Impact of a Strong Biomass Burning Event on the Radiative Forcing in the Arctic

The aim of the presented study was to investigate the impact on the radiation budget of a biomass burning plume, transported from Alaska to a high Arctic region (Ny-Ålesund, Svalbard), in early July 2015. This large aerosol load event is considered exceptional in the last 25 years, with mean aerosol optical depth increased by the factor of 10, in comparison to the average summer background values. We utilised in situ data with hygroscopic growth equations, as well as remote sensing measurements as inputs to radiative transfer models, with an objective to estimate biases associated with (i) hygroscopicity, (ii) 5 variability of single-scattering albedo profiles, and (iii) plane-parallel closure of the modeled atmosphere. A chemical weather model with satellite-derived biomass burning emissions was used to interpret the transport and transformations pathways. Provided MODTRAN radiative transfer model simulations for the smoke event (14:00 July 9 11:30 July 11) resulted in the mean aerosol direct radiative forcing on the level of -78.9 Wm−2 and -47.0 Wm−2, at the surface and the top of the atmosphere respectively, for the mean value of aerosol optical depth equal to 0.64 at 550 nm. It corresponded to the average 10 clear-sky direct radiative forcing of -43.3 Wm−2, estimated by radiometer and model simulations at the surface. Ultimately, uncertainty connected with the plane-parallel atmosphere approximation altered results by about 2 Wm−2. Furthermore, model-derived aerosol direct radiative forcing efficiency reached on average -126 Wm/τ550 and -71 Wm/τ550 at the surface and at the top of the atmosphere. Estimated heating rate up to 1.8 Kday−1 inside the BB plume, implied vertical mixing with turbulent kinetic energy of 0.3 m2s−2. 15

We added a short note in the figure caption to emphasize the above information.As this chapter is one of the main results in the paper, we rephrased it to state our result more clearly.
The major aim of the paper is only radiative effect, but that of BB plume.As for BB plume, we can only know very limited information from Fig. 2 (vertical distribution of extinction coefficients).I know that your group (including yourself as co-author) has already several papers related to this same BB and Markowicz et al. (2016a) shows comprehensive feature of BB plume.Even duplicated, some information be helpful to be shown in this paper also (for example, just like Fig. 2, 3, 4 or 10 in Markowicz et al., 2016a).
Thank you for this comment.We added a new (brief) chapter about the overall characteristics of the event in terms of aerosol optical properties.It is based on the similar figure to Fig. 10 from Markowicz et al., 2016a, highlighting the temporal variability of the optical and microphysical properties of the aerosol.

Specific comments
Ny-Alesund should be written "Ny-Ålesund" Corrected as suggested.
Fig. 1: Though the figure occupies whole page, the information it shows seems to be not so interesting for the reader.Also, what is "white-sky albedo"?
The figure was moved to the appendix and the meaning of white-sky albedo was explained.This is the method (also referred to as bihemispherical albedo), that applies the integration of BRDF over all viewing directions using only the diffuse irradiance.
P14, L5, 15, P16, L6: Relations with clouds are explained in several parts; however, we have no information on clouds in any figures.It is difficult to follow.
In the revised manuscript, Fig. 1b presents the occurrence of clouds in the respective period of time.
P15, L32, 34: What is Fcin or Fcout?There are no such symbols in Fig. 3. Fig. 3: Explanation/ figure caption of Fig. 3 is limited.What is the large gaps in observed radiative fluxes (P16, L7 says radiometer data are removed -not easy to understand).There are no flux or RF of "within the atmosphere (subscript atm)" in the figure!There is no results by Fu-Liou.What is "Rad"?There is no explanation in the caption.
Fcin or Fcout refer to the respective (total incoming and outgoing) fluxes calculated under the reference simulation (no aerosol load).As previously stated, the figure caption was ill-copied.In the revised manuscript this error was corrected.Large gaps in the radiometer data are indicated by the occurance of the clouds, we added additional information in the figure caption.
We would like to know the data of τ itself.
The information of the temporal variability of optical and microphysical properties (together with with τ ) were added in the Fig. 1 along with the according explanatory section.P17, L17: RFE appears first, but no explanation here (only shown afterwards in P19, L16).
Corrected as suggested.P20, L6: I have never heard of "Ny-Ålesund valley".Normally it is said as Ny-Ålesund fjord.
The referee is right.We used the Norwegian name (Kongsfjorden) instead.
We apologize for the typo.Nevertheless, the explanation of ILES is included when the EULAG model is described (section 2.1).In the line under consideration, we added the according note in the brackets "(see section 2.1)" after the abbreviation.Fig. 7: Is the wavy pattern in (a) meaningful?It seems to be rather artificial due to small change of vertical gradient of (T).
The vertical variability of a heating rate at a reference simulation (fig.7a) used for EULAG calculations is vastly connected with the shape of specific humidity profile, and accordingly with the water vapor absorption bands.The heating rate at the so called polluted simulation (fig.7d) additionally is a function of both the single-scattering albedo and aerosol load in the layers (the extinction coefficient profile).
Conclusion: Items of conclusion seems to be different from results and discussions.To indicate these conclusions, you need to add more discussions to connect to these conclusions.
We kindly disagree with the referee.The sentences used in the conclusions are almost a literal copy of the statements from the results.We may speculate that the fist version of the manuscript was a bit chaotic and therefore the main outcome from specific sections may have been missed.We hope, that in the revised manuscript, this issue doesn't appear any more.
P25, L 8: What is "the first" and "the latter".
Thank you, we rephrased the sentence.

