Light-absorbing impurities (LAIs) deposited in snow have the potential to substantially affect the snow radiation budget, with subsequent implications for snow melt. To more accurately quantify the snow albedo, the contribution from different LAIs needs to be assessed. Here we estimate the main LAI components, elemental carbon (EC) (as a proxy for black carbon) and mineral dust in snow from the Indian Himalayas and paired the results with snow samples from Arctic Finland. The impurities are collected onto quartz filters and are analyzed thermal–optically for EC, as well as with an additional optical measurement to estimate the light-absorption of dust separately on the filters. Laboratory tests were conducted using substrates containing soot and mineral particles, especially prepared to test the experimental setup. Analyzed ambient snow samples show EC concentrations that are in the same range as presented by previous research, for each respective region. In terms of the mass absorption cross section (MAC) our ambient EC surprisingly had about half of the MAC value compared to our laboratory standard EC (chimney soot), suggesting a less light absorptive EC in the snow, which has consequences for the snow albedo reduction caused by EC. In the Himalayan samples, larger contributions by dust (in the range of 50 % or greater for the light absorption caused by the LAI) highlighted the importance of dust acting as a light absorber in the snow. Moreover, EC concentrations in the Indian samples, acquired from a 120 cm deep snow pit (possibly covering the last five years of snow fall), suggest an increase in both EC and dust deposition. This work emphasizes the complexity in determining the snow albedo, showing that LAI concentrations alone might not be sufficient, but additional transient effects on the light-absorbing properties of the EC need to be considered and studied in the snow. Equally as imperative is the confirmation of the spatial and temporal representativeness of these data by comparing data from several and deeper pits explored at the same time.
The deposition of light-absorbing impurities (LAI) in snow influences the radiation budget and can cause enhanced melting via snow darkening (Warren and Wiscombe, 1980). This process affects regions with seasonal snow cover, leading to an earlier snow retreat, which has major implications for thawing and biogeochemical processes acting on the ground (AMAP, 2011). In mountainous areas with glaciers, the impurities perturb glacier properties and the hydrological cycle (e.g., Xu et al., 2009). The impact on snow reflectance (albedo) from black carbon (BC) aerosol particles is of particular interest. Being one of the most effective light-absorbing aerosols, BC enters the atmosphere by combustion of carbon-based fuels, in activities including forest fires and anthropogenic burning of bio- and fossil fuels (Bond et al., 2013). Because of its negative effect on snow albedo, considerable effort has been made to globally quantify BC in snow (e.g., Doherty et al., 2010; Ming et al., 2008; Schmitt et al., 2015), as well as in ice cores (e.g., McConnell et al., 2007; Ruppel et al., 2014; Xu et al., 2009).
The potential impact of LAI on snow and ice makes the Himalayas a region of special interest. It contains numerous glaciers which are in a general state of recession, although contrasting patterns have been reported in different areas (e.g., Bolch et al., 2012; Kääb et al., 2012). Himalayan glaciers act as freshwater sources for several major rivers in Asia, including the Indus, Ganges, Brahmaputra, Mekong, and Yangtze, thus affecting millions of people's lives (e.g., Immerzeel et al., 2010). The glaciers are especially susceptible to BC emissions, as India and China are located in close proximity and emit the most BC worldwide (Bond et al., 2013). A recent study by Ming et al. (2015) found a decreasing trend in albedo during the period of 2000–2011 on Himalayan glaciers, and suggested that rising air temperatures and the deposition of LAI are responsible for the decrease. In light of the vast area of the Himalayas, there is a lack of in situ measurements of LAI on glaciers, which are crucial for modeling work (Gertler et al., 2016). The lack of measurements is especially pronounced in the Indian Himalayas, since previous measurements of LAI in Himalayan snow and ice have largely been confined to China (e.g., Xu et al., 2006) and Nepal (e.g., Ginot et al., 2014; Kaspari et al., 2011, 2014; Ming et al., 2008).
