Black carbon and other light-absorbing impurities in snow in the Chilean Andes

Vertical profiles of black carbon (BC) and other light-absorbing impurities were measured in seasonal snow and permanent snowfields in the Chilean Andes during Austral winters 2015 and 2016, at 22 sites between latitudes 18°S and 41°S. The samples were analyzed for spectrally-resolved visible light absorption. For surface snow, the average mass mixing ratio of BC was 15 ng/g in northern Chile (18–33°S), 28 ng/g near Santiago (a major city near latitude 33°S, where urban pollution plays a significant role), and 13 ng/g in southern Chile (33–41°S). The regional average vertically-integrated loading of BC was 207 µg/m2 in the north, 780 µg/m2 near Santiago, and 2500 µg/m2 in the south, where the snow season was longer and the snow was deeper. For samples collected at locations where there had been no new snowfall for a week or more, the BC concentration in surface snow was high (~10–100 ng/g) and the sub-surface snow was comparatively clean, indicating the dominance of dry deposition of BC. Mean albedo reductions due to light-absorbing impurities were 0.0150, 0.0160, and 0.0077 for snow grain radii of 100 µm for northern Chile, the region near Santiago, and southern Chile; respective mean radiative forcings for the winter months were 2.8, 1.4, and 0.6 W/m2. In northern Chile, our measurements indicate that light-absorption by impurities in snow was dominated by dust rather than BC.

Sites in southern Chile. In contrast to northern and central Chile, southern Chile has frequent precipitation and abundant vegetation. Concentrations of mineral dust in snow are therefore expected to be low, relative to northern Chile. Southern Chile is considerably colder in winter (the average temperature in June is 17 °C in Arica and 7 °C in Osorno). Residential wood combustion is the main source of pollution for many cities in southern Chile 32 . Thus, a high incidence of black carbon is expected in snow in southern Chile, particularly relative to mineral dust. Sites are indicated on the map in Fig. 1.

Sample Collection and Filtration. Sample collection and filtration follows the methodology of Doherty
et al. 17,26,29 . Samples were collected at vertical intervals of 5 to 8 cm depth from the surface (exposed to the air) down to the solid ground. Typically, two parallel vertical profiles were collected, separated by a horizontal distance of 1 to 2 m. To minimize pollution due to nearby sources of transportation, sites were chosen to be approximately 1 km away from any roads when possible.
Samples were collected with a metal spatula into plastic food-handling bags, which were then packed in plastic Whirlpak bags. Food-handling bags were used for the inner bags because prior work indicated that BC adheres least to them, compared to other types of plastic 26,28 . To minimize the amount of particulates that adhered to the plastic bag with melting, the snow was stored in freezers or insulated containers, when possible, until immediately before filtration. The depth profile of snow density was also measured at each site.
Filtering was performed on a clean table or lab bench, pre-cleaned for dust and dirt. Gloves and lab coats were worn. Snow from Putre and San Pedro was filtered in temporary laboratories set up in local accommodations; snow from other locations was filtered in University laboratories. In southern Chile and near Santiago in 2016, bagged snow samples were stored in Styrofoam coolers to keep them at near-freezing temperatures during transportation to a freezer in a University; samples were thereafter kept frozen until they could be filtered in the University laboratory.
Because particulate matter adheres to glass less than it adheres to plastic, the snow was transferred to a glass beaker for melting, then the beaker was covered with a plastic bag to prevent external contamination. The snow was heated in a microwave oven until just melted, to avoid heating of the meltwater. A 0.4-µm Nuclepore filter was placed in a filter holder mounted on top of a side-arm flask. A stainless-steel funnel was clamped above the filter holder. The meltwater was then vacuum-filtered with a hand pump or electric pump through the filter. After filtering, filters were carefully placed in sterile petri dishes. The petri dishes were opened slightly to allow the filters to air dry, after which the petri dish lid was kept closed until optical analysis was performed. The mass of snow filtered was adjusted with the goal of producing filters with loadings within approximately 0.4 to 40 µg/cm 2 , to stay within the measurement-sensitivity range of the spectrophotometer (described in Section 2.3); filters determined to be outside the spectrophotometer measurement range were discarded. The meltwater volume was measured. The funnel, filter holder, and beaker were cleaned after each filtration with ultrapure water. optical Analysis. Optical analysis of filters is performed with the ISSW spectrophotometer described by Doherty et al. 29 , which is a modified version of the instrument developed by Grenfell et al. 25 . The ISSW configuration includes a sample filter sandwiched between two integrating spheres, positioned above and below the sample filter. Input light is fed via fiber optics through a side port into the lower sphere, and light exits through a side port in the upper sphere to the spectrometer via a fiber optic. Baffles prevent direct light from passing through the filter and prevent light from the filter from passing directly out through the side port; thus, only diffuse light is transmitted to the detector. The ISSW has a spectral range of 450-750 nm and has a measurement resolution of 1 nm which is then averaged to 10 nm.
