConC)finecoarse particle volume median radius (VolMedi-anRadFVolMedianRadC) finecoarse particle stan-dard deviation (StdDevFStdDevC)
Advances in Meteorology 3
50∘
45∘
40∘
35∘
30∘
25∘
20∘
50∘
45∘
40∘
35∘
30∘
25∘
20∘
75∘
80∘
85∘
90∘
95∘
100∘
105∘
110∘
115∘
120∘
125∘
130∘
135∘
75∘
80∘
85∘
90∘
95∘
100∘
105∘
110∘
115∘
120∘
125∘
130∘
135∘
(a)
Xianghe
Beijing Xinglong
Hebei ProvinceBohai Sea
TianjinCity
BeijingCity
43∘
42∘
41∘
40∘
39∘
38∘
37∘
36∘
43∘
42∘
41∘
40∘
39∘
38∘
37∘
36∘
114∘
115∘
116∘
117∘
118∘
119∘
114∘
115∘
116∘
117∘
118∘
119∘
(b)
Figure 1 Geographical location of the selected three AERONET sites
Table 1 Information of the selected three AERONET sites
Site name Observing period Number of records Longitude(degree)
Latitude(degree)
Elevation(meter)
Beijing 2001ndash2014 3418 116381 39977 92Xianghe 2001 2004ndash2012 2797 116962 39754 36Xinglong 2006ndash2012 517 117578 40396 970
It should be noted that the aerosol optical depth (AOD)fine fraction by volume (FFV) sphericity parameter (SP) andwater vapor (WV) are also used in later discussions Howeverthese four parameters are not applied in clustering
22 Fuzzy Clustering Cluster analysis is a statistical toolused for grouping the data sets into several clusters based onpredefined variables [14] The basic theory of cluster analysisis records in the same cluster which are more similar to eachother than to those in other clusters In fuzzy clustering everyrecord has a degree of belonging to each cluster rather thanjust completely belonging to one cluster [20] This degreeis represented by a parameter named membership degree(between 0 and 1) which indicates the confidence degreeof one record belonging to a cluster The records on theboundary of cluster will have smaller values than those nearthe center
In this study fuzzy 119888-mean (FCM) algorithm one ofthe most widely used fuzzy clustering algorithms is appliedGiven the data set containing 119899 records119883 = 1199091 1199092 119909119899and supposing there are 119888 clusters with the centers 119881 =V1 V2 119881119888 The FCM aims to minimize the objectivefunction
min 119869 (119906 V) =119899
sum119896=1
119888
sum119894=1
1205831198981198941198961198892119894119896 (1)
where 1198892119894119896 = 119909119896minusV1198942 represents the distance between record
119909119896 and center V119894 And 120583119898119894119896 is the degree to which record 119909119896belongs to cluster V119894 and is defined as
119906119894119896 =1
sum119888119895 (1003817100381710038171003817119909119896 minus V119894
1003817100381710038171003817 10038171003817100381710038171003817119909119896 minus V119895
10038171003817100381710038171003817)2(119898minus1) (2)
For a certain record sum119888119894 119906119894119896 = 1 The fuzzifier 119898 deter-mines the level of cluster fuzziness meaning the overlappingdegree of the cluster and 119898 ge 1 The center of the cluster isdefined as
V119894 =sum119899119896=1 119906
119898119894119896119909119896
sum119899119896=1 119906119898119894119896 (3)
The step of FCM is described as follows Firstly a randomfunction is used to determine the initial center of eachcluster Calculate membership degree and objective functionbased on functions (1) and (2) Meanwhile centers of eachcluster are calculated based on function (3) Repeat untilthe algorithm has converged that is the membership degreechange between two iterations is no more than the givensensitivity threshold It should be noted that before clusteringthe 22 parameters are normalized by the standard deviationto ensure that each parameter makes the same contributionto the distance calculation
3 Results and Discussions
31 Results By applying the FCM clustering a total of 6732records are classified into six clusters The characteristics ofoptical and microphysical properties of each cluster centerare shown in Table 2 Figure 2 displays aerosol optical depthsingle scattering albedo asymmetry parameter fine fractionby volume sphericity parameter and water vapor of eachcluster center Figure 3 shows the particle size distributions
4 Advances in Meteorology
Table 2 Results of the fuzzy cluster analysis
Properties Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
AOD
440 nm 125 156 079 076 073 087676 nm 077 094 045 043 044 064869 nm 054 066 031 031 033 0551020 nm 042 054 026 026 028 051
SSA
440 nm 095 090 090 084 088 089676 nm 096 092 091 087 091 094869 nm 095 091 090 085 090 0951020 nm 095 090 090 084 090 095
REFR
440 nm 144 150 143 148 151 150676 nm 144 151 146 151 153 153869 nm 144 152 147 152 154 1531020 nm 144 151 148 153 154 152
REFI
440 nm 0007 0015 0012 0023 0014 0007676 nm 0005 0009 0008 0014 0008 0003869 nm 0005 0009 0008 0015 0008 00031020 nm 0005 0010 0009 0016 0008 0003
ASYM
440 nm 074 071 071 068 069 072676 nm 070 065 064 062 064 068869 nm 067 062 061 061 063 0681020 nm 065 062 061 061 063 069
VolConF (120583m3120583m2) 016 017 011 009 008 007VolMedianRadF (120583m) 026 021 017 016 016 015
StdDevF 057 054 050 051 048 052VolConC (120583m3120583m2) 009 018 011 011 016 035VolMedianRadC (120583m) 294 260 273 279 271 235
StdDevC 055 057 060 063 065 060FFV 064 049 050 045 033 017SP () 5232 5783 5828 6427 2899 651
WV (gcm2) 218 113 134 064 091 091Number of records 917 1124 1188 1252 1394 857
Percentage to total number () 136 167 176 186 207 127
and Figure 4 represents the relationship between sphericityparameter and coarse fraction by volume
32 Characteristics of Aerosol Types
321 Fine Particle Nonabsorbing Aerosol As shown in Fig-ure 2(d) cluster 1 exhibits the largest value of fine fractionby volume (FFV) that is 064 Obviously cluster 1 is fineparticle dominated aerosol Figure 3 shows the particle sizedistribution of cluster 1 described by a fine radius andstandard deviation of 026120583m and 057 and a coarse radiusand standard deviation of 294 120583m and 055 Moreover thistype of aerosol displays clearly nonabsorbing properties asthe single scattering albedos (SSAs) at four wavelengths(440 676 869 and 1020 nm) are 095 096 095 and 095(Figure 2(b)) respectively As shown in Table 2 the real partsof refractive index (REFRs) of cluster 1 at fourwavelengths are144 which is lower than other clusters According to Brown(1984) lower value of real part of refractive index is possibly
associated with high hygroscopicity of aerosol [33] Whenthere is high relative humidity water is usually attached tothe surface of particle As a result it will enhance the abilityof particle scattering incident waveThis analysis is consistentwith large value of water vapor (218 gcm2) of cluster 1 whichis shown in Figure 2(f) Moreover this type of aerosol isusually observed in summer (Section 34) in which seasonBeijing is less affected by smoke and dust Therefore it islikely that these fine nonabsorbing particles are mainly theemissions of heavy traffic and industrial factories
There are similarities between cluster 1 and polluted con-tinental classified by Omar et al (2005) [14] But the formeris stronger in particle scattering and higher in fine fraction byvolume Consequently we identified cluster 1 as fine particlenonabsorbing aerosol To verify the relationship between fineparticle nonabsorbing aerosol and air pollution of Beijing wecollected the data of air quality index (AQI) over one year(October 2013 to October 2014) of Beijing (because the dataof AQI is not open to public only 2013-2014 is available to
Advances in Meteorology 5
400 500 600 700 800 900 1000 110000
03
06
09
12
15
Aero
sol o
ptic
al d
epth
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(a) AOD
400 500 600 700 800 900 1000 1100
084086088090092094096
Sing
le sc
atte
ring
albe
do
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(b) SSA
400 500 600 700 800 900 1000 1100
060063066069072075
Asy
mm
etry
par
amet
er
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(c) ASYM
Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 601
02
03
04
05
06
07
Fine
frac
tion
by v
olum
e
Cluster
(d) FMF
010203040506070
Sphe
ricity
par
amet
er (
)
ClusterCluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
(e) SP
00
04
08
12
16
20
24
Cluster
Wat
er v
apor
(g=G
2)
Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
(f) WV
Figure 2 The characteristics of six cluster centers (a) aerosol optical depth (b) single scattering albedo (c) asymmetry parameter (d) finefraction by volume (e) sphericity parameter (f) water vapor
us) The AQI is an index that indicates the pollution level ofthe atmosphere while a higher AQI value means a heavieratmospheric pollution The value of AQI level increase fromone to six represents air qualities of excellent good slightpollution moderate pollution heavy pollution and severepollution respectively Before comparison we collocate AQIdata with classified aerosol types and obtain 88 matchupsFigure 5 shows numbers of each aerosol type observed inper-AQI level over 2013-2014 at Beijing site It can be seenfrom Figure 5 that fine particle nonabsorbing aerosol is mostfrequently observed at AQI level 4 (moderate pollution)followed by AQI-levels 5 (heavy pollution) and 6 (severepollution)
322 Fine Particle Moderately Absorbing Aerosol The finefraction by volume of cluster 2 and cluster 3 is 049 and 050respectively Besides the SSAs at four wavelengths (440 676869 and 1020 nm) are 090 092 091 and 090 for cluster2 and 090 091 090 and 090 for cluster 3 Therefore bothclusters are considered as fine particle moderately absorbingaerosols (cluster 2 and cluster 3 named as fine-MA1 and fine-MA2 resp)
The fine-MA1 and fine-MA2 aerosols are mainly dis-tinguished by real parts of the refractive index (REFRs)and fine and coarse particle volume concentrations TheREFRs of fine-MA1 at four wavelengths are 150 151 152and 151 respectively Correspondingly the REFRs of fine-MA2 are much lower with REFRs being 143 146 147
6 Advances in Meteorology
000
005
010
015
020
025
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
Figure 3 The particle size distributions of six cluster centers
and 148 According to Bohren and Huffman (1983) [34]dry particle basically has a higher REFR usually being15ndash16 That is consistent with the fact that the water vaporof fine-MA1 (113 gcm2) is lower than that of fine-MA2(134 gcm2) which can be seen in Figure 2(f) Additionallythe fine particle volume concentration of fine-MA1 is thelargest one in six aerosol types (017 120583m3120583m2) and thecoarse particle volume concentration is also found being highlevel (018 120583m3120583m2) Correspondingly both fine and coarseparticle volume concentration of fine-MA2 are much lower(both are 011 120583m3120583m2) than fine-MA1
Compared with Omar et al (2005) [14] fine-MA1 is likepolluted continental aerosol and dirty pollution aerosol Butthe absorption is lower than the former and higher than thelatter It can be seen from Figure 5 that fine-MA1 is the mostfrequently observed at AQI levels 5 and 6 Besides fine-MA1is just a little lower than fine particle nonabsorbing aerosol atAQI level 4 Furthermore fine-MA1 is hardly observed atAQIlevels 1 (excellent) and 2 (good) On the contrary fine-MA2 ischaracterized by a low optical depth (Figure 2(a)) and a highfrequency of incidence at lowAQI level when the atmosphereis expected to be relatively clean The characteristics aremore likely the backgroundrural aerosol classified by Omaret al (2005) [14] Consequently fine-MA2 is possibly thebackground aerosol of Beijing
323 Fine Particle Highly Absorbing Aerosol Cluster 4 is ofinterest in view of its high absorption The SSAs at fourwavelengths (440 676 869 and 1020 nm) are 084 087 085and 084 (Figure 2(b)) respectively Besides the fine fractionby volume is 045 (Figure 2(d)) The fine particle volumeconcentration is 009120583m3120583m2 and the coarse particle vol-ume concentration is 011120583m3120583m2 In addition Figure 3shows the particle size distribution of cluster 4 describedby a fine radius and standard deviation of 016 120583m and 051and coarse radius and standard deviation of 279 120583m and
063 Therefore we identified cluster 4 as fine particle highlyabsorbing aerosol
Figure 2(e) shows that the sphericity parameter (thehigher value indicates the particle is closer to sphericity)of fine particle highly absorbing aerosol exhibits the largestvalue (6427) Moreover it can be clearly seen from Fig-ure 4(d) that this kind of aerosol has much more recordswith sphericity parameter higher than 90 It demonstratesthat fine particle highly absorbing aerosol is more likelya spherical aerosol These characteristics are like biomassburning aerosol classified by Omar et al (2005) [14] Addi-tionally it can be seen from Table 2 that the real parts ofthe refractive index at a high level are 148 151 152 and153 at four wavelengths According to the study result ofDubovik et al (2002) [11] higher REFRs and lower SSAs arepossibly associated with high concentrations of black carbonFurthermore this kind of aerosol usually occurs in winter(Section 34) in which season coal is frequently used forheating supply in northern China [23 28] Consequently fineparticle highly absorbing aerosol is possibly mainly producedby fossil fuel combustion
324 Polluted Dust Cluster 5 has the largest number ofrecords with 207 of total records being classified inthis type The SSAs at four wavelengths (440 676 869and 1020 nm) are 088 091 090 and 090 (Figure 2(b))respectively Besides the fine fraction by volume is 033(Figure 2(d)) The fine particle volume concentration is008120583m3120583m2 and coarse particle volume concentration is016 120583m3120583m2 Figure 3 shows the particle size distributionof cluster 5 described by a fine radius and standard devi-ation of 016 120583m and 048 and coarse radius and standarddeviation of 271 120583m and 065 As shown in Figure 4(e)most records of cluster 5 fall within sphericity parameterlower than 50 and coarse fraction by volume higher than05 Thus it indicates that cluster 5 is nonspherical coarseparticle dominated aerosol According to Okada et al (2001)nonspherical coarse particle is possibly contributed by dust[35] However there is no significant difference between SSAsat 440 nm and longer wavelengths But previous studies showthat when wavelength changes from 440 nm to longer theSSA will increase largely (Section 325) for dust aerosolBesides this kind of aerosol is frequently observed in spring(Section 34) in which season Beijing is affected by long-range transported dust Thus cluster 5 is possibly the dustmixes with anthropogenic aerosol As shown in Figure 4(e)cluster 5 also has some recordswith high sphericity parameteror low coarse fraction by volume Consequently we identifiedcluster 5 as polluted dust It should be noted that this polluteddust aerosol is the first time classified as a unique type bycluster analysis method The fact that polluted dust can beidentified by FCM demonstrates the practicability of fuzzyclustering in aerosol classification
325 Desert Dust Cluster 6 is significantly characterized bysmallest fine fraction by volume (017) and lowest sphericityparameter (651) Besides it shows the smallest fine particlevolume concentration (008120583m3120583m2) and largest coarseparticle volume concentration (016 120583m3120583m2) It indicates
Advances in Meteorology 7
0 25 50 75 10000
02
04
06
08
10C
oars
e fra
ctio
n by
vol
ume
Spherical parameter
(a) Cluster 1
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(b) Cluster 2
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(c) Cluster 3
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
eSpherical parameter
(d) Cluster 4
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(e) Cluster 5
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(f) Cluster 6
Figure 4 Relationships of sphericity parameter and coarse fraction by volume
that this is a completely coarse particle dominated aerosolIn addition as shown in Figure 4(f) most records of cluster6 concentrate in the range of sphericity parameter lowerthan 15 and coarse fraction by volume higher than 06 Itdemonstrates that cluster 6 is completely nonspherical coarseparticles dominated aerosol Moreover the SSAs of cluster 6exhibit a distinct feature with lower value at 440 nm (089)and a sharp increase to 094 095 and 095 at 660 nm 896 nmand 1020 nm (Figure 2(b)) respectively This is consistentwith the characteristic of desert dust derived by Levy et al(2007) [15] and Qin and Mitchell (2009) [16] Furthermoreaccording to Dubovik et al (2002) desert dust is possiblyabsorbed in the short spectral region due to the containediron oxide hematite [11] The SSA of dust particle might belower at short wavelength regions when it contains the FeO2Consequently these characteristics clearly certify that cluster6 is desert dust aerosol
326 Comparison with Other Results To assess the proposedfuzzy methodology we take a comparative study of our resultwith other published researches in aerosol classification
Considering study area and methodology two publishedresults are chosen global aerosol types by Omar et al (2005)[14] and East Asian aerosol types by Lee and Kim (2010) [17]
(1) Comparison with Global Aerosol TypesOmar et al (2005)applied cluster analysis to global ground-based observationsand obtained six categories backgroundrural industrialpollution dirty pollution biomass burning desert dustand polluted marine [14] Their characteristics are shownin Table 3 Five of the six aerosol types we classifiedshow similarities with Omarrsquos global aerosol types they arefine particle nonabsorbing aerosol to polluted continentalaerosol fine-MA2 to background aerosol fine particle highlyabsorbing aerosol to biomass burning aerosol and polluteddust and desert dust to dust
Figures 6 and 7 demonstrate comparisons of characteris-tics of aerosol types between global and Beijing aerosol typesincluding particle size distributions single scattering albedoand asymmetry parameter Particle size distributions (Fig-ure 6) show that the values of 119889119881119889 ln 119903 varied significantlybetween two researches The particle volume concentration
8 Advances in Meteorology
Table 3 Characteristics of global aerosol types by Omar et al (2005)
Properties (676 nm) Polluted continental Backgroundrural Biomass burning Dirty pollution Dust MarineSSA 092 088 080 072 093 093REFR 14098 14494 15202 14104 14520 13943REFI 00063 00092 00245 00337 00036 00044ASYM 0612 0580 0603 0594 0668 0711FFV 053 038 033 049 022 026
1 2 3 4 5 60
2
4
6
8
10
Num
ber o
f day
s
AQI-level
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 5 Numbers of each cluster observed in per-AQI level from2013 to 2014 at Beijing site
of Beijing is much higher than global aerosol types which isbecause Beijing usually suffers high concentration of aerosol[9] The SSAs of three fine particle aerosols of Beijing are095 091 and 087 which is larger than the value of globalaerosol 092 088 and 08 Interestingly all particle volumeconcentration SSA and asymmetry parameter of global dustare larger than polluted dust and smaller than desert dustwhich indicates that tow dust aerosol may be more ldquopurerdquothan dust in global aerosol types
(2) Comparison with East Asian Aerosol Types Lee and Kim(2010) performed a cluster analysis for AERONET recordsof East Asia which were divided into four fine-mode types(CAT (category) 1 to 4) and two coarse-mode types (CAT5and CAT6) [17] In addition the two coarse-modes werereferred to as dusty aerosols and the others were considereda mixed type of pollution Four of the six aerosol types inBeijing show similarities with Leersquos in East Asia they are fineparticle nonabsorbing aerosol to CAT2 fine particle highlyabsorbing aerosol to CAT3 and polluted dust and desert dustto two coarse-modes Table 4 shows the characteristics ofeach aerosol type identified by Lee et al (2010)
Figure 8 displays comparisons of characteristics of fouraerosol types between East Asia and Beijing Figures 8(a) and8(g) show that fine particle nonabsorbing aerosol of Beijingis stronger in particle scattering and lower in particle volume
concentration But the wavelength dependence of asymmetryparameter demonstrates similarities (Figure 8(d))The CAT3in East Asia and fine particle highly absorbing aerosol inBeijing exhibit similar particle size distributions and asym-metry parameter But the absorption of the latter is muchstronger than the former (Figure 8(b)) owing to the fossilfuel combustion in Beijing (see Section 323) It can beseen from Figures 8(c) 8(f) and 8(i) that the properties ofdesert dust in Beijing and CAT5 in East Asia are exactlysimilar to each other Meanwhile higher SSAs and particlevolume concentration of desert dust compared with CAT6probably due to sites close to dust source regions (like Yulinselected by Lee et al (2010)) were not included in thispaper It is interesting to note the pronounced differencesbetween polluted dust in Beijing and CAT5CAT6 in EastAsia indicating apparently polluted dust a unique aerosoltype
33 Variation of Optical Characteristics with MembershipDegrees The optical properties discussed in Section 32 arerepresented by the center values of each cluster But thedifferences between records at edge and near center ofcluster are inevitable Particularly for multiparticles mixtureaerosol only representing the aerosol by the center of clusterwill produce large errors As described in Section 22 themembership degree indicates the confidence degree of onerecord belonging to clusters which is an optimal parameterto analyze these particles at boundary ofmulticlustersThere-fore characteristics of aerosol (optical and microphysical)with different membership degree intervals are investigatedhere It should be noted thatwe focus on the internal variationof optical properties for each aerosol type
The membership degrees (MD) of six aerosol types aredivided into five intervals 017 le MD lt 027 027 le MD lt037 037 leMD lt 047 047 leMD lt 057 and 057 leMDWechose the value of 017 because when membership degree ofthe record to one type is larger than 16 (as there are six types)it means that this recordmay belong to that type Andwe takethe step of 01 to make sure that there is a similar amount ofrecords within each interval The numbers of records withineach interval are listed in Table 5
Figure 9 shows the AODs of six aerosol types varyingwith the membership degree and their mean values withineach membership degree interval at wavelength of 440 nmFine particle nonabsorbing aerosol (cluster 1 Figure 9(a))and fine-MA1 (cluster 2 Figure 9(b)) show large range ofAOD variation (from 0 to 4) The large range is reasonablebecause both types are frequently observed at polluted days(discussed in Section 32) Nevertheless the other four
Advances in Meteorology 9
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 1Continental
dV
dFH
r(
G3
G2)
(a)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 3Background
dV
dFH
r(
G3
G2)
(b)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 4Biomass
dV
dFH
r(
G3
G2)
(c)
000
005
010
015
020
025
030
Radius (G)
10minus1 100 101
Cluster 5Cluster 6Dust
dV
dFH
r(
G3
G2)
(d)
Figure 6 Size distribution comparison with global aerosol types
aerosol types show relatively smaller variation ranges (from0 to 2) And there are fewer that occurred during polluteddays Furthermore the AODs obviously concentrate in thelow varying range with increasing membership degree for allaerosol types
Figure 10 shows the mean values of AOD within the fivemembership degree intervals at four wavelengths (440 676869 and 1020 nm) Figures 10(a) and 10(b) show that theaverage AODs of fine particle moderately absorbing aerosolfor all wavelengths exhibit relatively flat behaviors Howeverthe other five aerosol types exhibit clearly declination withthe increase of themembership degree intervalMoreover thetrends are nearly the same between different wavelengths foreach aerosol type
The SSAs (440 nm) of six aerosol types varying withmembership degree and their mean values within eachmembership degree interval are plotted in Figure 11 Same
