Supplementary material for "From a polar to a marine environment: has thechanging Arctic led to a shift in aerosol light scattering properties?"by Heslin-Rees et al. (2020)

Contents

1 Data availability 25

2 Particle light scattering coefficients compared to ambient relative humidity 3

3 Back trajectory calculations (method) 5

4 The contribution of coarse mode particles to particle light scattering 6

5 Results of trajectory analysis with a fixed sea ice 7

6 Annual air mass contribution 810

7 The annual contribution of the scattering Ångström exponent with respect to the main air mass sectors 9

8 Table of results 10

1

1 Data availability

Figure S1 shows the available data of nephelometer measurements used in this study.

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

Years

0

2000

4000

6000

8000

Hou

rly c

ount

Raw Cleaned & hourly threshold applied

Figure S1. Data coverage: Grey bars denotes the number of hourly data points provided by the nephelometer before the quality controlprocedure. The green bars denote the number of hourly data points used.

2

2 Particle light scattering coefficients compared to ambient relative humidity15

Within this work, the effect of local cloud occurrence is explicitly addressed in the trend analysis. As a proxy for in-clouddata, the ambient relative humidity (RH) (see Fig. S2) is being used. Figure S2a shows that there is a slight influence from theambient relative humidity on the RH at the sample outlet of the nephelometer. From Fig. S2b it is shown that increasing ambientRH results in increased σsp up to a point, after which the high levels of ambient RH associated with the presence of clouds beginto reduce the inlet collection efficiency. The point at which inlet efficiency is negatively impacted by the presence of clouds is20approximately 95 %. The overall effect is deemed insignificant. Figure S2c shows that alpha declines with increasing ambientRH, suggesting that the aerosol particles take up water and thus increase in size despite the nephelometer measuring under dryconditions (i.e. sample outlet RH < 40%). After approximately 93% ambient RH, there is a slight upturn in α suggesting that thepresence of clouds at the inlet acts to remove a proportion of larger particles. However, the median α for cloud-free conditions(i.e. < 95%) is larger suggesting in fact that the size distribution is shifted closer to the smaller particles; this observation is put25down to the increasing uncertainties observed in the data.

3

30 40 50 60 70 80 90 100Ambient RH (%)

0

2

4

6

8

10

12

14

16

Inle

t Rel

ativ

e H

umid

ity (%

)

Median(no threshold i.e. clouds): 5.40

Median(cloud-free): 4.80

30 40 50 60 70 80 90 100Ambient RH (%)

0

1

2

3

4

5

6

7

sp (M

m1 )

Median(no threshold i.e. clouds): 2.36

Median(cloud-free): 2.50

30 40 50 60 70 80 90 100Ambient RH (%)

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

2.00

(-)

Median(no threshold i.e. clouds): 1.45

Median(cloud-free): 1.47

Figure S2. Effect of ambient relative humidity (RH): (a) The RH at the sample outlet vs. the ambient RH. (b) The scattering coefficient (σsp)vs. the ambient (RH). (c) The Ångström exponent (α) vs. the ambient (RH). The data is placed in bins of size 200 according to ascendingvalues of ambient RH. The medians for each consecutive bin are calculated for the variables. The dotted red line (at ambient RH = 95%)displays the cut-off point, whereby in-cloud scavenging and inlet losses, influence σsp significantly. (RH threshold of 92% perhaps)

4

3 Back trajectory calculations (method)

Figure S3 presents a detailed description of how the residence times over the three different surface types (land, ice and openwater) are calculated. Note that the proportion of time the back trajectories spend above the mixed layer is also presented, andfurthermore does not count towards the surface residence times.30

IceOpen waterLand

Above MLmean (lat, lon) =80.83, -22.22 (NW)

0.0

0.2

0.4

0.6

0.8

1.0

SIC

(-)

Figure S3. Example of the source attribution using back trajectory calculations: 168-h back trajectory arriving at the Zeppelin Observatoryon 2008-10-05 at 13:00:00. Sections of the back trajectory are coloured as follows: above the mixed layer (purple), and within the mixedlayer and above land (green), open water (blue), ice (white). The mean coordinates of the back trajectory, based only on the section thattraverses through the ML, is denoted by the star. The red cross displays the location of the Zeppelin Observatory. The sea ice concentrations(SIC) at the time of arrival are displayed, based on monthly 1° x 1°-resolution satellite-derived Met Office (2006) data. The meridian 11°89’Eand parallel of 78°91’N lines are presented for reference.