The spectrum of light attenuation recorded by the spectrometer needs to be calibrated in order to convert the measured signal to a loading of BC on the filter. This is done using a set of standards with a range of known loadings of fullerene, a synthetic black carbon (Alfa Aesar, Inc., Ward Hill, MA, USA), as described by Grenfell et al. 25 . The calibration curves express the light attenuation as a function of fullerene loading in µg/cm 2 . Assuming fullerene is a reasonable proxy for ambient black carbon, comparison of the light attenuation by a sample filter to the light  Table 1. Snow sampling sites in northern Chile. Depth indicates the total depth of the snowpack, except as noted. * These two sites were on glaciers or permanent snowfields, so for them we give the sampling depth rather than the total depth.
www.nature.com/scientificreports www.nature.com/scientificreports/ attenuation by the synthetic standards yields the BC-equivalent loading of the sample filter. Calibration is done separately for each 10-nm bin, based on separate calibration curves made from the fullerene standards. This is necessary because the sample filter usually contains non-BC light-absorbing particles as well as BC particles, and the spectral shapes of non-BC and BC particles differ. Thus the BC-equivalent loading generally varies with wavelength.
Starting from the wavelength-dependent BC-equivalent loading, we calculate the quantities given in Table 4. These are described below in turn (after ref. 29 ). C BC max is the average BC-equivalent concentration, calculated by assuming all absorption in the 650-700 nm wavelength band is due to BC. To convert from the measured fullerene-equivalent loading (µg BC-equivalent per cm 2 ) to concentration, the average 650-700 nm loading is multiplied by the exposed area on the filter (cm 2 ) and divided by the mass of snow water filtered (g), yielding µg/g. This is converted to ng/g, giving the mixing ratio as ng BC per g of snow water. C BC max represents the maximum possible black carbon in the sample. Also of interest is the amount of absorption due to non-BC particulate components. Here we use the absorption angstrom coefficient, Å tot , to discriminate between BC and non-BC components. Å tot is defined as:     Herein, Å tot is calculated as a linear fit to absorption vs. wavelength in log-log space between the wavelengths λ 1 = 450 nm and λ 2 = 600 nm. The absorption optical depth τ abs is the mass absorption coefficient of fullerene multiplied by the BC-equivalent loading. The mass absorption coefficient of fullerene is 8.9 m 2 /g at 450 nm and 6.5 m 2 /g at 600 nm. Pure BC has Å BC = 1.1. For non-BC, we assume that Å NBC = 5, based on prior work by Grenfell et al. 25 and Doherty et al. 26 . The deviation of Å tot from 1.1 is therefore an indication of the amount of absorption by non-BC components in the snow. C BC est is calculated following the method of Grenfell et al. 25 : The value r BC (λ 0 ) is the fraction of absorption due to BC at a given wavelength λ 0 , and it is determined by assuming that where Å tot is the measured absorption Angstrom exponent of all particles on the filter and, as given above, Å BC = 1.1 and Å NBC = 5.0, with r BC defined over the same wavelength range as the values of Å tot . Rearranging Eq. (4) yields r BC . The value for τ tot (λ 0 ) is L 0 β 0 , where L 0 is the BC-equivalent loading and β 0 is the mass absorption efficiency of fullerene at the given wavelength. Furthermore, τ as a function of wavelength is given by the first term on the right is replaced with τ tot (λ 0 ). C BC equiv is the mass of BC that would be needed to account for the integrated 300-750 nm solar absorption by all particulates on the filter. It is calculated by weighting the spectral, all-component (total) absorption by the downwelling solar irradiance spectrum 29 . C BC equiv differs from C BC max in that C BC max represents the maximum mass of BC possible, whereas C BC equiv represents the amount of absorbing material present, whether BC or not, expressed as an equivalent BC mass. C BC max is quantified at 650-700 nm, whereas C BC equiv is quantified for 300-750 nm; since non-BC components will absorb more strongly at shorter wavelengths, C BC equiv is always greater than or equal to C BC max . The fraction of solar absorption integrated across 300-750 nm by non-BC particles, f NBC est , is calculated by "(1) extrapolating the absorption spectra linearly from 450 nm down to 300 nm 25,26 ; (2) weighting the non-BC absorp- ) and all-component (total) absorption spectra (τ tot (λ 0 )) by a downwelling solar irradiance spectrum appropriate for clear-sky, mid-latitude winter; (3) integrating these weighted spectra across the visible range; and (4) taking the ratio of the two integrals" 29 .
