2.1 Free-running BC simulations in GCM

We evaluate BC surface concentration estimated from three freesimu performed with the LMD-ZT GCM (Hourdin and Armengaud, 1999; Li, 1999; Hourdin et al., 2006) and SPRINTARS (Takemura, 2012) models. The two freesimu with the LMD-ZT GCM comprise of (i) LMD-ZT GCM – with Indian emissions (GCM-indemiss), (ii) LMD-ZT GCM coupled to the Interaction of Chemistry and Aerosol (INCA) model (LMDORINCA) – with global emissions, or namely GCM-INCA. Aerosol simulations in GCMs are also referred as free-running aerosol simulations (freesimu) since simulated aerosol fields are not constrained by observations, unlike constrained simulations (described in Sect. 2.2). We provide only a very brief description of these models here, and detailed information could be obtained from references provided for models. Specific descriptions of aerosol treatment and atmospheric transport for GCM-indemiss are given by Boucher and Pham (2002), Reddy et al. (2004), and Verma et al. (2008); for LMDORINCA by Schulz et al. (2006); and for SPRINTARS by Takemura (2012). GCM-indemiss uses BC emissions over India from Reddy and Venkataraman (2002a, b), over Asia from Streets et al. (2003a), and global emissions of BC are from Cooke et al. (1999) and Cooke and Wilson (1996). BC simulation with GCM-indemiss is carried out with a zoom factor of 3 and 4 applied in latitude and longitude, respectively, having the zoom centralized at 15^∘ N and 75^∘ E, and extending from 5^∘ S to 35^∘ N in latitude, and from 50 to 100^∘ E in longitude, leading to a resolution of 0.8^∘ in latitude and 1^∘ in longitude over the zoomed region. The GCM-INCA global emission inventory for BC is from Generoso et al. (2003) and Bond et al. (2004). BC simulation with GCM-INCA is done in non-zoom mode at a horizontal resolution of 2.5^∘ in latitude and 3.75^∘ in longitude. In both GCM model variants 19 vertical layers with a hybrid sigma-pressure coordinate are used, having 5 layers below 600 hPa and 9 layers above 250 hPa. Model meteorological fields are nudged to 6-hourly data from European Centre for Medium-Range Weather Forecast (ECMWF) meteorological analysis data with a relaxation time of 0.1 day for the year of 2001 in GCM-indemiss, and for 2006 in GCM-INCA.

In SPRINTARS, global emission for BC is adopted from the standard inventories for the Coupled Model Intercomparison Project Phase 5 (CMIP5) (Lamarque et al., 2010; Moss et al., 2010). BC simulation with SPRINTARS is done at a horizontal resolution of approximately $\mathrm{2.8}{}^{\circ }\times \mathrm{2.8}{}^{\circ }$ with 20 vertical layers based on sigma-pressure levels (Takemura, 2012). These simulations are available for a long-term period from 1851 to 2010. The GCM-indemiss simulation for the year 2001 is compared with SPRINTARS output for 2001 (referred to as SP1). Long-term BC simulation output from SPRINTARS is used to compare model values (referred to as SP2) with observations corresponding to the year of observation. Also, these simulations are used to predict the annual rate of increase in BC concentration over the HKH glaciers as presented in Sect. 3. The 1σ uncertainty in mean model simulated atmospheric BC burden based on evaluation in 16 global aerosol models has been estimated as 42 % (Textor et al., 2006).

Among the three freesimu estimates, the prediction of the GCM-indemiss is found have better conformity with the measured than that of the rest other models (please refer to Sect. 3.1). The freesimu estimates from the GCM-indemiss are therefore used in the constrsimu approach (refer to Sect. 2.2). The freesimu of GCM-indemiss has also the highest conformity with the measurements at HA stations, and hence we utilize BC concentration simulated in GCM-indemiss to estimate BC concentration in snow and BC-induced SAR and their impact over the HKH region (refer to Sects. 2.3 and 3.2).

In order to examine sources of BC aerosol over the HKH region due to emissions from nearby and far-off regions as well as those from various source sectors (e.g., residential biofuel use (BF), open burning of biomass (OB), and fossil fuel (FF) combustion), region- and source-tagged simulations carried out in GCM-indemiss (Verma et al., 2011, 2008) are evaluated. The sectors for the BF source include wood and crop waste for residential cooking and heating, the sectors for OB include forest biomass and agricultural residues, and those for the FF source are coal-fired electric utilities, diesel transport, brick kilns, and the industrial, transportation, and domestic sectors (Reddy and Venkataraman, 2002a, b). Two sets of experiments were carried out, where aerosols were either tagged by source regions (region-tagged simulation) or source sectors (source-tagged simulation). In the region-tagged BC simulations, the BC aerosol transport and atmospheric processes are simulated for each geographical region with the emissions outside that region being switched off. These source regions are the following: (1) Indo-Gangetic Plain (IGP), (2) central India (CNI), (3) south India (SI), (4) northwest India (NWI), (5) Southeast Asia (SEA), (6) East Asia (EA), (7) Africa–west Asia (AFWA), and (8) rest of the world (ROW). In the source-tagged BC simulations, the BC aerosol transport and atmospheric processes are simulated for each of the source sectors – BF, FF, and OB.