as AODs points of SSA concentrate in the low varyingrange with increasing membership degree for all aerosoltypes It can be seen from Figure 11(a) that the SSAs offine particle nonabsorbing aerosol (cluster 1) fall within therange between 09 and 10 This characteristic is consistentwith the nonabsorbing aerosol (strong scattering) of cluster1 Besides it is interesting to find that there are somestrongly absorbing particles (SSA lt 08) at membershipdegrees between 027 and 047 in Figure 11(d) (cluster 4) Asdiscussed in Section 323 these particles are probably theblack carbon which are frequently observed in Beijing withhigh absorption [23 28]
The mean values of SSA within the five membershipdegree intervals at four wavelengths (440 676 869 and1020 nm) are plotted in Figure 12 Means of SSA showclear descending trend in Figure 12(a) which indicates thatwith the increase of membership degree scattering ability of
10 Advances in Meteorology
Cluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
010203040506070809101112
Sing
le sc
atte
ring
albe
do
Aerosol typeCluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
00
01
02
03
04
05
06
07
08
Asy
mm
etry
fact
or
Aerosol type
This paperOmar et al
This paperOmar et al
Figure 7 SSA and ASYM comparison with global aerosol types
Table 4 Characteristics of East Asian aerosol types by Lee et al (2010)
Properties CAT11 CAT2 CAT3 CAT4 CAT5 CAT6
SSA
440 nm 0915 0927 0904 0908 0893 0900676 nm 0927 0941 0908 0914 0939 0957869 nm 0918 0933 0897 0907 0945 09641020 nm 0912 0928 0889 0903 0948 0966
REFR
440 nm 1468 1478 1441 1454 1508 1549676 nm 1480 1483 1458 1472 1535 1549869 nm 1485 1483 1468 1482 1536 15371020 nm 1481 1476 1468 1481 1528 1525
REFI
440 nm 00119 00099 00127 00112 0007 00049676 nm 00086 00074 001 00088 00037 00024869 nm 00088 00078 00102 0009 00036 000231020 nm 00091 0008 00104 00092 00036 00025
ASYM
440 nm 0721 0729 0716 0713 073 0748676 nm 0664 0685 0654 0655 0693 0714869 nm 0636 0657 0626 0635 0691 07071020 nm 0626 0643 0618 0634 0696 0707
1CAT category
fine particle nonabsorbing aerosol (cluster 1) is enhancingThis characteristic confirms that the higher the membershipdegree the purer the nonabsorbing aerosol Figures 12(b) and12(c) show the mean values of SSA obviously descendingwhen membership degree increases from MDI1 to MDI2However the changes are hardly observed when member-ship degree interval increases from MDI2 to MDI5 Thesecharacteristics demonstrate that records at the edge of cluster2 (fine-M1) and cluster 3 (fine-M2) have stronger scattering(or lower absorbing) than those near center Figure 12(d)exhibits clearly ascending trend of SSA means which is amanifestation of the reasonability to identify cluster 4 ashighly absorbing aerosol The ascending trend also indicatesthat the higher membership degree denotes the purer highlyabsorbing aerosol
Figures 12(e) and 12(f) show some similarities betweenpolluted dust and desert dust at wavelengths of 440 nmFirstly mean values of SSA slightly decrease with the increaseof membership degree interval Furthermore mean values ofSSA at 440 nm are obviously lower than other wavelengthsAs discussed in Sections 324 and 325 these SSA trendsagree well with the relationships between SSA and dustcompositions which certify our identification that bothcluster 5 (polluted dust) and cluster 6 (desert dust) aredust-related aerosols However there are some differencesbetween two clusters SSAs at 440 nm are distinctly smallerthan at other wavelengths Besides Figures 12(e) and 12(f)show different trends at wavelengths longer than 440 nm theformer displays descending trend while the latter shows noclear variation These differences can be attributed to cluster
Advances in Meteorology 11
400
500
600
700
800
900
1000
1100
085
090
095
100Si
ngle
scat
terin
g al
bedo
Wavelengths (nm)
This paper cluster 1Lee et al CAT2
(a)
085
090
095
100
Sing
le sc
atte
ring
albe
do
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(b)
085
090
095
100
Sing
le sc
atte
ring
albe
do
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(c)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 1Lee et al CAT2
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(d)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(e)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(f)
000
005
010
015
020
025
This paper cluster 1Lee et al CAT2
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(g)
000
005
010
015
This paper cluster 4Lee et al CAT3
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(h)
00010203040506070809
Radius (G)
10minus1 100 101
This paper
This papercluster 6
cluster 5Lee et al
Lee et alCAT5
CAT6
dV
dFH
r(
G3
G2)
(i)
Figure 8 Comparison with East Asian aerosol types
Table 5 The record numbers within each membership degree interval
Membership degree intervals (MDI) C11 C2 C3 C4 C5 C6MDI1 017 le MD2 lt 027 156 131 60 117 43 86MDI2 027 leMD lt 037 508 569 298 603 404 308MDI3 037 leMD lt 047 305 326 240 456 413 234MDI4 047 leMD lt 057 115 128 197 181 275 153MDI5 057 leMD 40 34 122 37 117 761C cluster 2MD membership degree
12 Advances in Meteorology
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 9 AODs at 440 nmwith membership degrees asterisk represents mean value of AOD in each interval and the vertical bars representthe standard deviation
5 (polluted dust) containing absorbing anthropogenic aerosol(such as black carbon) while cluster 6 is dominated onlyby dust These characteristics validate that it is reasonable toidentify cluster 5 as polluted dust and cluster 6 as desert dust
Figure 13 is scatterplot of sphericity parameter versusmembership degree for six aerosol types Figures 13(a) 13(b)13(c) and 13(d) show high variation of sphericity parameterswhich indicates that the first four aerosol types show nocorrelation between sphericity parameters and membershipdegree However Figure 13(e) exhibits clearly descendingtrends and all mean values of sphericity parameter are lower
than 40 Moreover records in Figure 13(f) are more con-centrated with 95 of them falling below 20 These indicatethat both cluster 5 and cluster 6 are close to nonsphericityBecause most dust types are nonsphericity particles [35]these analyses again confirm our identification of polluteddust (cluster 5) and desert dust (cluster 6)
The above analyses confirm reasonability of our results ofclustering and identification of aerosol types Moreover theinternal variation of optical properties of aerosol type can bewell investigated with the help ofmembership degree interval(MDI)
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 3
50∘
45∘
40∘
35∘
30∘
25∘
20∘
50∘
45∘
40∘
35∘
30∘
25∘
20∘
75∘
80∘
85∘
90∘
95∘
100∘
105∘
110∘
115∘
120∘
125∘
130∘
135∘
75∘
80∘
85∘
90∘
95∘
100∘
105∘
110∘
115∘
120∘
125∘
130∘
135∘
(a)
Xianghe
Beijing Xinglong
Hebei ProvinceBohai Sea
TianjinCity
BeijingCity
43∘
42∘
41∘
40∘
39∘
38∘
37∘
36∘
43∘
42∘
41∘
40∘
39∘
38∘
37∘
36∘
114∘
115∘
116∘
117∘
118∘
119∘
114∘
115∘
116∘
117∘
118∘
119∘
(b)
Figure 1 Geographical location of the selected three AERONET sites
Table 1 Information of the selected three AERONET sites
Site name Observing period Number of records Longitude(degree)
Latitude(degree)
Elevation(meter)
Beijing 2001ndash2014 3418 116381 39977 92Xianghe 2001 2004ndash2012 2797 116962 39754 36Xinglong 2006ndash2012 517 117578 40396 970
It should be noted that the aerosol optical depth (AOD)fine fraction by volume (FFV) sphericity parameter (SP) andwater vapor (WV) are also used in later discussions Howeverthese four parameters are not applied in clustering
22 Fuzzy Clustering Cluster analysis is a statistical toolused for grouping the data sets into several clusters based onpredefined variables [14] The basic theory of cluster analysisis records in the same cluster which are more similar to eachother than to those in other clusters In fuzzy clustering everyrecord has a degree of belonging to each cluster rather thanjust completely belonging to one cluster [20] This degreeis represented by a parameter named membership degree(between 0 and 1) which indicates the confidence degreeof one record belonging to a cluster The records on theboundary of cluster will have smaller values than those nearthe center
In this study fuzzy 119888-mean (FCM) algorithm one ofthe most widely used fuzzy clustering algorithms is appliedGiven the data set containing 119899 records119883 = 1199091 1199092 119909119899and supposing there are 119888 clusters with the centers 119881 =V1 V2 119881119888 The FCM aims to minimize the objectivefunction
min 119869 (119906 V) =119899
sum119896=1
119888
sum119894=1
1205831198981198941198961198892119894119896 (1)
where 1198892119894119896 = 119909119896minusV1198942 represents the distance between record
119909119896 and center V119894 And 120583119898119894119896 is the degree to which record 119909119896belongs to cluster V119894 and is defined as
119906119894119896 =1
sum119888119895 (1003817100381710038171003817119909119896 minus V119894
1003817100381710038171003817 10038171003817100381710038171003817119909119896 minus V119895
10038171003817100381710038171003817)2(119898minus1) (2)
For a certain record sum119888119894 119906119894119896 = 1 The fuzzifier 119898 deter-mines the level of cluster fuzziness meaning the overlappingdegree of the cluster and 119898 ge 1 The center of the cluster isdefined as
V119894 =sum119899119896=1 119906
119898119894119896119909119896
sum119899119896=1 119906119898119894119896 (3)
The step of FCM is described as follows Firstly a randomfunction is used to determine the initial center of eachcluster Calculate membership degree and objective functionbased on functions (1) and (2) Meanwhile centers of eachcluster are calculated based on function (3) Repeat untilthe algorithm has converged that is the membership degreechange between two iterations is no more than the givensensitivity threshold It should be noted that before clusteringthe 22 parameters are normalized by the standard deviationto ensure that each parameter makes the same contributionto the distance calculation
3 Results and Discussions
31 Results By applying the FCM clustering a total of 6732records are classified into six clusters The characteristics ofoptical and microphysical properties of each cluster centerare shown in Table 2 Figure 2 displays aerosol optical depthsingle scattering albedo asymmetry parameter fine fractionby volume sphericity parameter and water vapor of eachcluster center Figure 3 shows the particle size distributions
4 Advances in Meteorology
Table 2 Results of the fuzzy cluster analysis
Properties Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
AOD
440 nm 125 156 079 076 073 087676 nm 077 094 045 043 044 064869 nm 054 066 031 031 033 0551020 nm 042 054 026 026 028 051
SSA
440 nm 095 090 090 084 088 089676 nm 096 092 091 087 091 094869 nm 095 091 090 085 090 0951020 nm 095 090 090 084 090 095
REFR
440 nm 144 150 143 148 151 150676 nm 144 151 146 151 153 153869 nm 144 152 147 152 154 1531020 nm 144 151 148 153 154 152
REFI
440 nm 0007 0015 0012 0023 0014 0007676 nm 0005 0009 0008 0014 0008 0003869 nm 0005 0009 0008 0015 0008 00031020 nm 0005 0010 0009 0016 0008 0003
ASYM
440 nm 074 071 071 068 069 072676 nm 070 065 064 062 064 068869 nm 067 062 061 061 063 0681020 nm 065 062 061 061 063 069
VolConF (120583m3120583m2) 016 017 011 009 008 007VolMedianRadF (120583m) 026 021 017 016 016 015
StdDevF 057 054 050 051 048 052VolConC (120583m3120583m2) 009 018 011 011 016 035VolMedianRadC (120583m) 294 260 273 279 271 235
StdDevC 055 057 060 063 065 060FFV 064 049 050 045 033 017SP () 5232 5783 5828 6427 2899 651
WV (gcm2) 218 113 134 064 091 091Number of records 917 1124 1188 1252 1394 857
Percentage to total number () 136 167 176 186 207 127
and Figure 4 represents the relationship between sphericityparameter and coarse fraction by volume
32 Characteristics of Aerosol Types
321 Fine Particle Nonabsorbing Aerosol As shown in Fig-ure 2(d) cluster 1 exhibits the largest value of fine fractionby volume (FFV) that is 064 Obviously cluster 1 is fineparticle dominated aerosol Figure 3 shows the particle sizedistribution of cluster 1 described by a fine radius andstandard deviation of 026120583m and 057 and a coarse radiusand standard deviation of 294 120583m and 055 Moreover thistype of aerosol displays clearly nonabsorbing properties asthe single scattering albedos (SSAs) at four wavelengths(440 676 869 and 1020 nm) are 095 096 095 and 095(Figure 2(b)) respectively As shown in Table 2 the real partsof refractive index (REFRs) of cluster 1 at fourwavelengths are144 which is lower than other clusters According to Brown(1984) lower value of real part of refractive index is possibly
associated with high hygroscopicity of aerosol [33] Whenthere is high relative humidity water is usually attached tothe surface of particle As a result it will enhance the abilityof particle scattering incident waveThis analysis is consistentwith large value of water vapor (218 gcm2) of cluster 1 whichis shown in Figure 2(f) Moreover this type of aerosol isusually observed in summer (Section 34) in which seasonBeijing is less affected by smoke and dust Therefore it islikely that these fine nonabsorbing particles are mainly theemissions of heavy traffic and industrial factories
There are similarities between cluster 1 and polluted con-tinental classified by Omar et al (2005) [14] But the formeris stronger in particle scattering and higher in fine fraction byvolume Consequently we identified cluster 1 as fine particlenonabsorbing aerosol To verify the relationship between fineparticle nonabsorbing aerosol and air pollution of Beijing wecollected the data of air quality index (AQI) over one year(October 2013 to October 2014) of Beijing (because the dataof AQI is not open to public only 2013-2014 is available to
Advances in Meteorology 5
400 500 600 700 800 900 1000 110000
03
06
09
12
15
Aero
sol o
ptic
al d
epth
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(a) AOD
400 500 600 700 800 900 1000 1100
084086088090092094096
Sing
le sc
atte
ring
albe
do
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(b) SSA
400 500 600 700 800 900 1000 1100
060063066069072075
Asy
mm
etry
par
amet
er
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(c) ASYM
Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 601
02
03
04
05
06
07
Fine
frac
tion
by v
olum
e
Cluster
(d) FMF
010203040506070
Sphe
ricity
par
amet
er (
)
ClusterCluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
(e) SP
00
04
08
12
16
20
24
Cluster
Wat
er v
apor
(g=G
2)
Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
(f) WV
Figure 2 The characteristics of six cluster centers (a) aerosol optical depth (b) single scattering albedo (c) asymmetry parameter (d) finefraction by volume (e) sphericity parameter (f) water vapor
us) The AQI is an index that indicates the pollution level ofthe atmosphere while a higher AQI value means a heavieratmospheric pollution The value of AQI level increase fromone to six represents air qualities of excellent good slightpollution moderate pollution heavy pollution and severepollution respectively Before comparison we collocate AQIdata with classified aerosol types and obtain 88 matchupsFigure 5 shows numbers of each aerosol type observed inper-AQI level over 2013-2014 at Beijing site It can be seenfrom Figure 5 that fine particle nonabsorbing aerosol is mostfrequently observed at AQI level 4 (moderate pollution)followed by AQI-levels 5 (heavy pollution) and 6 (severepollution)
322 Fine Particle Moderately Absorbing Aerosol The finefraction by volume of cluster 2 and cluster 3 is 049 and 050respectively Besides the SSAs at four wavelengths (440 676869 and 1020 nm) are 090 092 091 and 090 for cluster2 and 090 091 090 and 090 for cluster 3 Therefore bothclusters are considered as fine particle moderately absorbingaerosols (cluster 2 and cluster 3 named as fine-MA1 and fine-MA2 resp)
The fine-MA1 and fine-MA2 aerosols are mainly dis-tinguished by real parts of the refractive index (REFRs)and fine and coarse particle volume concentrations TheREFRs of fine-MA1 at four wavelengths are 150 151 152and 151 respectively Correspondingly the REFRs of fine-MA2 are much lower with REFRs being 143 146 147
6 Advances in Meteorology
000
005
010
015
020
025
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
Figure 3 The particle size distributions of six cluster centers
and 148 According to Bohren and Huffman (1983) [34]dry particle basically has a higher REFR usually being15ndash16 That is consistent with the fact that the water vaporof fine-MA1 (113 gcm2) is lower than that of fine-MA2(134 gcm2) which can be seen in Figure 2(f) Additionallythe fine particle volume concentration of fine-MA1 is thelargest one in six aerosol types (017 120583m3120583m2) and thecoarse particle volume concentration is also found being highlevel (018 120583m3120583m2) Correspondingly both fine and coarseparticle volume concentration of fine-MA2 are much lower(both are 011 120583m3120583m2) than fine-MA1
Compared with Omar et al (2005) [14] fine-MA1 is likepolluted continental aerosol and dirty pollution aerosol Butthe absorption is lower than the former and higher than thelatter It can be seen from Figure 5 that fine-MA1 is the mostfrequently observed at AQI levels 5 and 6 Besides fine-MA1is just a little lower than fine particle nonabsorbing aerosol atAQI level 4 Furthermore fine-MA1 is hardly observed atAQIlevels 1 (excellent) and 2 (good) On the contrary fine-MA2 ischaracterized by a low optical depth (Figure 2(a)) and a highfrequency of incidence at lowAQI level when the atmosphereis expected to be relatively clean The characteristics aremore likely the backgroundrural aerosol classified by Omaret al (2005) [14] Consequently fine-MA2 is possibly thebackground aerosol of Beijing
323 Fine Particle Highly Absorbing Aerosol Cluster 4 is ofinterest in view of its high absorption The SSAs at fourwavelengths (440 676 869 and 1020 nm) are 084 087 085and 084 (Figure 2(b)) respectively Besides the fine fractionby volume is 045 (Figure 2(d)) The fine particle volumeconcentration is 009120583m3120583m2 and the coarse particle vol-ume concentration is 011120583m3120583m2 In addition Figure 3shows the particle size distribution of cluster 4 describedby a fine radius and standard deviation of 016 120583m and 051and coarse radius and standard deviation of 279 120583m and
063 Therefore we identified cluster 4 as fine particle highlyabsorbing aerosol
Figure 2(e) shows that the sphericity parameter (thehigher value indicates the particle is closer to sphericity)of fine particle highly absorbing aerosol exhibits the largestvalue (6427) Moreover it can be clearly seen from Fig-ure 4(d) that this kind of aerosol has much more recordswith sphericity parameter higher than 90 It demonstratesthat fine particle highly absorbing aerosol is more likelya spherical aerosol These characteristics are like biomassburning aerosol classified by Omar et al (2005) [14] Addi-tionally it can be seen from Table 2 that the real parts ofthe refractive index at a high level are 148 151 152 and153 at four wavelengths According to the study result ofDubovik et al (2002) [11] higher REFRs and lower SSAs arepossibly associated with high concentrations of black carbonFurthermore this kind of aerosol usually occurs in winter(Section 34) in which season coal is frequently used forheating supply in northern China [23 28] Consequently fineparticle highly absorbing aerosol is possibly mainly producedby fossil fuel combustion
324 Polluted Dust Cluster 5 has the largest number ofrecords with 207 of total records being classified inthis type The SSAs at four wavelengths (440 676 869and 1020 nm) are 088 091 090 and 090 (Figure 2(b))respectively Besides the fine fraction by volume is 033(Figure 2(d)) The fine particle volume concentration is008120583m3120583m2 and coarse particle volume concentration is016 120583m3120583m2 Figure 3 shows the particle size distributionof cluster 5 described by a fine radius and standard devi-ation of 016 120583m and 048 and coarse radius and standarddeviation of 271 120583m and 065 As shown in Figure 4(e)most records of cluster 5 fall within sphericity parameterlower than 50 and coarse fraction by volume higher than05 Thus it indicates that cluster 5 is nonspherical coarseparticle dominated aerosol According to Okada et al (2001)nonspherical coarse particle is possibly contributed by dust[35] However there is no significant difference between SSAsat 440 nm and longer wavelengths But previous studies showthat when wavelength changes from 440 nm to longer theSSA will increase largely (Section 325) for dust aerosolBesides this kind of aerosol is frequently observed in spring(Section 34) in which season Beijing is affected by long-range transported dust Thus cluster 5 is possibly the dustmixes with anthropogenic aerosol As shown in Figure 4(e)cluster 5 also has some recordswith high sphericity parameteror low coarse fraction by volume Consequently we identifiedcluster 5 as polluted dust It should be noted that this polluteddust aerosol is the first time classified as a unique type bycluster analysis method The fact that polluted dust can beidentified by FCM demonstrates the practicability of fuzzyclustering in aerosol classification
325 Desert Dust Cluster 6 is significantly characterized bysmallest fine fraction by volume (017) and lowest sphericityparameter (651) Besides it shows the smallest fine particlevolume concentration (008120583m3120583m2) and largest coarseparticle volume concentration (016 120583m3120583m2) It indicates
Advances in Meteorology 7
0 25 50 75 10000
02
04
06
08
10C
oars
e fra
ctio
n by
vol
ume
Spherical parameter
(a) Cluster 1
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(b) Cluster 2
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(c) Cluster 3
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
eSpherical parameter
(d) Cluster 4
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(e) Cluster 5
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(f) Cluster 6
Figure 4 Relationships of sphericity parameter and coarse fraction by volume
that this is a completely coarse particle dominated aerosolIn addition as shown in Figure 4(f) most records of cluster6 concentrate in the range of sphericity parameter lowerthan 15 and coarse fraction by volume higher than 06 Itdemonstrates that cluster 6 is completely nonspherical coarseparticles dominated aerosol Moreover the SSAs of cluster 6exhibit a distinct feature with lower value at 440 nm (089)and a sharp increase to 094 095 and 095 at 660 nm 896 nmand 1020 nm (Figure 2(b)) respectively This is consistentwith the characteristic of desert dust derived by Levy et al(2007) [15] and Qin and Mitchell (2009) [16] Furthermoreaccording to Dubovik et al (2002) desert dust is possiblyabsorbed in the short spectral region due to the containediron oxide hematite [11] The SSA of dust particle might belower at short wavelength regions when it contains the FeO2Consequently these characteristics clearly certify that cluster6 is desert dust aerosol
326 Comparison with Other Results To assess the proposedfuzzy methodology we take a comparative study of our resultwith other published researches in aerosol classification
Considering study area and methodology two publishedresults are chosen global aerosol types by Omar et al (2005)[14] and East Asian aerosol types by Lee and Kim (2010) [17]
(1) Comparison with Global Aerosol TypesOmar et al (2005)applied cluster analysis to global ground-based observationsand obtained six categories backgroundrural industrialpollution dirty pollution biomass burning desert dustand polluted marine [14] Their characteristics are shownin Table 3 Five of the six aerosol types we classifiedshow similarities with Omarrsquos global aerosol types they arefine particle nonabsorbing aerosol to polluted continentalaerosol fine-MA2 to background aerosol fine particle highlyabsorbing aerosol to biomass burning aerosol and polluteddust and desert dust to dust
Figures 6 and 7 demonstrate comparisons of characteris-tics of aerosol types between global and Beijing aerosol typesincluding particle size distributions single scattering albedoand asymmetry parameter Particle size distributions (Fig-ure 6) show that the values of 119889119881119889 ln 119903 varied significantlybetween two researches The particle volume concentration
