Do contemporary (1980–2015) emissions determine the elemental carbon deposition trend at Holtedahlfonna glacier, Svalbard?

The climate impact of black carbon (BC) is notably amplified in the Arctic by its deposition that causes albedo decrease and subsequent earlier snow and ice spring melt. To comprehensively assess the climate impact of BC in the Arctic, information on both atmospheric BC concentrations and deposition are essential. Currently, Arctic BC deposition data are very scarce, while atmospheric BC concentrations have been shown to generally decrease since the 1990s. However, a 20 300-year Svalbard ice core showed a distinct increase in EC (elemental carbon, proxy for BC) deposition from 1970 to 2004 contradicting atmospheric measurements and modelling studies. Here, our objective was to decipher whether this increase has continued in the 21 st century, and to investigate the drivers of the observed EC deposition trends. For this, a shallow firn core was collected from the same Svalbard glacier, and a regional-to-meso-scale chemical transport model (SILAM) was run from 1980 to 2015. The ice and firn core data indicate peaking EC deposition values at the end of the 25 1990s, and lower values thereafter. The modelled BC deposition results generally support the observed glacier EC variations. However, the ice and firn core results clearly deviate from both measured and modelled atmospheric BC concentration trends, and the modelled BC deposition trend shows variations seemingly independent from BC emission or atmospheric BC concentration trends. Furthermore, according to the model ca. 99 % BC mass is wet-deposited at this Svalbard glacier, indicating that meteorological processes such as precipitation and scavenging efficiency have most 30 likely a stronger influence on the BC deposition trend than BC emission or atmospheric concentration trends. BC emission source sectors contribute differently to the modelled atmospheric BC concentrations and BC deposition, which further supports our conclusion that different processes affect atmospheric BC concentration and deposition trends. Consequently, Arctic BC deposition trends should not directly be inferred based on atmospheric BC measurements, and more observational BC deposition data are required to assess the climate impact of BC in Arctic snow. 35


Snow pit EC data
The EC variations in the snow pit covering the end of summer 2014 to April 2015 are shown in Figure 4a. The EC concentrations ranged between 4.7 and 20.3 µg L -1 which are generally similar to EC concentrations of 1. 4, 9.4 and 11.6 μg L −1 previously measured at the same site in spring surface snow of 2007and 2009, respectively (Forsström et 5 al., 2009Fig. 4b). The snow pit samples show a similar seasonal trend in EC concentrations as previously observed in Arctic snow packs with elevated concentrations during spring and summer and lower values in the autumn and winter (Doherty et al., 2010(Doherty et al., , 2013.

Firn core EC data
The EC concentrations of the shallow firn core are between 3.5 and 24.6 µg L -1 with an average of 10.4 µg L -1 (Fig. 4b). 10 The firn core EC concentrations match the snow pit EC observations for the overlapping part from 80 to 253 cm from the snow surface (Fig. 4b). The annual deposition of EC to the firn core was calculated using the dating (section 2.3) of the core. The EC deposition values in the firn core range from 2.8 to 19 mg m -2 yr -1 (on average 10 mg m -2 yr -1 ). Table 1 presents annual (averaged over calendar years) EC concentrations and deposition for 2006 to 2014.
Due to the comparably low temporal resolution of the EC samples no annual variation can be detected in the firn core, 15 although the observed EC variation may be partly caused by some samples covering more of the high BC laden spring to summer snow (e.g. two vs. zero spring layers) compared to cleaner winter snow (cf. Ruppel et al., 2014). The firn core is too short to indicate any clear temporal trend, but in general, the EC concentrations and deposition seem to be on a lower level from 2005 to 2011 and to increase to higher levels from 2012 to 2015 ( Fig. 4c and d, Table 1). The temporal trend of EC deposition is similar to the EC concentration trend observed in the core ( Fig. 4c and d).

