Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/295513914

Normalizingreflectancefromdifferentspectrometersandprotocolswithaninternalsoilstandard

ArticleinInternationalJournalofRemoteSensing·February2016

DOI:10.1080/01431161.2016.1148291

CITATIONS

3

READS

262

2authors:

Someoftheauthorsofthispublicationarealsoworkingontheserelatedprojects:

AnovelApproachforMOdelingSoilDEgradationusingsuper-SpectralorbitaldataViewproject

EstablishmentofanationalspectrallibraryofIsraelisoilsforquantitativeutilizationofregionalsoils

Viewproject

VeronikaKopackova

CzechGeologicalSurvey

49PUBLICATIONS115CITATIONS

SEEPROFILE

EyalBen-Dor

TelAvivUniversity

219PUBLICATIONS5,651CITATIONS

SEEPROFILE

AllcontentfollowingthispagewasuploadedbyVeronikaKopackovaon22February2016.

Theuserhasrequestedenhancementofthedownloadedfile.

Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=tres20

Download by: [Ceska Geologicka Sluzba], [Veronika Kopačková] Date: 22 February 2016, At: 01:46

International Journal of Remote Sensing

ISSN: 0143-1161 (Print) 1366-5901 (Online) Journal homepage: http://www.tandfonline.com/loi/tres20

Normalizing reflectance from differentspectrometers and protocols with an internal soilstandard

Veronika Kopačková & Eyal Ben-Dor

To cite this article: Veronika Kopačková & Eyal Ben-Dor (2016) Normalizing reflectance fromdifferent spectrometers and protocols with an internal soil standard, International Journal ofRemote Sensing, 37:6, 1276-1290

To link to this article: http://dx.doi.org/10.1080/01431161.2016.1148291

Published online: 21 Feb 2016.

Submit your article to this journal

View related articles

View Crossmark data

Normalizing reflectance from different spectrometersand protocols with an internal soil standardVeronika Kopačkováa and Eyal Ben-Dorb

aRemote Sensing Department, Czech Geological Survey, Prague, Czech Republic; bRemote SensingLaboratory, Tel Aviv University, Tel Aviv, Israel

ABSTRACTThe internal soil standard (ISS) concept in which a soil standardsample, exhibiting stable spectral performance, is used to normal-ize and align all other soil spectral measurements – was furtherexamined herein. Different spectrometers (Spectral Evolution andASD Spectral Pro) were used to measure a set of soil samples withthe soil standards sample as a reference. Two sand dune samplesserved as the ISS to align measurements made under differentconditions and the results were compared. It was shown that theISS method was able to correct the spectral information from onespectrometer to another; however, the differences in the resultsobtained when using the two different soil standards are dis-cussed. The main conclusion of this paper is that the soil spectraluser community should adopt the ISS method for the benefit ofall, and the sooner the better. This will allow much more effectiveexploitation of all data sets acquired on a daily basis by thegrowing soil spectral community that still lacks standardizationprocedures.

ARTICLE HISTORYReceived 17 June 2015Accepted 20 January 2016

1. Introduction

Soil spectroscopy has been shown to be a fast, cost-effective, environmentally friendly,non-destructive, reproducible, and repeatable analytical technique that has beenincreasingly used instead of more expensive soil analysis procedures (Malley, Martin,and Ben-Dor 2004). Based on soil spectroscopy, a proxy strategy using a chemometricsapproach has been developed for soils, along with the massive construction of soilspectral libraries worldwide (Viscarra Rossel 2009; Viscarra Rossel et al. 2006; ViscarraRossel et al. forthcoming). However, the community is still lacking agreed-upon stan-dards or protocols for reliable reflectance measurements in the laboratory and field. Thiswas emphasized by Ge et al. (2011) who showed that the visible, near-infrared, andshort-wave infrared proxy models are instrument/scanning environmental dependentwhile highlighting the need for cross-calibration between the instruments.

