Characterization of the cloud microphysical and optical properties and aerosol-cloud interaction in the Arctic from in situ ground-based measurements during the CLIMSLIP-NyA campaign, Svalbard

23 24 This study will focus on cloud microphysical and optical characterization of three different 25 types of episodes encountered during the ground based CLIMSLIP-NyA campaign performed 26 in Ny-Alesund, Svalbard: the Mixed Phase Cloud (MPC), snow precipitation and Blowing 27 Snow (BS) events. These in situ cloud measurements will be combined with aerosol 28 measurements and air mass backtrajectory simulations to qualify and parameterize the arctic 29 aerosol cloud interaction and to assess the influence of anthropogenic pollution transported into 30 the Arctic. 31 The results show a cloud bimodal distribution with the droplet mode at 10 μm and the crystal 32 mode centered at 250 μm, for the MPC cases. The precipitation cases presents a crystal 33 distribution centered around 350 μm with mostly of dendritic shape. The BS cases show a 34 higher concentration but smaller crystals, centered between 150 and 200 μm, with mainly 35 irregular crystals. 36 A “polluted” case, where aerosol properties are influenced by anthropogenic emission from 37 Europe and East Asia, was compared to a “clean” case with local aerosol sources. These 38 anthropogenic emissions seem to cause higher Black Carbon, aerosol and droplet 39 concentrations, a more pronounced accumulation mode, smaller droplet sizes and a higher 40 activation fraction Fa. Moreover, the activation diameter decreases as the droplet diameter 41 increases and Fa increases showing that smaller particles are activated and droplets grow when 42 the aerosol number decreases. This is in agreement with the first (Twomey) and second 43 (Albrecht) aerosol indirect effect. The quantification of the variations of droplet concentration 44 and size leads to IE (Indirect Effect) and NE (Nucleation Efficiency) coefficients values around 45 0.2 and 0.43, respectively. These values are close to those found by other studies in the arctic 46 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2017-672 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 1 November 2017 c © Author(s) 2017. CC BY 4.0 License.


24
This study will focus on cloud microphysical and optical characterization of three different 25 types of episodes encountered during the ground based CLIMSLIP-NyA campaign performed 26 in Ny-Alesund, Svalbard: the Mixed Phase Cloud (MPC), snow precipitation and Blowing 27 Snow (BS) events. These in situ cloud measurements will be combined with aerosol 28 measurements and air mass backtrajectory simulations to qualify and parameterize the arctic 29 aerosol cloud interaction and to assess the influence of anthropogenic pollution transported into 30 the Arctic. 31 The results show a cloud bimodal distribution with the droplet mode at 10 µm and the crystal 32 mode centered at 250 µm, for the MPC cases. The precipitation cases presents a crystal 33 distribution centered around 350 µm with mostly of dendritic shape. The BS cases show a 34 higher concentration but smaller crystals, centered between 150 and 200 µm, with mainly 35 irregular crystals. 36 A "polluted" case, where aerosol properties are influenced by anthropogenic emission from 37 Europe and East Asia, was compared to a "clean" case with local aerosol sources. These 38 anthropogenic emissions seem to cause higher Black Carbon, aerosol and droplet 39 concentrations, a more pronounced accumulation mode, smaller droplet sizes and a higher 40 activation fraction Fa. Moreover, the activation diameter decreases as the droplet diameter 41 increases and Fa increases showing that smaller particles are activated and droplets grow when 42 the aerosol number decreases. This is in agreement with the first (Twomey) and second 43 (Albrecht)  The rapid change in aerosol properties occurring in spring is known to cause changes in arctic 95 cloud properties, the so-called aerosol indirect