5

4 The contribution of coarse mode particles to particle light scattering

Coarse-mode particles contribute a significant proportion of the overall scattering observed at ZEP. To illustrate the importanceof the coarse mode (here defined for particles above 1µm), we have used particle size distribution measurements of fine andcoarse mode from 2008 and Mie theory assuming dry spherical particles with a refractive index of m= 1.51 and a wavelengthof λ= 550 nm (details on size distribution measurements and Mie code are given in Zieger et al., 2010). Figure S4 shows the35mean monthly average of the particle surface size distribution and the differential scattering coefficient (the integral will deliverσsp) for July to October. As can be seen, in 2008 the coarse mode contributed on average between 20 and 45 % (at λ= 550 nm)to the particle light scattering and as such is an important contributor to overall light scattering properties. In Zieger et al.(2010), the coarse mode was identified as sea salt using chemical and trajectory analysis.

101 102 103 1040

5

10

15x 106

d / nm

dS/d

logD

(nm

2 cm

)−3

Surface size distribution

JulyAugustSeptemberOctoberSMPS/OPSS sep.

101 102 103 1040

0.5

1

1.5

2

2.5

3x 10−6

d / nm

dσ/d

D (m

)

−2

Differential scattering coefficient

July (σCM=19.3%)

August (σCM=22.5%)

September (σCM=37.2%)

October (σCM=45.7%)

DMPS/OPSS sep.

a) b)

Figure S4. Monthly averages of (a) the particle surface size distribution and (b) the differential scattering coefficient (at λ= 550 nm)calculated using the SMPS (scanning mobility particle sizer) and OPSS (optical particle size spectrometer) measurements from Zieger et al.(2010). The percent contribution to light scattering of the coarse mode (CM) is given in the legend of panel b.

6

5 Results of trajectory analysis with a fixed sea ice40

Figure S5 shows the relative surface contribution of the air masses arriving at ZEP. To investigate the effect of retreating Arcticsea ice on the surface type contributions alone, the contributions are calculated by holding the mean SIC at the annual cycle for1999, The solid line denotes the surface type contributions based on the monthly sea ice concentrations (SIC). The dashed lineis the SIC fixed to the values for the year of 1999. It can be seen that the relative contribution from different surface types doesnot change significantly between the dashed and solid lines. This suggests that the retreat in sea ice is not the main contributor45to the increase spend the back trajectories spend over the open water.

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

Years

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

Rel

ativ

e co

ntrib

utio

n fr

om su

rfac

e ty

pes (

-)

land ice open water

Figure S5. Relative surface contributions: the solid line represents the relative seasonal presence of each different surface type. The dashedline presents the same information but with sea ice concentrations set at 1999 values. Note the contributions do not include parts of the backtrajectories above the mixed-layer (thus the sum is < 1).

7

6 Annual air mass contribution

The increased presence of south-west air masses in the summer is displayed in Fig. S6. The presence of SW air masses inthe summer is consistent with previous studies (Freud et al., 2017; Dall´Osto et al., 2017). The dominant flow from the northduring the winter and spring months is well documented by Stone et al. (2014).50

J F M A M J J A S O N DMonths

0.0

0.2

0.4

0.6

0.8

1.0

Rel

ativ

e co

ntrib

utio

n fr

om re

gion

(-)

NW NE SE SW

Figure S6. Annual variation of the monthly regional air mass contributions determined from the trajectory analysis for the entire observationperiod (1999-2016).

8

7 The annual contribution of the scattering Ångström exponent with respect to the main air mass sectors

In Figure S7, the hourly median α values are presented for each region, and for every year of observation, as a set of normalisedhistograms. The shift in air mass contributions is displayed in the relative histogram areas (A). The same years that featurepeaks in their open water contributions (i.e. 2006, 2011, and 2016 in Fig S5) display decreases in their respective α medians.α for the western air masses are distinctly lower, compared to the eastern originated air masses. The western air masses are55also more bimodal, suggesting the presence of more coarse-mode particles. It needs to be noted that the number of hourly datapoints for each season is not the same; spring and winter comprise of much more of the data (Tables S1 and S4) and thus annualmedians are weighted more in their favour.