P25. L19: What is "ILES"?
The explanation of ILES (Implicit Large Eddy Simulations) is included when the EULAG model is described (section 2.1).We added the reference note in the section 3.7 when it is mentioned for the first time in the results.
P25, L24: Impact on the atmospheric dynamics is not clearly described in the manuscript.-P25, L26-27: The meaning of the sentence "Thus, it is expected ... " is not clear.
Both sentences revealed shortcomings in English, we apologize for that.In the revised manuscript they were rephrased as following: In this study we have shown that long-range transport of wildfire aerosols from Alaska to European Arctic, certainly has a significant impact on radiative properties.Furthermore, our results also indicate an impact on atmospheric dynamics.We believe that the detailed studies on this topic are needed, especially considering a significant positive trend in mid-latitudes fire frequency during the summer season in the last 25 years; and therefore possibly more frequent advection over the Arctic region (Young et al., 2016) References: Descriptions are not complete in some, for example, Markowicz et al., 2002, or -2017b, Stone et al., 2008, Wang et al., 2006 Thank you, we improved this section.

The Anonymous Referee #2
We wanted to thank the reviewer for raising issues that limit the understanding of the paper as it helped us to improve the paper.We hope that the reviewer will be satisfied with the changes made to the new version of the paper.
Major comment [...] I found it difficult to fully assess the quality of the paper due to English language and grammar issues, which should be addressed before publication.Due to the number of such errors I am not able to point them all out in this review.[...] The paper went through a major reorganization regarding English shortcomings along with English-proof reading.