In addition to BC, other LAI can contribute significantly to the radiative balance of the cryosphere. Recent research has identified mineral dust and microorganisms as having a more important role than previously thought in the current decline in albedo of the Greenland ice sheet and other parts of the Arctic (e.g., Dumont et al., 2014; Lutz et al., 2016). Similarly, Kaspari et al. (2014) reported such high dust concentrations in the snow of Himalayan Nepal that the contribution of dust in lowering the snow albedo sometimes exceeded that of BC. The importance of dust has also been illustrated from other regions, for example the Colorado Rocky Mountains, US, where dust causes a significantly earlier peak in runoff (Painter et al., 2007). In the Arctic, Doherty et al. (2010) suggest that 30 to 50 % of sunlight absorbed in the snowpack by impurities is due to non-BC constituents. Evidently, dust has an important role in the cryospheric radiative balance. However, differentiating between the different impurities in the snow is not trivial and requires more than one analytical technique (Doherty et al., 2016). Traditionally, dust in snow has been quantified by gravimetrically measuring filters (e.g., Aoki et al., 2006; Painter et al., 2012). Other methods consist of using transmitted light microscopy (Thevenon et al., 2009), a microparticle counter to measure the insoluble dust (Ginot et al., 2014), or mass spectrometry (using iron as a proxy for dust) (Kaspari et al., 2014).
At present, three primary methods are used to measure BC in snow and ice (see Qian et al., 2015, in which they are extensively presented). Out of the three methods, two utilize filters to collect impurities in a melted sample. The first filter method measures optically the spectrally resolved absorption by the impurities using an integrating sphere sandwich spectrophotometer (ISSW) (e.g., Doherty et al., 2010; Grenfell et al., 2011). The second filter method is the thermal–optical analysis of filters (e.g., Forsström et al., 2009; Hagler et al., 2007). The third, non-filter-based method, uses laser-induced incandescence with a single particle soot photometer (SP2) (e.g., McConnell et al., 2007; Schwarz et al., 2012).
Each measurement method has benefits and drawbacks. The SP2 is specific to refractory BC and is able to provide estimates on the size of the BC particles. However, the SP2 has a size range limitation (roughly 70–600 nm, depending on the instrument settings and nebulizer setup), which may result in the underestimation of BC mass since particles in snow have been reported to be larger (Schwarz et al., 2012, 2013). Conversely, the use of filters can provide a practical logistics advantage for the collection of LAI in remote locations, as it is difficult to maintain the necessary frozen chain for the snow samples from the field to the laboratory for analysis. Particulate losses can be very significant if a sample is not kept frozen and thus will not provide accurate results. Filtering of liquid samples can be conducted in the field, and the substrates are more easily stored and transported to the laboratory. The ISSW method has the advantage that it measures light-absorbing constituents on the filter indiscriminately. Thus, the ISSW method is not specific to BC, and requires interpretation of the spectral response to determine the BC component. The thermal–optical method (TOM) provides an actual measurement of elemental carbon (EC) that is instrumentally defined. EC is assumed to be the dominant light-absorbing component of BC, and often EC and BC are used interchangeably in literature. The sampling efficiency of quartz filters used in TOM is not well characterized for small particles (Lim et al., 2014). However, smaller particles normally contribute little to the total particulate mass (Hinds, 1999). Thus, each method for measuring BC in snow has both advantages and disadvantages.
Map of sampling locations and sites discussed in text.
Here we present observations of LAI in snow from two glaciers in the Sunderdhunga valley in the Indian Himalayas, which have not – to our knowledge – been explored previously with respect to LAI in snow. Using a measuring approach whereby the TOM is combined with a custom-built particle soot absorption photometer (PSAP), we perform laboratory tests to provide a correct interpretation of the results. Our Himalayan observations are further compared to samples from Arctic Finland for their LAI content.
Snow pit filter samples taken from Sunderdhunga in 2015. Durga Kot glacier snow pits are A–D and Bhanolti glacier snow pit E.
Snow samples were collected in September 2015, during the Indian
post-monsoon season, from two adjacent glaciers in the Sunderdhunga valley
(Fig. 1). The Bhanolti and Durga Kot glaciers (30
At the Durga Kot glacier four snow pits with varying depths were dug at different elevations, while at the Bhanolti glacier one snow pit was dug (see Table 1 for snow pits and sample details). Snow samples were collected with a metal spatula in Nasco Whirl-Pak bags, and thereafter brought to base camp where the snow was melted and filtered. As it was not possible to maintain the crucial frozen chain for the snow samples during transport back to the laboratory, this approach of melting in the field was used for the glacier snow samples. The snow was melted gently over a camping stove in enclosed glassware to avoid contamination. The liquid samples were subsequently filtered through quartz fiber filters (Munktell, 55 mm, grade T 293), in accordance with previous work (e.g., Forsström et al., 2009; Svensson et al., 2013). Filters were dried in ambient conditions in petri dishes and thereafter transported to the laboratory for analysis (described in Sect. 2.2).