In addition to reporting concentrations in given snow layers, the vertically-integrated BC mass, "Int BC", was calculated as follows: First, the snow water equivalent is calculated in each sample layer by multiplying the snow density (g/cm 3 ) by the layer depth (cm). The BC amount within each sampled layer of snow is then calculated by multiplying C BC est (ng/g) by the snow water equivalent (g/cm 2 ), yielding a BC mass per unit area (ng/cm 2 ). This is done for each sampled layer, and the BC mass from all layers summed together to give integrated BC for the snow column, which is reported in µg/m 2 . Because it is difficult to sample all the way to the ground without contamination by soil, the BC amount for the last few centimeters is typically estimated from C BC est and the density of the layer sampled immediately above. For a given sample site, measurements within a few hundred meters, if made on the same day, are averaged. In such cases, a nearby measurement may be used to fill in missing values (e.g. in a case where filter loading from one sample was too high for an accurate measurement). estimating albedo reduction and radiative forcing. The albedo reduction and radiative forcing due to light-absorbing impurities in snow were calculated using the method of Dang et al. 33 , and making use of the  Table 4. Quantities calculated from ISSW measurements. The wavelength range given is that used to calculate the quantity, as described in the text. Mass ratios are ratios with respect to snow water mass.
www.nature.com/scientificreports www.nature.com/scientificreports/ parameterization of Dang, Brandt, and Warren 19 , as follows. The all-sky broadband albedo reduction was calculated according to cloudy clear where CF is the cloud fraction. Monthly mean cloud fraction was determined from the CERES Terra + Aqua Edition 4A 34 . Monthly means for each of the three winter months (June, July, August) were averaged from 2000 to 2017 (Fig. 4a). The albedo reduction was determined from the parameterization of Dang et al. 19 , which gives the albedo reduction for clear and overcast skies for a range of BC mass mixing ratios and snow grain radii. To get the albedo reduction due to BC, the albedo reduction was interpolated to the BC mass mixing ratio for a given value of C BC est . This was done for cloudy skies (yielding δα cloudy ) and for clear skies (yielding δα clear ), after which Eq. (6) was applied. These steps were repeated for C BC equiv to calculate the albedo reduction for all light-absorbing impurities. Albedo reductions due to BC and due to all light-absorbing impurities were calculated for the months of June-July and for snow grain radii of 100 and 1000 µm.
The radiative forcing is calculated as the albedo reduction times the all-sky shortwave flux, where the latter is also taken from CERES (see above and Fig. 4b) and averaged over the years 2000 to 2017.
Albedo reductions and radiative forcings were not calculated for every site given in Tables 1-3, but rather for median values of C BC est and C BC equiv for each location (e.g. the median for all the Putre sites, etc.), except when the locations were separated by a large altitude difference or when the site was revisited on different days.