8 Advances in Meteorology
Table 3 Characteristics of global aerosol types by Omar et al (2005)
Properties (676 nm) Polluted continental Backgroundrural Biomass burning Dirty pollution Dust MarineSSA 092 088 080 072 093 093REFR 14098 14494 15202 14104 14520 13943REFI 00063 00092 00245 00337 00036 00044ASYM 0612 0580 0603 0594 0668 0711FFV 053 038 033 049 022 026
1 2 3 4 5 60
2
4
6
8
10
Num
ber o
f day
s
AQI-level
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 5 Numbers of each cluster observed in per-AQI level from2013 to 2014 at Beijing site
of Beijing is much higher than global aerosol types which isbecause Beijing usually suffers high concentration of aerosol[9] The SSAs of three fine particle aerosols of Beijing are095 091 and 087 which is larger than the value of globalaerosol 092 088 and 08 Interestingly all particle volumeconcentration SSA and asymmetry parameter of global dustare larger than polluted dust and smaller than desert dustwhich indicates that tow dust aerosol may be more ldquopurerdquothan dust in global aerosol types
(2) Comparison with East Asian Aerosol Types Lee and Kim(2010) performed a cluster analysis for AERONET recordsof East Asia which were divided into four fine-mode types(CAT (category) 1 to 4) and two coarse-mode types (CAT5and CAT6) [17] In addition the two coarse-modes werereferred to as dusty aerosols and the others were considereda mixed type of pollution Four of the six aerosol types inBeijing show similarities with Leersquos in East Asia they are fineparticle nonabsorbing aerosol to CAT2 fine particle highlyabsorbing aerosol to CAT3 and polluted dust and desert dustto two coarse-modes Table 4 shows the characteristics ofeach aerosol type identified by Lee et al (2010)
Figure 8 displays comparisons of characteristics of fouraerosol types between East Asia and Beijing Figures 8(a) and8(g) show that fine particle nonabsorbing aerosol of Beijingis stronger in particle scattering and lower in particle volume
concentration But the wavelength dependence of asymmetryparameter demonstrates similarities (Figure 8(d))The CAT3in East Asia and fine particle highly absorbing aerosol inBeijing exhibit similar particle size distributions and asym-metry parameter But the absorption of the latter is muchstronger than the former (Figure 8(b)) owing to the fossilfuel combustion in Beijing (see Section 323) It can beseen from Figures 8(c) 8(f) and 8(i) that the properties ofdesert dust in Beijing and CAT5 in East Asia are exactlysimilar to each other Meanwhile higher SSAs and particlevolume concentration of desert dust compared with CAT6probably due to sites close to dust source regions (like Yulinselected by Lee et al (2010)) were not included in thispaper It is interesting to note the pronounced differencesbetween polluted dust in Beijing and CAT5CAT6 in EastAsia indicating apparently polluted dust a unique aerosoltype
33 Variation of Optical Characteristics with MembershipDegrees The optical properties discussed in Section 32 arerepresented by the center values of each cluster But thedifferences between records at edge and near center ofcluster are inevitable Particularly for multiparticles mixtureaerosol only representing the aerosol by the center of clusterwill produce large errors As described in Section 22 themembership degree indicates the confidence degree of onerecord belonging to clusters which is an optimal parameterto analyze these particles at boundary ofmulticlustersThere-fore characteristics of aerosol (optical and microphysical)with different membership degree intervals are investigatedhere It should be noted thatwe focus on the internal variationof optical properties for each aerosol type
The membership degrees (MD) of six aerosol types aredivided into five intervals 017 le MD lt 027 027 le MD lt037 037 leMD lt 047 047 leMD lt 057 and 057 leMDWechose the value of 017 because when membership degree ofthe record to one type is larger than 16 (as there are six types)it means that this recordmay belong to that type Andwe takethe step of 01 to make sure that there is a similar amount ofrecords within each interval The numbers of records withineach interval are listed in Table 5
Figure 9 shows the AODs of six aerosol types varyingwith the membership degree and their mean values withineach membership degree interval at wavelength of 440 nmFine particle nonabsorbing aerosol (cluster 1 Figure 9(a))and fine-MA1 (cluster 2 Figure 9(b)) show large range ofAOD variation (from 0 to 4) The large range is reasonablebecause both types are frequently observed at polluted days(discussed in Section 32) Nevertheless the other four
Advances in Meteorology 9
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 1Continental
dV
dFH
r(
G3
G2)
(a)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 3Background
dV
dFH
r(
G3
G2)
(b)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 4Biomass
dV
dFH
r(
G3
G2)
(c)
000
005
010
015
020
025
030
Radius (G)
10minus1 100 101
Cluster 5Cluster 6Dust
dV
dFH
r(
G3
G2)
(d)
Figure 6 Size distribution comparison with global aerosol types
aerosol types show relatively smaller variation ranges (from0 to 2) And there are fewer that occurred during polluteddays Furthermore the AODs obviously concentrate in thelow varying range with increasing membership degree for allaerosol types
Figure 10 shows the mean values of AOD within the fivemembership degree intervals at four wavelengths (440 676869 and 1020 nm) Figures 10(a) and 10(b) show that theaverage AODs of fine particle moderately absorbing aerosolfor all wavelengths exhibit relatively flat behaviors Howeverthe other five aerosol types exhibit clearly declination withthe increase of themembership degree intervalMoreover thetrends are nearly the same between different wavelengths foreach aerosol type
The SSAs (440 nm) of six aerosol types varying withmembership degree and their mean values within eachmembership degree interval are plotted in Figure 11 Same
as AODs points of SSA concentrate in the low varyingrange with increasing membership degree for all aerosoltypes It can be seen from Figure 11(a) that the SSAs offine particle nonabsorbing aerosol (cluster 1) fall within therange between 09 and 10 This characteristic is consistentwith the nonabsorbing aerosol (strong scattering) of cluster1 Besides it is interesting to find that there are somestrongly absorbing particles (SSA lt 08) at membershipdegrees between 027 and 047 in Figure 11(d) (cluster 4) Asdiscussed in Section 323 these particles are probably theblack carbon which are frequently observed in Beijing withhigh absorption [23 28]
The mean values of SSA within the five membershipdegree intervals at four wavelengths (440 676 869 and1020 nm) are plotted in Figure 12 Means of SSA showclear descending trend in Figure 12(a) which indicates thatwith the increase of membership degree scattering ability of
10 Advances in Meteorology
Cluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
010203040506070809101112
Sing
le sc
atte
ring
albe
do
Aerosol typeCluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
00
01
02
03
04
05
06
07
08
Asy
mm
etry
fact
or
Aerosol type
This paperOmar et al
This paperOmar et al
Figure 7 SSA and ASYM comparison with global aerosol types
Table 4 Characteristics of East Asian aerosol types by Lee et al (2010)
Properties CAT11 CAT2 CAT3 CAT4 CAT5 CAT6
SSA
440 nm 0915 0927 0904 0908 0893 0900676 nm 0927 0941 0908 0914 0939 0957869 nm 0918 0933 0897 0907 0945 09641020 nm 0912 0928 0889 0903 0948 0966
REFR
440 nm 1468 1478 1441 1454 1508 1549676 nm 1480 1483 1458 1472 1535 1549869 nm 1485 1483 1468 1482 1536 15371020 nm 1481 1476 1468 1481 1528 1525
REFI
440 nm 00119 00099 00127 00112 0007 00049676 nm 00086 00074 001 00088 00037 00024869 nm 00088 00078 00102 0009 00036 000231020 nm 00091 0008 00104 00092 00036 00025
ASYM
440 nm 0721 0729 0716 0713 073 0748676 nm 0664 0685 0654 0655 0693 0714869 nm 0636 0657 0626 0635 0691 07071020 nm 0626 0643 0618 0634 0696 0707
1CAT category
fine particle nonabsorbing aerosol (cluster 1) is enhancingThis characteristic confirms that the higher the membershipdegree the purer the nonabsorbing aerosol Figures 12(b) and12(c) show the mean values of SSA obviously descendingwhen membership degree increases from MDI1 to MDI2However the changes are hardly observed when member-ship degree interval increases from MDI2 to MDI5 Thesecharacteristics demonstrate that records at the edge of cluster2 (fine-M1) and cluster 3 (fine-M2) have stronger scattering(or lower absorbing) than those near center Figure 12(d)exhibits clearly ascending trend of SSA means which is amanifestation of the reasonability to identify cluster 4 ashighly absorbing aerosol The ascending trend also indicatesthat the higher membership degree denotes the purer highlyabsorbing aerosol
Figures 12(e) and 12(f) show some similarities betweenpolluted dust and desert dust at wavelengths of 440 nmFirstly mean values of SSA slightly decrease with the increaseof membership degree interval Furthermore mean values ofSSA at 440 nm are obviously lower than other wavelengthsAs discussed in Sections 324 and 325 these SSA trendsagree well with the relationships between SSA and dustcompositions which certify our identification that bothcluster 5 (polluted dust) and cluster 6 (desert dust) aredust-related aerosols However there are some differencesbetween two clusters SSAs at 440 nm are distinctly smallerthan at other wavelengths Besides Figures 12(e) and 12(f)show different trends at wavelengths longer than 440 nm theformer displays descending trend while the latter shows noclear variation These differences can be attributed to cluster
Advances in Meteorology 11
400
500
600
700
800
900
1000
1100
085
090
095
100Si
ngle
scat
terin
g al
bedo
Wavelengths (nm)
This paper cluster 1Lee et al CAT2
(a)
085
090
095
100
Sing
le sc
atte
ring
albe
do
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(b)
085
090
095
100
Sing
le sc
atte
ring
albe
do
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(c)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 1Lee et al CAT2
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(d)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(e)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(f)
000
005
010
015
020
025
This paper cluster 1Lee et al CAT2
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(g)
000
005
010
015
This paper cluster 4Lee et al CAT3
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(h)
00010203040506070809
Radius (G)
10minus1 100 101
This paper
This papercluster 6
cluster 5Lee et al
Lee et alCAT5
CAT6
dV
dFH
r(
G3
G2)
(i)
Figure 8 Comparison with East Asian aerosol types
Table 5 The record numbers within each membership degree interval
Membership degree intervals (MDI) C11 C2 C3 C4 C5 C6MDI1 017 le MD2 lt 027 156 131 60 117 43 86MDI2 027 leMD lt 037 508 569 298 603 404 308MDI3 037 leMD lt 047 305 326 240 456 413 234MDI4 047 leMD lt 057 115 128 197 181 275 153MDI5 057 leMD 40 34 122 37 117 761C cluster 2MD membership degree
12 Advances in Meteorology
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 9 AODs at 440 nmwith membership degrees asterisk represents mean value of AOD in each interval and the vertical bars representthe standard deviation
5 (polluted dust) containing absorbing anthropogenic aerosol(such as black carbon) while cluster 6 is dominated onlyby dust These characteristics validate that it is reasonable toidentify cluster 5 as polluted dust and cluster 6 as desert dust
Figure 13 is scatterplot of sphericity parameter versusmembership degree for six aerosol types Figures 13(a) 13(b)13(c) and 13(d) show high variation of sphericity parameterswhich indicates that the first four aerosol types show nocorrelation between sphericity parameters and membershipdegree However Figure 13(e) exhibits clearly descendingtrends and all mean values of sphericity parameter are lower
than 40 Moreover records in Figure 13(f) are more con-centrated with 95 of them falling below 20 These indicatethat both cluster 5 and cluster 6 are close to nonsphericityBecause most dust types are nonsphericity particles [35]these analyses again confirm our identification of polluteddust (cluster 5) and desert dust (cluster 6)
The above analyses confirm reasonability of our results ofclustering and identification of aerosol types Moreover theinternal variation of optical properties of aerosol type can bewell investigated with the help ofmembership degree interval(MDI)
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
4 Advances in Meteorology
Table 2 Results of the fuzzy cluster analysis
Properties Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
AOD
440 nm 125 156 079 076 073 087676 nm 077 094 045 043 044 064869 nm 054 066 031 031 033 0551020 nm 042 054 026 026 028 051
SSA
440 nm 095 090 090 084 088 089676 nm 096 092 091 087 091 094869 nm 095 091 090 085 090 0951020 nm 095 090 090 084 090 095
REFR
440 nm 144 150 143 148 151 150676 nm 144 151 146 151 153 153869 nm 144 152 147 152 154 1531020 nm 144 151 148 153 154 152
REFI
440 nm 0007 0015 0012 0023 0014 0007676 nm 0005 0009 0008 0014 0008 0003869 nm 0005 0009 0008 0015 0008 00031020 nm 0005 0010 0009 0016 0008 0003
ASYM
440 nm 074 071 071 068 069 072676 nm 070 065 064 062 064 068869 nm 067 062 061 061 063 0681020 nm 065 062 061 061 063 069
VolConF (120583m3120583m2) 016 017 011 009 008 007VolMedianRadF (120583m) 026 021 017 016 016 015
StdDevF 057 054 050 051 048 052VolConC (120583m3120583m2) 009 018 011 011 016 035VolMedianRadC (120583m) 294 260 273 279 271 235
StdDevC 055 057 060 063 065 060FFV 064 049 050 045 033 017SP () 5232 5783 5828 6427 2899 651
WV (gcm2) 218 113 134 064 091 091Number of records 917 1124 1188 1252 1394 857
Percentage to total number () 136 167 176 186 207 127
and Figure 4 represents the relationship between sphericityparameter and coarse fraction by volume
32 Characteristics of Aerosol Types
321 Fine Particle Nonabsorbing Aerosol As shown in Fig-ure 2(d) cluster 1 exhibits the largest value of fine fractionby volume (FFV) that is 064 Obviously cluster 1 is fineparticle dominated aerosol Figure 3 shows the particle sizedistribution of cluster 1 described by a fine radius andstandard deviation of 026120583m and 057 and a coarse radiusand standard deviation of 294 120583m and 055 Moreover thistype of aerosol displays clearly nonabsorbing properties asthe single scattering albedos (SSAs) at four wavelengths(440 676 869 and 1020 nm) are 095 096 095 and 095(Figure 2(b)) respectively As shown in Table 2 the real partsof refractive index (REFRs) of cluster 1 at fourwavelengths are144 which is lower than other clusters According to Brown(1984) lower value of real part of refractive index is possibly
associated with high hygroscopicity of aerosol [33] Whenthere is high relative humidity water is usually attached tothe surface of particle As a result it will enhance the abilityof particle scattering incident waveThis analysis is consistentwith large value of water vapor (218 gcm2) of cluster 1 whichis shown in Figure 2(f) Moreover this type of aerosol isusually observed in summer (Section 34) in which seasonBeijing is less affected by smoke and dust Therefore it islikely that these fine nonabsorbing particles are mainly theemissions of heavy traffic and industrial factories
There are similarities between cluster 1 and polluted con-tinental classified by Omar et al (2005) [14] But the formeris stronger in particle scattering and higher in fine fraction byvolume Consequently we identified cluster 1 as fine particlenonabsorbing aerosol To verify the relationship between fineparticle nonabsorbing aerosol and air pollution of Beijing wecollected the data of air quality index (AQI) over one year(October 2013 to October 2014) of Beijing (because the dataof AQI is not open to public only 2013-2014 is available to
Advances in Meteorology 5
400 500 600 700 800 900 1000 110000
03
06
09
12
15
Aero
sol o
ptic
al d
epth
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(a) AOD
400 500 600 700 800 900 1000 1100
084086088090092094096
Sing
le sc
atte
ring
albe
do
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(b) SSA
400 500 600 700 800 900 1000 1100
060063066069072075
Asy
mm
etry
par
amet
er
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(c) ASYM
Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 601
02
03
04
05
06
07
Fine
frac
tion
by v
olum
e
Cluster
(d) FMF
010203040506070
Sphe
ricity
par
amet
er (
)
ClusterCluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
(e) SP
00
04
08
12
16
20
24
Cluster
Wat
er v
apor
(g=G
2)
Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
(f) WV
Figure 2 The characteristics of six cluster centers (a) aerosol optical depth (b) single scattering albedo (c) asymmetry parameter (d) finefraction by volume (e) sphericity parameter (f) water vapor
us) The AQI is an index that indicates the pollution level ofthe atmosphere while a higher AQI value means a heavieratmospheric pollution The value of AQI level increase fromone to six represents air qualities of excellent good slightpollution moderate pollution heavy pollution and severepollution respectively Before comparison we collocate AQIdata with classified aerosol types and obtain 88 matchupsFigure 5 shows numbers of each aerosol type observed inper-AQI level over 2013-2014 at Beijing site It can be seenfrom Figure 5 that fine particle nonabsorbing aerosol is mostfrequently observed at AQI level 4 (moderate pollution)followed by AQI-levels 5 (heavy pollution) and 6 (severepollution)
322 Fine Particle Moderately Absorbing Aerosol The finefraction by volume of cluster 2 and cluster 3 is 049 and 050respectively Besides the SSAs at four wavelengths (440 676869 and 1020 nm) are 090 092 091 and 090 for cluster2 and 090 091 090 and 090 for cluster 3 Therefore bothclusters are considered as fine particle moderately absorbingaerosols (cluster 2 and cluster 3 named as fine-MA1 and fine-MA2 resp)
The fine-MA1 and fine-MA2 aerosols are mainly dis-tinguished by real parts of the refractive index (REFRs)and fine and coarse particle volume concentrations TheREFRs of fine-MA1 at four wavelengths are 150 151 152and 151 respectively Correspondingly the REFRs of fine-MA2 are much lower with REFRs being 143 146 147
6 Advances in Meteorology
000
005
010
015
020
025
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
Figure 3 The particle size distributions of six cluster centers
and 148 According to Bohren and Huffman (1983) [34]dry particle basically has a higher REFR usually being15ndash16 That is consistent with the fact that the water vaporof fine-MA1 (113 gcm2) is lower than that of fine-MA2(134 gcm2) which can be seen in Figure 2(f) Additionallythe fine particle volume concentration of fine-MA1 is thelargest one in six aerosol types (017 120583m3120583m2) and thecoarse particle volume concentration is also found being highlevel (018 120583m3120583m2) Correspondingly both fine and coarseparticle volume concentration of fine-MA2 are much lower(both are 011 120583m3120583m2) than fine-MA1
Compared with Omar et al (2005) [14] fine-MA1 is likepolluted continental aerosol and dirty pollution aerosol Butthe absorption is lower than the former and higher than thelatter It can be seen from Figure 5 that fine-MA1 is the mostfrequently observed at AQI levels 5 and 6 Besides fine-MA1is just a little lower than fine particle nonabsorbing aerosol atAQI level 4 Furthermore fine-MA1 is hardly observed atAQIlevels 1 (excellent) and 2 (good) On the contrary fine-MA2 ischaracterized by a low optical depth (Figure 2(a)) and a highfrequency of incidence at lowAQI level when the atmosphereis expected to be relatively clean The characteristics aremore likely the backgroundrural aerosol classified by Omaret al (2005) [14] Consequently fine-MA2 is possibly thebackground aerosol of Beijing
323 Fine Particle Highly Absorbing Aerosol Cluster 4 is ofinterest in view of its high absorption The SSAs at fourwavelengths (440 676 869 and 1020 nm) are 084 087 085and 084 (Figure 2(b)) respectively Besides the fine fractionby volume is 045 (Figure 2(d)) The fine particle volumeconcentration is 009120583m3120583m2 and the coarse particle vol-ume concentration is 011120583m3120583m2 In addition Figure 3shows the particle size distribution of cluster 4 describedby a fine radius and standard deviation of 016 120583m and 051and coarse radius and standard deviation of 279 120583m and
063 Therefore we identified cluster 4 as fine particle highlyabsorbing aerosol
Figure 2(e) shows that the sphericity parameter (thehigher value indicates the particle is closer to sphericity)of fine particle highly absorbing aerosol exhibits the largestvalue (6427) Moreover it can be clearly seen from Fig-ure 4(d) that this kind of aerosol has much more recordswith sphericity parameter higher than 90 It demonstratesthat fine particle highly absorbing aerosol is more likelya spherical aerosol These characteristics are like biomassburning aerosol classified by Omar et al (2005) [14] Addi-tionally it can be seen from Table 2 that the real parts ofthe refractive index at a high level are 148 151 152 and153 at four wavelengths According to the study result ofDubovik et al (2002) [11] higher REFRs and lower SSAs arepossibly associated with high concentrations of black carbonFurthermore this kind of aerosol usually occurs in winter(Section 34) in which season coal is frequently used forheating supply in northern China [23 28] Consequently fineparticle highly absorbing aerosol is possibly mainly producedby fossil fuel combustion
324 Polluted Dust Cluster 5 has the largest number ofrecords with 207 of total records being classified inthis type The SSAs at four wavelengths (440 676 869and 1020 nm) are 088 091 090 and 090 (Figure 2(b))respectively Besides the fine fraction by volume is 033(Figure 2(d)) The fine particle volume concentration is008120583m3120583m2 and coarse particle volume concentration is016 120583m3120583m2 Figure 3 shows the particle size distributionof cluster 5 described by a fine radius and standard devi-ation of 016 120583m and 048 and coarse radius and standarddeviation of 271 120583m and 065 As shown in Figure 4(e)most records of cluster 5 fall within sphericity parameterlower than 50 and coarse fraction by volume higher than05 Thus it indicates that cluster 5 is nonspherical coarseparticle dominated aerosol According to Okada et al (2001)nonspherical coarse particle is possibly contributed by dust[35] However there is no significant difference between SSAsat 440 nm and longer wavelengths But previous studies showthat when wavelength changes from 440 nm to longer theSSA will increase largely (Section 325) for dust aerosolBesides this kind of aerosol is frequently observed in spring(Section 34) in which season Beijing is affected by long-range transported dust Thus cluster 5 is possibly the dustmixes with anthropogenic aerosol As shown in Figure 4(e)cluster 5 also has some recordswith high sphericity parameteror low coarse fraction by volume Consequently we identifiedcluster 5 as polluted dust It should be noted that this polluteddust aerosol is the first time classified as a unique type bycluster analysis method The fact that polluted dust can beidentified by FCM demonstrates the practicability of fuzzyclustering in aerosol classification
325 Desert Dust Cluster 6 is significantly characterized bysmallest fine fraction by volume (017) and lowest sphericityparameter (651) Besides it shows the smallest fine particlevolume concentration (008120583m3120583m2) and largest coarseparticle volume concentration (016 120583m3120583m2) It indicates
Advances in Meteorology 7
0 25 50 75 10000
02
04
06
08
10C
oars
e fra
ctio
n by
vol
ume
Spherical parameter
(a) Cluster 1
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(b) Cluster 2
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(c) Cluster 3
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
eSpherical parameter
(d) Cluster 4