The problem with sharing soil spectral libraries that were acquired using differentprotocols and conditions has already been discussed (Ben Dor, Ong, and Lau 2015;

CONTACT Veronika Kopačková veronika.kopackova@seznam.cz Remote Sensing Department, Czech GeologicalSurvey, Klárov 3, Prague 118 21, Czech Republic

INTERNATIONAL JOURNAL OF REMOTE SENSING, 2016VOL. 37, NO. 6, 1276–1290http://dx.doi.org/10.1080/01431161.2016.1148291

© 2016 Taylor & Francis

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

Jung, Götze, and Gläßer 2009; Nocita et al. 2015; Pimstein, Ben Dor, and Notesco2010; Viscarra Rossel 2009; Wetterlind and Stenberg 2010). Nowadays, not only is thenumber of different protocols on the rise, but so is the number of spectrometermodels, resulting in significant variation among data sets. This obviously makes futuredata-sharing, comparisons, and analyses quite complicated, if not impossible.Therefore, this issue is becoming more and more important for today’s growingspectral community (Nocita et al. 2015) as more and more users are entering thisfield and generating soil spectral libraries at both national (e.g. Knadel et al. 2012)and continental (e.g. the Land Use/Cover Area frame Statistical Survey in Europe;Stevens et al. 2013) levels.

Recently, Ben Dor, Ong, and Lau (2015) suggested using the internal soil standard(ISS) concept to minimize the systematic effects encountered with any laboratorycondition or measurement protocol. The idea behind the ISS concept is that the samesample (or samples), which has been measured under standard (motherhood; alsotermed ‘benchmark’) conditions, will be measured under the user’s conditions and thedifference will be attributed to the effects of the systems in both measurement set-ups. The ISS sample (motherhood sample) should be a homogeneous material thathas grain sizes and shapes similar to soil particles, and it should be stable in spaceand time (Pimstein, Ben Dor, and Notesco 2010; Willis 1972).

To correct most of the system effects within the soil spectral measurements (e.g.dirt on the white plate, lamp deterioration, user drift, spectrometer performance,measurement geometry, and other unknown factors), a correction factor can becalculated between the ISS samples acquired under the two sets of conditions andconsequently applied to all of the user’s spectral readings. Assuming that the correc-tion will bring all of the measurements close to the motherhood set-up, spectrallibraries that are collected under different conditions can be merged into a biggerdatabase and then compared and utilized. Furthermore, by employing the ISSmethod, not only can a correction factor be extracted, but also user monitoring ofthe spectrometer’s stability and performance in the short and long run becomespossible.

In the previous study (Ben Dor, Ong, and Lau 2015), the ISS concept has beenexamined using four different ASD spectrometers under different external conditions(protocols, geographical location, climate, and ambient conditions) and with differentusers. As the ISS-based method was developed on a particular company’s spectro-meters (ASD, with the same wavelength range, spectral resolution, bulb, and assem-bly), it was interesting to further check this concept’s performance with a differentmodel of spectrometer and spectral configuration (from a different company) alongwith different measurement protocols. Therefore, this paper demonstrates how thesuggested ISS protocol can be employed under different spectrometer makes. Themotherhood spectrometer was the ASD from the Commonwealth Scientific andIndustrial Research Organisation (CSIRO, Perth, Australia) while the SpectralEvolution spectrometer from the Czech Geological Survey (CGS, Prague, CzechRepublic) and the ASD spectrometer from Tel Aviv University (TAU, Tel Aviv, Israel)were selected to gather the user’s spectra for this study. The newly obtained resultsfurther contribute to a verification of the ISS concept, which claims that an ISS canalign measurements made under different conditions.

INTERNATIONAL JOURNAL OF REMOTE SENSING 1277

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

2. Materials and methods

The experiment was performed at the CGS in Prague, Czech Republic, and remote sensinglaboratory, TAU, Israel. Two Australian sand dune samples with a grain size of less than250 µm served as the ISS to align soil spectral measurements in this experiment. Thesetwo samples, namely Wylie Bay (WB) and Lucky Bay (LB), fulfilled the condition defined byPimstein, Ben Dor, and Notesco (2010). They had grain property (shape, size, and nature)as similar as possible to natural soil, they exhibited a stable spectral performance acrossthe whole spectral region, and they were as spectrally featureless as possible (see Ben Dor,Ong, and Lau 2015 for the exact geographical locations and a description of the chemicalattributes). The user’s soil samples consist of two sets:

● The same five soil samples used by Ben Dor, Ong, and Lau (2015) in their work(from Western Australia)