effect. Increase in aerosol concentration with 96 constant Liquid Water Path (LWP) is known to increase cloud droplet concentration and cloud 97 optical thickness but decrease droplet size (Twomey, 1974(Twomey, , 1977, decrease the precipitation 98 efficiency and increase the cloud lifetime (Albrecht, 1989). Also, in a temperature rise scenario, 99 the cloud height is expected to increase (Pincus and Baker, 1994). The impacts of anthropogenic 100 aerosol transported to the Arctic on clouds are not fully understood, but Garrett and Zhao (2006) 101 showed that the cloud emissivity is higher for polluted case, contributing to the arctic warming. 102 103 In the case of artic MPC where liquid and ice phases coexist, the aerosol-cloud interaction is 104 complexified by the addition of the ice phase and several interaction mechanisms have been 105 assumed. Lohmann (2002aLohmann ( , 2002b proposed that an increase in ice nuclei could increase the 106 cloud ice content at the expense of the liquid content. This so-called glaciation indirect effect 107 would mean, as the precipitation is more efficient for the ice phase, a decrease in cloud cover 108 in lifetime. The riming indirect effect predicts a riming efficiency decrease due to the 109 supercooled droplet size decrease. Thus, an increase in Cloud Condensation Nuclei (CCN) 110 could lead to a decrease in Ice Water Content (IWC) and ice particles concentration (Borys et 111 al., 2003). According to the data of the two measurement campaigns ISDAC (Indirect and Semi The work presented here is included in the frame of the project CLIMSLIP (CLimate IMpacts  120 of Short-LIved Pollutants in the polar region). The main objective of this project is to reduce the 121 uncertainties of the radiative forcing due to the anthropogenic emissions of tropospheric ozone, 122 methane and aerosol including Black Carbon (BC The cloud ground based instrumentation used during CLIMSLIP-NyA was installed on a 161 measurement pole and is presented in Figure 2. The cloud optical and microphysical properties 162 were thus assessed by three independent instruments: a PMS Forward Scattering Spectrometer 163 Probe (FSSP-100), a Cloud Particle Imager (CPI) and a Polar Nephelometer (PN). They were 164 all connected to the same pump by plastic tubes, leading to the sampling volume indicated on 165 Figure 2. They were operated approximately 2 m above the platform level and mounted on a 166 tilting and rotating mast, allowing them to be moved manually in the prevailing wind direction. 167 The proper alignment of their inlet with the flow was based on the wind direction measurements 168 performed by a mechanical and ultrasonic anemometer. 169 170 The FSSP-100 measures the number and the size of particles going through the sampling 171 volume, from the forward scattering of a 632.8 nm wavelength laser beam (Knollenberg, 1981, 172 Dye and Baumgardner, 1984). Using the Mie theory, this instrument is dedicated to droplets. 173 The Particle Size Distribution (PSD) is thus computed in 15 adjustable size classes with 174 uncertainties on the effective diameter and LWC of respectively 2 µm and 30 % (Febvre et al.,175 2012). 176 The CPI is an imager and takes pictures of the particles when going through the detection 177 volume with 256 grey levels, thanks to a CCD camera with a resolution of 1024 × 1024 pixels. 178 These images allow computing the particles size and so the PSD, but several morphological 179 The three cloud instruments operated at a one Hz resolution. The data processing has followed 225 the conclusions of the cloud instrumentation study presented in Guyot et al. (2015). This paper 226 highlights the biases that can exist between the instruments and the need of an Ensemble of 227 Particles Probe (EPP) to standardize the data. In the case of the CLIMSLIP campaign, such 228 correction was not possible for two reasons not developed further.
(1) Strong discrepancies of 229 the EPP Nezvorov probe, probably because of a too low sampling speed.