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(e) Cluster 5
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(f) Cluster 6
Figure 4 Relationships of sphericity parameter and coarse fraction by volume
that this is a completely coarse particle dominated aerosolIn addition as shown in Figure 4(f) most records of cluster6 concentrate in the range of sphericity parameter lowerthan 15 and coarse fraction by volume higher than 06 Itdemonstrates that cluster 6 is completely nonspherical coarseparticles dominated aerosol Moreover the SSAs of cluster 6exhibit a distinct feature with lower value at 440 nm (089)and a sharp increase to 094 095 and 095 at 660 nm 896 nmand 1020 nm (Figure 2(b)) respectively This is consistentwith the characteristic of desert dust derived by Levy et al(2007) [15] and Qin and Mitchell (2009) [16] Furthermoreaccording to Dubovik et al (2002) desert dust is possiblyabsorbed in the short spectral region due to the containediron oxide hematite [11] The SSA of dust particle might belower at short wavelength regions when it contains the FeO2Consequently these characteristics clearly certify that cluster6 is desert dust aerosol
326 Comparison with Other Results To assess the proposedfuzzy methodology we take a comparative study of our resultwith other published researches in aerosol classification
Considering study area and methodology two publishedresults are chosen global aerosol types by Omar et al (2005)[14] and East Asian aerosol types by Lee and Kim (2010) [17]
(1) Comparison with Global Aerosol TypesOmar et al (2005)applied cluster analysis to global ground-based observationsand obtained six categories backgroundrural industrialpollution dirty pollution biomass burning desert dustand polluted marine [14] Their characteristics are shownin Table 3 Five of the six aerosol types we classifiedshow similarities with Omarrsquos global aerosol types they arefine particle nonabsorbing aerosol to polluted continentalaerosol fine-MA2 to background aerosol fine particle highlyabsorbing aerosol to biomass burning aerosol and polluteddust and desert dust to dust
Figures 6 and 7 demonstrate comparisons of characteris-tics of aerosol types between global and Beijing aerosol typesincluding particle size distributions single scattering albedoand asymmetry parameter Particle size distributions (Fig-ure 6) show that the values of 119889119881119889 ln 119903 varied significantlybetween two researches The particle volume concentration
8 Advances in Meteorology
Table 3 Characteristics of global aerosol types by Omar et al (2005)
Properties (676 nm) Polluted continental Backgroundrural Biomass burning Dirty pollution Dust MarineSSA 092 088 080 072 093 093REFR 14098 14494 15202 14104 14520 13943REFI 00063 00092 00245 00337 00036 00044ASYM 0612 0580 0603 0594 0668 0711FFV 053 038 033 049 022 026
1 2 3 4 5 60
2
4
6
8
10
Num
ber o
f day
s
AQI-level
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 5 Numbers of each cluster observed in per-AQI level from2013 to 2014 at Beijing site
of Beijing is much higher than global aerosol types which isbecause Beijing usually suffers high concentration of aerosol[9] The SSAs of three fine particle aerosols of Beijing are095 091 and 087 which is larger than the value of globalaerosol 092 088 and 08 Interestingly all particle volumeconcentration SSA and asymmetry parameter of global dustare larger than polluted dust and smaller than desert dustwhich indicates that tow dust aerosol may be more ldquopurerdquothan dust in global aerosol types
(2) Comparison with East Asian Aerosol Types Lee and Kim(2010) performed a cluster analysis for AERONET recordsof East Asia which were divided into four fine-mode types(CAT (category) 1 to 4) and two coarse-mode types (CAT5and CAT6) [17] In addition the two coarse-modes werereferred to as dusty aerosols and the others were considereda mixed type of pollution Four of the six aerosol types inBeijing show similarities with Leersquos in East Asia they are fineparticle nonabsorbing aerosol to CAT2 fine particle highlyabsorbing aerosol to CAT3 and polluted dust and desert dustto two coarse-modes Table 4 shows the characteristics ofeach aerosol type identified by Lee et al (2010)
Figure 8 displays comparisons of characteristics of fouraerosol types between East Asia and Beijing Figures 8(a) and8(g) show that fine particle nonabsorbing aerosol of Beijingis stronger in particle scattering and lower in particle volume
concentration But the wavelength dependence of asymmetryparameter demonstrates similarities (Figure 8(d))The CAT3in East Asia and fine particle highly absorbing aerosol inBeijing exhibit similar particle size distributions and asym-metry parameter But the absorption of the latter is muchstronger than the former (Figure 8(b)) owing to the fossilfuel combustion in Beijing (see Section 323) It can beseen from Figures 8(c) 8(f) and 8(i) that the properties ofdesert dust in Beijing and CAT5 in East Asia are exactlysimilar to each other Meanwhile higher SSAs and particlevolume concentration of desert dust compared with CAT6probably due to sites close to dust source regions (like Yulinselected by Lee et al (2010)) were not included in thispaper It is interesting to note the pronounced differencesbetween polluted dust in Beijing and CAT5CAT6 in EastAsia indicating apparently polluted dust a unique aerosoltype
33 Variation of Optical Characteristics with MembershipDegrees The optical properties discussed in Section 32 arerepresented by the center values of each cluster But thedifferences between records at edge and near center ofcluster are inevitable Particularly for multiparticles mixtureaerosol only representing the aerosol by the center of clusterwill produce large errors As described in Section 22 themembership degree indicates the confidence degree of onerecord belonging to clusters which is an optimal parameterto analyze these particles at boundary ofmulticlustersThere-fore characteristics of aerosol (optical and microphysical)with different membership degree intervals are investigatedhere It should be noted thatwe focus on the internal variationof optical properties for each aerosol type
The membership degrees (MD) of six aerosol types aredivided into five intervals 017 le MD lt 027 027 le MD lt037 037 leMD lt 047 047 leMD lt 057 and 057 leMDWechose the value of 017 because when membership degree ofthe record to one type is larger than 16 (as there are six types)it means that this recordmay belong to that type Andwe takethe step of 01 to make sure that there is a similar amount ofrecords within each interval The numbers of records withineach interval are listed in Table 5
Figure 9 shows the AODs of six aerosol types varyingwith the membership degree and their mean values withineach membership degree interval at wavelength of 440 nmFine particle nonabsorbing aerosol (cluster 1 Figure 9(a))and fine-MA1 (cluster 2 Figure 9(b)) show large range ofAOD variation (from 0 to 4) The large range is reasonablebecause both types are frequently observed at polluted days(discussed in Section 32) Nevertheless the other four
Advances in Meteorology 9
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 1Continental
dV
dFH
r(
G3
G2)
(a)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 3Background
dV
dFH
r(
G3
G2)
(b)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 4Biomass
dV
dFH
r(
G3
G2)
(c)
000
005
010
015
020
025
030
Radius (G)
10minus1 100 101
Cluster 5Cluster 6Dust
dV
dFH
r(
G3
G2)
(d)
Figure 6 Size distribution comparison with global aerosol types
aerosol types show relatively smaller variation ranges (from0 to 2) And there are fewer that occurred during polluteddays Furthermore the AODs obviously concentrate in thelow varying range with increasing membership degree for allaerosol types
Figure 10 shows the mean values of AOD within the fivemembership degree intervals at four wavelengths (440 676869 and 1020 nm) Figures 10(a) and 10(b) show that theaverage AODs of fine particle moderately absorbing aerosolfor all wavelengths exhibit relatively flat behaviors Howeverthe other five aerosol types exhibit clearly declination withthe increase of themembership degree intervalMoreover thetrends are nearly the same between different wavelengths foreach aerosol type
The SSAs (440 nm) of six aerosol types varying withmembership degree and their mean values within eachmembership degree interval are plotted in Figure 11 Same
as AODs points of SSA concentrate in the low varyingrange with increasing membership degree for all aerosoltypes It can be seen from Figure 11(a) that the SSAs offine particle nonabsorbing aerosol (cluster 1) fall within therange between 09 and 10 This characteristic is consistentwith the nonabsorbing aerosol (strong scattering) of cluster1 Besides it is interesting to find that there are somestrongly absorbing particles (SSA lt 08) at membershipdegrees between 027 and 047 in Figure 11(d) (cluster 4) Asdiscussed in Section 323 these particles are probably theblack carbon which are frequently observed in Beijing withhigh absorption [23 28]
The mean values of SSA within the five membershipdegree intervals at four wavelengths (440 676 869 and1020 nm) are plotted in Figure 12 Means of SSA showclear descending trend in Figure 12(a) which indicates thatwith the increase of membership degree scattering ability of
10 Advances in Meteorology
Cluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
010203040506070809101112
Sing
le sc
atte
ring
albe
do
Aerosol typeCluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
00
01
02
03
04
05
06
07
08
Asy
mm
etry
fact
or
Aerosol type
This paperOmar et al
This paperOmar et al
Figure 7 SSA and ASYM comparison with global aerosol types
Table 4 Characteristics of East Asian aerosol types by Lee et al (2010)
Properties CAT11 CAT2 CAT3 CAT4 CAT5 CAT6
SSA
440 nm 0915 0927 0904 0908 0893 0900676 nm 0927 0941 0908 0914 0939 0957869 nm 0918 0933 0897 0907 0945 09641020 nm 0912 0928 0889 0903 0948 0966
REFR
440 nm 1468 1478 1441 1454 1508 1549676 nm 1480 1483 1458 1472 1535 1549869 nm 1485 1483 1468 1482 1536 15371020 nm 1481 1476 1468 1481 1528 1525
REFI
440 nm 00119 00099 00127 00112 0007 00049676 nm 00086 00074 001 00088 00037 00024869 nm 00088 00078 00102 0009 00036 000231020 nm 00091 0008 00104 00092 00036 00025
ASYM
440 nm 0721 0729 0716 0713 073 0748676 nm 0664 0685 0654 0655 0693 0714869 nm 0636 0657 0626 0635 0691 07071020 nm 0626 0643 0618 0634 0696 0707
1CAT category
fine particle nonabsorbing aerosol (cluster 1) is enhancingThis characteristic confirms that the higher the membershipdegree the purer the nonabsorbing aerosol Figures 12(b) and12(c) show the mean values of SSA obviously descendingwhen membership degree increases from MDI1 to MDI2However the changes are hardly observed when member-ship degree interval increases from MDI2 to MDI5 Thesecharacteristics demonstrate that records at the edge of cluster2 (fine-M1) and cluster 3 (fine-M2) have stronger scattering(or lower absorbing) than those near center Figure 12(d)exhibits clearly ascending trend of SSA means which is amanifestation of the reasonability to identify cluster 4 ashighly absorbing aerosol The ascending trend also indicatesthat the higher membership degree denotes the purer highlyabsorbing aerosol
Figures 12(e) and 12(f) show some similarities betweenpolluted dust and desert dust at wavelengths of 440 nmFirstly mean values of SSA slightly decrease with the increaseof membership degree interval Furthermore mean values ofSSA at 440 nm are obviously lower than other wavelengthsAs discussed in Sections 324 and 325 these SSA trendsagree well with the relationships between SSA and dustcompositions which certify our identification that bothcluster 5 (polluted dust) and cluster 6 (desert dust) aredust-related aerosols However there are some differencesbetween two clusters SSAs at 440 nm are distinctly smallerthan at other wavelengths Besides Figures 12(e) and 12(f)show different trends at wavelengths longer than 440 nm theformer displays descending trend while the latter shows noclear variation These differences can be attributed to cluster
Advances in Meteorology 11
400
500
600
700
800
900
1000
1100
085
090
095
100Si
ngle
scat
terin
g al
bedo
Wavelengths (nm)
This paper cluster 1Lee et al CAT2
(a)
085
090
095
100
Sing
le sc
atte
ring
albe
do
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(b)
085
090
095
100
Sing
le sc
atte
ring
albe
do
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(c)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 1Lee et al CAT2
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(d)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(e)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(f)
000
005
010
015
020
025
This paper cluster 1Lee et al CAT2
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(g)
000
005
010
015
This paper cluster 4Lee et al CAT3
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(h)
00010203040506070809
Radius (G)
10minus1 100 101
This paper
This papercluster 6
cluster 5Lee et al
Lee et alCAT5
CAT6
dV
dFH
r(
G3
G2)
(i)
Figure 8 Comparison with East Asian aerosol types
Table 5 The record numbers within each membership degree interval
Membership degree intervals (MDI) C11 C2 C3 C4 C5 C6MDI1 017 le MD2 lt 027 156 131 60 117 43 86MDI2 027 leMD lt 037 508 569 298 603 404 308MDI3 037 leMD lt 047 305 326 240 456 413 234MDI4 047 leMD lt 057 115 128 197 181 275 153MDI5 057 leMD 40 34 122 37 117 761C cluster 2MD membership degree
12 Advances in Meteorology
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 9 AODs at 440 nmwith membership degrees asterisk represents mean value of AOD in each interval and the vertical bars representthe standard deviation
5 (polluted dust) containing absorbing anthropogenic aerosol(such as black carbon) while cluster 6 is dominated onlyby dust These characteristics validate that it is reasonable toidentify cluster 5 as polluted dust and cluster 6 as desert dust
Figure 13 is scatterplot of sphericity parameter versusmembership degree for six aerosol types Figures 13(a) 13(b)13(c) and 13(d) show high variation of sphericity parameterswhich indicates that the first four aerosol types show nocorrelation between sphericity parameters and membershipdegree However Figure 13(e) exhibits clearly descendingtrends and all mean values of sphericity parameter are lower
than 40 Moreover records in Figure 13(f) are more con-centrated with 95 of them falling below 20 These indicatethat both cluster 5 and cluster 6 are close to nonsphericityBecause most dust types are nonsphericity particles [35]these analyses again confirm our identification of polluteddust (cluster 5) and desert dust (cluster 6)
The above analyses confirm reasonability of our results ofclustering and identification of aerosol types Moreover theinternal variation of optical properties of aerosol type can bewell investigated with the help ofmembership degree interval(MDI)
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 5
400 500 600 700 800 900 1000 110000
03
06
09
12
15
Aero
sol o
ptic
al d
epth
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(a) AOD
400 500 600 700 800 900 1000 1100
084086088090092094096
Sing
le sc
atte
ring
albe
do
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(b) SSA
400 500 600 700 800 900 1000 1100
060063066069072075
Asy
mm
etry
par
amet
er
Wavelength (nm)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
(c) ASYM
Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 601
02
03
04
05
06
07
Fine
frac
tion
by v
olum
e
Cluster
(d) FMF
010203040506070
Sphe
ricity
par
amet
er (
)
ClusterCluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
(e) SP
00
04
08
12
16
20
24
Cluster
Wat
er v
apor
(g=G
2)
Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6
(f) WV
Figure 2 The characteristics of six cluster centers (a) aerosol optical depth (b) single scattering albedo (c) asymmetry parameter (d) finefraction by volume (e) sphericity parameter (f) water vapor
us) The AQI is an index that indicates the pollution level ofthe atmosphere while a higher AQI value means a heavieratmospheric pollution The value of AQI level increase fromone to six represents air qualities of excellent good slightpollution moderate pollution heavy pollution and severepollution respectively Before comparison we collocate AQIdata with classified aerosol types and obtain 88 matchupsFigure 5 shows numbers of each aerosol type observed inper-AQI level over 2013-2014 at Beijing site It can be seenfrom Figure 5 that fine particle nonabsorbing aerosol is mostfrequently observed at AQI level 4 (moderate pollution)followed by AQI-levels 5 (heavy pollution) and 6 (severepollution)
322 Fine Particle Moderately Absorbing Aerosol The finefraction by volume of cluster 2 and cluster 3 is 049 and 050respectively Besides the SSAs at four wavelengths (440 676869 and 1020 nm) are 090 092 091 and 090 for cluster2 and 090 091 090 and 090 for cluster 3 Therefore bothclusters are considered as fine particle moderately absorbingaerosols (cluster 2 and cluster 3 named as fine-MA1 and fine-MA2 resp)
The fine-MA1 and fine-MA2 aerosols are mainly dis-tinguished by real parts of the refractive index (REFRs)and fine and coarse particle volume concentrations TheREFRs of fine-MA1 at four wavelengths are 150 151 152and 151 respectively Correspondingly the REFRs of fine-MA2 are much lower with REFRs being 143 146 147
6 Advances in Meteorology
000
005
010
015
020
025
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
Figure 3 The particle size distributions of six cluster centers
and 148 According to Bohren and Huffman (1983) [34]dry particle basically has a higher REFR usually being15ndash16 That is consistent with the fact that the water vaporof fine-MA1 (113 gcm2) is lower than that of fine-MA2(134 gcm2) which can be seen in Figure 2(f) Additionallythe fine particle volume concentration of fine-MA1 is thelargest one in six aerosol types (017 120583m3120583m2) and thecoarse particle volume concentration is also found being highlevel (018 120583m3120583m2) Correspondingly both fine and coarseparticle volume concentration of fine-MA2 are much lower(both are 011 120583m3120583m2) than fine-MA1
Compared with Omar et al (2005) [14] fine-MA1 is likepolluted continental aerosol and dirty pollution aerosol Butthe absorption is lower than the former and higher than thelatter It can be seen from Figure 5 that fine-MA1 is the mostfrequently observed at AQI levels 5 and 6 Besides fine-MA1is just a little lower than fine particle nonabsorbing aerosol atAQI level 4 Furthermore fine-MA1 is hardly observed atAQIlevels 1 (excellent) and 2 (good) On the contrary fine-MA2 ischaracterized by a low optical depth (Figure 2(a)) and a highfrequency of incidence at lowAQI level when the atmosphereis expected to be relatively clean The characteristics aremore likely the backgroundrural aerosol classified by Omaret al (2005) [14] Consequently fine-MA2 is possibly thebackground aerosol of Beijing
323 Fine Particle Highly Absorbing Aerosol Cluster 4 is ofinterest in view of its high absorption The SSAs at fourwavelengths (440 676 869 and 1020 nm) are 084 087 085and 084 (Figure 2(b)) respectively Besides the fine fractionby volume is 045 (Figure 2(d)) The fine particle volumeconcentration is 009120583m3120583m2 and the coarse particle vol-ume concentration is 011120583m3120583m2 In addition Figure 3shows the particle size distribution of cluster 4 describedby a fine radius and standard deviation of 016 120583m and 051and coarse radius and standard deviation of 279 120583m and
063 Therefore we identified cluster 4 as fine particle highlyabsorbing aerosol
Figure 2(e) shows that the sphericity parameter (thehigher value indicates the particle is closer to sphericity)of fine particle highly absorbing aerosol exhibits the largestvalue (6427) Moreover it can be clearly seen from Fig-ure 4(d) that this kind of aerosol has much more recordswith sphericity parameter higher than 90 It demonstratesthat fine particle highly absorbing aerosol is more likelya spherical aerosol These characteristics are like biomassburning aerosol classified by Omar et al (2005) [14] Addi-tionally it can be seen from Table 2 that the real parts ofthe refractive index at a high level are 148 151 152 and153 at four wavelengths According to the study result ofDubovik et al (2002) [11] higher REFRs and lower SSAs arepossibly associated with high concentrations of black carbonFurthermore this kind of aerosol usually occurs in winter(Section 34) in which season coal is frequently used forheating supply in northern China [23 28] Consequently fineparticle highly absorbing aerosol is possibly mainly producedby fossil fuel combustion
324 Polluted Dust Cluster 5 has the largest number ofrecords with 207 of total records being classified inthis type The SSAs at four wavelengths (440 676 869and 1020 nm) are 088 091 090 and 090 (Figure 2(b))respectively Besides the fine fraction by volume is 033(Figure 2(d)) The fine particle volume concentration is008120583m3120583m2 and coarse particle volume concentration is016 120583m3120583m2 Figure 3 shows the particle size distributionof cluster 5 described by a fine radius and standard devi-ation of 016 120583m and 048 and coarse radius and standarddeviation of 271 120583m and 065 As shown in Figure 4(e)most records of cluster 5 fall within sphericity parameterlower than 50 and coarse fraction by volume higher than05 Thus it indicates that cluster 5 is nonspherical coarseparticle dominated aerosol According to Okada et al (2001)nonspherical coarse particle is possibly contributed by dust[35] However there is no significant difference between SSAsat 440 nm and longer wavelengths But previous studies showthat when wavelength changes from 440 nm to longer theSSA will increase largely (Section 325) for dust aerosolBesides this kind of aerosol is frequently observed in spring(Section 34) in which season Beijing is affected by long-range transported dust Thus cluster 5 is possibly the dustmixes with anthropogenic aerosol As shown in Figure 4(e)cluster 5 also has some recordswith high sphericity parameteror low coarse fraction by volume Consequently we identifiedcluster 5 as polluted dust It should be noted that this polluteddust aerosol is the first time classified as a unique type bycluster analysis method The fact that polluted dust can beidentified by FCM demonstrates the practicability of fuzzyclustering in aerosol classification
325 Desert Dust Cluster 6 is significantly characterized bysmallest fine fraction by volume (017) and lowest sphericityparameter (651) Besides it shows the smallest fine particlevolume concentration (008120583m3120583m2) and largest coarseparticle volume concentration (016 120583m3120583m2) It indicates
Advances in Meteorology 7
0 25 50 75 10000
02
04
06
08
10C
oars
e fra
ctio
n by
vol
ume
Spherical parameter
(a) Cluster 1
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(b) Cluster 2
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(c) Cluster 3
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
eSpherical parameter
(d) Cluster 4
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(e) Cluster 5
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(f) Cluster 6
Figure 4 Relationships of sphericity parameter and coarse fraction by volume
that this is a completely coarse particle dominated aerosolIn addition as shown in Figure 4(f) most records of cluster6 concentrate in the range of sphericity parameter lowerthan 15 and coarse fraction by volume higher than 06 Itdemonstrates that cluster 6 is completely nonspherical coarseparticles dominated aerosol Moreover the SSAs of cluster 6exhibit a distinct feature with lower value at 440 nm (089)and a sharp increase to 094 095 and 095 at 660 nm 896 nmand 1020 nm (Figure 2(b)) respectively This is consistentwith the characteristic of desert dust derived by Levy et al(2007) [15] and Qin and Mitchell (2009) [16] Furthermoreaccording to Dubovik et al (2002) desert dust is possiblyabsorbed in the short spectral region due to the containediron oxide hematite [11] The SSA of dust particle might belower at short wavelength regions when it contains the FeO2Consequently these characteristics clearly certify that cluster6 is desert dust aerosol