0

1

2: 1.974 A: 44.20: 1.977 A: 24.28: 1.937 A: 18.84: 1.900 A: 11.96

1999

0.0

0.5

1.0: 2.330 A: 34.07: 2.504 A: 17.43: 2.299 A: 25.75: 2.320 A: 13.13

2000

0.0

0.5

1.0: 2.304 A: 28.77: 2.337 A: 16.89: 2.298 A: 37.02: 2.320 A: 15.38

2001

0.0

0.5

1.0: 1.174 A: 17.79: 1.138 A: 15.50: 1.577 A: 34.17: 1.215 A: 31.00

2002

0.0

0.5

1.0: 1.378 A: 23.08: 1.157 A: 14.31: 1.248 A: 40.57: 1.283 A: 22.03

2003

0.0

0.5

1.0: 1.926 A: 32.53: 1.896 A: 14.26: 1.834 A: 38.38: 1.783 A: 14.10

2004

0.0

0.5

1.0: 1.823 A: 31.71: 1.576 A: 14.15: 1.884 A: 36.16: 1.605 A: 17.21

2005

0.0

0.5

1.0: 1.580 A: 21.69: 1.599 A: 19.40: 1.714 A: 35.30: 1.663 A: 22.75

2006

0.0

0.5

1.0: 1.297 A: 21.07: 1.317 A: 11.92: 1.342 A: 45.45: 1.278 A: 20.38

2007

0.0

0.5

1.0: 1.392 A: 16.79: 1.469 A: 19.24: 1.440 A: 41.03: 1.477 A: 22.35

2008

0.0

0.5

1.0: 1.366 A: 27.24: 1.311 A: 7.96: 1.436 A: 46.59: 1.092 A: 17.83

2009

0.0

0.5

1.0: 1.291 A: 24.67: 0.931 A: 15.93: 1.175 A: 38.18: 1.244 A: 20.12

2010

0 20.0

0.5

1.0: 1.327 A: 35.25: 1.055 A: 22.02: 1.263 A: 20.16: 1.182 A: 21.03

2011

0 20.0

0.5

1.0: 1.117 A: 31.04: 0.957 A: 17.90: 1.224 A: 33.33: 1.059 A: 15.98

2012

0 20.0

0.5

1.0: 1.379 A: 23.84: 1.210 A: 26.91: 1.317 A: 29.09: 1.015 A: 19.56

2013

0 20.0

0.5

1.0: 1.275 A: 21.88: 1.239 A: 25.89: 1.307 A: 29.97: 1.407 A: 21.14

2014

0 20.0

0.5

1.0: 1.159 A: 22.34: 0.750 A: 22.79: 1.340 A: 29.25: 1.032 A: 23.36

2015

0 20.0

0.5

1.0: 1.272 A: 19.02: 0.576 A: 35.50: 1.427 A: 20.08: 0.584 A: 18.51

2016

Cardinal directionsSE SW NE NW (-)

Nor

mal

ised

cou

nt

Figure S7. Annual histograms for hourly scattering Ångström exponent (α) medians of the four source regions namely, North-east (NE),South-east (SE), North-west (NW) and South-west (SW). The histograms are normalised by the total of each respective year. The respectiveα medians for each region are given, along with the region’s relative contribution (A (%)). The area left of the vertical dashed line, where(α< 1), represents the approximate contribution from coarse-mode aerosol particles. Note the scale for the year 1999 is different.

9

8 Table of results

Tables S1, S2, S3 and S4 present the seasonal trends in the aerosol optical properties for both statistical tests (i.e. least mean60square and seasonal Mann-Kendall test). The median for each season is also given, along with the number of hourly data pointscomprised of the points used in the trend analysis.

Table S1. Trends in scattering coefficient (σsp) for daily medians and seasonal medians: Theil-Sen Slope estimator (TS) is presented asthe change in Mm−1 per year and the relative change (% yr−1). Least Mean Squares (LMS) is the percentage change based on a logtransformation. For daily medians, the Man-Kendall test (MK test) is based on a "prewhitened" time series; all seasons denotes the result ofthe seasonal MK test. Decreasing (D) and increasing (I) statistically significant trends are signified.

Period Hourly Data σsp TS TS LMS MK LMSpoints points (Mm−1) (Mm−1yr−1) (%yr−1) (%yr−1)

Daily spring 14930 792 3.38 0.04 1.24 1.52 I Isummer 6875 470 1.11 0.03 2.66 1.81 I Iautumn 10137 630 1.49 0.07 4.44 3.50 I Iwinter 13411 690 3.10 0.19 6.03 6.10 I Itotal 45353 2582 2.30 0.06 2.43 2.90 I I

Seasonal spring 14930 14 3.40 0.01 0.22 -0.24 - -summer 6875 15 1.09 0.02 2.06 0.76 - -autumn 10137 12 1.64 0.09 5.24 3.74 I Iwinter 13411 11 3.24 0.21 6.50 4.60 I Itotal 45353 52 2.09 0.05 2.21 1.27 I -

Table S2. Trends in backscattering coefficient (σbsp) for daily medians and seasonal medians: Theil-Sen Slope estimator (TS) is presentedas the change in Mm−1 per year and the relative change (% yr−1). Least Mean Squares (LMS) is the percentage change based on a logtransformation. For daily medians, the Man-Kendall test (MK test) is based on a "prewhitened" time series; all seasons denotes the result ofthe seasonal MK test. Decreasing (D) and increasing (I) statistically significant trends are signified.