Snow samples collected in Finland originated from the seasonal snowpack of
Sodankylä (67
To estimate the contribution to the reduction in transmission on the filter sample substrate due to minerals, we compared the light transmission through the filter using the PSAP before and after heating the sample as part of the TOM analysis. Since it is difficult to gravimetrically determine the dust content on quartz filters, we decided to use this combined instrument approach to estimate the dust content. A custom built PSAP (Krecl et al., 2007) was used for the optical measurements, and for the TOM a Sunset Laboratory organic carbon elemental carbon (OCEC) analyzer was used to determine EC. Brief descriptions of the OCEC analyzer and the PSAP are given below in Sect. 2.2.1 and 2.2.2, respectively.
The approach of measuring light transmission before and after heat treatment
to estimate the different light-absorbing components has been previously
used for airborne sampled aerosol (e.g., Hansen et al., 1993). In Hansen et
al. (1993), filter samples were optically analyzed before and after being
treated in a 600
From a 10 cm
Uncertainties associated with the TOM method are mainly associated with uneven filter loading, loss of particles to filtering containers, and the inefficiency of the filters, capturing the impurities (undercatch) (Forsström et al., 2013; Lim et al., 2014). For our filtering set-up, the undercatch has been estimated as ca. 22 % (Forsström et al., 2013), and is most likely significant for smaller sized particles, since undercatch tests have indicated an inefficiency for smaller sized particles (Lim et al., 2014). During OCEC-analysis, an artifact from samples with a high fraction of pyrolysis OC (Lim et al., 2014), and the interference of an accurate split point determination from filters containing a high dust load, can also be considerable for the TOM method (Wang et al., 2012). Generally, mineral dust may contain carbonate carbon (CC) that may interfere with the OC and EC measurements unless it is chemically removed prior to analysis. However, unlike in other OCEC analysis protocols (such as IMPROVE), here chemical removal of CC was unnecessary, as in the EUSAAR_2 protocol CC evolves during the fourth temperature step of the OC analysis (Cavalli et al., 2010; Karanasiou et al., 2011), and consequently CC does not interfere with our EC quantification. In cases where CC is present in very high concentrations on the filters, ca. 5 % of CC may only evolve during the EC analysis step causing potential overestimation of EC (Karanasiou et al., 2011). As none of our filters indicated high CC concentrations during the fourth temperature step of the OC analysis we assume only minor potential overestimation of our EC results due to CC. Refraining from acid pretreatment of the samples is also advisable, as incomplete volatilization of residual acid is known to cause irreversible damage to the measurement instruments. Furthermore, the acid treatment has been shown to cause intense charring phenomena, which may lead to severe overestimation of EC (Jankowski et al., 2008).
The PSAP uses a single diode at 526 nm as a light source. The light is split by two light-pipes, which illuminate two areas of 3.1 mm in diameter. The filter substrate is placed over these areas and individual detectors below the filter measure the transmitted light. During normal operation, when measuring BC in air, these two signals are used as sample and reference points. The reference point is exposed to particle-free air and the sample spot is exposed to particles present in the ambient air. In this experiment both signals are used to measure the change in transmission by comparing the signal before and after the filter has been analyzed using TOM. The signal change is related to the transmission from a particle free filter (filtered using Milli-Q (MQ) water and dried; the measurement procedure is further explained in Sect. 2.3).
The corrections required for the PSAP when used for air sampling is well
documented (e.g., Bond et al., 1999; Virkkula et al., 2005), in particular
concerning enhanced absorption from the filter itself, through multiple
scattering effects from the filter fibers and particle loading effects
(shadowing and reduction in multiple scattering). However, these corrections
are essentially uncharacterized for melted snow samples and the quartz fiber
filters used. The fiber filters used are substantially thicker compared to
what is normally used for PSAP measurements (Pallflex cellulose membrane
filter) or the ISSW measurements (Nuclepore filter). Moreover, the filter
substrate is very large in terms of surface area compared to the particles
sampled. The geometry is very complex and in relation to a particle the
substrate is more of a three dimensional web or sponge rather than a flat
surface area on a filter. An example of a blank filter sample obtained by a
scanning electron microscope is presented in Fig. 2. The horizontal scale
of 500
The basis for the optical attenuation measurements is the exponential
attenuation of light as it passes through some medium, often described by
the Bouguer–Lambert–Beer law (Eq. 1) as follows:
Electron microscope image of a blank quartz-fiber filter used in this study.
A series of laboratory tests using the OCEC analyzer and the PSAP
combination were conducted before initiating analysis of the field samples.