326 Comparison with Other Results To assess the proposedfuzzy methodology we take a comparative study of our resultwith other published researches in aerosol classification
Considering study area and methodology two publishedresults are chosen global aerosol types by Omar et al (2005)[14] and East Asian aerosol types by Lee and Kim (2010) [17]
(1) Comparison with Global Aerosol TypesOmar et al (2005)applied cluster analysis to global ground-based observationsand obtained six categories backgroundrural industrialpollution dirty pollution biomass burning desert dustand polluted marine [14] Their characteristics are shownin Table 3 Five of the six aerosol types we classifiedshow similarities with Omarrsquos global aerosol types they arefine particle nonabsorbing aerosol to polluted continentalaerosol fine-MA2 to background aerosol fine particle highlyabsorbing aerosol to biomass burning aerosol and polluteddust and desert dust to dust
Figures 6 and 7 demonstrate comparisons of characteris-tics of aerosol types between global and Beijing aerosol typesincluding particle size distributions single scattering albedoand asymmetry parameter Particle size distributions (Fig-ure 6) show that the values of 119889119881119889 ln 119903 varied significantlybetween two researches The particle volume concentration
8 Advances in Meteorology
Table 3 Characteristics of global aerosol types by Omar et al (2005)
Properties (676 nm) Polluted continental Backgroundrural Biomass burning Dirty pollution Dust MarineSSA 092 088 080 072 093 093REFR 14098 14494 15202 14104 14520 13943REFI 00063 00092 00245 00337 00036 00044ASYM 0612 0580 0603 0594 0668 0711FFV 053 038 033 049 022 026
1 2 3 4 5 60
2
4
6
8
10
Num
ber o
f day
s
AQI-level
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 5 Numbers of each cluster observed in per-AQI level from2013 to 2014 at Beijing site
of Beijing is much higher than global aerosol types which isbecause Beijing usually suffers high concentration of aerosol[9] The SSAs of three fine particle aerosols of Beijing are095 091 and 087 which is larger than the value of globalaerosol 092 088 and 08 Interestingly all particle volumeconcentration SSA and asymmetry parameter of global dustare larger than polluted dust and smaller than desert dustwhich indicates that tow dust aerosol may be more ldquopurerdquothan dust in global aerosol types
(2) Comparison with East Asian Aerosol Types Lee and Kim(2010) performed a cluster analysis for AERONET recordsof East Asia which were divided into four fine-mode types(CAT (category) 1 to 4) and two coarse-mode types (CAT5and CAT6) [17] In addition the two coarse-modes werereferred to as dusty aerosols and the others were considereda mixed type of pollution Four of the six aerosol types inBeijing show similarities with Leersquos in East Asia they are fineparticle nonabsorbing aerosol to CAT2 fine particle highlyabsorbing aerosol to CAT3 and polluted dust and desert dustto two coarse-modes Table 4 shows the characteristics ofeach aerosol type identified by Lee et al (2010)
Figure 8 displays comparisons of characteristics of fouraerosol types between East Asia and Beijing Figures 8(a) and8(g) show that fine particle nonabsorbing aerosol of Beijingis stronger in particle scattering and lower in particle volume
concentration But the wavelength dependence of asymmetryparameter demonstrates similarities (Figure 8(d))The CAT3in East Asia and fine particle highly absorbing aerosol inBeijing exhibit similar particle size distributions and asym-metry parameter But the absorption of the latter is muchstronger than the former (Figure 8(b)) owing to the fossilfuel combustion in Beijing (see Section 323) It can beseen from Figures 8(c) 8(f) and 8(i) that the properties ofdesert dust in Beijing and CAT5 in East Asia are exactlysimilar to each other Meanwhile higher SSAs and particlevolume concentration of desert dust compared with CAT6probably due to sites close to dust source regions (like Yulinselected by Lee et al (2010)) were not included in thispaper It is interesting to note the pronounced differencesbetween polluted dust in Beijing and CAT5CAT6 in EastAsia indicating apparently polluted dust a unique aerosoltype
33 Variation of Optical Characteristics with MembershipDegrees The optical properties discussed in Section 32 arerepresented by the center values of each cluster But thedifferences between records at edge and near center ofcluster are inevitable Particularly for multiparticles mixtureaerosol only representing the aerosol by the center of clusterwill produce large errors As described in Section 22 themembership degree indicates the confidence degree of onerecord belonging to clusters which is an optimal parameterto analyze these particles at boundary ofmulticlustersThere-fore characteristics of aerosol (optical and microphysical)with different membership degree intervals are investigatedhere It should be noted thatwe focus on the internal variationof optical properties for each aerosol type
The membership degrees (MD) of six aerosol types aredivided into five intervals 017 le MD lt 027 027 le MD lt037 037 leMD lt 047 047 leMD lt 057 and 057 leMDWechose the value of 017 because when membership degree ofthe record to one type is larger than 16 (as there are six types)it means that this recordmay belong to that type Andwe takethe step of 01 to make sure that there is a similar amount ofrecords within each interval The numbers of records withineach interval are listed in Table 5
Figure 9 shows the AODs of six aerosol types varyingwith the membership degree and their mean values withineach membership degree interval at wavelength of 440 nmFine particle nonabsorbing aerosol (cluster 1 Figure 9(a))and fine-MA1 (cluster 2 Figure 9(b)) show large range ofAOD variation (from 0 to 4) The large range is reasonablebecause both types are frequently observed at polluted days(discussed in Section 32) Nevertheless the other four
Advances in Meteorology 9
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 1Continental
dV
dFH
r(
G3
G2)
(a)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 3Background
dV
dFH
r(
G3
G2)
(b)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 4Biomass
dV
dFH
r(
G3
G2)
(c)
000
005
010
015
020
025
030
Radius (G)
10minus1 100 101
Cluster 5Cluster 6Dust
dV
dFH
r(
G3
G2)
(d)
Figure 6 Size distribution comparison with global aerosol types
aerosol types show relatively smaller variation ranges (from0 to 2) And there are fewer that occurred during polluteddays Furthermore the AODs obviously concentrate in thelow varying range with increasing membership degree for allaerosol types
Figure 10 shows the mean values of AOD within the fivemembership degree intervals at four wavelengths (440 676869 and 1020 nm) Figures 10(a) and 10(b) show that theaverage AODs of fine particle moderately absorbing aerosolfor all wavelengths exhibit relatively flat behaviors Howeverthe other five aerosol types exhibit clearly declination withthe increase of themembership degree intervalMoreover thetrends are nearly the same between different wavelengths foreach aerosol type
The SSAs (440 nm) of six aerosol types varying withmembership degree and their mean values within eachmembership degree interval are plotted in Figure 11 Same
as AODs points of SSA concentrate in the low varyingrange with increasing membership degree for all aerosoltypes It can be seen from Figure 11(a) that the SSAs offine particle nonabsorbing aerosol (cluster 1) fall within therange between 09 and 10 This characteristic is consistentwith the nonabsorbing aerosol (strong scattering) of cluster1 Besides it is interesting to find that there are somestrongly absorbing particles (SSA lt 08) at membershipdegrees between 027 and 047 in Figure 11(d) (cluster 4) Asdiscussed in Section 323 these particles are probably theblack carbon which are frequently observed in Beijing withhigh absorption [23 28]
The mean values of SSA within the five membershipdegree intervals at four wavelengths (440 676 869 and1020 nm) are plotted in Figure 12 Means of SSA showclear descending trend in Figure 12(a) which indicates thatwith the increase of membership degree scattering ability of
10 Advances in Meteorology
Cluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
010203040506070809101112
Sing
le sc
atte
ring
albe
do
Aerosol typeCluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
00
01
02
03
04
05
06
07
08
Asy
mm
etry
fact
or
Aerosol type
This paperOmar et al
This paperOmar et al
Figure 7 SSA and ASYM comparison with global aerosol types
Table 4 Characteristics of East Asian aerosol types by Lee et al (2010)
Properties CAT11 CAT2 CAT3 CAT4 CAT5 CAT6
SSA
440 nm 0915 0927 0904 0908 0893 0900676 nm 0927 0941 0908 0914 0939 0957869 nm 0918 0933 0897 0907 0945 09641020 nm 0912 0928 0889 0903 0948 0966
REFR
440 nm 1468 1478 1441 1454 1508 1549676 nm 1480 1483 1458 1472 1535 1549869 nm 1485 1483 1468 1482 1536 15371020 nm 1481 1476 1468 1481 1528 1525
REFI
440 nm 00119 00099 00127 00112 0007 00049676 nm 00086 00074 001 00088 00037 00024869 nm 00088 00078 00102 0009 00036 000231020 nm 00091 0008 00104 00092 00036 00025
ASYM
440 nm 0721 0729 0716 0713 073 0748676 nm 0664 0685 0654 0655 0693 0714869 nm 0636 0657 0626 0635 0691 07071020 nm 0626 0643 0618 0634 0696 0707
1CAT category
fine particle nonabsorbing aerosol (cluster 1) is enhancingThis characteristic confirms that the higher the membershipdegree the purer the nonabsorbing aerosol Figures 12(b) and12(c) show the mean values of SSA obviously descendingwhen membership degree increases from MDI1 to MDI2However the changes are hardly observed when member-ship degree interval increases from MDI2 to MDI5 Thesecharacteristics demonstrate that records at the edge of cluster2 (fine-M1) and cluster 3 (fine-M2) have stronger scattering(or lower absorbing) than those near center Figure 12(d)exhibits clearly ascending trend of SSA means which is amanifestation of the reasonability to identify cluster 4 ashighly absorbing aerosol The ascending trend also indicatesthat the higher membership degree denotes the purer highlyabsorbing aerosol
Figures 12(e) and 12(f) show some similarities betweenpolluted dust and desert dust at wavelengths of 440 nmFirstly mean values of SSA slightly decrease with the increaseof membership degree interval Furthermore mean values ofSSA at 440 nm are obviously lower than other wavelengthsAs discussed in Sections 324 and 325 these SSA trendsagree well with the relationships between SSA and dustcompositions which certify our identification that bothcluster 5 (polluted dust) and cluster 6 (desert dust) aredust-related aerosols However there are some differencesbetween two clusters SSAs at 440 nm are distinctly smallerthan at other wavelengths Besides Figures 12(e) and 12(f)show different trends at wavelengths longer than 440 nm theformer displays descending trend while the latter shows noclear variation These differences can be attributed to cluster
Advances in Meteorology 11
400
500
600
700
800
900
1000
1100
085
090
095
100Si
ngle
scat
terin
g al
bedo
Wavelengths (nm)
This paper cluster 1Lee et al CAT2
(a)
085
090
095
100
Sing
le sc
atte
ring
albe
do
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(b)
085
090
095
100
Sing
le sc
atte
ring
albe
do
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(c)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 1Lee et al CAT2
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(d)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(e)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(f)
000
005
010
015
020
025
This paper cluster 1Lee et al CAT2
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(g)
000
005
010
015
This paper cluster 4Lee et al CAT3
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(h)
00010203040506070809
Radius (G)
10minus1 100 101
This paper
This papercluster 6
cluster 5Lee et al
Lee et alCAT5
CAT6
dV
dFH
r(
G3
G2)
(i)
Figure 8 Comparison with East Asian aerosol types
Table 5 The record numbers within each membership degree interval
Membership degree intervals (MDI) C11 C2 C3 C4 C5 C6MDI1 017 le MD2 lt 027 156 131 60 117 43 86MDI2 027 leMD lt 037 508 569 298 603 404 308MDI3 037 leMD lt 047 305 326 240 456 413 234MDI4 047 leMD lt 057 115 128 197 181 275 153MDI5 057 leMD 40 34 122 37 117 761C cluster 2MD membership degree
12 Advances in Meteorology
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 9 AODs at 440 nmwith membership degrees asterisk represents mean value of AOD in each interval and the vertical bars representthe standard deviation
5 (polluted dust) containing absorbing anthropogenic aerosol(such as black carbon) while cluster 6 is dominated onlyby dust These characteristics validate that it is reasonable toidentify cluster 5 as polluted dust and cluster 6 as desert dust
Figure 13 is scatterplot of sphericity parameter versusmembership degree for six aerosol types Figures 13(a) 13(b)13(c) and 13(d) show high variation of sphericity parameterswhich indicates that the first four aerosol types show nocorrelation between sphericity parameters and membershipdegree However Figure 13(e) exhibits clearly descendingtrends and all mean values of sphericity parameter are lower
than 40 Moreover records in Figure 13(f) are more con-centrated with 95 of them falling below 20 These indicatethat both cluster 5 and cluster 6 are close to nonsphericityBecause most dust types are nonsphericity particles [35]these analyses again confirm our identification of polluteddust (cluster 5) and desert dust (cluster 6)
The above analyses confirm reasonability of our results ofclustering and identification of aerosol types Moreover theinternal variation of optical properties of aerosol type can bewell investigated with the help ofmembership degree interval(MDI)
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
6 Advances in Meteorology
000
005
010
015
020
025
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
Figure 3 The particle size distributions of six cluster centers
and 148 According to Bohren and Huffman (1983) [34]dry particle basically has a higher REFR usually being15ndash16 That is consistent with the fact that the water vaporof fine-MA1 (113 gcm2) is lower than that of fine-MA2(134 gcm2) which can be seen in Figure 2(f) Additionallythe fine particle volume concentration of fine-MA1 is thelargest one in six aerosol types (017 120583m3120583m2) and thecoarse particle volume concentration is also found being highlevel (018 120583m3120583m2) Correspondingly both fine and coarseparticle volume concentration of fine-MA2 are much lower(both are 011 120583m3120583m2) than fine-MA1
Compared with Omar et al (2005) [14] fine-MA1 is likepolluted continental aerosol and dirty pollution aerosol Butthe absorption is lower than the former and higher than thelatter It can be seen from Figure 5 that fine-MA1 is the mostfrequently observed at AQI levels 5 and 6 Besides fine-MA1is just a little lower than fine particle nonabsorbing aerosol atAQI level 4 Furthermore fine-MA1 is hardly observed atAQIlevels 1 (excellent) and 2 (good) On the contrary fine-MA2 ischaracterized by a low optical depth (Figure 2(a)) and a highfrequency of incidence at lowAQI level when the atmosphereis expected to be relatively clean The characteristics aremore likely the backgroundrural aerosol classified by Omaret al (2005) [14] Consequently fine-MA2 is possibly thebackground aerosol of Beijing
323 Fine Particle Highly Absorbing Aerosol Cluster 4 is ofinterest in view of its high absorption The SSAs at fourwavelengths (440 676 869 and 1020 nm) are 084 087 085and 084 (Figure 2(b)) respectively Besides the fine fractionby volume is 045 (Figure 2(d)) The fine particle volumeconcentration is 009120583m3120583m2 and the coarse particle vol-ume concentration is 011120583m3120583m2 In addition Figure 3shows the particle size distribution of cluster 4 describedby a fine radius and standard deviation of 016 120583m and 051and coarse radius and standard deviation of 279 120583m and
063 Therefore we identified cluster 4 as fine particle highlyabsorbing aerosol
Figure 2(e) shows that the sphericity parameter (thehigher value indicates the particle is closer to sphericity)of fine particle highly absorbing aerosol exhibits the largestvalue (6427) Moreover it can be clearly seen from Fig-ure 4(d) that this kind of aerosol has much more recordswith sphericity parameter higher than 90 It demonstratesthat fine particle highly absorbing aerosol is more likelya spherical aerosol These characteristics are like biomassburning aerosol classified by Omar et al (2005) [14] Addi-tionally it can be seen from Table 2 that the real parts ofthe refractive index at a high level are 148 151 152 and153 at four wavelengths According to the study result ofDubovik et al (2002) [11] higher REFRs and lower SSAs arepossibly associated with high concentrations of black carbonFurthermore this kind of aerosol usually occurs in winter(Section 34) in which season coal is frequently used forheating supply in northern China [23 28] Consequently fineparticle highly absorbing aerosol is possibly mainly producedby fossil fuel combustion
324 Polluted Dust Cluster 5 has the largest number ofrecords with 207 of total records being classified inthis type The SSAs at four wavelengths (440 676 869and 1020 nm) are 088 091 090 and 090 (Figure 2(b))respectively Besides the fine fraction by volume is 033(Figure 2(d)) The fine particle volume concentration is008120583m3120583m2 and coarse particle volume concentration is016 120583m3120583m2 Figure 3 shows the particle size distributionof cluster 5 described by a fine radius and standard devi-ation of 016 120583m and 048 and coarse radius and standarddeviation of 271 120583m and 065 As shown in Figure 4(e)most records of cluster 5 fall within sphericity parameterlower than 50 and coarse fraction by volume higher than05 Thus it indicates that cluster 5 is nonspherical coarseparticle dominated aerosol According to Okada et al (2001)nonspherical coarse particle is possibly contributed by dust[35] However there is no significant difference between SSAsat 440 nm and longer wavelengths But previous studies showthat when wavelength changes from 440 nm to longer theSSA will increase largely (Section 325) for dust aerosolBesides this kind of aerosol is frequently observed in spring(Section 34) in which season Beijing is affected by long-range transported dust Thus cluster 5 is possibly the dustmixes with anthropogenic aerosol As shown in Figure 4(e)cluster 5 also has some recordswith high sphericity parameteror low coarse fraction by volume Consequently we identifiedcluster 5 as polluted dust It should be noted that this polluteddust aerosol is the first time classified as a unique type bycluster analysis method The fact that polluted dust can beidentified by FCM demonstrates the practicability of fuzzyclustering in aerosol classification
325 Desert Dust Cluster 6 is significantly characterized bysmallest fine fraction by volume (017) and lowest sphericityparameter (651) Besides it shows the smallest fine particlevolume concentration (008120583m3120583m2) and largest coarseparticle volume concentration (016 120583m3120583m2) It indicates
Advances in Meteorology 7
0 25 50 75 10000
02
04
06
08
10C
oars
e fra
ctio
n by
vol
ume
Spherical parameter
(a) Cluster 1
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(b) Cluster 2
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(c) Cluster 3
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
eSpherical parameter
(d) Cluster 4
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(e) Cluster 5
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(f) Cluster 6
Figure 4 Relationships of sphericity parameter and coarse fraction by volume
that this is a completely coarse particle dominated aerosolIn addition as shown in Figure 4(f) most records of cluster6 concentrate in the range of sphericity parameter lowerthan 15 and coarse fraction by volume higher than 06 Itdemonstrates that cluster 6 is completely nonspherical coarseparticles dominated aerosol Moreover the SSAs of cluster 6exhibit a distinct feature with lower value at 440 nm (089)and a sharp increase to 094 095 and 095 at 660 nm 896 nmand 1020 nm (Figure 2(b)) respectively This is consistentwith the characteristic of desert dust derived by Levy et al(2007) [15] and Qin and Mitchell (2009) [16] Furthermoreaccording to Dubovik et al (2002) desert dust is possiblyabsorbed in the short spectral region due to the containediron oxide hematite [11] The SSA of dust particle might belower at short wavelength regions when it contains the FeO2Consequently these characteristics clearly certify that cluster6 is desert dust aerosol
326 Comparison with Other Results To assess the proposedfuzzy methodology we take a comparative study of our resultwith other published researches in aerosol classification
Considering study area and methodology two publishedresults are chosen global aerosol types by Omar et al (2005)[14] and East Asian aerosol types by Lee and Kim (2010) [17]
(1) Comparison with Global Aerosol TypesOmar et al (2005)applied cluster analysis to global ground-based observationsand obtained six categories backgroundrural industrialpollution dirty pollution biomass burning desert dustand polluted marine [14] Their characteristics are shownin Table 3 Five of the six aerosol types we classifiedshow similarities with Omarrsquos global aerosol types they arefine particle nonabsorbing aerosol to polluted continentalaerosol fine-MA2 to background aerosol fine particle highlyabsorbing aerosol to biomass burning aerosol and polluteddust and desert dust to dust
Figures 6 and 7 demonstrate comparisons of characteris-tics of aerosol types between global and Beijing aerosol typesincluding particle size distributions single scattering albedoand asymmetry parameter Particle size distributions (Fig-ure 6) show that the values of 119889119881119889 ln 119903 varied significantlybetween two researches The particle volume concentration
8 Advances in Meteorology
Table 3 Characteristics of global aerosol types by Omar et al (2005)
Properties (676 nm) Polluted continental Backgroundrural Biomass burning Dirty pollution Dust MarineSSA 092 088 080 072 093 093REFR 14098 14494 15202 14104 14520 13943REFI 00063 00092 00245 00337 00036 00044ASYM 0612 0580 0603 0594 0668 0711FFV 053 038 033 049 022 026
1 2 3 4 5 60
2
4
6
8
10
Num
ber o
f day
s
AQI-level
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 5 Numbers of each cluster observed in per-AQI level from2013 to 2014 at Beijing site
of Beijing is much higher than global aerosol types which isbecause Beijing usually suffers high concentration of aerosol[9] The SSAs of three fine particle aerosols of Beijing are095 091 and 087 which is larger than the value of globalaerosol 092 088 and 08 Interestingly all particle volumeconcentration SSA and asymmetry parameter of global dustare larger than polluted dust and smaller than desert dustwhich indicates that tow dust aerosol may be more ldquopurerdquothan dust in global aerosol types
(2) Comparison with East Asian Aerosol Types Lee and Kim(2010) performed a cluster analysis for AERONET recordsof East Asia which were divided into four fine-mode types(CAT (category) 1 to 4) and two coarse-mode types (CAT5and CAT6) [17] In addition the two coarse-modes werereferred to as dusty aerosols and the others were considereda mixed type of pollution Four of the six aerosol types inBeijing show similarities with Leersquos in East Asia they are fineparticle nonabsorbing aerosol to CAT2 fine particle highlyabsorbing aerosol to CAT3 and polluted dust and desert dustto two coarse-modes Table 4 shows the characteristics ofeach aerosol type identified by Lee et al (2010)
Figure 8 displays comparisons of characteristics of fouraerosol types between East Asia and Beijing Figures 8(a) and8(g) show that fine particle nonabsorbing aerosol of Beijingis stronger in particle scattering and lower in particle volume