Results
Latitude dependence of light-absorbing impurities. C BC est , Å, f NBC est , and Integrated BC (Int BC) are shown as functions of latitude in Fig. 5. Sampling sites are color-coded and listed in the legend, from north to south; the color coding is the same as for the site labels in the maps shown in Figs 1 and 2. One point is plotted for each layer sampled; typically these are averages of two samples made for the same layer, although sometimes more or fewer samples were taken of a given layer. At some sites, samples were taken at different locations within the site area, or the site was revisited on a different date, as indicated in Tables 1-3. For Fig. 5d (lower right), which shows the estimate of snow-column-integrated black carbon mass, values for sites within a few hundred meters and at a similar altitude were averaged. Multiple points are due to sample collections separated by a considerable distance, altitude, or for different conditions at the same site (e.g. La Ola, where one set of samples was made in a snow drift) or due to revisiting the site at a later date (e.g. Valle Nevado). Points are missing where snow density was not measured (e.g. Putre) or where there was no surface measurement. surface snow. Table 5 gives values for Å tot , C BC est , C BC max , and C BC equiv for surface samples at each location. The layer thickness of the surface sample is also shown in the table. Medians were calculated across all surface samples from a given location, except when the locations were separated by a large altitude difference or when the site was revisited on different days. (These values are given for each layer in each sampling site in Tables S1-S3 of the Supplemental).
Regional results. Regional averages and standard deviations for measurements in northern Chile, the Santiago region, and southern Chile are given in Table 6. In northern and southern Chile, where there were sometimes large distances between sites, the average and standard deviation for each site were geographically weighted by the increment of latitude that the site was judged to represent. Values shown in the table include surface measurements of Å tot , C BC est , C BC max , and C BC equiv (as well as the subsurface values for C ) BC est , the total snow depth, snow water equivalent, integrated black carbon, and average black carbon. For subsurface and snowpack measurements, averages in the north exclude Putre, where snow density was not measured. Also shown in Table 6 are the albedo reductions due to light-absorbing impurities in snow for BC alone and for all light-absorbing impurities and the resulting radiative forcing; these are discussed below.
Albedo reduction and radiative forcing. Figure 4a,b show the cloud fraction and shortwave flux for the Austral winter months June, July, and August. While cloud fraction does not change much with month, it increases substantially from north to south. Shortwave flux changes considerably with both latitude and month. Figure 4c shows the albedo reduction due to all light-absorbing impurities in snow for each site shown in Table 5 (for the values of C BC equiv given in the table) assuming snow grain radii r = 100 µm (small symbols) and r = 1000 µm (large symbols). The choice of these representative grain sizes is explained by Wiscombe and Warren 35 and Warren 36 . Because new snow is typically fine-grained, the albedo reduction is plotted for r = 100 µm for the beginning of winter (June in the north; July in the south). By contrast, melting snow is typically coarse-grained; thus the albedo reduction is plotted for r = 1000 µm for the melt season (July in the north, August in the south). The albedo reduction and radiative forcing are greatest in the far north and in the Santiago region.
Albedo reductions and radiative forcings for regional averages are given in Table 6 for 100 and 1000 µm snow grains and for BC-only and all light-absorbing impurities.

Discussion
Latitude dependence of non-BC light-absorbing impurities. The contribution of non-black carbon constituents to particulate absorption shows a trend with latitude, decreasing southward. The fraction of absorption from non-BC decreases from around 0.9 near San Pedro to 0.1 at the southern-most site of Osorno (Fig. 5c). The far north is a desert where large amounts of dust may be lofted into the air by natural and human causes (see, e.g. Fig. 6). In contrast, the southern region is characterized by frequent precipitation and the atmosphere is therefore less dusty.
www.nature.com/scientificreports www.nature.com/scientificreports/ How much the Angstrom exponent, Å tot, deviates from 1.1 is an indication of the relative contribution of non-BC particles to the light-absorption. Starting at the north end of Chile and moving south, Å tot first increases from about 2.5 at Putre (Site 1), to a maximum of ~5 at La Ola (Site 3; Fig. 5b), then decreases to a minimum of 1.1-1.9 at Osorno (Site 22), indicating minor contribution from non-BC at Osorno. This geographic trend in Å tot holds for samples from all layers (Fig. 5b). For surface samples, the trend is also the same, except that Å tot peaks at Site 2 above San Pedro (Å tot = 4.5; Table 5).