For this purpose, the following filter sets were created:
A set of filter samples ( The second set of filters ( The last set of laboratory solutions made contained various mixtures of SiC
mineral and chimney soot (
The procedure to analyze all three sets of filters samples was identical.
After the filter substrates had dried, one punch (1 cm
The change in optical depth as a function of analyzed EC using our two
standard types of BC particles (filter set no. 1) is shown in Fig. 3. Both
materials behave optically similar and the slopes are within 15 % of each
other, with chimney soot having a slope of 39.8
Comparison of the optical depth (at
Figure 4 shows results analogous to Fig. 3, but for the two mineral aerosol
solutions (filter set no. 2). The slope of the optical depth of SiC versus
measured SiC amount is more than a factor of one hundred smaller
(0.23
From the analysis of chimney and NIST soot (Fig. 3) and SiC and stone crush
dust (Fig. 4) the experiments were extended to comprise mixtures of soot and
dust. Using the MAC of chimney soot (see Fig. 3), we estimate the EC content
of the third set of filters, containing a mixture of SiC and chimney soot.
The estimated EC (eEC) is based on the difference between the optical
thickness before TOM analysis (
The optical depth (at
In addition to chimney soot, the mineral SiC is the second absorbing
component on the third set of filters. In Fig. 6 the optically estimated
SiC content, based on the SiC slope in Fig. 4 and
EC amount observed by the TOM (EC
Based on the relations established for EC and SiC individually in Figs. 3 and 4, respectively, it is possible to retrieve their separate concentrations from a mixture based on the change in filter transmission before and after heating the filter. The consistent results from these laboratory tests gives confidence in the applicability of applying this method on our ambient samples from India and Finland.
Comparison between the weighed SiC amounts added to the water and the optically derived SiC density on the substrate. The data is for Chimney soot and SiC mixtures, with two alternative slopes; one containing all data points (1.02), and one excluding three data points in the top right of graph (0.88).
In all of the Sunderdhunga snow pits, a distinct layer with concentrated
impurities was observed. These impurity layers always had the highest EC
concentrations (exceeding 300
Previous studies of BC in snow and ice from the Himalayas have shown seasonal variation. At the Mera glacier in Nepal Ginot et al. (2014) showed that BC concentrations peak during the pre-monsoon in a shallow ice core. From the same glacier, Kaspari et al. (2014) observed similar seasonal peaks of BC concentration in snow and firn samples taken above the equilibrium line altitude, where the snow had not undergone any significant summer melt. Interestingly, dust did not show the same strong seasonality as BC in their studies (Ginot et al., 2014; Kaspari et al., 2014).
Measurements of BC in snow taken closest to Sunderdhunga, as reported in the
literature, are from about 140 km east-north-east (78
For reference, the EC concentration in the surface snow from the Finnish
Arctic were in the range of 6.2–102
In the site specific derived MAC values there is a significant difference.
In Fig. 7 the optical depth of EC (
The optical depth
In our case, if the laboratory generated BC consists of smaller particles
compared to the snow samples this could lead to a larger MAC value for the
lab-standards. The size distribution of the BC particles in the filters are
unknown to us, but as suggested by the modeled MAC curve, presented in
Fig. 8, this size dependence can play a role. The modeled MAC for
theoretical BC particles demonstrates a decrease in MAC with particle size,
particularly for particles larger than about 130 nm. The absorption
efficiencies were calculated for
Another hypothesis is related to the fact that the samples are liquids and that the matrix is strongly light scattering and rather thick. It is likely that the liquid will embed the particles deeper into the filter than what is typical for air samples (e.g., Chen et al., 2004). In air and on filter surfaces, BC mixed with a scattering medium shows enhanced absorption. On the samples presented in Table 1, about 90 to 95 % of the carbon is water insoluble organic carbon, whereas the laboratory BC was essentially free from OC. This difference could explain the lower MAC for the ambient samples if the net effect of the added OC actually made the BC a less efficient absorber in this particular matrix. Further tests are required to confirm this hypothesis.
Modeled mass absorption coefficient (MAC) of single BC particles as
a function of particle diameter at
Because the ambient mineral dust MAC value is unknown for our snow samples,
it is not possible to use the SiC or stone crush MAC values to estimate the
dust content on the filters. Instead, we use the fraction of minerals
(
Studies from the Nepalese Himalayas concluded that dust may be responsible for about 40 % of the snow albedo reduction (Kaspari et al., 2014). Similarly, Qu et al. (2014) observed that the contribution of dust to albedo reduction can reach as much as 56 % on a glacier in the Tibetan plateau. Our dust estimate, as a fraction of the optical depth of LAI on the filter, shows similar results or an even greater fraction of dust than these previous studies, highlighting the importance of dust (see also Fig. 10a) causing albedo reduction in this region of the Himalayas.