(2) The 230 standardization according to the extinction coefficient of the PN is not consistent with the 231 aerosol data (there are more droplets than CCN). Thus, this study will not provide quantitative 232 results but qualitative ones based on case comparisons and variation studies. 233 According to Guyot el al. (2015), measurements with an angle between the instruments 234 orientation and the wind direction higher than 30° can modify the PSD due to changes in the 235 sampling conditions. Those measurements were therefore not taken into account for the study. 236 Moreover, the ground based low sampling speed induces low sampling rate, especially for the 237 CPI with values between 0.5 and 20 sampled particles per minute. This doesn't allow us to 238 work on low time resolution scale. To get sufficient particle statistics, the minimum average 239 time resolution will be 1 minute for the FSSP and one day for the CPI. 240 241 During the aircraft campaign, a cloud particle can break on impaction with the inlet due to the 242 high sampling speed corresponding to the plane speed. This results in more numerous and 243 smaller droplets or crystals and creates artifacts in the PSD (Rogers et al., 2006). Due to the 244 low sampling speed, ground based measurements has the advantage to avoid this effect, but at 245 the expense of the sampling rate. 246 247 248 In the second case, the ceilometer locates the liquid layer around 1 km altitude or more ( Figure  281 3.c). No droplets are sampled. The station is so below the mixed layer, within the ice 282

Identification and characterization of the study cases
precipitation. This layer has a variable extinction coefficient depending on the crystal density 283 but the laser beam is not completely attenuated. The relative humidity shows high values around 284 90 % but remains lower than the MPC cases. 285 Moreover, the temperature varied between -20 to -1 °C, so it remains always below the 286 solidification point, liquid particles were always supercooled droplets. The Blowing Snow 287 episodes will be discussed in annex. 288 289 In the following, the LMPL and precipitation layer cases will be microphysically and optically 290 characterized. These characterizations will be useful to determine futures measurements that 291 are not completed with visual observations (e.g., remote sensing measurements). Moreover, 292 combined with other measurement campaign in the Arctic, we hope to increase knowledge 293 about growth processes in low level mixed phase arctic clouds. 294 295 296

298
Arctic MPC can be characterized by a succession of layers with liquid or ice dominance. The 299 phase heterogeneity is both horizontal and vertical. Because of the fixed position of the 300 measuring station, we could not control the location of the measurements within the cloud 301 system. However, a characterization of the mean parameters is possible. 302 303 The determination of the thermodynamic phase of a cloud can be based on microphysical and 304 optical criteria. Figure 4 presents the occurrence number of the MPC liquid fraction Fliq and the 305 asymmetry parameter g. Fliq is computed as : (1) 308

309
The results show a higher observation frequency for extreme Fliq values (close to 0 or 100 %). 310 The minimum frequency is between 20 and 70 %. This means that the low level mixed phase 311 cloud layers are preferentially with liquid or ice dominance for the spatial resolution of our 312 measurements. This confirms the conclusions from the scientific literature (e.g., Gayet and al.,313 2009; Korolev and Isaac, 2006). 314 Moreover, g shows a more or less linear relation with Fliq. This highlights the relation between 315 the optical properties and the microphysical properties. Therefore, the knowledge of the MPC 316 microphysical properties is a key parameter to reliably assess the radiative transfer in the Arctic. 317 The g variability is significantly larger for Fliq below 50 %. This tends to show a more complex 318 optical behavior for ice dominating layers. minute resolution corresponding to a spatial resolution of 800 meters. 324 325 326 Figure 5 shows the average PSD, from 3 µm to 2.3 mm, obtained with the FSSP and the CPI 327 for the four LMPL cases. The mean Fliq is also indicated. The four PSD show similar trends, 328 i.e. two modes centered at 10 µm for droplets and around 250 µm for ice crystals. 329 According to Costa et al. (2014), these PSD correspond to the coexistence regime characterized 330 by RHw (relative humidity according to liquid water) and RHi (relative humidity according to 331 ice) > 100 % and stable coexistence of crystals and supercooled liquid droplets with the droplet 332 PSD 10 6 higher than the crystal PSD. This is opposite to the Bergeron regime where RHw < 100 333 % and RHi > 100 %, so the crystals grow in expense of the droplets (Costa et al., 2014). This 334 reveals that the Wegener-Bergeron-Findeisen process doesn't alone explain the formation and 335 growth of ice crystals. 