concentration But the wavelength dependence of asymmetryparameter demonstrates similarities (Figure 8(d))The CAT3in East Asia and fine particle highly absorbing aerosol inBeijing exhibit similar particle size distributions and asym-metry parameter But the absorption of the latter is muchstronger than the former (Figure 8(b)) owing to the fossilfuel combustion in Beijing (see Section 323) It can beseen from Figures 8(c) 8(f) and 8(i) that the properties ofdesert dust in Beijing and CAT5 in East Asia are exactlysimilar to each other Meanwhile higher SSAs and particlevolume concentration of desert dust compared with CAT6probably due to sites close to dust source regions (like Yulinselected by Lee et al (2010)) were not included in thispaper It is interesting to note the pronounced differencesbetween polluted dust in Beijing and CAT5CAT6 in EastAsia indicating apparently polluted dust a unique aerosoltype
33 Variation of Optical Characteristics with MembershipDegrees The optical properties discussed in Section 32 arerepresented by the center values of each cluster But thedifferences between records at edge and near center ofcluster are inevitable Particularly for multiparticles mixtureaerosol only representing the aerosol by the center of clusterwill produce large errors As described in Section 22 themembership degree indicates the confidence degree of onerecord belonging to clusters which is an optimal parameterto analyze these particles at boundary ofmulticlustersThere-fore characteristics of aerosol (optical and microphysical)with different membership degree intervals are investigatedhere It should be noted thatwe focus on the internal variationof optical properties for each aerosol type
The membership degrees (MD) of six aerosol types aredivided into five intervals 017 le MD lt 027 027 le MD lt037 037 leMD lt 047 047 leMD lt 057 and 057 leMDWechose the value of 017 because when membership degree ofthe record to one type is larger than 16 (as there are six types)it means that this recordmay belong to that type Andwe takethe step of 01 to make sure that there is a similar amount ofrecords within each interval The numbers of records withineach interval are listed in Table 5
Figure 9 shows the AODs of six aerosol types varyingwith the membership degree and their mean values withineach membership degree interval at wavelength of 440 nmFine particle nonabsorbing aerosol (cluster 1 Figure 9(a))and fine-MA1 (cluster 2 Figure 9(b)) show large range ofAOD variation (from 0 to 4) The large range is reasonablebecause both types are frequently observed at polluted days(discussed in Section 32) Nevertheless the other four
Advances in Meteorology 9
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 1Continental
dV
dFH
r(
G3
G2)
(a)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 3Background
dV
dFH
r(
G3
G2)
(b)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 4Biomass
dV
dFH
r(
G3
G2)
(c)
000
005
010
015
020
025
030
Radius (G)
10minus1 100 101
Cluster 5Cluster 6Dust
dV
dFH
r(
G3
G2)
(d)
Figure 6 Size distribution comparison with global aerosol types
aerosol types show relatively smaller variation ranges (from0 to 2) And there are fewer that occurred during polluteddays Furthermore the AODs obviously concentrate in thelow varying range with increasing membership degree for allaerosol types
Figure 10 shows the mean values of AOD within the fivemembership degree intervals at four wavelengths (440 676869 and 1020 nm) Figures 10(a) and 10(b) show that theaverage AODs of fine particle moderately absorbing aerosolfor all wavelengths exhibit relatively flat behaviors Howeverthe other five aerosol types exhibit clearly declination withthe increase of themembership degree intervalMoreover thetrends are nearly the same between different wavelengths foreach aerosol type
The SSAs (440 nm) of six aerosol types varying withmembership degree and their mean values within eachmembership degree interval are plotted in Figure 11 Same
as AODs points of SSA concentrate in the low varyingrange with increasing membership degree for all aerosoltypes It can be seen from Figure 11(a) that the SSAs offine particle nonabsorbing aerosol (cluster 1) fall within therange between 09 and 10 This characteristic is consistentwith the nonabsorbing aerosol (strong scattering) of cluster1 Besides it is interesting to find that there are somestrongly absorbing particles (SSA lt 08) at membershipdegrees between 027 and 047 in Figure 11(d) (cluster 4) Asdiscussed in Section 323 these particles are probably theblack carbon which are frequently observed in Beijing withhigh absorption [23 28]
The mean values of SSA within the five membershipdegree intervals at four wavelengths (440 676 869 and1020 nm) are plotted in Figure 12 Means of SSA showclear descending trend in Figure 12(a) which indicates thatwith the increase of membership degree scattering ability of
10 Advances in Meteorology
Cluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
010203040506070809101112
Sing
le sc
atte
ring
albe
do
Aerosol typeCluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
00
01
02
03
04
05
06
07
08
Asy
mm
etry
fact
or
Aerosol type
This paperOmar et al
This paperOmar et al
Figure 7 SSA and ASYM comparison with global aerosol types
Table 4 Characteristics of East Asian aerosol types by Lee et al (2010)
Properties CAT11 CAT2 CAT3 CAT4 CAT5 CAT6
SSA
440 nm 0915 0927 0904 0908 0893 0900676 nm 0927 0941 0908 0914 0939 0957869 nm 0918 0933 0897 0907 0945 09641020 nm 0912 0928 0889 0903 0948 0966
REFR
440 nm 1468 1478 1441 1454 1508 1549676 nm 1480 1483 1458 1472 1535 1549869 nm 1485 1483 1468 1482 1536 15371020 nm 1481 1476 1468 1481 1528 1525
REFI
440 nm 00119 00099 00127 00112 0007 00049676 nm 00086 00074 001 00088 00037 00024869 nm 00088 00078 00102 0009 00036 000231020 nm 00091 0008 00104 00092 00036 00025
ASYM
440 nm 0721 0729 0716 0713 073 0748676 nm 0664 0685 0654 0655 0693 0714869 nm 0636 0657 0626 0635 0691 07071020 nm 0626 0643 0618 0634 0696 0707
1CAT category
fine particle nonabsorbing aerosol (cluster 1) is enhancingThis characteristic confirms that the higher the membershipdegree the purer the nonabsorbing aerosol Figures 12(b) and12(c) show the mean values of SSA obviously descendingwhen membership degree increases from MDI1 to MDI2However the changes are hardly observed when member-ship degree interval increases from MDI2 to MDI5 Thesecharacteristics demonstrate that records at the edge of cluster2 (fine-M1) and cluster 3 (fine-M2) have stronger scattering(or lower absorbing) than those near center Figure 12(d)exhibits clearly ascending trend of SSA means which is amanifestation of the reasonability to identify cluster 4 ashighly absorbing aerosol The ascending trend also indicatesthat the higher membership degree denotes the purer highlyabsorbing aerosol
Figures 12(e) and 12(f) show some similarities betweenpolluted dust and desert dust at wavelengths of 440 nmFirstly mean values of SSA slightly decrease with the increaseof membership degree interval Furthermore mean values ofSSA at 440 nm are obviously lower than other wavelengthsAs discussed in Sections 324 and 325 these SSA trendsagree well with the relationships between SSA and dustcompositions which certify our identification that bothcluster 5 (polluted dust) and cluster 6 (desert dust) aredust-related aerosols However there are some differencesbetween two clusters SSAs at 440 nm are distinctly smallerthan at other wavelengths Besides Figures 12(e) and 12(f)show different trends at wavelengths longer than 440 nm theformer displays descending trend while the latter shows noclear variation These differences can be attributed to cluster
Advances in Meteorology 11
400
500
600
700
800
900
1000
1100
085
090
095
100Si
ngle
scat
terin
g al
bedo
Wavelengths (nm)
This paper cluster 1Lee et al CAT2
(a)
085
090
095
100
Sing
le sc
atte
ring
albe
do
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(b)
085
090
095
100
Sing
le sc
atte
ring
albe
do
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(c)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 1Lee et al CAT2
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(d)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(e)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(f)
000
005
010
015
020
025
This paper cluster 1Lee et al CAT2
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(g)
000
005
010
015
This paper cluster 4Lee et al CAT3
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(h)
00010203040506070809
Radius (G)
10minus1 100 101
This paper
This papercluster 6
cluster 5Lee et al
Lee et alCAT5
CAT6
dV
dFH
r(
G3
G2)
(i)
Figure 8 Comparison with East Asian aerosol types
Table 5 The record numbers within each membership degree interval
Membership degree intervals (MDI) C11 C2 C3 C4 C5 C6MDI1 017 le MD2 lt 027 156 131 60 117 43 86MDI2 027 leMD lt 037 508 569 298 603 404 308MDI3 037 leMD lt 047 305 326 240 456 413 234MDI4 047 leMD lt 057 115 128 197 181 275 153MDI5 057 leMD 40 34 122 37 117 761C cluster 2MD membership degree
12 Advances in Meteorology
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 9 AODs at 440 nmwith membership degrees asterisk represents mean value of AOD in each interval and the vertical bars representthe standard deviation
5 (polluted dust) containing absorbing anthropogenic aerosol(such as black carbon) while cluster 6 is dominated onlyby dust These characteristics validate that it is reasonable toidentify cluster 5 as polluted dust and cluster 6 as desert dust
Figure 13 is scatterplot of sphericity parameter versusmembership degree for six aerosol types Figures 13(a) 13(b)13(c) and 13(d) show high variation of sphericity parameterswhich indicates that the first four aerosol types show nocorrelation between sphericity parameters and membershipdegree However Figure 13(e) exhibits clearly descendingtrends and all mean values of sphericity parameter are lower
than 40 Moreover records in Figure 13(f) are more con-centrated with 95 of them falling below 20 These indicatethat both cluster 5 and cluster 6 are close to nonsphericityBecause most dust types are nonsphericity particles [35]these analyses again confirm our identification of polluteddust (cluster 5) and desert dust (cluster 6)
The above analyses confirm reasonability of our results ofclustering and identification of aerosol types Moreover theinternal variation of optical properties of aerosol type can bewell investigated with the help ofmembership degree interval(MDI)
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 7
0 25 50 75 10000
02
04
06
08
10C
oars
e fra
ctio
n by
vol
ume
Spherical parameter
(a) Cluster 1
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(b) Cluster 2
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(c) Cluster 3
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
eSpherical parameter
(d) Cluster 4
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(e) Cluster 5
0 25 50 75 10000
02
04
06
08
10
Coa
rse f
ract
ion
by v
olum
e
Spherical parameter
(f) Cluster 6
Figure 4 Relationships of sphericity parameter and coarse fraction by volume
that this is a completely coarse particle dominated aerosolIn addition as shown in Figure 4(f) most records of cluster6 concentrate in the range of sphericity parameter lowerthan 15 and coarse fraction by volume higher than 06 Itdemonstrates that cluster 6 is completely nonspherical coarseparticles dominated aerosol Moreover the SSAs of cluster 6exhibit a distinct feature with lower value at 440 nm (089)and a sharp increase to 094 095 and 095 at 660 nm 896 nmand 1020 nm (Figure 2(b)) respectively This is consistentwith the characteristic of desert dust derived by Levy et al(2007) [15] and Qin and Mitchell (2009) [16] Furthermoreaccording to Dubovik et al (2002) desert dust is possiblyabsorbed in the short spectral region due to the containediron oxide hematite [11] The SSA of dust particle might belower at short wavelength regions when it contains the FeO2Consequently these characteristics clearly certify that cluster6 is desert dust aerosol
326 Comparison with Other Results To assess the proposedfuzzy methodology we take a comparative study of our resultwith other published researches in aerosol classification
Considering study area and methodology two publishedresults are chosen global aerosol types by Omar et al (2005)[14] and East Asian aerosol types by Lee and Kim (2010) [17]
(1) Comparison with Global Aerosol TypesOmar et al (2005)applied cluster analysis to global ground-based observationsand obtained six categories backgroundrural industrialpollution dirty pollution biomass burning desert dustand polluted marine [14] Their characteristics are shownin Table 3 Five of the six aerosol types we classifiedshow similarities with Omarrsquos global aerosol types they arefine particle nonabsorbing aerosol to polluted continentalaerosol fine-MA2 to background aerosol fine particle highlyabsorbing aerosol to biomass burning aerosol and polluteddust and desert dust to dust
Figures 6 and 7 demonstrate comparisons of characteris-tics of aerosol types between global and Beijing aerosol typesincluding particle size distributions single scattering albedoand asymmetry parameter Particle size distributions (Fig-ure 6) show that the values of 119889119881119889 ln 119903 varied significantlybetween two researches The particle volume concentration
8 Advances in Meteorology
Table 3 Characteristics of global aerosol types by Omar et al (2005)
Properties (676 nm) Polluted continental Backgroundrural Biomass burning Dirty pollution Dust MarineSSA 092 088 080 072 093 093REFR 14098 14494 15202 14104 14520 13943REFI 00063 00092 00245 00337 00036 00044ASYM 0612 0580 0603 0594 0668 0711FFV 053 038 033 049 022 026
1 2 3 4 5 60
2
4
6
8
10
Num
ber o
f day
s
AQI-level
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 5 Numbers of each cluster observed in per-AQI level from2013 to 2014 at Beijing site
of Beijing is much higher than global aerosol types which isbecause Beijing usually suffers high concentration of aerosol[9] The SSAs of three fine particle aerosols of Beijing are095 091 and 087 which is larger than the value of globalaerosol 092 088 and 08 Interestingly all particle volumeconcentration SSA and asymmetry parameter of global dustare larger than polluted dust and smaller than desert dustwhich indicates that tow dust aerosol may be more ldquopurerdquothan dust in global aerosol types
(2) Comparison with East Asian Aerosol Types Lee and Kim(2010) performed a cluster analysis for AERONET recordsof East Asia which were divided into four fine-mode types(CAT (category) 1 to 4) and two coarse-mode types (CAT5and CAT6) [17] In addition the two coarse-modes werereferred to as dusty aerosols and the others were considereda mixed type of pollution Four of the six aerosol types inBeijing show similarities with Leersquos in East Asia they are fineparticle nonabsorbing aerosol to CAT2 fine particle highlyabsorbing aerosol to CAT3 and polluted dust and desert dustto two coarse-modes Table 4 shows the characteristics ofeach aerosol type identified by Lee et al (2010)
Figure 8 displays comparisons of characteristics of fouraerosol types between East Asia and Beijing Figures 8(a) and8(g) show that fine particle nonabsorbing aerosol of Beijingis stronger in particle scattering and lower in particle volume
concentration But the wavelength dependence of asymmetryparameter demonstrates similarities (Figure 8(d))The CAT3in East Asia and fine particle highly absorbing aerosol inBeijing exhibit similar particle size distributions and asym-metry parameter But the absorption of the latter is muchstronger than the former (Figure 8(b)) owing to the fossilfuel combustion in Beijing (see Section 323) It can beseen from Figures 8(c) 8(f) and 8(i) that the properties ofdesert dust in Beijing and CAT5 in East Asia are exactlysimilar to each other Meanwhile higher SSAs and particlevolume concentration of desert dust compared with CAT6probably due to sites close to dust source regions (like Yulinselected by Lee et al (2010)) were not included in thispaper It is interesting to note the pronounced differencesbetween polluted dust in Beijing and CAT5CAT6 in EastAsia indicating apparently polluted dust a unique aerosoltype
33 Variation of Optical Characteristics with MembershipDegrees The optical properties discussed in Section 32 arerepresented by the center values of each cluster But thedifferences between records at edge and near center ofcluster are inevitable Particularly for multiparticles mixtureaerosol only representing the aerosol by the center of clusterwill produce large errors As described in Section 22 themembership degree indicates the confidence degree of onerecord belonging to clusters which is an optimal parameterto analyze these particles at boundary ofmulticlustersThere-fore characteristics of aerosol (optical and microphysical)with different membership degree intervals are investigatedhere It should be noted thatwe focus on the internal variationof optical properties for each aerosol type
The membership degrees (MD) of six aerosol types aredivided into five intervals 017 le MD lt 027 027 le MD lt037 037 leMD lt 047 047 leMD lt 057 and 057 leMDWechose the value of 017 because when membership degree ofthe record to one type is larger than 16 (as there are six types)it means that this recordmay belong to that type Andwe takethe step of 01 to make sure that there is a similar amount ofrecords within each interval The numbers of records withineach interval are listed in Table 5
Figure 9 shows the AODs of six aerosol types varyingwith the membership degree and their mean values withineach membership degree interval at wavelength of 440 nmFine particle nonabsorbing aerosol (cluster 1 Figure 9(a))and fine-MA1 (cluster 2 Figure 9(b)) show large range ofAOD variation (from 0 to 4) The large range is reasonablebecause both types are frequently observed at polluted days(discussed in Section 32) Nevertheless the other four
Advances in Meteorology 9
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 1Continental
dV
dFH
r(
G3
G2)
(a)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 3Background
dV
dFH
r(
G3
G2)
(b)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 4Biomass
dV
dFH
r(
G3
G2)
(c)
000
005
010
015
020
025
030
Radius (G)
10minus1 100 101
Cluster 5Cluster 6Dust
dV
dFH
r(
G3
G2)
(d)
Figure 6 Size distribution comparison with global aerosol types
aerosol types show relatively smaller variation ranges (from0 to 2) And there are fewer that occurred during polluteddays Furthermore the AODs obviously concentrate in thelow varying range with increasing membership degree for allaerosol types
Figure 10 shows the mean values of AOD within the fivemembership degree intervals at four wavelengths (440 676869 and 1020 nm) Figures 10(a) and 10(b) show that theaverage AODs of fine particle moderately absorbing aerosolfor all wavelengths exhibit relatively flat behaviors Howeverthe other five aerosol types exhibit clearly declination withthe increase of themembership degree intervalMoreover thetrends are nearly the same between different wavelengths foreach aerosol type
The SSAs (440 nm) of six aerosol types varying withmembership degree and their mean values within eachmembership degree interval are plotted in Figure 11 Same
as AODs points of SSA concentrate in the low varyingrange with increasing membership degree for all aerosoltypes It can be seen from Figure 11(a) that the SSAs offine particle nonabsorbing aerosol (cluster 1) fall within therange between 09 and 10 This characteristic is consistentwith the nonabsorbing aerosol (strong scattering) of cluster1 Besides it is interesting to find that there are somestrongly absorbing particles (SSA lt 08) at membershipdegrees between 027 and 047 in Figure 11(d) (cluster 4) Asdiscussed in Section 323 these particles are probably theblack carbon which are frequently observed in Beijing withhigh absorption [23 28]
The mean values of SSA within the five membershipdegree intervals at four wavelengths (440 676 869 and1020 nm) are plotted in Figure 12 Means of SSA showclear descending trend in Figure 12(a) which indicates thatwith the increase of membership degree scattering ability of
10 Advances in Meteorology
Cluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
010203040506070809101112
Sing
le sc
atte
ring
albe
do
Aerosol typeCluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
00
01
02
03
04
05
06
07
08
Asy
mm
etry
fact
or
Aerosol type
This paperOmar et al
This paperOmar et al
Figure 7 SSA and ASYM comparison with global aerosol types
Table 4 Characteristics of East Asian aerosol types by Lee et al (2010)
Properties CAT11 CAT2 CAT3 CAT4 CAT5 CAT6
SSA
440 nm 0915 0927 0904 0908 0893 0900676 nm 0927 0941 0908 0914 0939 0957869 nm 0918 0933 0897 0907 0945 09641020 nm 0912 0928 0889 0903 0948 0966
REFR
440 nm 1468 1478 1441 1454 1508 1549676 nm 1480 1483 1458 1472 1535 1549869 nm 1485 1483 1468 1482 1536 15371020 nm 1481 1476 1468 1481 1528 1525
REFI
440 nm 00119 00099 00127 00112 0007 00049676 nm 00086 00074 001 00088 00037 00024869 nm 00088 00078 00102 0009 00036 000231020 nm 00091 0008 00104 00092 00036 00025
ASYM
440 nm 0721 0729 0716 0713 073 0748676 nm 0664 0685 0654 0655 0693 0714869 nm 0636 0657 0626 0635 0691 07071020 nm 0626 0643 0618 0634 0696 0707
1CAT category
fine particle nonabsorbing aerosol (cluster 1) is enhancingThis characteristic confirms that the higher the membershipdegree the purer the nonabsorbing aerosol Figures 12(b) and12(c) show the mean values of SSA obviously descendingwhen membership degree increases from MDI1 to MDI2However the changes are hardly observed when member-ship degree interval increases from MDI2 to MDI5 Thesecharacteristics demonstrate that records at the edge of cluster2 (fine-M1) and cluster 3 (fine-M2) have stronger scattering(or lower absorbing) than those near center Figure 12(d)exhibits clearly ascending trend of SSA means which is amanifestation of the reasonability to identify cluster 4 ashighly absorbing aerosol The ascending trend also indicatesthat the higher membership degree denotes the purer highlyabsorbing aerosol
Figures 12(e) and 12(f) show some similarities betweenpolluted dust and desert dust at wavelengths of 440 nmFirstly mean values of SSA slightly decrease with the increaseof membership degree interval Furthermore mean values ofSSA at 440 nm are obviously lower than other wavelengthsAs discussed in Sections 324 and 325 these SSA trendsagree well with the relationships between SSA and dustcompositions which certify our identification that bothcluster 5 (polluted dust) and cluster 6 (desert dust) aredust-related aerosols However there are some differencesbetween two clusters SSAs at 440 nm are distinctly smallerthan at other wavelengths Besides Figures 12(e) and 12(f)show different trends at wavelengths longer than 440 nm theformer displays descending trend while the latter shows noclear variation These differences can be attributed to cluster
Advances in Meteorology 11
400
500
600
700
800
900
1000
1100
085
090
095
100Si
ngle
scat
terin
g al
bedo
Wavelengths (nm)
This paper cluster 1Lee et al CAT2
(a)
085
090
095
100
Sing
le sc
atte
ring
albe
do
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(b)
085
090
095
100
Sing
le sc
atte
ring
albe
do
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(c)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 1Lee et al CAT2
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(d)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(e)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(f)
000
005
010
015
020
025
This paper cluster 1Lee et al CAT2
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(g)
000
005
010
015
This paper cluster 4Lee et al CAT3
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(h)
00010203040506070809
Radius (G)
10minus1 100 101
This paper
This papercluster 6
cluster 5Lee et al
Lee et alCAT5
CAT6
dV
dFH
r(
G3
G2)
(i)
Figure 8 Comparison with East Asian aerosol types