An average of the vertical profiles from pits C, D, and E is presented for
EC and
The variables plotted in Fig. 10b display layers of enhanced amounts of both dust and EC, located between ice layers, and additionally high values at the top of the pit above the first ice layer. These layers are interpreted as indicators for seasonal variation at this location, with alternating melt and refreezing periods marked by the ice layers. As the ice layers and the enhancements in LAI are interleaved it suggests that the impurities were deposited on the glacier mainly in-between the melt and refreeze periods. In addition, the melting seems to take place in a relative shallow layer at the surface and does not protrude deeply, which may cause the annual layers to mix (Doherty et al., 2013). The observed variation in EC and dust values could correspond to the findings of Ginot et al. (2014) and Kaspari et al. (2014) who showed annually peaking BC concentrations in the pre-monsoon in Himalayan ice cores. However, between the ice layers at ca. 65 and 85 cm, no clear peak is observed in EC or dust values (Fig. 10b), which could either indicate that no peak occurred during that particular year, or an ice layer formed at ca. 65 cm in the middle of the year, as potentially occurred at ca. 105 cm.
Frequency of the occurrence of dust optical thickness fractions at the three sampling sites.
The snow pit potentially covers five years of snow accumulation which is
certainly too short a time period to make any conclusions on a temporal
trend of LAI variations at the site. However, an evident increase in LAI is
present, especially in the top 20 cm. Due to the time span of the snow pit
we cannot know for certain whether this increase presents a short term
pollution event or indicates increasing LAI at the site over a longer time
period. We have two hypotheses for the observed increase in EC
concentrations and the fraction of dust occurring in the top layer of the
snow pit. The higher values may be a consequence of increased ambient EC and
dust concentrations in the area, causing increased dry and wet deposition
fluxes of these impurities to the glacier, even when assuming constant
precipitation. Moreover, as it is
Here observations of LAI in snow originating from two glaciers in the Indian Himalayas are first presented with a method not used widely before to determine LAI in snow. Consisting of a custom built PSAP and an OCEC analyzer, the attenuation of light is studied on quartz filters, providing estimates on the fraction of light-absorbance caused by non-EC constituents in LAI. Himalayan data display a much greater light-absorbance by dust in the LAI compared to filter samples originating from the seasonal snowpack of Arctic Finland. The role of dust in reducing the snow albedo in this part of Himalayan glaciers needs to be further evaluated, as our results suggest that it might be the dominating LAI in the snow. Our measurements further reveal that the optical properties of EC are different for laboratory-generated soot compared to EC deposited on snow. Our finding of a MAC value of about half of the laboratory EC for the ambient EC particles could have implications for the snow albedo reduction caused by EC. Over approximately the last five year period in the Himalayas, EC concentrations in the snow display signs of increase in the top part of the snow pit compared to deeper layers. Additional work on the optical properties of EC in snow are needed to enable more accurate estimates of albedo reduction caused by EC in snow, both spatially and temporally. This should be done by measuring the EC particles light-absorption properties while in the snow, as the ambient conditions could be different than laboratory settings.
Data are available upon request from the listed contact author.
The authors declare that they have no conflict of interest.
This work has been supported by the Academy of Finland projects: “Absorbing Aerosols and Fate of Indian Glaciers” (AAFIG; project number 268004), “Greenhouse gas, aerosol and albedo variations in the changing Arctic” (project number 269095), and “Novel Assessment of Black Carbon in the Eurasian Arctic: From Historical Concentrations and Sources to Future Climate Impacts” (NABCEA (project number 296302). The Academy of Finland Center of Excellence program (project number 272041), as well as the Nordic research and innovation initiative “Cryosphere-Atmosphere Interactions in a Changing Arctic Climate” have also supported this work. Jonas Svensson is thankful for the support from Svenska Kulturfonden. We would like to thank the providers of the soot Consti Taloteknikka, and Göran Lidén at SU Luftlab for the mineral samples. The ACES department at Stockholm University, is part of the Bolin Centre for Climate Research. Finally we would like to thank the participants of the AAFIG 2015 expedition, including Sherpas and guides from Real Adventure, for their work during the expedition. Edited by: Mingjin Tang Reviewed by: three anonymous referees