336 However, the March 11 th and 29 th PSDs show differences with the other cases with a high 337 concentration for the smallest CPI classes. This is due to big droplets sampled by the CPI. The 338 FSSP doesn't show such consequent differences in droplet PSD or diameter. We also point out 339 that the absolute values should not be taken into account. Indeed, in addition to instrumental 340 issues (see Guyot el al., 2015), the results and the differences between the cases are largely 341 dependent on the station residence time within the liquid or mixed layer which cannot be 342 controlled. Similar PSDs were observed at the Mount-Zeppelin station by Uchiyama et al. 343 (2013) in 2011. This publication concludes that the liquid/ice distribution is a function of the 344 cloud evolution stage; we highlight here the importance of the station position inside the cloud 345 system for our data analysis. The shape classification performed by the CPI is presented Figure 6.a. The high droplet 355 concentrations of the smallest CPI classes observed on March 11 th and 29 th (see Figure 5) are 356 responsible of the strong number dominance of the droplets with a value of around 85 %. 357 However, liquid water represents a very small proportion in mass and surface fractions. For 358 these two quantities, side planes and irregular shapes dominate. 359 The assessment of the crystal growth mode is confronted to the fact that the measurement 360 station can change its position in the cloud. An evolution in the CPI PSD is so not necessary 361 due to particles growth. However, the crystal shape, accurately measured by the CPI, is a good 362 indicator for the growth mode and the high percentage of regular shape would indicate a growth Besides, the temperature is higher with an average value of -5 °C. This could reveal an influence 399 of the mixed layer and/or temperature effect. 400 However, the ceilometer located the cloud base at an altitude of approximately 1000 m for the 401 three days, which would indicate that the station position doesn't explain the differences. The 402 temperature differences could lead to different growth processes and so different sizes. 403 This information can be provided by the CPI image classification presented Figure 8.  To conclude, the results were limited by the low particle sampling rate and the uncontrolled 437 position of the station inside the cloud system. However, differences between LMPL and 438 precipitation layer have been explicated and allow a quick recognition without visual 439 observations in future studies. These results will be compared to other measurement campaign 440 for a better understanding of the microphysical processes and feedbacks that take place in low 441 level mixed phase arctic clouds. The objective of this part is to quantify the effects of the aerosol properties on the cloud 447 properties observed during the CLIMSLIP campaign. To do this, we will in a first step compare 448 the two experiments of March 11 th and 29 th that will be the "clean" and "polluted" cases, 449 respectively. In a second step, several aerosol cloud-interaction processes will be evaluated and 450 in situ measurements will be used to assess quantities that are required in parametrization of the 451 arctic aerosol-cloud interaction. 452 453

455
This analysis will be supported by results from the lagrangian particle dispersion model (PSC) expressed in kg of tracer per air kg. In this study, we will focus on the CO tracer which 476 gives an assessment on the origin of the anthropogenic pollution transported to Svalbard. 477 478 The aerosol cloud interaction study will also be supported by two The March 11 th case, just like March 29 th , presents a stable atmosphere with a low level mixed 497 phase cloud. The liquid layer was sampled but, unfortunately, the ceilometer beam was almost 498 entirely attenuated within the first 500 meters, avoiding to assess cloud top and base altitude. 499 The sounding balloon show the inversion layer around 925 mb (700 m) for both days. 500 Figure 9 shows the time evolution of the DMPS, CPC and FSSP concentration, the activation 501 fraction and the average BC concentration. The DMPS ceased to work from 7:30 until the 502 following day, but, as the CPC and aethalometer parameters show almost constant values until 503 12:00, the DMPS concentration is assumed to do the same. The DMPS concentration is plotted 504 for different particles sizes (total, > 50 nm and > 100 nm). The DMPS PSD shows a bimodal 505 distribution with a pronounced Aitken mode which is as important as the accumulation mode 506 (not shown). This is obvious in Figure 9 where the accumulation mode concentration, i.e. 507 particles sizes larger than 100 nm, equals half the total concentration. 