● An additional 10 soil samples from Israel

Both data sets were measured at the CGS and TAU at different dates.The set of five soil samples from Western Australia, which went through a laboratory

procedure (air-dried and sieved: grain size less than 2 mm), were used as a spectral soilset which the ISS method was first applied to. These samples had dominant quartzcontent but differed in haematite, goethite, and feldspar content as well as organic andmineral carbon contents (for a detailed description, see also Ben Dor, Ong, and Lau2015). In order to further validate the ISS performance results, the same ISS routinewas employed using an additional soil set – 10 Israeli soil samples. These 10 soilsamples were taken from the TAU soil library stock (<2 mm grain size, Ben-Dor andBanin’s (1995)) and spectrally measured (see the details latter). The soils went throughwet laboratory analyses using traditional soil science methods as described in Ben-Dorand Banin (1995). The soil attributes are provided in Table 1.

The complete set of soil spectra (2 ISS samples from Australia, 5 soil samples fromAustralia, and 10 validation soils from Israel) was spectrally measured at both the CGSand the TAU laboratories using each laboratory protocol. At the CGS, the measurementswere performed using the Spectral Evolution spectrometer (model: SR 2500; spectralresolution: 3.5 nm (350–1000 nm)/22 nm (1500 nm)/22 nm (2100 nm); 768 bands;calibration: National Institute of Science and Technology (NIST) Traceable radiance

Table 1. The attributes of the Israeli soil samples as obtained by wet analyses described in Ben-Dorand Banin (1995).Soil type Clay (%) CaCO3 (%) OM (%) SSA (mg2 g−1)

E6 20.4 0.63 0.56 83.7C11 30.1 26.28 5.92 205.4H8 28.7 23.00 2.26 172.6E9 6.3 3.21 0.09 10.8H9 22.3 26.43 1.69 179.0A5 54.2 8.43 2.52 369.0S4 10.2 5.03 1.00 54.9S10 19.2 25.8 1.44 96.2H4 31.8 34.2 3.78 239.3B3 32.9 19.73 8.71 256.2

OM, organic matter; SSA, specific surface area.

1278 V. KOPAČKOVÁ AND E. BEN-DOR

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

calibration ±5% at 400 nm; ±4% at 700 nm; reflectance contact probe with 5 W tungstenhalogen lamp). The spectral measurements of the five Australian soil samples weremeasured on 19 May 2015 and the Israeli soil on 1 December 2015. In both cases, thesoil spectra were acquired at ambient (room) conditions (19 May: temperature 22°C andrelative humidity (RH) 73% outside; 1 December: temperature 23°C and RH 70% outside)and diffuse daylight (10 am–14 pm).

At the TAU, the measurements were performed using the ASD spectrometer (ASDFieldSpec Pro FR SN:6232) with a spectral resolution of 1.4 nm for 350–700 nm, and 2 nmfor 1000–2500 nm, 2151 bands, calibration: NIST traceable with spectral bands at 1-nmsampling intervals, and a reflectance contact probe with a 6.5 W tungsten halogen lamp.The spectral measurements of the Australian samples were acquired at the TAU on 24October and the Israeli soils on 4 November 2015. All measurements took place underambient (room) conditions (24 October: 25°C and RH 46% outside; 4 November: tem-perature 28°C and RH 40% outside) and diffuse daylight (10 am–14 pm).

At both laboratories, the CGS and the TAU, the soil spectra were measured using acontact probe assembly. At CGS, the probe used was held over the sample by a steadyhand while in the TAU a stable construction was used. Furthermore, the measurementsfollowed the same protocol suggested by CSIRO Perth laboratory (Ben Dor, Ong, and Lau2014): the spectrometer was warmed up for 2 h and the bulb for 1 h prior to measure-ments. A contact-probe assembly (commercially available for both spectrometers) armedwith a tungsten halogen lamp and sapphire window was used to measure a whitereference plate (Spectralon® Reflectance Standards, Labsphere, North Sutton, NH, USA)and then after the soil and the ISS samples.