Specific comments
Along with Myhre et al. (2013), the manuscript could include a citation to Sand et al., (2017), who investigated specifically the radiative forcing of aerosols in the Arctic in the AeroCom phase II models.
Indeed, both papers were significant for the section, thank you.
True, we meant 'data coverage', thank you.P.4, ll. 2 -10: I think authors should clearly indicate here their own new/original contributions in this paper , and what work (e.g.simulations) was already performed for previous studies such as Markowicz et al. (2017b).
Corrected as requested.In the revised paper the section is written as follows: Previously presented by scientific papers, and characterized in this research, was the study of smoke transport over the Arctic during July 2015.Markowicz et al., 2016a reported the temporal and spatial variability of aerosol single-scattering properties measured by in situ and ground-based remote sensing instruments over Svalbard and in Andenes, Norway.Moroni et al., 2017, discussed morphochemical characteristics and mixing state of smoke particles in Ny-Ålesund as indicated by DEKATI 12-stage low volume impactor, combined with scanning electron microscopy.Markowicz et al., 2017b on the other hand, presented a comprehensive description of smoke radiative and optical properties on a regional scale.The paper examined ageing processes of the smoke plume under study, while transported from the source region across the High Arctic.Simple Fu-Liou radiative transfer model, combined with NAAPS aerosol transport model, were used to determine the spatial distribution of aerosol single-scattering properties and RF s for the period of 5-15 July 2015, in the area to the north of 55 o N, where the transport of BB aerosol was observed.
In this paper, we utilise MODTRAN radiative transfer simulations and aerosol optical properties obtained from in situ and ground-based remote sensing instruments, to retrieve clear-sky direct RF over the area close to Ny-Ålesund.The research aims to estimate the biases connected with (i) hygroscopicity, (ii) variability of ω profiles, and (iii) plane-parallel closure of the modeled atmosphere.The main outcome of this research is the implementation of new methodology to retrieve the profile of ω at ambient conditions, utilising in situ measurements and lidar profiles (section 3.2).Simulated RF s were compared to simple radiative transfer model (section 3.5).Section 3.6 shows an example of RF distribution at the surface, in the vicinity of Kongsfjorden.The last part presents the influence of unstably stratified biomass burning air masses on the turbulence development, which is shown in section 3.7.Additionally, we confirmed the source region of the BB plume.A chemical weather model with satellite-derived biomass burning emissions was used to interpret the transport and transformations pathways.P7., section 2.2 : If I understand correctly, the in -situ measurements (e.g.SMPS, PSAP), are performed at the surface.Can you give reasons why these values are representative of the whole column , since the plumes extends at relatively high altitudes, and the Arctic surface and free troposphere are often decoupled.
In the revised manuscript, we added an explanatory section concerning our assumptions to ω and g retrieval, quoted below: Vertical profiles of single-scattering properties at ambient conditions are used as input parameters to MODTRAN and Monte Carlo calculations.The retrieval is based on the in situ singlescattering properties, measured at the surface in dry conditions (denoted later on as superscript 'd'), and on vertical profiles of σ a ext , as well as RH at ambient conditions (hereinafter superscript 'a') from KARL lidar and radio-sounding data.
In the reference to temporal variability of range-corrected signal, measured at 532 nm by Micropulse Lidar, Markowicz et al, 2016a, characterize smoke plume as a rather well-mixed layer of BB aerosol extending from around 4 -6 km on 9 th to 0 -3.5 km later on.Both contributions of BB-like aerosol in the NAAPS AOD, estimated on the level as high as 80%, and the similarity between columnar and in situ aerosol extensive properties such as α (Markowicz et al, 2016a), suggest that smoke plume may have crossed PBL and mixed with the lowermost part of the troposphere.Additionally, very little aerosol load existing above smoke plume plays a minor role in affecting the radiative properties of the atmosphere and therefore may be neglected.This is why, in the presented methodology, we assume no changes in chemical composition vertically, so that most of the possible vertical variability of ω a at ambient conditions, is attributed to changes in RH.Therefore, we approximate initial profiles of ω d and R d ef f by setting them up to the values of in situ measurements and consider them constant with altitude.By introducing hygroscopic growth model for particles with known size distribution, one may obtain ω a profile as well as g a .P.7, l. 12: The a and d superscripts should be explained there, when they are first introduced, and not on page 8.

Corrected as suggested.
P. 9, equations 7 and 8: The text mentions RFnet and RFrel, but the equations give Fnet and frel.
Thank you, this was our mistake while copy-pasting to latex.P.9, l.15 : If this product is from MODIS, this should be indicated.Indeed, thank you.P.9 l. 22: The "BRDF" acronym should be explained here.
Thank you for your helpful comment, we referred also to Lund Myhre et al. (2007) in the section under consideration.
P. 12 l.20: Is PM10 really reported in ppb, not µgm −3 ?Thank you, the text was corrected to "the mass mixing ratio".P. 15 l. 10 " and additional no change in the irradiances from the reference simulation" it is not clear what you mean by this sentence.
Indeed, we rephrased the sentence.P. 15, l.16: what do you mean here by a "real" value of albedo?Indeed, we rephrased the sentence.
Figure 3.There are several issues with this figure.First, the caption does not seem to match the contents, as the "Rad" quantities, which seem to be observations, are not explained in the caption.The caption mentions Fu -Liou results that are apparently not shown.The quantities do not seem to be daily mean values.In addition, RF quantities in panel b should use different colors/symbols than the F results in panel a , as the current choices is very confusing.
Thank you, we didn't notice that the caption was ill-copied.In the revised manuscript, the caption matches the figure.
We changed the colors/symbols in the b panel for the clarity.
Figure 3: What are the reasons for the differences between F and ModF results at the end of the period, after 12h on 11 July ?
This difference is a result of low cloud appearance at around noon 11 th July, as explained in the section 3.1.In the revised version of the manuscript we removed all cloud-contaminated data from this figure, also the F in after 11:30 July 11. . 15, 16: This section should include more paragraphs breaks to better separate the different ideas.