Particles with Å tot near 1.0 are essentially black, indicating their light-absorbing component is nearly all black carbon (Bond et al. 16 ), indicating a combustion source for particulate matter. The low values of Å tot at our southern sites are consistent with a combustion source for the snow particulate matter. Added to the knowledge that wood-burning is the main source of heating in this region and the lack of dust sources we infer that light-absorbing particles in snow are dominated by carbonaceous aerosols. To the north, the high values of Å tot of 3 to >5 indicate a much larger contribution from non-BC sources. Previous measurements with the ISSW coupled with chemical analyses (refs 26,28,29 ) indicate that these high values of Å tot could be due to very high brown carbon to black carbon ratios in combustion-sourced aerosol or, more likely for the highest values seen here (>3), due to dust. Dust as the dominant source of light-absorbing particles to snow in northern Chile is consistent with the low population density, the proximity of dust sources and the low snow-cover fraction.
Precise attribution of absorption to sources using Å tot is not possible. We have assumed that Å NBC is 5.0. If the particulate absorption at Site 3 is in fact essentially all due to non-BC (e.g. dust), this supports the validity of using Å NBC = 5.0 for this region. If, in fact, there is a significant contribution to absorption by BC even at Site 3, Å NBC www.nature.com/scientificreports www.nature.com/scientificreports/ should be higher; in this case, our reported values of C BC est will be too low and, correspondingly, our attribution of absorption to brown carbon and dust will be too high. Regardless, trends in Å tot are strongly indicative of compositional trends in light-absorbing particulate matter in snow. Fig. 5a, C BC est decreased southward to a minimum above Elqui (Site 4), then increased continuing south to the Santiago area (Sites 7-11), and finally decreased gradually southward to Osorno (Site 22). In the north, this trend is consistent with the trend for non-BC, suggesting that the snow becomes overall cleaner moving south from the far north of Chile (Putre) to Elqui.

Latitude dependence of black carbon. As shown in
The trend for C BC est described above is in keeping with the regional averages, for which C BC est is highest near Santiago. Near Santiago there is a lot of variation in C BC est from site to site, which may be linked to the influence of dry deposition, as will be discussed below.
The integrated black carbon is a metric for total BC deposition over the period of snowpack accumulation. For regional averages, both integrated black carbon and snow depth increased southward. C BC est for surface samples in the Santiago region is higher than the value for southern Chile; however Santiago has considerably less integrated BC. The values for average BC are the same. The explanation is that there can be high BC concentrations but low column-integrated BC mass if the snowpack is thin.
Wet vs. dry deposition. The above observations alone do not permit unambiguous identification of the dominant mechanism of BC deposition (dry deposition, in which BC settles onto snow from the atmosphere, or wet deposition, in which BC is transferred to snow via precipitation), because of a number of complicating factors. One complication is that the sets of numbers used to compute C BC est , the integrated BC, and the subsurface BC vary slightly due to missing data and variations in snow depth (e.g. there was not always a subsurface layer to sample). This is particularly important for the Santiago area (as well as in the north), where snow depth was typically thinner and where there were large variations in C BC est (110%) and integrated BC (66%). In southern Chile the variations were smaller, 70% and 52%, respectively, indicating a more uniform snowpack. It is furthermore worth noting that if snow cover is intermittent at a given location, then the column-integrated BC mass will not reflect total BC deposition for the winter, whereas if there is no loss of snow particulate matter to melting, the column-integrated BC will reflect the total BC deposition since snowfall started. There are, however, situational factors that help constrain this distinction; an exploration of wet vs. dry deposition using a few case studies follows.