Modelled BC data
The results of modelled atmospheric BC concentrations and BC deposition at Holtedahlfonna are presented in Figure 5.
The modelled annual atmospheric BC concentrations decrease quite constantly from 1990 onwards after notably higher values modelled for the 1980s. The modelled BC deposition on the other hand shows significant variation from year to year with no clear decadal trend. The modelled atmospheric BC concentration and deposition trend correlate only weakly 25 (r = 0.29, p = 0.08) over the study period. The model results suggest that 98.7 % of BC is wet-deposited at Holtedahlfonna There are notable differences in the source contributions for the modelled BC deposition and atmospheric BC concentrations at Holtedahlfonna (Fig. 6). Over the period of 1980 to 2015 transport and domestic emissions are the most important sources for BC deposited at Holtedahlfonna (Fig. 6a), both with ca. 30 % contribution, while the domestic 30 sector (43 % on average) is the most important emission source for atmospheric BC concentrations at the glacier, followed by the industry and transport sectors (Fig. 6b). For both the modelled atmospheric BC concentrations and deposition the contribution of domestic emissions has decreased during the investigated time period while the contribution of transport, including shipping, and natural fires has increased, and the contribution of industry and other sectors has stayed quite constant.

Comparison of the snow and firn core EC data with the 2005 ice core
Previous EC concentrations from surface snow in 2007and 2009(Forsström et al., 2013 and the snow pit and firn core data collected from the Holtedahlfonna 2015 coring site (Stake 10), corroborate each other (Fig. 4b). However, the firn core EC concentrations measured at Stake 10 (an average of 10.4 µg L -1 ) are notably lower than recorded in the 300-

5
year ice core collected from a different site on the same glacier in 2005 (on average 35.8 µg L -1 ) (Ruppel et al., 2014). On the other hand, the overall annual EC deposition in the firn core (on average 10 mg m -2 yr -1 ) compares quite well to the EC deposition recorded in the 300-year ice core (on average 11.2 mg m -2 yr -1 ). Yet, there is a notable drop of a factor of 2.5 in the EC deposition values from the last data point in the 300-year ice core (of 23.7 mg m -2 yr -1 deposited in the sample covering ca. 2001 to 2003), to the first sample in the firn core (9.3 mg m -2 yr -1 deposited ca. between mid-2005 to 10 early-2006) (Fig. 7). Regrettably, the new firn and old ice core do not temporally overlap, and therefore it cannot be confirmed whether the measurements at the separate coring locations are directly comparable. In the following we discuss the hypothesis of an actual rapid drop in EC deposition having occurred between the end of 2003 and mid-2005 at Holtedahlfonna, as suggested by the current data. Secondly, we explore the hypothesis that this difference in EC deposition is caused by local post-depositional factors at the coring sites, impeding the comparison of the cores. In 15 addition, the sources for the deposited EC are examined.

Post-depositional processes affecting EC deposition at the two Holtedahlfonna coring sites
EC concentrations in snow and ice are strongly affected by numerous additional factors to atmospheric BC concentrations, such as EC scavenging efficiencies, precipitation amounts, and post-depositional processes of wind drift, sublimation and melt, that may dilute or concentrate EC in the snow (e.g. Doherty et al., 2010Doherty et al., , 2013. Snow accumulation 20 rates and post-depositional factors may vary locally, potentially causing differences in EC concentrations between the two Holtedahlfonna coring sites located 2.8 km apart (Fig. 1). Previous results on EC or BC concentrations in surface snow and full vertical snow profiles have shown that EC concentrations and column loads can vary substantially (commonly of a factor of two, but even a factor of 16 has been reported) even on a meter to meter scale due to post-depositional processes (e.g. Doherty et al., 2010Doherty et al., , 2016Svensson et al., 2013;Forsström et al., 2013;Delaney et al., 2015). scavenging is at a specific time at Holtedahlfonna, and thereby how much atmospheric and in-cloud EC present at Holtedahfonna is actually deposited. These processes may vary temporally with notable effects on the observed EC deposition trend at Holtedahlfonna. As according to our model results almost 99 % of BC is wet-deposited at Holtedahlfonna, the significance of meteorological processes and their variation in comparison to sole BC emissions for the observed EC deposition trend have to be considered. Moreover, it would be a gross oversimplification to assume that 25 the EC deposition trend at Holtedahlfonna would solely reflect BC emission trends in source areas and/or atmospheric BC concentration trends, since local and regional meteorological processes affect the EC deposition rate notably. As a possible example of the consequences of disregarding temporal meteorological variation, previous modelling results of historical BC deposition in Finland using constant (year 1997) meteorology since 1850 show that the modelled BC deposition trend closely follows the inventory BC emission trend, while the observed BC deposition trend clearly 30 diverged from the modelled trend (Ruppel et al., 2015). One possible explanation for the described discrepancy are variations in meteorological processes affecting BC scavenging efficiencies that were unaccounted for in the model. Thus, to produce generally more plausible modelled data, atmospheric BC deposition at Holtedahlfonna was here modelled only beginning from 1980 since when reliable meteorological data has been available.
Atmospheric BC concentration trends, on the other hand, have been generally observed to follow BC emission trends in 35 the Arctic (e.g. Sharma et al., 2013). In our results the modelled atmospheric BC concentration decreases between 1980 and 2015 ( Fig. 5), as has also been observed between 1990 and 2009 at the three long-term Arctic BC monitoring stations in Alert, Barrow and Ny-Ålesund (Sharma et al., 2013), and in a 47-year weekly measurement record from northern Notably, however, the modelled annual BC deposition does not clearly follow (or correlate to) the declining BC emission ( Fig. 3) or modelled and measured atmospheric BC concentration trend (Fig. 5). Instead, the modelled BC deposition shows significant variation from year to year. The modelled BC deposition trend follows the wet deposition pattern at the 10 site which varies mostly irrespective of peaks or minima in the atmospheric BC concentrations. Consequently, the modelled BC deposition seems for the most part to be driven by other parameters, for example by meteorological processes, rather than atmospheric BC concentrations. On a 35-year perspective, on the other hand, the modelled BC deposition trend is decreasing similar to the modelled atmospheric BC concentration trend, although the rate of the deposition decrease may not be as evident as for the concentrations due to strong yearly variations (Fig. 5).