Pp
The paragraph breaks were added.P. 16, l. 6: How would increase turbulence lead to higher variability in F in ?
We apologize for this linguistic shortcoming.The higher variability of F in on 10 th is a direct effect of the appearance of cumulus clouds.They, in turn, result from: (1) the aerosol activation based on the most common mechanism of cloud formation and (2) the instability of the atmospheric dynamics, as this is the reason why cumulus clouds are formed rather than other clouds.
After rephrasing, this sentence should be as follows: We may expect that higher variability of Rad F in , visible by comparison to the 9 th July, together with an appearance of clouds inside the smoke plume, are likely to result from both a possible BB aerosol activation and increased turbulence.Further to this, a number of high-and mid-level cumulus clouds are reported around noon and in the afternoon (Markowicz et al., 2016).P. 17, l. 17: Explain the meaning of "RFE" when it is first introduced.For what reason is RFE a more accurate quantity for intercomparisons?
Corrected as suggested.RFE is a more accurate quantity for inter-comparison only when intrinsic properties of the plume are taken under consideration as it was stated in the further part of the sentence.However, in the revised version of the manuscript this sentence, after rephrasing of the paragraph was omitted.P. 17, l. 31: It is not clear here for someone unfamiliar with these codes that DISORT is included with in MODTRAN and not a standalone radiative transfer model.Consider rephrasing this sentence.

Corrected as suggested.
P. 17, l. 31 and elsewhere: Can you explain what you mean by "robust" when referring to Fu -Liou?Do you mean more detailed?
We apologize for this ill-translation.We meant 'fast' and 'less-complicated' in terms of solvers of the radiative transfer equations.It was improved in the revised manuscript.
Pp. 18 -19: This section should include more paragraphs breaks to better separate the different ideas.
Corrected as suggested.
P. 18, ll. 13 -18: I do not think it is needed here to remind the meaning of the different colors on Figure 4, since they are already explained on the Figure.
We agree, thank you for this suggestion.P. 19, l. 4 and elsewhere: The correct reference is Lund Myhre et al. (2007), not Myhre et al., since "Lund Myhre" is the last name of the first author.
Corrected as suggested.
P. 19, l. 12 -15 : This section would be clearer if the analysis of Figure 4 started with this remark , since the most obvious result from Figure 4 is that there is a very good agreement for RF between MODTRAN and Fu -Liou.
We agree with the reviewer and changed the text accordingly.The main reason for the modeled discrepancies in RF E are (1) the differences in inputs to models, in particular the assumed aerosol optical properties and secondarily PW as well as (2) the distinction between solvers of the radiative transfer equations used in both models, that may give different results even though the exact inputs are assumed.The latter issue is more widely described in the following paper: Myhre, G. et al,2009:Intercomparison of radiative forcing calculations of stratospheric water vapour and contrails, METEOROL Z, 18(6), pp585-596.
Note, this part of the section was moved to the appendix B. This was requested by the Referee 1 being concerned that the inter-comparison between RTM models was not the main subject of the manuscript and additionally unnecessarily lengthened the paper.P. 20, l. 7: " In the previous sections, we discussed the RF computed for a single cell " maybe this should also be mentioned explicitly in the beginning of the previous sections, e.g. at the beginning of 3.2.
We decided to add this information in the description of models.P. 20, l l .13 -14: Why not show RF directly, instead of this relative value?This should maybe be explained when the equations are discussed.
In the revised manuscript, we added the following information to the 2.3.4 section with 3D Monte Carlo equations: The results from 3D Monte Carlo model, as mentioned earlier, are used to characterise spatial variability of RF and therefore to diagnose possible uncertainties resulting from using single-column radiative transfer models, represented by MODTRAN and Fu-Liou codes.Taking into account the above goals, we resigned from performing time-consuming simulations of daily mean broadband RF s for the model domain; and instead we relied on the relative value of RF calculated for 1 λ, with respect to its value at TOA at a given zenith angle.Such an approach allowed for defining higher spatial resolution.
Figure 6: There are also several issues with this figure.First, the colorbar should include a label.Since values go from negative to positive, it would be a lot clearer to use a divergence colormap where 0 is indicated by a special color, for example white.It is also unclear to a reader unfamiliar with the "ICA" terminology what is the exact difference between panels a and b.I understand that the point is to study the effect of e.g.topog raphy on the RF calculations, but consider writing a more explicit caption, and consider including in the text an explanation of the difference between these two calculations and the aim of this 2 -panel comparison.
The label to the colorbar was added.Regarding the divergence colormap, we kindly disagree with the referee, as this would limit the number of colors used for the negative RFs.As the area of a positive RF is very small, we feel that this change is not of a great importance.Instead we added a black line to the colorbar highlighting 0 value.Sand et al., 2017) often showed a positive RF of BB aerosols over snow and ice.Is this due to the high single -scattering albedo here ?To a relatively low surface albedo c ompared to typical snow and ice -co vered surfaces in the Arctic ?This averages refer to the BB event, in particular 14:00 July 9 th -11:30 July 11 th .We changed the sentence accordingly.P. 25, l .7: Are you really comparing modelled RF to observations in this study?
We apologize for this shortcoming in English.We have changed the sentence to match the actual meaning.Nevertheless, note that we also added a comparison of modeled and measured Fs.