At both laboratories (the CGS and the TAU), the samples were placed in a black(quartz) Petri dish and the probe was brought into complete contact with them. The twoISS samples were measured at the beginning of the experiment (after the whitereference) and then again after each of five soil measurements were conducted. Thisprocedure was set up to check the stability of the sensors during the measurementsperiod. Each soil measurement was taken identically using the same protocol at eachlaboratory. The spectrometer was programmed to provide a single measurement from20 (CSG) and 30 (TAU) spectra, which were then averaged together. Three replicateswere carried out for each soil sample in both CGS and TAU. In each replicate, the soilsample was mixed before the measurement, homogenized without any surface smooth-ing, and then touched with the contact-probe glass. After inspecting the spectraldeviation of the three replicates (less than 4%), the spectra were averaged to a repre-sentative spectrum for further work.

The five Australian soil spectra acquired with the Spectral Evolution spectrometer(CGS) and with the ASD spectrometer (TAU) were compared to those acquired by theASD spectrometer (ASD FieldSpec Pro FR SN:6232) at CSIRO, Perth, in 2012, using adetailed protocol (CSIRO-0) that was considered to be the benchmark (motherhood) set-up for this work (Ben Dor, Ong, and Lau 2014). The 10 Israeli soils that were measured atthe CGS and TAU were compared before and after applying the ISS correction.

To enable a precise comparison between the CSIRO, CGS, and TAU spectra that wereacquired under different spectral configurations, the soils (the motherhood fiveAustralian soils, the ISS spectra, and the 10 Israeli soils) were resampled into the CGSspectral configuration of the Spectral Evolution spectrometer.

INTERNATIONAL JOURNAL OF REMOTE SENSING 1279

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

2.1. Spectral correction to the soil benchmark measurements(motherhood set-up)

To apply the ISS correction, Pimstein, Ben Dor, and Notesco (2010) were followed tocorrect the CGS to the CSIRO measurements as follows:

Cλ ¼ 1� ððρS;λ � ρSBM; λÞ=ρS; λÞ; (1)

Rc;λ ¼ Ro;λ � Cλ; (2)

where λ is a given wavelength, ρS, λ is the reflectance of the slave ISS reference (WB andLB; sand samples measured at the CGS), ρSBM,λ is the reflectance of the soil benchmark(SBM) ISS reference (WB and LB; sand samples measured at CSIRO with CSIRO-0 set-up),Ro,λ is the original soil sample reflectance (the soil spectrum measured at the CGS andTAU), Rc,λ is the corrected soil sample reflectance (the soil spectrum measured at the CGSand TAU normalized to the CSIRO-0 set-up), and Cλ is a correction factor.

2.2. Correction performance assessment

To quantitatively examine the correction performance, the average sumof deviations squared(ASDS) index suggested by Ben Dor, Kindel, and Goetz (2004) was calculated as follows:

ASDS ¼X

ðρSIS;λ � 1Þ2=nÞ; (3)

ρSIS;λ ¼ ρRC;λ=ρRM; λ; (4)

where n is the number of spectral channels, ρSIS,λ is the reflectance value (at a givenwavelength) of the examined sample (e.g. original CGS and ISS-corrected spectra in caseof the five Australian soils) relative to the reflectance value (at a given wavelength) ofthe same soil with the motherhood (benchmark) set-up, and ρRM,λ is the CSIRO-0spectrum. In the case of the 10 Israeli soils, the ρRC,λ is the reflectance value (at agiven wavelength) of the examined sample (e.g. original TAU) relative to the reflectancevalue (at a given wavelength) of the same soil at CGS. This was also calculated for theISS-corrected TAU versus CGS.

Lower ASDS values indicate better ISS correction performance. To judge the correc-tion power of the ISS method of the five soils, the normalized deviation ASDS value (ND-ASDS) was calculated according to

ND� ASDS ¼ ASDSb�ASDSað Þ=ASDSb; (5)

where ASDSb is the ASDS calculated before ISS correction and ASDSa is the ASDScalculated after ISS correction.

To judge the ISS correction of the 10 soils measured at CGS and TAU, the ‘improve-ment correction percentages’ (ICP) of the ISS correction was also calculated using thefollowing equation:

ICP ¼ ðASDSTAU=CGSa�ASDSTAU=CGSbÞ=ASDSTAU=CGSa; (6)

1280 V. KOPAČKOVÁ AND E. BEN-DOR

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

where ASDSTAU/CGSa is the ASDS value calculated between the TAU and CGS spectrabefore the ISS correction and ASDSTAU/CGSb is the ASDS values calculated between TAUand CGS after the ISS correction.