15
The modelled BC deposition at Holtedahlfonna is ca. a magnitude lower than the measured EC deposition in the ice and firn cores (Fig. 8). Similar notable under-estimations in modelled BC values compared to observations have been previously reported in the Arctic both for atmospheric BC concentrations (e.g. Dutkiewicz et al., 2014) and BC deposition (e.g. Ruppel et al., 2013Ruppel et al., , 2015. The modelled BC deposition trend at Holtedahlfonna does not show clear consistency with the observed EC deposition in the ice and firn cores, although some similarities can be observed. The 20 notable variation in the measured ice/firn core EC deposition from data point to the next in addition to the year to year variation in the modelled BC deposition highlights the significance of wet deposition patterns and underlying varying meteorological processes to the surface deposition trends. In addition, the modelled BC deposition trend seems to support a notable drop in BC deposition observed between the ice and firn core. The 300-year ice core recorded an average EC  (Fig. 7). Furthermore, the modelled deposition does not show an 35 increasing trend from ca. 2005 to 2015 as indicated by the firn core measurements (Fig. 8).
Consequently, the model results support some features of the ice and firn core observations, such as higher EC deposition in the 1980s and 90s and a drop in deposition thereafter, but these variations are smoothed and lowered by the model in comparison to the ice and firn core values (Fig. 8) changing the precipitation velocities, and the secondmay influence the mixing of the lower troposphere. In addition, the anthropogenic inventory emissions are available as monthly or annual emissions for only every 5 or 10 years (Fig. 3), and the data is linearly interpolated between these data points. Thus, the scenario-based emission datasets may smooth out modelled BC variations in comparison to the ice and firn core observations, and consequently the year-to-year variations

20
Meanwhile, the model could be improved by including a temperature dependency to the scavenging efficiency of BC, as Cozic et al. (2007) showed that the scavenging efficiency of BC increases significantly from temperatures of -20 (~10 % BC scavenged in mixed phase clouds) to 0 °C (60 % scavenged in liquid clouds).

Sources contributing to modelled BC deposition and atmospheric concentrations at Holtedahlfonna
In Ruppel et al. (2014) it was hypothesized that the observed increase in the Holtedahlfonna ice core EC deposition from That area is a major source for BC in Svalbard (e.g. Hirdman et al, 2010;Stohl et al., 2013), and according to Stohl et al. (2013) flaring may contribute to 20-40 % of annual mean surface BC concentrations in Svalbard, but these emissions have been strongly under-estimated or even disregarded in emission inventories (Stohl et al., 2013;Huang et al., 2015).
However, our current model results suggest a significantly lower contribution of flaring to the BC values on