Impact of a Strong Biomass Burning Event on the Radiative Forcing in the Arctic
transported from Alaska to : a : high Arctic region (Ny-Ålesund , Svalbard), in early July 2015.This large aerosol load event is considered exceptional in the last 25 years, with mean aerosol optical depth increased by the factor of 10, in comparison to the average summer background values.We utilised in situ data with hygroscopic growth equations, as well as remote sensing measurements as inputs to radiative transfer models, with an objective to estimate biases associated with (i) hygroscopicity, (ii) variability of ω :::::::::::::: single-scattering ::::: albedo ::::::: profiles, : and (iii) plane-parallel closure of the modeled atmosphere.A chemical weather model with satellite-derived biomass burning emissions was used to interpret the transport and transformations pathways.

Introduction
Wildfires are considered significant sources of carbon in the atmosphere.It is estimated that up to 2.0 Pg of carbon aerosol is released into the atmosphere each year (Van der Werf et al., 2010).In the past 100 years, an intensification of fires in mid-latitudes is observed appreciably affecting radiative and optical properties of the atmosphere (Mtetwa and McCormick, 2003).The emitted particles from biomass burning (BB) sources mainly consist of organic and black carbon (IPCC, 2001) of which 90% are built of the fine mode regarding aerosol size distribution (Dubovik et al., 2002).The impact of the plume on the atmospheric instability conditions and its rather small particle radius might result in a rapid transport on an intercontinental scale, within just several days (Nikonovas et al., 2015).Thus, it is likely that the biomass burning aerosol considerably affects the optical and radiative properties of the atmosphere in the substantial part of the globe.The influence of BB aerosol is manifested by heating the air layer where transport takes place.Regarding the columnar properties however, it implies weak cooling effects at the top of the atmosphere (TOA) due to predominant scattering properties of the plume (Hansen et al., 2004).
The magnitude of its impact is nevertheless strongly dependent on the chemical composition that results from the adversative radiative response of the atmosphere exposed to black and organic carbon, being negative for the latter (Myhre et al., 2013a).