Meteorological data and observations suggest a lack of snowfall in northern and central Chile from 1 to 6 July 2015, with the exception of a light snowfall on 4 July at Site 1, Putre (observed by the sampling group). On 5 and 6 July a cold front may have resulted in some snowfall in the mountainous region east of Santiago. Another front brought snowfall to the Santiago area on 11 and 12 July and to Elqui on 13 July. Blizzard conditions with snow and wind were experienced at La Ola on 13 July, the day before sampling was performed. On 16 July a less intense front brought some snowfall to the mountains east of Site 4, Elqui, but no precipitation was measured at the station near the Elqui sampling site. There was no evidence of additional snowfall for the remainder of the 2015 field season. Further details are given in the Supplemental.
Based on the above as well as reports from locals near the field sites, Table 7 gives estimates of days since last snowfall, C BC est for both the surface and sub-surface snow, and integrated BC for many of the sites for the 2015 field season.
In northern Chile and the Santiago area, the date of the most recent snowfall could be estimated from observations and weather reports. As shown in Table 7 and in Fig. 7, the contamination of the surface layer and the integrated black carbon mass both generally increased with time. Given that the subsurface samples are relatively clean everywhere, proximity to Santiago seems unlikely to be a major factor. Instead, the relative dirtiness of the surface snow suggests that dry deposition plays a major role. In addition, consolidation of BC in the surface layer with snow melt 17 may be playing a role, as well as sublimation of the surface snow at high elevations in the dry Andes 37 .  Table 5. Median values at each location for the surface layer of snow.
Data from meteorological stations suggest that most of the precipitation from 27.2 to 33.4°S was due to the passage of a front on 12 July, with small amounts of precipitation in this latitude range the following day for stations from 33.3 to 33.4°S. However, precipitation was measured at Santo Domingo at 33.39°S on 5 July. Thus it is possible some of the subsurface snow sampled at Valle Nevado and Valle Maipo was seven days older, consistent with the dirtier subsurface snow observed there.
Near Santiago, measurements made on sequential days more clearly indicate the importance of dry deposition. At Yerba Loca, near Santiago, the full column of the snowpack was sampled two days in a row (25 and 26 July 2016) at nearby locations, allowing for determination of the dry-deposition rate. The snow was melting and the depth decreased from 10 to 5 cm, while the snow density increased from 0.19 to 0.44 g/cm 3 . The snow-column-integrated BC increased from 31 to 200 µg/m 2 , indicating that the new snowfall in the Santiago area was relatively clean, and light-absorbing impurities arose mainly from dry deposition.   www.nature.com/scientificreports www.nature.com/scientificreports/ Aerosol deposition rate near santiago. The freshly fallen snow of 25 July 2016 at Yerba Loca was quite clean both in terms of the BC concentration and in terms of the non-BC contribution to absorption, which was negligible. On 26 July particulate absorption included a significant contribution by non-BC components, as Å tot increased from ~1.1 to 2.0 and f NBC est increased from a negligible value to 0.3. Since no snow fell between these two days, this contribution must have been via dry deposition.
The deposition rate is estimated as the difference in integrated BC mass divided by the time difference, or 17.3 ng/cm 2 -day. This is equivalent to 5.2 mg/m 2 -month, which is roughly comparable to the estimated inter-monsoon dry-deposition rate of BC in the Himalayas (see Fig. 3a in Menegoz et al. 38 ).  Table 7. Estimated number of days since the last snowfall before sampling, and median surface and subsurface black carbon concentration s, in northern and central Chile. www.nature.com/scientificreports www.nature.com/scientificreports/ Trend in BC with distance from a major highway. To determine whether significantly more BC is deposited close to major highways, measurements were made at varying distances from the Trans-Andean highway to the southeast of Juncal, in the Juncal National Park. Measurements were made 3.64 km, 1.89 km, 1.05 km, and 0.561 km from the highway (Sites 8a-d). Moving along this transect toward the highway, the surface black-carbon estimate was 15, 15, 14, and 18 ng/g. Thus black carbon from vehicles does not seem to be preferentially deposited near the road.
Albedo reduction and radiative forcing. Albedo reduction and radiative forcing due to impurities in snow in the Chilean Andes are found to be greatest in the far north and in the Santiago region. For example, under all-sky conditions, mean albedo reductions due to light-absorbing impurities in snow in the Chilean Andes are estimated to be 0.0150, 0.0160, and 0.0077 for northern Chile (in June), the region near Santiago (in June), and southern Chile (in July), respectively, for snow grain radii of 100 µm (representative of new snow). Likewise, mean radiative forcings are estimated to be 2.8, 1.4, and 0.6 W/m 2 , respectively.