3. Results and discussion

Figures 1(a) and (b) present the WB and LB spectra as measured in the CSIRO-0, CGS, andTAU set-ups. Both CGS and TAU measurement samples showed a different albedoresponse than that obtained with the CSIRO-0 set-up. In the WB sample, this albedoshift was less pronounced than in the LB sample in both laboratories (CGS and TAU).Based on this observation, the CGS and TAU samples were corrected by using the LB andWB samples separately to extract the correction factor, rather than averaging them intoone factor as done by Ben Dor, Ong, and Lau (2015).

To visually examine the ISS method’s performance in correcting the CGS and TAUspectral readings into the CSIRO-0 motherhood set-up, the spectra of each soil beforeand after the ISS correction for LB and WB separately (Figures 2–5) are provided along-side the corresponding soil spectra as measured in the CSIRO-0 set-up. As seen in the LB

Figure 1. The spectra of the internal soil standard (ISS) samples WB and LB as measured by (a) theCSIRO-0 and CGS set-up and (b) the CSIRO-0 and TAU set-up.

INTERNATIONAL JOURNAL OF REMOTE SENSING 1281

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

correction (Figure 3), both the corrected CGS and TAU samples were shifted (in terms ofalbedo) to almost match the CSIRO-0 spectra, whereas in the WB correction (Figure 2),the corrected CGS and TAU samples were less shifted (but still held a correction towardsthe CSIRO-0 position). These results agreed with the shift observed in Figures 1(a) and(b) between the CGS, TAU, and the CSIRO-0 ISS spectra.

More positive ND-ASDS values indicated better ISS correction performance, whichcan also be an indication of the correction percentages. Figures 6–9 provide theASDS and ND-ASDS representation of both LB and WB corrections in the CGS andTAU laboratories. These figures confirm the previous (visible) inspection shown inFigures 2–5. The LB has corrected the CGS spectra by up to 70% and the TAU by upto 88%, whereas for the WB, it is only up to 27% for the CGS and up to 15% for theTAU. This finding shows that the LB sample is more appropriate to correct for boththe laboratories used for the set-ups: the Spectral Evolution spectrometer and the

Figure 2. Spectra of the five soil samples from Western Australia as measured by CSIRO-0 (CSIRO),CGS, and CGS corrected to CSIRO-0 using the internal soil standard (ISS) (CGS-ISS-CORRECTED)approach with WB as the standard (see Soil 1 for line definition).

1282 V. KOPAČKOVÁ AND E. BEN-DOR

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

CGS protocol, and the ASD spectrometer and the TAU protocol in order to align allreadings into the CSIRO-0 (standard) configuration.

Comparing these values with those of Ben Dor, Ong, and Lau (2015), ISS correctionresults (comparing different ASD spectrometers under different conditions such asgeometry, laboratory, relative humidity, users, and white references) show that theresults obtained here are at the same magnitude as those obtained by Ben Dor, Ong,and Lau (2015) using the same method and samples. This can be seen by both the ASDSobtained here (Figures 6(a) and (b)), which varied between 0.006 and 0.033 for CGS andbetween 0.0036 and 0.021 for TAU, compared to the ASDS of Ben Dor, Ong, and Lau(2015), which varied between 0.0030 and 0.0490, after the ISS correction took place.Furthermore, comparing the ND-ASDS values (Figures 7(a) and (b)) in both studies

Figure 3. Spectra of the five soil samples from Western Australia as measured by CSIRO-0 (CSIRO),CGS, and CGS corrected to CSIRO-0 using the internal soil standard (ISS) (CGS-ISS-CORRECTED)approach with LB as a standard (see Soil 1 for line definition).

INTERNATIONAL JOURNAL OF REMOTE SENSING 1283

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

provided the ‘correction percentages’ of the ISS method and yielded values that variedbetween 58% and 75% (the LB standard used) compared to 27% and 70% in Ben Dor,Ong, and Lau (2014) report. This suggests that for correcting the spectra of differentspectrometers, the ISS method performs at a satisfactory level, as demonstrated by themethod’s developer.