Methodology
This chapter consists of a few subsections dedicated to a brief description of all data and models used in this research.In paragraph 2.1 we will focus on characterization of all models used to track the transport of smoke, as well as to calculate the impact of the BB plume :: on ::::::: radiative :::: and :::::::: dynamical ::::::::: properties :: of ::: the :::::::::: atmosphere. 3
The Fu-Liou v. 200503 (Fu andLiou, 1992, 1993) :::: RTM : uses the δ 2/4 stream solver, applied for 6 shortwave and 12 longwave spectral bands.The optical properties of the atmosphere are calculated by the correlated-k distribution method, defined for each spectral band (Fu and Liou, 1992).The optical properties of aerosols, as well as thermodynamical properties of the atmosphere, were based on the results provided by the NAAPS re-analysis (Lynch et al., 2016).Fu-Liou was used to determine the spatial distribution of RF for the period 5 -15 July 2015 in the area to the north of 55N, where the transport of BB aerosol was observed.The results of these simulations are presented in Markowicz et al. (2017b).In this paper, we compare the results of our retrieval from MODTRAN simulations and robust Fu-Liou code over the area close to Ny-Alesund.
The :::::: NAAPS : (Navy Aerosol Analysis and Prediction System(NAAPS) is the semi-lagrangian aerosol transport model run on a 1x 1grid and 6 temporal resolution and with 28 vertical layers (Hogan and Rosmond, 1991).NAAPS model utilises 5 basic processes including emission from the source, mixing and diffusion within the PBL, dispersion and advection due to the wind as well as wet and dry deposition (Lynch et al., 2016).Recently, a new version of NAAPS model was implemented, which assimilates quality controlled τ values retrieved from MODIS and MISR products (Hyer et al., 2011;Shi et al., 2014).
3D effects of the RF were calculated using 3D forward Monte Carlo code (Marshak et al., 1995), which utilises : a : maximum cross-section method to compute photon paths in the three-dimensional model of the atmosphere (Marchuk et al., 2013).A number of modifications were made to the original setup of the code, including such phenomena as absorption of photons by atmospheric gases , reflection and absorption at the undulated Earth's surface (Rozwadowska andGórecka, 2012, 2017).The proposed by Ricchiazzi and Gautier (1998).
Implicit large-eddy simulations (ILESs) were performed using the 3D nonhydrostatic anelastic Eulerian-semi-Lagrangian (EULAG) model (Prusa et al., 2008), to estimate the dynamical response of the atmosphere induced by the BB plume.The EULAG model is set up to solve for the three velocity components u, v, and w in the x-, y-, and z-directions (i.e.W-E, S-N, and vertical directions), as well as the potential temperature (θ).The governing equations are solved in an Eulerian framework without explicit subgrid-scale (SGS) terms included.The nonoscillatory forward-in-time integration is performed with the Multidimensional Positive definite Advection Transport Algorithm (MPDATA; Smolarkiewicz, 2006).We rely on the ability of the MPDATA to implicitly account for the effect of unresolved turbulence on the resolved flow, through the truncation terms associated with the algorithm.For more details on ILES, see e.g.Grinstein et al. (2007).The horizontal grid spacing was set to 200 m and the vertical grid spacing to 50 m.The size of the computational domain was set to 19 km in the horizontal directions and 20 km in the vertical direction.The uppermost 5 km is a sponge layer included to prevent reflection of gravity waves at the top of the domain.The upper boundary of the domain is impermeable with a free slip condition, and the lower boundary is impermeable with partial slip conditions, characterized by a specified drag coefficient of 0.001.The flow is periodic across the lateral boundaries of the domain.The EULAG simulations were calculated for 12:00 UTC, 10 th July 2015, in Ny-Ålesund, using results obtained from the radiative transfer model and radio-sounding data.
For a more detailed specification, one is encouraged to read a section concerning instrumentation in Markowicz et al. (2016a).
Extinction profiles were retrieved from KARL Raman lidar.The instrument uses Nd:Yag laser pulse at 355, 532, 1064 nm with the power of 10 W at each wavelength, to obtain backscatter and extinction coefficients.Also, the depolarization is measured at water vapour channels (407, 660 nm).The detection is carried out by 70 cm mirror with a 1.75 mrad field of view, :::: and the overlap issue is fulfilled at 700 m a.g.l.Further details :::: may be found in Hoffmann (2011) and Ritter et al. (2016).
Continuous measurements of radiation fluxes are provided in Ny-Ålesund under the Baseline Surface Radiation Network (BSRN).Ball-shaded CMP22 by Kipp&Zonen installed on solar tracker by Schulz & Partner, measures total incoming and reflected solar radiation at 200 -3600 nm (Maturilli et al., 2015).and Osborne (2000).
The size distributions of aerosols at ambient conditions were estimated by introducing the hygroscopic growth factor χ(RH), related to growth of particles due to water uptake, yielding: where D is the diameter of the particle at the : a : certain RH (Zieger et al., 2010).
The short-wave direct aerosol radiative forcing (spectral relative radiative forcing) , RF rel (λ), is expressed as: where superscript 'aer' stands for a clear-sky conditions with an aerosol included and superscript '0' for a clear-sky without one.We can also define RF with respect to the cell surface S c instead of the actual surface within a given column S s : 3 Results
Model calculations usually overestimate Mod RF surf values, which on average deviate from Rad RF surf by around 32.9 %, possibly related to all-sky conditions which increase diffusive flux.This result should be additionally interpreted with caution since RF surf from radiometers might be cloud contaminated and cause increased variability of the obtained quantity.
High single-scattering albedo values and negative RF toa clearly show that scattering is dominant with respect to the light absorption contribution.Indeed, absorption species (mainly BC) are able to mitigate the cooling effect of the BB event into :: in the atmosphere, but not sufficient to change the RF sign.This means that BC particles play a minor role with respect to scattering particles (sulfate, OC :::::: organic :::::: carbon ::::: (OC), etc.).That could be demonstrated also ::: also ::: be ::::::::::: demonstrated : by the changes in atmospheric concentrations of BC, OC, sulfate and oxalate ::: and :::::: sulfate : measured at Gruvebadet.In particular, the relative concentrations increase of about 20 times for BC (and EC), OCand oxalate ::: and ::: OC, and about 10 times for non-sea-salt sulfate during the BB event, with respect to the background level.In spite of the BC and OC, relative increases are similar, : ; the absolute concentrations of OC are significantly higher than BC (and EC), reaching values as high as 4500 i.e. more than 10 times higher than atmospheric concentration of BC (around 300 ngm −3 ; Moroni et al., 2017).
For the fjord surface, an absolute value of RF is smaller and weakly variable over the fjord surface, mean RF cell rel is equal to mean RF rel and reaches -0.2632 ± 0.0092.Its coefficient of variation is 3.5%.The actual value of RF variability over the sea may be even lower, because the noise of Monte Carlo method may enhance it.The land ::: RF :: is ::::::::::: characterized : with both RF cell rel and RF rel less negative mean values of -0.1395 ± 0.1180, as well as -0.1326 ± 0.1084 respectively, and much stronger surface variability.The respective coefficients of variation are 84.6% and 81.7%.In our simulation, the variability of RF over the sea downward by the atmosphere.The net horizontal transport is observed for both atmospheres, with and without aerosols, but in each case the effective height of reflection is different.An appearance of dense, low-lying aerosol layer reduces the effective reflectance of the atmosphere, compared to the case without aerosols.Thus the gradient in irradiance, with distance from the reflective land is stronger in the aerosol case.The atmosphere without aerosols acts similarly to a very thin cloud located higher over the Earth's surface, while aerosol layer can be compared to a thicker cloud with its base at a lower height (Rozwadowska and Górecka, 2012).
The main factors influencing RF and its variability over land in the vicinity of :::::::::: Ny-Ålesund ::::::: comprise :: of : reflective properties of the land surface, slope exposition concerning the sun, and shading of the sun by the mountains.The impact of photons reflected from nearby sunlit slopes and horizontal photon transport due to multiple reflections between the sky and the surface on RF variability, are of secondary importance over the land.
In the analysed case, the highest magnitude of negative RF was found for sun-facing slopes of white sky albedo of around 0.2.In such places, the effective solar zenith angle is relatively low and a high contribution of the direct solar radiation to the total irradiance results in a substantial reduction in the surface irradiance due to the presence of aerosols, ::::: hence, ::: an : RF rel of about -0.39.:: In ::: the ::::: main, ::: for slopes lit by diffused radiation, the RF is positive, i.e. presence of aerosols increases the amount of radiation absorbed by the surface.In shaded places with the effective solar zenith angle of :::::::::::: approximately 90 o and white sky albedo of around 0.4, RF rel can be as high as 0.07 in our simulation.
Table 3. Mean relative radiative forcing RF calculated concerning the actual surface, RF rel , and the horizontal cell surface RF cell rel using the 3D Monte Carlo model.RF pp rel is RF computed, using the Independent Column Approximation approach.Computations were done for λ=469 nm, solar zenith angle=57 o , solar azimuth=173 o , and aerosol properties of the 191 st day of 2015.