30
Holtedahlfonna between 1980 and 2015: ca. 7 % for the atmospheric concentrations and 2 % for the deposited BC (Fig.   6). Only in sporadic years, such as 1982 and 2010, the flaring contribution is suggested to have increased to over 10 % of the total BC deposited. Interestingly, this modelled contribution of flaring matches well with state of the art dual-carbon isotope source apportionment measurements of atmospheric EC from Tiksi, north-eastern Russia, which suggested flaring to contribute only to 6 % of annual atmospheric EC concentrations at the site (Winiger et al., 2017). No increase in the Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2017-357, 2017 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 5 May 2017 c Author(s) 2017. CC-BY 3.0 License.
As seen in Figure 6 there are notable differences in the source contributions for the modelled BC deposition and atmospheric concentrations. While transport and domestic emissions appear to be the most important sources for BC deposited at Holtedahlfonna, the domestic sector seems to be the most important emission source for atmospheric BC concentrations at the glacier. This difference in the source contribution to the modelled BC deposition vs. atmospheric concentration can be explained by the difference in emission location, injection height, transport pathways, and removal 5 of BC from the atmosphere. In the current setting of chemical transport models, such as SILAM, the physical properties of the emitted particles (type, size, hygroscopic properties) are characterized on a low description level, e.g. no aerosol dynamics, and thus no substantial difference in physical properties between the different emission sectors is present.
Nevertheless, in long-term assessments of BC, the meteorology is key to determine transport pathways and scavenging of the particles from the atmosphere, and may thereby affect the differences between source contributions of modelled 10 atmospheric and deposited BC.
For the modelled BC deposition, the contribution of domestic emissions has decreased while transport emissions have generally increased from 1980 to 2015, particularly when including shipping (Fig. 6a). The north of 40° N BC emissions from the transport sector have first increased from 1980 to ca. 2000 and then decreased (Fig. 3). A similar trend is also identifiable in the modelled source contribution of BC deposition at Holtedahlfonna, although in the 2010s the 15 contribution of transport increases again (Fig. 6a). While this temporal evolution of emissions and modelled BC deposition from the transport sector resemble to some extent the observed EC deposition trend in the Holtedahlfonna ice and firn cores, the fraction of transport emissions to the total BC deposited at Holtedahlfonna seems based on the model data too low to solely explain the recorded ice and firn core EC deposition trends. Furthermore, the model results show that between 1980 and 2015 the contribution of natural fire emissions to both the atmospheric BC concentrations and BC 20 deposition has increased (Fig. 6), as also suggested by their increasing emissions (Fig. 3). Interestingly however, natural fires constitute 24 % of total BC emissions north of 40° N between 2010 and 2015 in the used emission data, but their contribution to the modelled atmospheric BC concentrations and BC deposition at Holtedahlfonna is significantly lower, ca. 5 % for both atmospheric composition and deposition in this time period. This may suggest that natural fire BC emissions are prone to be washed out of the atmosphere before reaching Svalbard. BC emissions from natural fires appear 25 mostly in spring and summer, and are the dominant source contributor in this season at Holtedahlfonna, but their contribution to annual deposition increases only seldom to notable values at the glacier.
In summary, emissions from the domestic and transport sector, followed by industry, seem to affect the BC values at Holtedahlfonna the most. None of the anthropogenic or natural fire emissions have varied independently or together in a manner that could solely explain the observed EC variation in the Holtedahlfonna ice and firn cores. Furthermore, the 30 amount of BC emissions from individual sectors (Fig. 3) does not equal the modelled contribution of these emission sectors to the atmospheric BC concentrations or especially BC deposition at Holtedahlfonna (Fig. 6). Consequently, it seems most likely that meteorological processes affecting wet deposition patters at the glacier (and during transport) have had a stronger influence on the EC deposition trends at Holtedahlfonna than the BC emission trends.

35
According to a shallow firn core collected in spring 2015 from Holtedahlfonna glacier, Svalbard, EC concentrations and deposition have dropped to lower values in the 21 st century after rapidly increasing values recorded from 1970 to 2004 at the glacier in a 300-year ice core (Ruppel et al., 2014). Neither the increasing trend from 1970 nor the rapid drop in EC Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2017-357, 2017 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 5 May 2017 c Author(s) 2017. CC-BY 3.0 License.        Ruppel et al., 2014) and in the shallow firn core (red curve) collected from different sites (see Fig. 1) on