Note, however, that in the CSIRO measurements, jitter at around 980 and 1800 nmoccurred due to detector changes, whereas in CGS measurements, these ‘noises’ are notobtained. Although the ISS-corrected spectra of the CGS measurements are still infectedby these jitters, the good correction performance, either qualitatively or quantitatively asseen in Figures 1–5, still demonstrates the conceptual power of ISS.

In general, we can say that the ISS method showed very promising results with thetwo different spectrometers and with the five Australian soils. The ISS samples success-fully corrected both the user measurements (CGS and TAU) to the CSIRO (motherhood)set-up. As discussed, it is interesting to note that the WB sample was less able to correctfor both the CGS and TAU samples than the LB sample, demonstrating that the WB

Figure 4. Spectra of the five soil samples from Western Australia as measured by CSIRO-0 (CSIRO),TAU, and TAU corrected to CSIRO-0 using the internal soil standard (ISS) (TAU-ISS-CORRECTED)approach with WB as the standard (see Soil 1 for line definition).

1284 V. KOPAČKOVÁ AND E. BEN-DOR

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

sample may not be an ideal ISS for this case. In Ben Dor, Ong, and Lau (2015), bothsamples (LB and WB) were used together to extract an average correction factor; theywere not investigated separately. It can be speculated that because the WB sample hasonly 90% quartz and 10% aragonite whereas the LB sample has 99% quartz, the formeris more sensitive to ambient (external) conditions and thus performs weakly.Furthermore, as the WB consists of very strong absorption features around 1.4 and1.9 µm (see Figures 1(a) and (b)), it may mask out any small spectral changes obtainedby the spectrometer’s performance or the protocol which the ISS method aims tocorrect. It is, therefore, suggested that in any future ISS procedure, only LB should beused.

In order to demonstrate the ISS approach as a key tool for sharing soil spectrallibraries in terms of standardization as well as to validate the results, the 10 Israelisoils at both CGS and TAU were used, as already discussed, along with the favourable ISSsample (LB) as previously proven. These soils (Table 1) cover all ranges of clay (6.3–

Figure 5. Spectra of the five soil samples from Western Australia as measured by CSIRO-0 (CSIRO),TAU, and TAU corrected to CSIRO-0 using the internal soil standard (ISS) (TAU-ISS-CORRECTED)approach with LB as a standard (see Soil 1 for line definition).

INTERNATIONAL JOURNAL OF REMOTE SENSING 1285

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

54.2%), calcium carbonates (0.62–34.2%), organic matter (0.56–8.71%), and specific sur-face areas; SSA (10.8–369.0 m2 g−1) which represents a good variety of soils withdifferent chromophores.

In Figure 8, the 10 Israeli soil spectra measured at CGS and TAU before and afterapplying the LB ISS correction procedure are shown. As can be seen, the LB ISS correctionenables the spectra of all soils at both CGS and TAU to better match between each other.To judge it quantitatively, the ICPs were calculated, as already discussed in Equation (6).Figure 9 presents these results as a histogram plot. This plot provides good confirmation ofthe visual observation in Figure 8, which was discussed previously. The LB-ISS correctionimproves the matching between the two laboratories from 5% (soil C11) to 52% (soil E9).This demonstrates the capability of the LB-ISS method to enable better matching of thesame spectra acquired by different spectrometers and under different external conditions.

In practice, a large amount of the same ISS samples (both LB and WB) are availablefrom CSIRO and ready to be shipped to those interested at no cost. The call for scientistsinterested in getting the ISS samples was already announced by Ben Dor, Ong, and Lau

Figure 6. Histogram showing the ASDS values (see Equation (3)) of each soil relative to the CSIRO-0set-up with WB and LB as the internal soil standards (ISS) (before and after applying the ISScorrection protocol): (a) the CGS ASDS values and (b) TAU ASDS values.

1286 V. KOPAČKOVÁ AND E. BEN-DOR

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

(2015) while providing contact for information details. It is strongly felt that using oneacceptable ISS sample, as available in CSIRO and suggested herein, will provide thecommunity with a better comparison of the spectral information and, above all, that itwill provide an effective means of sharing soil spectral libraries worldwide.