All cells
Water Land Using the ICA approach to RF estimation results in an underestimation of the surface variability in the RF, also results in biased domain mean values of the RF .In the case under study, the mean difference between the more accurate RF for the horizontal cell surface and the RF calculated using the plane-parallel approach, RF cell rel and RF pp rel are -0.0032± 0.0699, which is 1.9 % of the mean RF cell rel .This, in conversion to daily mean shortwave RF, gives the average error not exceeding 2 Wm −2 while using the plane-parallel approach.Thus, it is almost as high as the effect of ω d translation to ambient conditions considered in our study.Additionally, the mean bias is higher for the sea than for the land.However, for individual cells/columns, the variability of deviations from the real value of RF cell rel is much larger for the land, where the standard deviation of the difference RF cell rel -RF pp rel equals 63.8% of the mean RF cell rel .The negative bias with the largest magnitude, of 0.247 was found for the case of sun-facing slopes discussed above.For shaded inclined areas, the plane-parallel approach seriously underestimates radiative forcing ::::: where the mean bias equals 0.233.applied heating rates ::::: given, ::: by: where ::::::: whereby ρ is air density and C is a specific heat capacity defined both for short-and longwave, irradiances are obtained from MODTRAN simulations for 10 th July 11:30 UTC.The r h profiles for the reference case (Fig. 7a) and the aerosol polluted case (Fig. 7d) both show a thin layer near the surface (z<0.5 km) : with significant heating; : : 2.7 and 3.4 Kday −1 respectively.
in the lower part of this layer is around 0.2 Kday −1 , while in the upper part it reaches values of up to 1.8 Kday −1 .The two simulations have the same initial profile of θ which is represented by the navy blue lines in Figure 7b,e.There is a layer between altitudes of 2 and 3 km with nearly constant initial θ, but in general, it decreases with altitude.Due to the stable initial stratification and the lack of, e.g.strong surface heating, turbulence develops slowly in the performed simulations (see TKE profiles in Fig. 7c,f).After 16 h, a turbulent layer starts to develop near the surface in both simulations.The TKE in ::: this ::::: layer reaches values of around 0.1 m 2 s −2 ::: and :: it ::::::: extends : up to 0.5 km at the time t=48 h.After 24 h, a second turbulent layer starts to develop in the polluted case, at an altitude of approximately 3.4 km.The thickness of this layer increases with time, and at t=48 h, it covers altitudes between 2.5 and 4.2 km with maximum TKE values of 0.3 m 2 s −2 and updrafts/downdrafts with vertical velocities of around 1 ms −1 .By contrast, the flow in the reference case remains almost non-turbulent above 0.5 km, with vertical velocities close to zero throughout the simulation period.In the regions with relatively high TKE, θ becomes nearly constant with altitude, and the polluted simulation indicates that the initially well-mixed layer around z=2.5 km expands and moves upwards over time.
Outside the clearly turbulent regions very little vertical mixing takes place, and the potential temperature is approximately given by: where 'z' symbolises altitude and 't' time.