Table 5 The record numbers within each membership degree interval
Membership degree intervals (MDI) C11 C2 C3 C4 C5 C6MDI1 017 le MD2 lt 027 156 131 60 117 43 86MDI2 027 leMD lt 037 508 569 298 603 404 308MDI3 037 leMD lt 047 305 326 240 456 413 234MDI4 047 leMD lt 057 115 128 197 181 275 153MDI5 057 leMD 40 34 122 37 117 761C cluster 2MD membership degree
12 Advances in Meteorology
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 9 AODs at 440 nmwith membership degrees asterisk represents mean value of AOD in each interval and the vertical bars representthe standard deviation
5 (polluted dust) containing absorbing anthropogenic aerosol(such as black carbon) while cluster 6 is dominated onlyby dust These characteristics validate that it is reasonable toidentify cluster 5 as polluted dust and cluster 6 as desert dust
Figure 13 is scatterplot of sphericity parameter versusmembership degree for six aerosol types Figures 13(a) 13(b)13(c) and 13(d) show high variation of sphericity parameterswhich indicates that the first four aerosol types show nocorrelation between sphericity parameters and membershipdegree However Figure 13(e) exhibits clearly descendingtrends and all mean values of sphericity parameter are lower
than 40 Moreover records in Figure 13(f) are more con-centrated with 95 of them falling below 20 These indicatethat both cluster 5 and cluster 6 are close to nonsphericityBecause most dust types are nonsphericity particles [35]these analyses again confirm our identification of polluteddust (cluster 5) and desert dust (cluster 6)
The above analyses confirm reasonability of our results ofclustering and identification of aerosol types Moreover theinternal variation of optical properties of aerosol type can bewell investigated with the help ofmembership degree interval(MDI)
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
8 Advances in Meteorology
Table 3 Characteristics of global aerosol types by Omar et al (2005)
Properties (676 nm) Polluted continental Backgroundrural Biomass burning Dirty pollution Dust MarineSSA 092 088 080 072 093 093REFR 14098 14494 15202 14104 14520 13943REFI 00063 00092 00245 00337 00036 00044ASYM 0612 0580 0603 0594 0668 0711FFV 053 038 033 049 022 026
1 2 3 4 5 60
2
4
6
8
10
Num
ber o
f day
s
AQI-level
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 5 Numbers of each cluster observed in per-AQI level from2013 to 2014 at Beijing site
of Beijing is much higher than global aerosol types which isbecause Beijing usually suffers high concentration of aerosol[9] The SSAs of three fine particle aerosols of Beijing are095 091 and 087 which is larger than the value of globalaerosol 092 088 and 08 Interestingly all particle volumeconcentration SSA and asymmetry parameter of global dustare larger than polluted dust and smaller than desert dustwhich indicates that tow dust aerosol may be more ldquopurerdquothan dust in global aerosol types
(2) Comparison with East Asian Aerosol Types Lee and Kim(2010) performed a cluster analysis for AERONET recordsof East Asia which were divided into four fine-mode types(CAT (category) 1 to 4) and two coarse-mode types (CAT5and CAT6) [17] In addition the two coarse-modes werereferred to as dusty aerosols and the others were considereda mixed type of pollution Four of the six aerosol types inBeijing show similarities with Leersquos in East Asia they are fineparticle nonabsorbing aerosol to CAT2 fine particle highlyabsorbing aerosol to CAT3 and polluted dust and desert dustto two coarse-modes Table 4 shows the characteristics ofeach aerosol type identified by Lee et al (2010)
Figure 8 displays comparisons of characteristics of fouraerosol types between East Asia and Beijing Figures 8(a) and8(g) show that fine particle nonabsorbing aerosol of Beijingis stronger in particle scattering and lower in particle volume
concentration But the wavelength dependence of asymmetryparameter demonstrates similarities (Figure 8(d))The CAT3in East Asia and fine particle highly absorbing aerosol inBeijing exhibit similar particle size distributions and asym-metry parameter But the absorption of the latter is muchstronger than the former (Figure 8(b)) owing to the fossilfuel combustion in Beijing (see Section 323) It can beseen from Figures 8(c) 8(f) and 8(i) that the properties ofdesert dust in Beijing and CAT5 in East Asia are exactlysimilar to each other Meanwhile higher SSAs and particlevolume concentration of desert dust compared with CAT6probably due to sites close to dust source regions (like Yulinselected by Lee et al (2010)) were not included in thispaper It is interesting to note the pronounced differencesbetween polluted dust in Beijing and CAT5CAT6 in EastAsia indicating apparently polluted dust a unique aerosoltype
33 Variation of Optical Characteristics with MembershipDegrees The optical properties discussed in Section 32 arerepresented by the center values of each cluster But thedifferences between records at edge and near center ofcluster are inevitable Particularly for multiparticles mixtureaerosol only representing the aerosol by the center of clusterwill produce large errors As described in Section 22 themembership degree indicates the confidence degree of onerecord belonging to clusters which is an optimal parameterto analyze these particles at boundary ofmulticlustersThere-fore characteristics of aerosol (optical and microphysical)with different membership degree intervals are investigatedhere It should be noted thatwe focus on the internal variationof optical properties for each aerosol type
The membership degrees (MD) of six aerosol types aredivided into five intervals 017 le MD lt 027 027 le MD lt037 037 leMD lt 047 047 leMD lt 057 and 057 leMDWechose the value of 017 because when membership degree ofthe record to one type is larger than 16 (as there are six types)it means that this recordmay belong to that type Andwe takethe step of 01 to make sure that there is a similar amount ofrecords within each interval The numbers of records withineach interval are listed in Table 5
Figure 9 shows the AODs of six aerosol types varyingwith the membership degree and their mean values withineach membership degree interval at wavelength of 440 nmFine particle nonabsorbing aerosol (cluster 1 Figure 9(a))and fine-MA1 (cluster 2 Figure 9(b)) show large range ofAOD variation (from 0 to 4) The large range is reasonablebecause both types are frequently observed at polluted days(discussed in Section 32) Nevertheless the other four
Advances in Meteorology 9
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 1Continental
dV
dFH
r(
G3
G2)
(a)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 3Background
dV
dFH
r(
G3
G2)
(b)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 4Biomass
dV
dFH
r(
G3
G2)
(c)
000
005
010
015
020
025
030
Radius (G)
10minus1 100 101
Cluster 5Cluster 6Dust
dV
dFH
r(
G3
G2)
(d)
Figure 6 Size distribution comparison with global aerosol types
aerosol types show relatively smaller variation ranges (from0 to 2) And there are fewer that occurred during polluteddays Furthermore the AODs obviously concentrate in thelow varying range with increasing membership degree for allaerosol types
Figure 10 shows the mean values of AOD within the fivemembership degree intervals at four wavelengths (440 676869 and 1020 nm) Figures 10(a) and 10(b) show that theaverage AODs of fine particle moderately absorbing aerosolfor all wavelengths exhibit relatively flat behaviors Howeverthe other five aerosol types exhibit clearly declination withthe increase of themembership degree intervalMoreover thetrends are nearly the same between different wavelengths foreach aerosol type
The SSAs (440 nm) of six aerosol types varying withmembership degree and their mean values within eachmembership degree interval are plotted in Figure 11 Same
as AODs points of SSA concentrate in the low varyingrange with increasing membership degree for all aerosoltypes It can be seen from Figure 11(a) that the SSAs offine particle nonabsorbing aerosol (cluster 1) fall within therange between 09 and 10 This characteristic is consistentwith the nonabsorbing aerosol (strong scattering) of cluster1 Besides it is interesting to find that there are somestrongly absorbing particles (SSA lt 08) at membershipdegrees between 027 and 047 in Figure 11(d) (cluster 4) Asdiscussed in Section 323 these particles are probably theblack carbon which are frequently observed in Beijing withhigh absorption [23 28]
The mean values of SSA within the five membershipdegree intervals at four wavelengths (440 676 869 and1020 nm) are plotted in Figure 12 Means of SSA showclear descending trend in Figure 12(a) which indicates thatwith the increase of membership degree scattering ability of
10 Advances in Meteorology
Cluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
010203040506070809101112
Sing
le sc
atte
ring
albe
do
Aerosol typeCluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
00
01
02
03
04
05
06
07
08
Asy
mm
etry
fact
or
Aerosol type
This paperOmar et al
This paperOmar et al
Figure 7 SSA and ASYM comparison with global aerosol types
Table 4 Characteristics of East Asian aerosol types by Lee et al (2010)
Properties CAT11 CAT2 CAT3 CAT4 CAT5 CAT6
SSA
440 nm 0915 0927 0904 0908 0893 0900676 nm 0927 0941 0908 0914 0939 0957869 nm 0918 0933 0897 0907 0945 09641020 nm 0912 0928 0889 0903 0948 0966
REFR
440 nm 1468 1478 1441 1454 1508 1549676 nm 1480 1483 1458 1472 1535 1549869 nm 1485 1483 1468 1482 1536 15371020 nm 1481 1476 1468 1481 1528 1525
REFI
440 nm 00119 00099 00127 00112 0007 00049676 nm 00086 00074 001 00088 00037 00024869 nm 00088 00078 00102 0009 00036 000231020 nm 00091 0008 00104 00092 00036 00025
ASYM
440 nm 0721 0729 0716 0713 073 0748676 nm 0664 0685 0654 0655 0693 0714869 nm 0636 0657 0626 0635 0691 07071020 nm 0626 0643 0618 0634 0696 0707
1CAT category
fine particle nonabsorbing aerosol (cluster 1) is enhancingThis characteristic confirms that the higher the membershipdegree the purer the nonabsorbing aerosol Figures 12(b) and12(c) show the mean values of SSA obviously descendingwhen membership degree increases from MDI1 to MDI2However the changes are hardly observed when member-ship degree interval increases from MDI2 to MDI5 Thesecharacteristics demonstrate that records at the edge of cluster2 (fine-M1) and cluster 3 (fine-M2) have stronger scattering(or lower absorbing) than those near center Figure 12(d)exhibits clearly ascending trend of SSA means which is amanifestation of the reasonability to identify cluster 4 ashighly absorbing aerosol The ascending trend also indicatesthat the higher membership degree denotes the purer highlyabsorbing aerosol
Figures 12(e) and 12(f) show some similarities betweenpolluted dust and desert dust at wavelengths of 440 nmFirstly mean values of SSA slightly decrease with the increaseof membership degree interval Furthermore mean values ofSSA at 440 nm are obviously lower than other wavelengthsAs discussed in Sections 324 and 325 these SSA trendsagree well with the relationships between SSA and dustcompositions which certify our identification that bothcluster 5 (polluted dust) and cluster 6 (desert dust) aredust-related aerosols However there are some differencesbetween two clusters SSAs at 440 nm are distinctly smallerthan at other wavelengths Besides Figures 12(e) and 12(f)show different trends at wavelengths longer than 440 nm theformer displays descending trend while the latter shows noclear variation These differences can be attributed to cluster
Advances in Meteorology 11
400
500
600
700
800
900
1000
1100
085
090
095
100Si
ngle
scat
terin
g al
bedo
Wavelengths (nm)
This paper cluster 1Lee et al CAT2
(a)
085
090
095
100
Sing
le sc
atte
ring
albe
do
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(b)
085
090
095
100
Sing
le sc
atte
ring
albe
do
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(c)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 1Lee et al CAT2
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(d)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(e)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(f)
000
005
010
015
020
025
This paper cluster 1Lee et al CAT2
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(g)
000
005
010
015
This paper cluster 4Lee et al CAT3
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(h)
00010203040506070809
Radius (G)
10minus1 100 101
This paper
This papercluster 6
cluster 5Lee et al
Lee et alCAT5
CAT6
dV
dFH
r(
G3
G2)
(i)
Figure 8 Comparison with East Asian aerosol types
Table 5 The record numbers within each membership degree interval
Membership degree intervals (MDI) C11 C2 C3 C4 C5 C6MDI1 017 le MD2 lt 027 156 131 60 117 43 86MDI2 027 leMD lt 037 508 569 298 603 404 308MDI3 037 leMD lt 047 305 326 240 456 413 234MDI4 047 leMD lt 057 115 128 197 181 275 153MDI5 057 leMD 40 34 122 37 117 761C cluster 2MD membership degree
12 Advances in Meteorology
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 9 AODs at 440 nmwith membership degrees asterisk represents mean value of AOD in each interval and the vertical bars representthe standard deviation
5 (polluted dust) containing absorbing anthropogenic aerosol(such as black carbon) while cluster 6 is dominated onlyby dust These characteristics validate that it is reasonable toidentify cluster 5 as polluted dust and cluster 6 as desert dust
Figure 13 is scatterplot of sphericity parameter versusmembership degree for six aerosol types Figures 13(a) 13(b)13(c) and 13(d) show high variation of sphericity parameterswhich indicates that the first four aerosol types show nocorrelation between sphericity parameters and membershipdegree However Figure 13(e) exhibits clearly descendingtrends and all mean values of sphericity parameter are lower
than 40 Moreover records in Figure 13(f) are more con-centrated with 95 of them falling below 20 These indicatethat both cluster 5 and cluster 6 are close to nonsphericityBecause most dust types are nonsphericity particles [35]these analyses again confirm our identification of polluteddust (cluster 5) and desert dust (cluster 6)
The above analyses confirm reasonability of our results ofclustering and identification of aerosol types Moreover theinternal variation of optical properties of aerosol type can bewell investigated with the help ofmembership degree interval(MDI)
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 9
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 1Continental
dV
dFH
r(
G3
G2)
(a)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 3Background
dV
dFH
r(
G3
G2)
(b)
000
005
010
015
Radius (G)
10minus1 100 101
Cluster 4Biomass
dV
dFH
r(
G3
G2)
(c)
000
005
010
015
020
025
030
Radius (G)
10minus1 100 101
Cluster 5Cluster 6Dust
dV
dFH
r(
G3
G2)
(d)
Figure 6 Size distribution comparison with global aerosol types
aerosol types show relatively smaller variation ranges (from0 to 2) And there are fewer that occurred during polluteddays Furthermore the AODs obviously concentrate in thelow varying range with increasing membership degree for allaerosol types
Figure 10 shows the mean values of AOD within the fivemembership degree intervals at four wavelengths (440 676869 and 1020 nm) Figures 10(a) and 10(b) show that theaverage AODs of fine particle moderately absorbing aerosolfor all wavelengths exhibit relatively flat behaviors Howeverthe other five aerosol types exhibit clearly declination withthe increase of themembership degree intervalMoreover thetrends are nearly the same between different wavelengths foreach aerosol type
The SSAs (440 nm) of six aerosol types varying withmembership degree and their mean values within eachmembership degree interval are plotted in Figure 11 Same
as AODs points of SSA concentrate in the low varyingrange with increasing membership degree for all aerosoltypes It can be seen from Figure 11(a) that the SSAs offine particle nonabsorbing aerosol (cluster 1) fall within therange between 09 and 10 This characteristic is consistentwith the nonabsorbing aerosol (strong scattering) of cluster1 Besides it is interesting to find that there are somestrongly absorbing particles (SSA lt 08) at membershipdegrees between 027 and 047 in Figure 11(d) (cluster 4) Asdiscussed in Section 323 these particles are probably theblack carbon which are frequently observed in Beijing withhigh absorption [23 28]
The mean values of SSA within the five membershipdegree intervals at four wavelengths (440 676 869 and1020 nm) are plotted in Figure 12 Means of SSA showclear descending trend in Figure 12(a) which indicates thatwith the increase of membership degree scattering ability of
10 Advances in Meteorology
Cluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
010203040506070809101112
Sing
le sc
atte
ring
albe
do
Aerosol typeCluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
00
01
02
03
04
05
06
07
08
Asy
mm
etry
fact
or
Aerosol type
This paperOmar et al
This paperOmar et al
Figure 7 SSA and ASYM comparison with global aerosol types
Table 4 Characteristics of East Asian aerosol types by Lee et al (2010)
Properties CAT11 CAT2 CAT3 CAT4 CAT5 CAT6
SSA
440 nm 0915 0927 0904 0908 0893 0900676 nm 0927 0941 0908 0914 0939 0957869 nm 0918 0933 0897 0907 0945 09641020 nm 0912 0928 0889 0903 0948 0966
REFR
440 nm 1468 1478 1441 1454 1508 1549676 nm 1480 1483 1458 1472 1535 1549869 nm 1485 1483 1468 1482 1536 15371020 nm 1481 1476 1468 1481 1528 1525
REFI
440 nm 00119 00099 00127 00112 0007 00049676 nm 00086 00074 001 00088 00037 00024869 nm 00088 00078 00102 0009 00036 000231020 nm 00091 0008 00104 00092 00036 00025
ASYM
440 nm 0721 0729 0716 0713 073 0748676 nm 0664 0685 0654 0655 0693 0714869 nm 0636 0657 0626 0635 0691 07071020 nm 0626 0643 0618 0634 0696 0707
1CAT category
fine particle nonabsorbing aerosol (cluster 1) is enhancingThis characteristic confirms that the higher the membershipdegree the purer the nonabsorbing aerosol Figures 12(b) and12(c) show the mean values of SSA obviously descendingwhen membership degree increases from MDI1 to MDI2However the changes are hardly observed when member-ship degree interval increases from MDI2 to MDI5 Thesecharacteristics demonstrate that records at the edge of cluster2 (fine-M1) and cluster 3 (fine-M2) have stronger scattering(or lower absorbing) than those near center Figure 12(d)exhibits clearly ascending trend of SSA means which is amanifestation of the reasonability to identify cluster 4 ashighly absorbing aerosol The ascending trend also indicatesthat the higher membership degree denotes the purer highlyabsorbing aerosol
Figures 12(e) and 12(f) show some similarities betweenpolluted dust and desert dust at wavelengths of 440 nmFirstly mean values of SSA slightly decrease with the increaseof membership degree interval Furthermore mean values ofSSA at 440 nm are obviously lower than other wavelengthsAs discussed in Sections 324 and 325 these SSA trendsagree well with the relationships between SSA and dustcompositions which certify our identification that bothcluster 5 (polluted dust) and cluster 6 (desert dust) aredust-related aerosols However there are some differencesbetween two clusters SSAs at 440 nm are distinctly smallerthan at other wavelengths Besides Figures 12(e) and 12(f)show different trends at wavelengths longer than 440 nm theformer displays descending trend while the latter shows noclear variation These differences can be attributed to cluster
Advances in Meteorology 11
400
500
600
700
800
900
1000
1100
085
090
095
100Si
ngle
scat
terin
g al
bedo
Wavelengths (nm)
This paper cluster 1Lee et al CAT2
(a)
085
090
095
100
Sing
le sc
atte
ring
albe
do
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(b)
085
090
095
100
Sing
le sc
atte
ring
albe
do
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(c)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 1Lee et al CAT2
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(d)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(e)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(f)
000
005
010
015
020
025
This paper cluster 1Lee et al CAT2
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(g)
000
005
010
015
This paper cluster 4Lee et al CAT3
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(h)
00010203040506070809
Radius (G)
10minus1 100 101
This paper
This papercluster 6
cluster 5Lee et al
Lee et alCAT5
CAT6
dV
dFH
r(
G3
G2)
(i)
Figure 8 Comparison with East Asian aerosol types
Table 5 The record numbers within each membership degree interval
Membership degree intervals (MDI) C11 C2 C3 C4 C5 C6MDI1 017 le MD2 lt 027 156 131 60 117 43 86MDI2 027 leMD lt 037 508 569 298 603 404 308MDI3 037 leMD lt 047 305 326 240 456 413 234MDI4 047 leMD lt 057 115 128 197 181 275 153MDI5 057 leMD 40 34 122 37 117 761C cluster 2MD membership degree
12 Advances in Meteorology
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 9 AODs at 440 nmwith membership degrees asterisk represents mean value of AOD in each interval and the vertical bars representthe standard deviation
5 (polluted dust) containing absorbing anthropogenic aerosol(such as black carbon) while cluster 6 is dominated onlyby dust These characteristics validate that it is reasonable toidentify cluster 5 as polluted dust and cluster 6 as desert dust
Figure 13 is scatterplot of sphericity parameter versusmembership degree for six aerosol types Figures 13(a) 13(b)13(c) and 13(d) show high variation of sphericity parameterswhich indicates that the first four aerosol types show nocorrelation between sphericity parameters and membershipdegree However Figure 13(e) exhibits clearly descendingtrends and all mean values of sphericity parameter are lower
than 40 Moreover records in Figure 13(f) are more con-centrated with 95 of them falling below 20 These indicatethat both cluster 5 and cluster 6 are close to nonsphericityBecause most dust types are nonsphericity particles [35]these analyses again confirm our identification of polluteddust (cluster 5) and desert dust (cluster 6)
The above analyses confirm reasonability of our results ofclustering and identification of aerosol types Moreover theinternal variation of optical properties of aerosol type can bewell investigated with the help ofmembership degree interval(MDI)
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
10 Advances in Meteorology
Cluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
010203040506070809101112
Sing
le sc
atte
ring
albe
do
Aerosol typeCluster 1 Cluster 3 Cluster 4 Cluster 5 Cluster 6
00
01
02
03
04
05
06
07
08
Asy
mm
etry
fact
or
Aerosol type
This paperOmar et al
This paperOmar et al
Figure 7 SSA and ASYM comparison with global aerosol types
Table 4 Characteristics of East Asian aerosol types by Lee et al (2010)
Properties CAT11 CAT2 CAT3 CAT4 CAT5 CAT6
SSA
440 nm 0915 0927 0904 0908 0893 0900676 nm 0927 0941 0908 0914 0939 0957869 nm 0918 0933 0897 0907 0945 09641020 nm 0912 0928 0889 0903 0948 0966
REFR
440 nm 1468 1478 1441 1454 1508 1549676 nm 1480 1483 1458 1472 1535 1549869 nm 1485 1483 1468 1482 1536 15371020 nm 1481 1476 1468 1481 1528 1525
REFI
440 nm 00119 00099 00127 00112 0007 00049676 nm 00086 00074 001 00088 00037 00024869 nm 00088 00078 00102 0009 00036 000231020 nm 00091 0008 00104 00092 00036 00025
ASYM
440 nm 0721 0729 0716 0713 073 0748676 nm 0664 0685 0654 0655 0693 0714869 nm 0636 0657 0626 0635 0691 07071020 nm 0626 0643 0618 0634 0696 0707
1CAT category
fine particle nonabsorbing aerosol (cluster 1) is enhancingThis characteristic confirms that the higher the membershipdegree the purer the nonabsorbing aerosol Figures 12(b) and12(c) show the mean values of SSA obviously descendingwhen membership degree increases from MDI1 to MDI2However the changes are hardly observed when member-ship degree interval increases from MDI2 to MDI5 Thesecharacteristics demonstrate that records at the edge of cluster2 (fine-M1) and cluster 3 (fine-M2) have stronger scattering(or lower absorbing) than those near center Figure 12(d)exhibits clearly ascending trend of SSA means which is amanifestation of the reasonability to identify cluster 4 ashighly absorbing aerosol The ascending trend also indicatesthat the higher membership degree denotes the purer highlyabsorbing aerosol