Period Hourly Data σbsp TS TS LMS MK LMSpoints points (Mm−1) (Mm−1yr−1) (%yr−1) (%yr−1)

Daily spring 14704 791 0.44 0.01 1.32 0.34 I -summer 6063 469 0.18 0.00 2.01 0.52 I Iautumn 9054 629 0.21 0.01 3.32 0.85 I Iwinter 12781 691 0.39 0.02 5.27 1.86 I Itotal 42602 2580 0.30 0.01 2.14 0.62 I I

Seasonal spring 14704 14 0.44 0.00 0.73 -0.01 - -summer 6063 15 0.19 0.00 2.53 0.46 I Iautumn 9054 12 0.26 0.01 3.93 0.87 I Iwinter 12781 11 0.42 0.02 5.78 1.44 I -total 42602 52 0.30 0.01 2.30 0.43 I -

10

Table S3. Trend in hemispheric backscattering fraction b (-) for daily medians and seasonal medians: Theil-Sen Slope estimator (TS) ispresented as the change per year and the relative change (% yr−1). Least Mean Squares (LMS) is the relative change based on a log trans-formation. For daily medians, the Man-Kendall test (MK test) is based on a "prewhitened" time series; all seasons denotes the result of theseasonal MK test. Decreasing (D) and increasing (I) statistically significant trends are signified.

Period Hourly Data b TS TS LMS MK LMSpoints points (-) (yr−1) (%yr−1) (%yr−1)

Daily spring 14490 791 0.1296 0.0000 0.0320 -0.0140 - -summer 5654 467 0.1490 -0.0008 -0.5194 -0.0820 D Dautumn 8508 622 0.1323 -0.0012 -0.9020 -0.1540 D Dwinter 12515 690 0.1206 -0.0009 -0.7734 -0.1470 D Dtotal 41167 2570 0.1301 -0.0004 -0.3249 -0.0710 D D

Seasonal spring 14490 14 0.1276 0.0009 0.7199 0.0610 - -summer 5654 15 0.1464 -0.0005 -0.3575 -0.0620 - -autumn 8508 12 0.1319 -0.0007 -0.5545 -0.1110 - -winter 12515 11 0.1196 -0.0001 -0.0739 -0.1110 - -total 41167 52 0.1318 -0.0002 -0.1674 -0.0400 - -

Table S4. Trend in scattering Ångström exponent (-) for daily medians and seasonal medians: Theil-Sen Slope estimator (TS) is presentedas the change per year and the relative change (% yr−1). Least Mean Squares (LMS) is the percentage change based on a log transformation.For daily medians, the Man-Kendall test (MK test) is based on a "prewhitened" time series; all seasons denotes the result of the seasonal MKtest. Decreasing (D) and increasing (I) statistically significant trends are signified.

Period Hourly Data α TS TS LMS MK LMSpoints points (-) (yr−1) (%yr−1) (%yr−1)

Daily spring 14930 792 1.59 -0.06 -3.51 -4.89 D Dsummer 6875 470 1.68 -0.07 -4.42 -4.91 D Dautumn 10137 630 1.17 -0.09 -7.29 -7.61 D Dwinter 13411 690 1.30 -0.08 -5.84 -7.31 D Dtotal 45353 2582 1.44 -0.07 -4.92 -6.42 D D

Seasonal spring 14930 14 1.51 -0.05 -3.10 -1.58 D Dsummer 6875 15 1.74 -0.07 -3.95 -2.56 D Dautumn 10137 12 1.20 -0.10 -8.68 -3.54 D Dwinter 13411 11 1.22 -0.06 -4.86 -2.99 D Dtotal 45353 52 1.36 -0.07 -5.05 -2.65 D D

11

References

Dall´Osto, M., Beddows, D., Tunved, P., Krejci, R., Ström, J., Hansson, H.-C., Yoon, Y., Park, K.-T., Becagli, S., Udisti, R., et al.: Arctic seaice melt leads to atmospheric new particle formation, Sci. Rep., 7, 3318, 2017.65

Freud, E., Krejci, R., Tunved, P., Leaitch, R., Nguyen, Q. T., Massling, A., Skov, H., and Barrie, L.: Pan-Arctic aerosol number size distribu-tions: seasonality and transport patterns, Atmos. Chem. Phys., 17, 8101–8128, 2017.

Met Office, H. C.: HadISST 1.1 - Global sea-Ice coverage and SST (1870-Present), http://badc.nerc.ac.uk/view/badc.nerc.ac.uk, accessed:2019-04-30, 2006.

Stone, R. S., Sharma, S., Herber, A., Eleftheriadis, K., and Nelson, D. W.: A characterization of Arctic aerosols on the basis of aerosol optical70depth and black carbon measurements, Elementa., 2014.

Zieger, P., Fierz-Schmidhauser, R., Gysel, M., Ström, J., Henne, S., Yttri, K. E., Baltensperger, U., and Weingartner, E.: Effects of relativehumidity on aerosol light scattering in the Arctic, Atmos. Chem. Phys., 10, 3875–3890, 2010.

12