Figure 5 :
Figure 5: What are the reasons for the strong differences in RFE between MODTRAN and Fu -Liou for 9 July ?

Figure 6 :
Figure 6: Results seem to show a negative RF over high -albedo surfaces.Other studies (e.g.Sand et al., 2017) often showed a positive RF of BB aerosols over snow and ice.Is this due to the high single -scattering albedo here ?To a relatively low surface albedo c ompared to typical snow and ice -co vered surfaces in the Arctic ?

Fig. 6
Fig.6in the manuscript and the work bySand et al. (2017) present radiative forcing at different levels.The figure shows aerosol radiative forcing at the surface whileSand et al (2017) at the top of the atmosphere.Aerosol radiative forcing at the surface is typically negative.

Table 1 .
Description of the instruments installed in Ny-Ålesund used as input data for atmospheric radiative transfer model.
th July to -142.4 /τ 550 on 10 th July according to Fu-Liou model.The RF E atm is between 39.6 and 71.2 /τ 550 on 9 th July regarding MODTRAN simulations, while Fu-Liou results show its values being on average higher ranging from 42.5 -66.3 /τ 550 .RF E toa varies from -71.0 /τ 550 to -86.6 /τ 550 for MODTRAN calculations as well as from -61.5 /τ 550 to -76.3 /τ 550 for Fu-Liou model.The mean daily values of radiative forcing efficiency (RFE) calculated by means of