4. Conclusions

This study verified the ISS concept, which claims that an ISS can align measurementsmade under different conditions. It was shown that this concept can work for differentspectrometer models. This adds to the other systematic effects already reported by BenDor, Ong, and Lau (2015). The ISS samples chosen by those authors, LB and WB, werenot equal in their performance; it is assumed that the LB sample is more stable, reliable,and appropriate for these conditions. This is most likely due to the fact that it has 99%quartz which makes it stable under diverse conditions and small absorption featureswith bright albedo which still contains the same particles as in the original soil sample

Figure 7. Histogram showing the normalized deviation CGS (ND) ASDS (see Equation (5)) of each soilrelative to CSIRO-0 for WB and LB as the internal soil standards (ISS): (a) CGS (ND) ASDS values and(b) TAU (ND) ASDS values.

INTERNATIONAL JOURNAL OF REMOTE SENSING 1287

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

(e.g. grain size, shape, and natural smoothing). As the WB consists of very strongabsorption features around 1.4 and 1.9 mm, it may mask out any small spectral changesobtained by the spectrometer’s performance or the protocol which the ISS method aimsto correct, and thus it is not recommended for future use. In this direction, it issuggested that in any future ISS procedure, only the LB sample should be used.

It is highly recommended that the soil spectral community takes this opportunity tostart using the ISS concept. This will enable making large (soil spectral) data uniform andaccordingly will enable better exploitation of such a database for the benefit of all. It couldthen be merged, compared, and analysed in a large-scale format, giving better accuracy of

Figure 8. The spectra of 10 Israeli soils as measured in CGS (grey lines) and TAU (black lines): before(dashed) and after (solid) the LB-ISS correction.

1288 V. KOPAČKOVÁ AND E. BEN-DOR

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

proximate soil attribute models as well as a more precise view of the worldwide spectralinformation. In an era where all Earth Observation measurements require standards andprotocols (e.g. Group on Earth Observations 2015 declaration), this article serves as aprecursor to finally standardize the growing numbers of soil reflectance measurements.

Acknowledgement

We would like to thank Shy Ashkenazy for conducting the measurement in Israel.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This research was funded by the Czech–Israel Binational Foundation 2013–2015 and the CzechMinistry of Education and Sports [grant LH 13266].

References

Ben-Dor, E., and A. Banin. 1995. “Near-Infrared Analysis as a Rapid Method to SimultaneouslyEvaluate Several Soil Properties.” Proceedings of Soil Science Society of America 59: 364–372.

Ben Dor, E., C. Ong, and I. C. Lau. 2014. “Soil Reflectance Measurement Protocol. v2. CSIRO.” DataCollection. doi:10.4225/08/54992DB18BB6E.

Ben Dor, E., C. Ong, and I. C. Lau. 2015. “Reflectance Measurements of Soils in the Laboratory:Standards and Protocols.” Geoderma 245–246: 112–124. doi:10.1016/j.geoderma.2015.01.002.

Ben Dor, E., B. Kindel, and A. F. H. Goetz. 2004. “Quality Assessment of Several Methods to RecoverSurface Reflectance using Synthetic Imaging Spectroscopy (IS) Data.” Remote Sensing ofEnvironment 90: 389–404. doi:10.1016/j.rse.2004.01.014.

Figure 9. The ‘improvement percentages’ of the LB-ISS correction (see text for detailed explanation).

INTERNATIONAL JOURNAL OF REMOTE SENSING 1289

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

Ge, Y., C. L. S. Morgan, S. Grunwald, D. J. Brown, and D. V. Sarkhot. 2011. “Comparison of SoilReflectance Spectra and Calibration Models Obtained Using Multiple Spectrometers.” Geoderma161: 202–211. doi:10.1016/j.geoderma.2010.12.020.

Jung, A., C. Götze, and C. Gläßer. 2009. “A Comparison of Four Spectrometers and Their Effect onthe Similarity of Spectral Libraries.” Proc. of the 6th Workshop on Imaging Spectroscopy: AnInnovative Tool for Scientific and Commercial Applications, Tel Aviv, Israel, March 2009.

Knadel, M., F. Deng, A. Thomsen, and M. H. Greve. 2012. “Development of a Danish National VIS-NIR Soil Spectral Library for Soil Organic Carbon Determination.” Digital Soil Assessments andBeyond: Proceedings of the 5th Global Workshop on Digital Soil Mapping, Sydney, Australia,July 2012. Vol. 403. CRC Press.