508 The CPC displays an aerosol concentration (> 3 nm) relatively stable and weak between 100 509 and 130 cm -3 . The average BC concentration reaches 22.6 ng m -3 during the liquid episode. The 510 FSSP shows a droplet concentration up to 100 cm -3 , which leads to Fa values between 60 and 511 80 % for the sections clearly in the densest zone of the MPC liquid layer. 512 The mass arriving at the station during the liquid episode of the March 11 th described in Figure 9. 530 The FPES shows that the aerosol sources are mainly located in the north of Scandinavia and so 531 that the long-range transport of anthropogenic aerosols is relatively limited. Indeed, over the 532 FLEXPART-WRF time computation of 12 days, the air masses come principally from Svalbard 533 and Scandinavia surrounding, showing very slow move. The CO PSC map presents an 534 anthropogenic origin dominated by North Europa: Scandinavia, north of Germany, Netherland, 535 Belgium and north of France. 536 The closer air masses origin makes this case the "clean" case. The important contribution of 537 local aerosol sources, mainly composed of gaseous precursors for the arctic region during this 538 period of the year (Quinn et al., 2007), explains the relative small aerosol mean diameter and 539 the high Aitken mode concentration observed by the DMPS (see Figure 9). 540 541 542 4.2 The "Polluted" case of March 29 th 543 544 Figure 11 displays the same time series as Figure 9 for the liquid episode of March 29 th . The 545 CPC and DMPS total concentration are decreasing going respectively from 220 cm -3 to 120 cm -546 3 and from 175 cm -3 to 80 cm -3 , due to the scavenging by ice precipitation. The FSSP droplet 547 concentrating reaches 150 cm -3 and the average BC concentration 65.8 ng m -3 . Comparing to 548 the March 11 th case, these four concentration are all higher during the March 29 th . The activation 549 fraction is also higher on March 29 th with values between 80 and 100 % in the liquid layer and 550 Fa increases as the aerosol concentration decreases. Moreover, the DMPS PSD shows that 90 % of the aerosol concentration is included in the 557 accumulation mode, with an effective diameter almost constant at 300 nm. Therefore, the 558 aerosol diameter is larger and the droplet diameter is smaller for the March 29 th case compared 559 to the March 11 th case. 560 561 562 Figure 11: Same as Figure 9 for the March 29 th case 563 564 The differences observed between the two days can be explained by the air masses origin. 565 Figure 12 shows the same FLEXPART-WRF FPES and CO PSC for the air mass arriving at 566 the station during the liquid episode of the March 29 th . Backtrajectories distinguish clearly two 567 origin regions. The first one is Western Europa. The second air mass shows higher values of 568 time residence and comes from northeast Asia: northeast China and extreme east Russia. The 569 particularity of March 29 th consists thus in this air mass coming from Asia which is the region 570 generally accepted to emit the highest aerosol concentration compared to the others regions of 571 the world (Boucher et al., 2013). 572 Therefore, compared to the "clean" case of March 11 th , March 29 th shows long range transport 573 of anthropogenically influenced air masses, leading to higher aerosol concentration in the Arctic 574 with especially a BC mass concentration 3 times higher. Thus, March 29 th constitute the 575 "polluted" case. According to Quennehen et al. (2012), during the route, the Aitken mode 576 concentration quickly decreases by coagulation, for the benefit of the accumulation mode, 577 increasing the average effective diameter. This explains the accumulation mode dominance 578 observed in Figure 11 and the increase of the average DMPS effective diameter, and confirms 579 the strong influence of the lower latitudes emissions during the "polluted" case. On the contrary, 580 the "clean" case shows local sources composed of fresh particles, for at least half the 581 concentration. 582 This long range anthropogenic pollution has also strong influence on cloud properties. Indeed, 583 CCN abilities being mainly due to the aerosol size in the Arctic (Mc Farquhar et  This qualitative study has to be completed with quantitative parameters that can be found in the 594 scientific literature. Therefore, the next section will focus on the quantitative variations of 595 droplet concentration and size according to aerosol properties. Moreover, glaciation and riming 596 indirect effect will be assessed. The sensitivity of cloud diameter and concentration according to aerosol haze will be assessed 601 from two parameters, called the Indirect Effect parameter (IE) and the Nucleation Efficiency 602 (NE) and defined as follows (Feingold et al., 2001  LWP was not measured and we assumed that the LWP is effectively constant. This is reasonable 619 since the sampled clouds were all low level mixed phase arctic clouds and from the same season.