In northern Chile, BC plays a smaller role in albedo reduction than non-BC. For example, albedo reduction for 100 µm snow grains due to BC alone is only about 43% of that for all light-absorbing impurities. By comparison, these albedo reductions are 53% and 82% near Santiago and in southern Chile, where a greater share of light absorption is due to BC.
Comparison to other measurements. The regional means and standard deviations of C BC equiv for northern Chile measured in this work (refer back to Tables 5 and 6) are higher than values for measurements made in the Cordillera Blanca in Peru 23 . At an altitude of ~4900 m, Schmitt et al. 23 measured mean effective black carbon values of 10 ± 7 ng/g and 17 ± 12 ng/g. By contrast, for the three sample sites at around 5000 m for this work, the mean value of C BC equiv was 112 ± 29 ng/g. The mean for northern Chile, corresponding to altitudes of 1800-5300 m, was 72 ± 17 ng/g.
Regional averages of C BC est in the top 5 cm of snow in the Arctic and North America were 27 ng/g and 29 ng/g (ref. 33 ). This is comparable to the regional average near Santiago (28 ng/g; refer to Table 6), but is about twice as large as the regional averages for northern and southern Chile (~14 ng/g). The subsurface snow was clean in the Andes (7 to 15 ng/g) compared to snow in the Arctic and North America (21 and 30 ng/g; ref. 33 ).
In northern Chile, a large fraction of the computed albedo reduction is due to non-BC. When only black carbon is considered, the albedo reduction due to BC is about 43% of that for all light-absorbing impurities ( Table 6). This value and the value for snow near Santiago (53%) are roughly comparable to values for the Arctic, western North America, and China (53%, 49 to 78%, and 41%). By contrast, in southern Chile, black carbon accounts for a greater share of the albedo reduction (82%).

Conclusions
Based on snow-sampling measurements made in 2015 and 2016, the black-carbon mass-mixing ratio in snow in the Chilean Andes decreased southward from Putre (18.1°S) to the Elqui Valley above La Serena (29.9°S), after which it increased continuing southward to the Santiago area (33.3°S), then again decreased from the Santiago area south to Osorno (41.1°S).
The contribution of non-BC light-absorbing components to particulate absorption in snow decreased dramatically from north to south, as expected given the transition from the dusty, vegetation-sparse Atacama Desert to the wetter, vegetation-rich south of Chile.
Regional results for Chile are as follows: for surface snow, the average mass mixing ratio of BC was 15 ± 4 ng/g in the north, 28 ± 35 ng/g near Santiago, and 13 ± 3 ng/g in the south. The regional average vertically-integrated loading of BC was 207 ± 46 µg/m 2 in the north, 780 ± 510 µg/m 2 near Santiago, and 2500 ± 440 µg/m 2 in the south, where the snow season was longer and the snow was deeper.
The estimated black carbon mass mixing ratio in surface snow generally increased with time since the last snowfall for a transect from northern to central Chile from 26.5 to 33.5°S, indicating that dry deposition plays a larger role than wet deposition. Sampling on two consecutive days near Santiago indicated a dry deposition rate of 5.2 mg/m 2 -month, which is roughly comparable to the inter-monsoon dry deposition rate in the Himalayas.
Mean albedo reductions due to light-absorbing impurities in snow in the Chilean Andes are estimated to be 0.0150, 0.0160, and 0.0077 for northern Chile (in June), the region near Santiago (in June), and southern Chile (in July), respectively, for snow grain radii of 100 µm (representative of new snow). Corresponding mean radiative forcings are 2.8, 1.4, and 0.6 W/m 2 , respectively. In the north, BC plays a smaller role than non-BC; for example, the albedo reduction for 100 µm snow grains due to BC alone is only about 43% of that for all light-absorbing impurities. By comparison, these albedo reductions are 53% and 82% near Santiago and in southern Chile, where a greater share of light absorption is due to BC.