Malley, D. P., D. Martin, and E. Ben-Dor. 2004. “Application in Analysis of Soils.” Chapter 26 in NearInfrared Spectroscopy in Agriculture, edited by R. Craig, R. Windham, and J. Workman, 729–784.Madison, WI: Soil Science Society of America, Publishers.

Nocita, M., A. Stevens, B. van Wesemael, M. Aitkenhead, M. Bachmann, B. Barthès, E. Ben Dor, D. J.Brown, M. Clairotte, A. Csorba, P. Dardenne, J. A. M. Demattê, V. Genot, C. Guerrero, M. Knadel, L.Montanarella, C. Noon, L. Ramirez-Lopez, J. Robertson, H. Sakai, J. M. Soriano-Disla, K. D.Shepherd, B. Stenberg, E. K. Towett, Ro. Vargas, and J. Wetterlind. 2015. “Chapter Four – SoilSpectroscopy: An Alternative to Wet Chemistry for Soil Monitoring.” In Advances in Agronomy,edited by D. L. Sparks, 139–159. Cambridge, MA: Academic Press. Vol. 132. ISSN 0065-2113, ISBN9780128021354. doi:10.1016/bs.agron.2015.02.002.

Pimstein, A., E. Ben Dor, and G. Notesco. 2010. “Performance of Three Identical Spectrometers inRetrieving Soil Reflectance under Laboratory Conditions.” Soil Science Society of America Journal75: 110–174.

Stevens, A., M. Nocita, G. Tóth, L. Montanarella, and B. Van Wesemael. 2013. “Prediction of SoilOrganic Carbon at the European Scale by Visible and Near InfraRed Reflectance Spectroscopy.”PLoS One 8 (6): e66409. doi:10.1371/journal.pone.0066409.

Viscarra Rossel, R. A. 2009. “The Soil Spectroscopy Group and the Development of a Global SoilSpectral Library.” EGU General Assembly Conference Abstracts, 14021, Vienna, April 19–24.Vol. 11.

Viscarra Rossel, R. A., D. J. J. Walvoort, A. B. McBratney, L. J. Janik, and J. O. Skjemstad. 2006.“Visible, Near Infrared, Mid Infrared or Combined Diffuse Reflectance Spectroscopy forSimultaneous Assessment of Various Soil Properties.” Geoderma 131 (1–2): 59–75. doi:10.1016/j.geoderma.2005.03.007.

Viscarra Rossel, R. A., T. Behrens, E. Ben-Dor, D. J. Brown, J. A. M. Dematte, K. D. Shepherd, Z. Shi, B.Stenberg, A. Stevens, V. Adamchuk, H. Aïchi, B. G. Barth`es, H. M. Bartholomeus, A. D. Bayer, M.Bernoux, K. B¨ottcher, L. Brodsk´y, C. W. Du, A. Chappell, Y. Fouad, V. Genot, C. Gomez, S.Grunwald, A. Gubler, C. Guerrero, C. B. Hedley, M. Knadel, H. J. M. Morr´as, M. Nocita, L. Ramirez-Lopez, P. Roudier, E. M. Rufasto Campos, P. Sanborn,V. M. Sellitto, K. A. Sudduth, B. G. Rawlins, C.Walter, L. A. Winowiecki, S. Y. Hong, and W. Ji. Forthcoming. “A Global Spectral Library toCharacterize the World’s Soil.“ Earth-Science Reviews. doi:10.1016/j.earscirev.2016.01.012.

Wetterlind, J., and B. Stenberg. 2010. “Near-Infrared Spectroscopy for Within-Field SoilCharacterization: Small Local Calibrations Compared with National Libraries Spiked with LocalSamples.” European Journal of Soil Science 61 (6): 823–843. doi:10.1111/ejs.2010.61.issue-6.

Willis, D. E. 1972. “Internal Standard Method Calculations.” Chromatographia 5 (1): 42–43.doi:10.1007/BF02270655.

1290 V. KOPAČKOVÁ AND E. BEN-DOR

Dow

nloa

ded

by [

Ces

ka G

eolo

gick

a Sl

uzba

], [

Ver

onik

a K

opak

ová]

at 0

1:46

22

Febr

uary

201

6

View publication statsView publication stats