20
The glaciation (Lohmann, 2002a(Lohmann, , 2002b and the riming indirect effect (Borys et al., 2003) were 648 evaluated during the CLIMSLIP campaign thanks to the CPI and the nephelometer 649 measurements. The comparison between the crystal concentration and IWC with σ (or nuclei 650 concentration) is displayed in Figure 14. showed an increase of the lateral scattering as Fliq decreases. 712 According to Guyot et al. (2015), only isoaxial measurements with a wind speed higher than 5 713 m/s are selected. This deleted a non-negligible amount of data and so limited the analysis, 714 especially for the precipitation cases where the particle statistics were the weaker. Moreover, 715 the position of the station within the cloud system was approximate despite the ceilometer 716 measurements. 717 A study by comparison of the effects of the anthropological aerosols transported to the Arctic 718 was performed. According to the FLEXPART/WRF simulations, the "polluted" case of March 719 29 th showed air masses from Europe and East Asia whereas the aerosol sources during the 720 "clean" case of March 11 th were closer (mainly from Scandinavia) and the anthropogenic 721 contribution doesn't exceed northern Europe. 722 Thus, the polluted case presents higher Black Carbon, aerosol and droplets concentrations, a 723 more important accumulation mode, smaller droplet sizes and higher activation fraction Fa. The 724 March 29 th activation diameter Da decreased when the droplet diameter increased and Fa 725 increased, proving that smaller aerosol particles are activated and droplets grow up when the 726 aerosol number decreases. These results confirm the first and second aerosol indirect effects 727 with the coefficients IE and NE respectively around 0.2 and 0.43. These values are very close 728 to those found by Garrett et al. (2004), which performed measurement at Barrow in Alaska, and 729 are so good candidates to be used to parameterize arctic aerosol-cloud interaction in climate 730 models. Furthermore, the crystal concentration and IWC do not show any correlation with the 731 aerosol properties, which indicates that the glaciation and riming indirect effects are not 732 highlighted during the CLIMSLIP-NyA campaign. 733 734 735 Acknoledgements. This work was supported by the ANR project CLIMSLIP and the conseil 736 general de l'Allier. We also thank the AVI for providing the ceilometer data and the ITM and 737 NILU for monitoring the Mount Zeppelin station. We are grateful to scientists, engineers and 738 technicians that make this study possible. Boris Quennehen acknowledges the IPSL 739 CICLAD/CLIMSERV mesocenter for providing computing resources.

Annex: Characterization of the Blowing Snow (BS) cases
During ground based measurements, some snow was collected that was suspended in the 748 atmosphere due to wind. This is the so-called Blowing Snow (BS). This annex aims at the 749 microphysics characterization in order to recognize this kind of episode and the optical 750 properties measurements, especially the phase function, that can be used as a reference to 751 develop new parameterizations of the snow simple scattering properties (Räisänen et al., 2015). 752 753 754 Figure 15: Same as Figure 5 for the BS cases. 755 756 When the BS occurs, the sky is clear as observed by the ceilometer. However, crystal particles 757 are sampled. They are snowflakes initially resting on the ground but getting suspended in the 758 air by the wind. 759 Figure 15 shows the average PSDs measured by the CPI for the two BS cases. The shape and 760 the amplitude are similar for the two PSDs, with a mean diameter between 150 and 200 µm. for 761 a maximum class concentration around 10 -2 L -1 µm -1 . The CPI shape classification, plotted in 762 Figure 16.a, shows a large prevalence of irregular crystals, as well in number, surface or mass 763 (i.e. volume), with a percentage around 90% of the crystals. These two characteristics constitute 764 the microphysics signature of the BS. The difference between the BS and MPC (see Even if the resuspension of crystals in the air can modify the shape by impacts, we consider 773 that the sampled crystals are similar to the deposited precipitations and aged for several days. 774 Thus, the BS events during CLIMSLIP were excellent occasions to measure the arctic snow 775 properties. 776 Figure 16.b displays the average phase functions of the BS cases. The shape of the curves are 777 very similar to the precipitation cases, typical of ice particles, but with lower g values. These 778 measurements constitute a unique database to develop parameterizations of the arctic snow 779 optical properties. Indeed, in most of the climate and weather forecast models, the computation 780 of the snow albedo uses the approximation of spherical snow grain (Wang and Zeng, 2010).