Figures 12(e) and 12(f) show some similarities betweenpolluted dust and desert dust at wavelengths of 440 nmFirstly mean values of SSA slightly decrease with the increaseof membership degree interval Furthermore mean values ofSSA at 440 nm are obviously lower than other wavelengthsAs discussed in Sections 324 and 325 these SSA trendsagree well with the relationships between SSA and dustcompositions which certify our identification that bothcluster 5 (polluted dust) and cluster 6 (desert dust) aredust-related aerosols However there are some differencesbetween two clusters SSAs at 440 nm are distinctly smallerthan at other wavelengths Besides Figures 12(e) and 12(f)show different trends at wavelengths longer than 440 nm theformer displays descending trend while the latter shows noclear variation These differences can be attributed to cluster
Advances in Meteorology 11
400
500
600
700
800
900
1000
1100
085
090
095
100Si
ngle
scat
terin
g al
bedo
Wavelengths (nm)
This paper cluster 1Lee et al CAT2
(a)
085
090
095
100
Sing
le sc
atte
ring
albe
do
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(b)
085
090
095
100
Sing
le sc
atte
ring
albe
do
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(c)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 1Lee et al CAT2
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(d)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(e)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(f)
000
005
010
015
020
025
This paper cluster 1Lee et al CAT2
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(g)
000
005
010
015
This paper cluster 4Lee et al CAT3
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(h)
00010203040506070809
Radius (G)
10minus1 100 101
This paper
This papercluster 6
cluster 5Lee et al
Lee et alCAT5
CAT6
dV
dFH
r(
G3
G2)
(i)
Figure 8 Comparison with East Asian aerosol types
Table 5 The record numbers within each membership degree interval
Membership degree intervals (MDI) C11 C2 C3 C4 C5 C6MDI1 017 le MD2 lt 027 156 131 60 117 43 86MDI2 027 leMD lt 037 508 569 298 603 404 308MDI3 037 leMD lt 047 305 326 240 456 413 234MDI4 047 leMD lt 057 115 128 197 181 275 153MDI5 057 leMD 40 34 122 37 117 761C cluster 2MD membership degree
12 Advances in Meteorology
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 9 AODs at 440 nmwith membership degrees asterisk represents mean value of AOD in each interval and the vertical bars representthe standard deviation
5 (polluted dust) containing absorbing anthropogenic aerosol(such as black carbon) while cluster 6 is dominated onlyby dust These characteristics validate that it is reasonable toidentify cluster 5 as polluted dust and cluster 6 as desert dust
Figure 13 is scatterplot of sphericity parameter versusmembership degree for six aerosol types Figures 13(a) 13(b)13(c) and 13(d) show high variation of sphericity parameterswhich indicates that the first four aerosol types show nocorrelation between sphericity parameters and membershipdegree However Figure 13(e) exhibits clearly descendingtrends and all mean values of sphericity parameter are lower
than 40 Moreover records in Figure 13(f) are more con-centrated with 95 of them falling below 20 These indicatethat both cluster 5 and cluster 6 are close to nonsphericityBecause most dust types are nonsphericity particles [35]these analyses again confirm our identification of polluteddust (cluster 5) and desert dust (cluster 6)
The above analyses confirm reasonability of our results ofclustering and identification of aerosol types Moreover theinternal variation of optical properties of aerosol type can bewell investigated with the help ofmembership degree interval(MDI)
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 11
400
500
600
700
800
900
1000
1100
085
090
095
100Si
ngle
scat
terin
g al
bedo
Wavelengths (nm)
This paper cluster 1Lee et al CAT2
(a)
085
090
095
100
Sing
le sc
atte
ring
albe
do
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(b)
085
090
095
100
Sing
le sc
atte
ring
albe
do
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(c)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 1Lee et al CAT2
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(d)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
This paper cluster 4Lee et al CAT3
400
500
600
700
800
900
1000
1100
Wavelengths (nm)
(e)
050
055
060
065
070
075
080
Asy
mm
etry
par
amet
er
400
500
600
700
800
900
1000
1100
Wavelengths (nm)This paper
This paper Lee et al
Lee et al
cluster 6
cluster 5 CAT5
CAT6(f)
000
005
010
015
020
025
This paper cluster 1Lee et al CAT2
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(g)
000
005
010
015
This paper cluster 4Lee et al CAT3
Radius (G)
10minus1 100 101
dV
dFH
r(
G3
G2)
(h)
00010203040506070809
Radius (G)
10minus1 100 101
This paper
This papercluster 6
cluster 5Lee et al
Lee et alCAT5
CAT6
dV
dFH
r(
G3
G2)
(i)
Figure 8 Comparison with East Asian aerosol types
Table 5 The record numbers within each membership degree interval
Membership degree intervals (MDI) C11 C2 C3 C4 C5 C6MDI1 017 le MD2 lt 027 156 131 60 117 43 86MDI2 027 leMD lt 037 508 569 298 603 404 308MDI3 037 leMD lt 047 305 326 240 456 413 234MDI4 047 leMD lt 057 115 128 197 181 275 153MDI5 057 leMD 40 34 122 37 117 761C cluster 2MD membership degree
12 Advances in Meteorology
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 9 AODs at 440 nmwith membership degrees asterisk represents mean value of AOD in each interval and the vertical bars representthe standard deviation
5 (polluted dust) containing absorbing anthropogenic aerosol(such as black carbon) while cluster 6 is dominated onlyby dust These characteristics validate that it is reasonable toidentify cluster 5 as polluted dust and cluster 6 as desert dust
Figure 13 is scatterplot of sphericity parameter versusmembership degree for six aerosol types Figures 13(a) 13(b)13(c) and 13(d) show high variation of sphericity parameterswhich indicates that the first four aerosol types show nocorrelation between sphericity parameters and membershipdegree However Figure 13(e) exhibits clearly descendingtrends and all mean values of sphericity parameter are lower
than 40 Moreover records in Figure 13(f) are more con-centrated with 95 of them falling below 20 These indicatethat both cluster 5 and cluster 6 are close to nonsphericityBecause most dust types are nonsphericity particles [35]these analyses again confirm our identification of polluteddust (cluster 5) and desert dust (cluster 6)
The above analyses confirm reasonability of our results ofclustering and identification of aerosol types Moreover theinternal variation of optical properties of aerosol type can bewell investigated with the help ofmembership degree interval(MDI)
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
12 Advances in Meteorology
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
02 03 04 05 06 07 080
1
2
3
4
Membership degree
Aero
sol o
ptic
al d
epth
at440HG
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 9 AODs at 440 nmwith membership degrees asterisk represents mean value of AOD in each interval and the vertical bars representthe standard deviation
5 (polluted dust) containing absorbing anthropogenic aerosol(such as black carbon) while cluster 6 is dominated onlyby dust These characteristics validate that it is reasonable toidentify cluster 5 as polluted dust and cluster 6 as desert dust
Figure 13 is scatterplot of sphericity parameter versusmembership degree for six aerosol types Figures 13(a) 13(b)13(c) and 13(d) show high variation of sphericity parameterswhich indicates that the first four aerosol types show nocorrelation between sphericity parameters and membershipdegree However Figure 13(e) exhibits clearly descendingtrends and all mean values of sphericity parameter are lower
than 40 Moreover records in Figure 13(f) are more con-centrated with 95 of them falling below 20 These indicatethat both cluster 5 and cluster 6 are close to nonsphericityBecause most dust types are nonsphericity particles [35]these analyses again confirm our identification of polluteddust (cluster 5) and desert dust (cluster 6)
The above analyses confirm reasonability of our results ofclustering and identification of aerosol types Moreover theinternal variation of optical properties of aerosol type can bewell investigated with the help ofmembership degree interval(MDI)
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 13
MDI1 MDI2 MDI3 MDI4 MDI500
05
10
15
20
25
30Ae
roso
l opt
ical
dep
th
Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(a) Cluster 1
00
05
10
15
20
25
30
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
0002040608101214
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
00
05
10
15
20
25
Aero
sol o
ptic
al d
epth
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 10 Variation of AODs with membership degree intervals at four wavelengths the vertical bars represent the standard deviation
34 Seasonal Variation of Aerosol Types The seasonal vari-ations of aerosol types over Beijing are investigated in thissection Figure 14 shows the monthly distributions of sixaerosol types during 2001ndash2014 in Beijing It can be seen fromFigure 14 that the fine particle nonabsorbing aerosol (cluster1) is mostly observed from June to September (65 of totalnumber) These four months generally experience the largestatmospheric humidity in Beijing Due to highly relativehumidity condition the growth aerosol hygroscopicity couldresult in increasing of scattering [33] This is consistent withthe fact that fine particle nonabsorbing aerosol shows largestwater vapor (Figure 2(d))
Fine-MA1 is hardly observed from July to Septem-ber while in other months it is of frequent occurrenceThe fine-MA2 generally occurs throughout the year As
discussed in Section 322 fine-MA2 is likely the back-ground aerosol therefore the similar monthly frequencyof incidence is reasonable The fine particle highly absorb-ing aerosol is with the greatest frequency of occurrencein winter (November to January) As discussed in Sec-tion 323 this type of aerosol may be attributed toburning of coal in winter for heating supply over northChina
Polluted dust and desert dust both are frequently detectedduring spring (March to May) In spring Beijing is generallyaffected by Gobi Desert When transported dust mixes withanthropogenic aerosol it exhibits characteristics of polluteddust On the contrary when there is low-level or no mixturewith anthropogenic particle it will exhibit characteristics ofdesert dust
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
14 Advances in Meteorology
075
080
085
090
095
100
105
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
Sing
le sc
atte
ring
albe
doat
440HG
(a) Cluster 1
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(b) Cluster 2
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(c) Cluster 3
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(d) Cluster 4
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(e) Cluster 5
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
075
080
085
090
095
100
105
Sing
le sc
atte
ring
albe
doat
440HG
(f) Cluster 6
Figure 11 Same as Figure 9 but for SSAs at 440 nm
Consequently Beijing is affected by various aerosol typesin different seasons The dominant aerosol types are polluteddust and desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbing aerosolin winterThe fine particle moderately absorbing aerosols canbe observed throughout the year
4 Conclusions
In Beijing the dominant aerosol types and their characteris-tics are still unclear In this paper we conduct a fuzzy clusteranalysis based on fourteen-year (2001ndash2014) AERONET dataset to obtain dominant aerosol types in Beijing Fuzzy 119888-mean
algorithm is applied to classify a total of 6732 records intosix aerosol types fine particle nonabsorbing two kindsof fine particle moderately absorbing (fine-MA1 and fine-MA2) fine particle highly absorbing polluted dust anddust aerosol Following the clustering detailed propertieswithin different membership degree intervals (MDI) areanalyzedMeanwhile temporal variations of aerosol types arealso investigated The main findings can be summarized asfollows
(1) There are large variations of optical characteris-tics between different aerosol types in Beijing Fineparticle nonabsorbing aerosol exhibits high fine
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 15
090
092
094
096
098
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
Sing
le sc
atte
ring
albe
do
(a) Cluster 1
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(b) Cluster 2
086
088
090
092
094
096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(c) Cluster 3
078080082084086088090092
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(d) Cluster 4
084086088090092094096
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(e) Cluster 5
086088090092094096098100
Sing
le sc
atte
ring
albe
do
MDI1 MDI2 MDI3 MDI4 MDI5Membership degree interval
440 HG
676 HG
869 HG
1020 HG
(f) Cluster 6
Figure 12 Same as Figure 10 but for SSAs
fraction by volume (064) and strong scattering prop-erty the SSA at 440 nm is 095 Two kinds of fineparticlemoderately absorbing aerosols (fine-MA1 andfine-MA2) performmoderate absorptionThe SSAs at440 nm both are 090 Fine particle highly absorbingaerosol displays strong absorbability the SSA at440 nm is 084 The polluted dust aerosol is for thefirst time classified as a unique type by cluster analysismethod and its SSA at 440 nm is 088 Desert dustis dominated by nonspherical coarse particles theSSAs at four selected wavelengths (440 676 869 and1020 nm) are 089 094 095 and 095 Furthermorethe results indicate that fuzzy clustering is capable
of identifying aerosol types in regions where aerosolparticles are contributed by complex components
(2) The optical properties of six aerosol types exhibitdifferent internal variations These variations can bewell investigated with the help of membership degreeinterval (MDI)
(3) Aerosol types of Beijing display clearly seasonal vari-ation The dominant aerosol types are polluted dustand desert dust in spring fine particle nonabsorbingaerosol in summer and fine particle highly absorbingaerosol inwinterThefine particlemoderately absorb-ing aerosol can be observed throughout the year
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
16 Advances in Meteorology
0
20
40
60
80
100Sp
heric
ity p
aram
eter
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(a) Cluster 1
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(b) Cluster 2
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(c) Cluster 3
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(d) Cluster 4
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(e) Cluster 5
0
20
40
60
80
100
Sphe
ricity
par
amet
er
02 03 04 05 06 07 08Membership degree
MDI1MDI2MDI3
MDI4MDI5
(f) Cluster 6
Figure 13 Scatterplots between sphericity parameter and membership degree
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Authorsrsquo Contributions
All authors assisted in analyzing and editing of the paperWenhao Zhang is the main author who processed the dataanalyzed the results and wrote the manuscript
Acknowledgments
This study was supported by National Natural ScienceFoundation of China (Grants 41401422 and 41501404) theChina High Resolution Earth Observation Project (noY6D0060038) the Major State Basic Research DevelopmentProgram of China (no Y070070070) the Civil Space Pre-Research Project in 13th Five-Year Plan (no Y7K00100KJ)and the Innovative Projects of Institute of Remote Sens-ing and Digital Earth Chinese Academy of Sciences (noY7SG0600CX) The AERONET data are downloaded from
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Advances in Meteorology 17
1 2 3 4 5 6 7 8 9 10 11 120
50
100
150
200
250
300
Num
ber o
f rec
ords
(Month)
Cluster 1Cluster 2Cluster 3
Cluster 4Cluster 5Cluster 6
Figure 14 The monthly distributions for six clusters (2001ndash2014)
httpaeronetgsfcnasagov The authors acknowledge allthe AERONET investigators for maintaining CIMEL instru-ments and providing high quality aerosol products
References
[1] Y J Kaufman D Tanre and O Boucher ldquoA satellite view ofaerosols in the climate systemrdquo Nature vol 419 no 6903 pp215ndash223 2002
[2] C E Adler and G Hirsch Hadorn ldquoThe IPCC and treatmentof uncertainties topics and sources of dissensusrdquo Wiley Inter-disciplinary Reviews Climate Change vol 5 no 5 pp 663ndash6762014
[3] M O Andreae C D Jones and P M Cox ldquoStrong present-dayaerosol cooling implies a hot futurerdquo Nature vol 435 no 7046pp 1187ndash1190 2005
[4] H Yu Y J Kaufman M Chin et al ldquoA review of measurement-based assessments of the aerosol direct radiative effect andforcingrdquo Atmospheric Chemistry and Physics vol 6 no 3 pp613ndash666 2006
[5] O Torres P K Bhartia J R Herman Z Ahmad and J GleasonldquoDerivation of aerosol properties from satellite measurementsof backscattered ultraviolet radiation theoretical basisrdquo Journalof Geophysical Research D vol 103 no 14 pp 17099ndash17110 1998
[6] J HansenM Sato and R Ruedy ldquoRadiative forcing and climateresponserdquo Journal of Geophysical Research D Atmospheres vol102 no 6 pp 6831ndash6864 1997
[7] Y J Kaufman D Tanre O Dubovik A Karnieli and L ARemer ldquoAbsorption of sunlight by dust as inferred from satelliteand ground-based remote sensingrdquo Geophysical Research Let-ters vol 28 no 8 pp 1479ndash1482 2001
[8] M-K Kim W K M Lau K-M Kim and W-S Lee ldquoA GCMstudy of effects of radiative forcing of sulfate aerosol on largescale circulation and rainfall in East Asia during boreal springrdquoGeophysical Research Letters vol 34 no 24 Article ID L247012007
[9] A Pozzer A De Meij J Yoon H Tost A K Georgoulias andM Astitha ldquoAOD trends during 2001-2010 from observations
and model simulationsrdquo Atmospheric Chemistry and Physicsvol 15 no 10 pp 5521ndash5535 2015
[10] T F Eck B N Holben J S Reid et al ldquoWavelength dependenceof the optical depth of biomass burning urban and desert dustaerosolsrdquo Journal of Geophysical Research vol 104 no 24 pp31333ndash31349 1999
[11] O Dubovik B Holben T F Eck et al ldquoVariability of absorptionand optical properties of key aerosol types observed in world-wide locationsrdquo Journal of the Atmospheric Sciences vol 59 no3 pp 590ndash608 2002
[12] M Balarabe K Abdullah andMNawawi ldquoSeasonal Variationsof Aerosol Optical Properties and Identification of DifferentAerosol Types Based on AERONET Data over Sub-SaharaWest-Africardquo Atmospheric and Climate Sciences vol 06 no 01pp 13ndash28 2016
[13] J Lee J Kim C H Song et al ldquoCharacteristics of aerosol typesfrom AERONET sunphotometer measurementsrdquo AtmosphericEnvironment vol 44 no 26 pp 3110ndash3117 2010
[14] A H Omar J Won D M Winker S Yoon O Dubovikand M P McCormick ldquoDevelopment of global aerosol modelsusing cluster analysis of Aerosol Robotic Network (AERONET)measurementsrdquo Journal of Geophysical Research vol 110 no 10pp 1ndash14 2005
[15] R C Levy L A Remer and O Dubovik ldquoGlobal aerosoloptical properties and application to Moderate ResolutionImaging Spectroradiometer aerosol retrieval over landrdquo Journalof Geophysical Research Atmospheres vol 112 no 13 Article IDD13210 2007
[16] Y Qin and RMMitchell ldquoCharacterisation of episodic aerosoltypes over the Australian continentrdquo Atmospheric Chemistryand Physics vol 9 no 6 pp 1943ndash1956 2009
[17] K H Lee and Y J Kim ldquoSatellite remote sensing of Asianaerosols a case study of clean polluted and Asian dust stormdaysrdquo Atmospheric Measurement Techniques vol 3 no 6 pp1771ndash1784 2010
[18] P J Sheridan D J Delene and J A Ogren ldquoFour years of con-tinuous surface aerosol measurements from the Department ofEnergyrsquos Atmospheric RadiationMeasurement Program South-ern Great Plains Cloud and Radiation Testbed siterdquo Journal ofGeophysical Research Atmospheres vol 106 no 18 Article ID2001JD000785 pp 20735ndash20747 2001
[19] W Zhang X Gu H Xu T Yu and F Zheng ldquoAssessment ofOMInear-UVaerosol optical depth overCentral andEastAsiardquoJournal of Geophysical Research Atmospheres vol 121 no 1 pp382ndash398 2016
[20] L Wu and Q-C Zeng ldquoClassifying Asian dust aerosols andtheir columnar optical properties using fuzzy clusteringrdquo Jour-nal of Geophysical Research Atmospheres vol 119 no 5 pp2529ndash2542 2014
[21] S D Xie Z Liu T Chen and L Hua ldquoSpatiotemporal varia-tions of ambient PM10 source contributions in Beijing in 2004using positivematrix factorizationrdquoAtmospheric Chemistry andPhysics vol 8 no 10 pp 2701ndash2716 2008
[22] X Zhao X Zhang W Pu W Meng and X Xu ldquoScatteringproperties of the atmospheric aerosol in Beijing Chinardquo Atmo-spheric Research vol 101 no 3 pp 799ndash808 2011
[23] X Zheng X Liu F Zhao F Duan T Yu and H CachierldquoSeasonal characteristics of biomass burning contribution toBeijing aerosolrdquo Science in China Series B Chemistry vol 48no 5 pp 481ndash488 2005
[24] B Guinot H Cachier J Sciare Y Tong W Xin and Y JianhualdquoBeijing aerosol atmospheric interactions and new trendsrdquo
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
18 Advances in Meteorology
Journal of Geophysical Research Atmospheres vol 112 no 14Article ID D14314 2007
[25] X He C C Li A K Lau et al ldquoAn intensive study of aerosoloptical properties in Beijing urban areardquoAtmospheric Chemistryand Physics vol 9 no 22 pp 8903ndash8915 2009
[26] P Liu C Zhao Q Zhang et al ldquoAircraft study of aerosol verticaldistributions over Beijing and their optical propertiesrdquo Tellus Bvol 61 no 5 2009
[27] X Yu B Zhu Y Yin J Yang Y Li and X Bu ldquoA comparativeanalysis of aerosol properties in dust and haze-fog days in aChinese urban regionrdquo Atmospheric Research vol 99 no 2 pp241ndash247 2011
[28] X Yu C Shi J Ma et al ldquoAerosol optical properties duringfirework biomass burning and dust episodes in Beijingrdquo Atmo-spheric Environment vol 81 pp 475ndash484 2013
[29] HChe X Xia J Zhu et al ldquoAerosol optical properties under thecondition of heavy haze over an urban site of Beijing ChinardquoEnvironmental science and pollution research international vol22 no 2 pp 1043ndash1053 2015
[30] J C Bezdek Pattern Recognition with Fuzzy Objective FunctionAlgorithms Plenum Press New York NY USA 1981
[31] B N Holben T F Eck I Slutsker et al ldquoAERONETmdashafederated instrument network and data archive for aerosolcharacterizationrdquo Remote Sensing of Environment vol 66 no1 pp 1ndash16 1998
[32] O Dubovik A Smirnov B N Holben et al ldquoAccuracyassessments of aerosol optical properties retrieved fromAerosolRobotic Network (AERONET) Sun and sky radiance measure-mentsrdquo Journal of Geophysical ResearchD Atmospheres vol 105no 8 pp 9791ndash9806 2000
[33] G S Brown ldquoThe validity of shadowing corrections in roughsurface scatteringrdquo Radio Science vol 19 no 6 pp 1461ndash14681984
[34] C F Bohren and D R Huffman Absorption and Scattering ofLight by Small ParticlesWiley-VCHWeinheimGermany 1998
[35] K Okada J Heintzenberg K Kai and Y Qin ldquoShape ofatmospheric mineral particles collected in three Chinese arid-regionsrdquo Geophysical Research Letters vol 28 no 16 pp 3123ndash3126 2001
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in
Submit your manuscripts athttpswwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ClimatologyJournal of
EcologyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
EarthquakesJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mining
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporation httpwwwhindawicom Volume 201
International Journal of
OceanographyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal ofPetroleum Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Atmospheric SciencesInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MineralogyInternational Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MeteorologyAdvances in
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geological ResearchJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Geology Advances in