Recently, Wang et al. (2014) introduced an explicit aerosol tagging technique, implemented in the Community Atmosphere Model (CAM5), in which BC emitted from 14 independent source regions was tagged and explicitly tracked to quantify source-region-resolved characteristics of BC. This tagging technique has been applied to evaluate the source of origin of BC over the region of Arctic and that over the Himalayas and Tibetan Plateau (HTP) (Wang et al., 2014; Zhang et al., 2015). The tagging technique in CAM5 appears similar to that applied for the region-tagged simulation in GCM-indemiss (Verma et al., 2007, 2008, 2011), though the classified regions of interest in CAM5 and GCM-indemiss are different. The source regions implemented in GCM-indemiss zoom grid were classified on the basis of differences in composition of their aerosol emission fluxes and their proximity to the Indian Ocean and the subcontinent (Verma et al., 2007). The masked regions on the GCM zoom grid are shown as Fig. S1 in the Supplement provided with this paper. The tagged region of “South Asia” (SAS) in Zhang et al. (2015) includes most of the parts of the Indian subcontinent (including together the tagged regions, IGP, CNI, SI, and NWI in GCM-indemiss). The tagged region of “AFWA” in GCM-indemiss includes the combined tagged regions of sub-Saharan Africa (SAF), North Africa (NAF), and the Middle East (MDE) in Zhang et al. (2015).

2.2 Constrained BC simulation

The constrained simulation (constrsimu) approach takes into account the influence of a possible discrepancy in base emissions, the localized effect of high emission flux (from the plains near LA), and the combined effect of interannual variability in meteorological effects (Kumar et al., 2018). This is used to establish an alternate approach for estimating the atmospheric BC concentration by surpassing the error induced specifically due to emissions in source regions which prevail in case of the free-running aerosol simulations. In constrained simulations, GCM-indemiss AOD is constrained by the observed AOD. An inversion algorithm is formulated to obtain the surface concentration of BC from constrained AOD of aerosol constituents. It may be notable that constrained simulations are considered only for the LA stations. For the HA stations which are far away from the source of emissions, the impact of regional emission bias is postulated to be minor, and aerosol pollutants are primarily governed by the atmospheric transport processes (simulated atmospheric residence time) from the surrounding regions. Our postulate is supported by the fact that BC from the freesimu of GCM-indemiss that are underestimated by a large factor at LA stations (refer Sect. 3.1 and Table 4) match consistently well with available observations at HA stations (refer to Sect. 3.1 and Fig. 2). Hence, in the present study, simulated BC from GCM-indemiss is considered a suitable choice (based on analysis of the performance of the model with measurements, discussed in Sect. 3.1) for evaluation of BC-induced SAR and its impact on annual snowmelt runoff from the glaciers under study.

Table 3Input parameters for SNICAR-Online model.

NOAA: National Oceanic and Atmospheric Administration solar position calculator. ${\mathit{\tau }}_{ij\stackrel{\mathrm{‾}}{t}}=$ glacierwise seasonal mean value of snowpack thickness. ${\mathit{\rho }}_{ij\stackrel{\mathrm{‾}}{t}}=$ glacierwise seasonal mean value of snowpack density. ${\mathrm{BC}}_{\mathrm{c}ij\stackrel{\mathrm{‾}}{t}}$ = glacierwise seasonal mean value of BC_c.

Download Print Version | Download XLSX

Table 4Comparison of model output range of atmospheric BC concentration with the observed values at different stations.

¹ NMB (normalized mean bias, %): $\frac{\mathrm{Observed}\mathrm{BC}\mathrm{concentration}\mathrm{value}\mathrm{from}\mathrm{literature}-\mathrm{Modeled}\mathrm{value}\mathrm{of}\mathrm{BC}\mathrm{concentration}}{\mathrm{Observed}\mathrm{BC}\mathrm{concentration}\mathrm{value}\mathrm{from}\mathrm{literature}}\times \mathrm{100}$.
² Factor: $\frac{\mathrm{Observed}\mathrm{BC}\mathrm{concentration}\mathrm{value}\mathrm{from}\mathrm{literature}\mathrm{for}\mathrm{LA}\mathrm{stations}}{\mathrm{Modeled}\mathrm{value}\mathrm{of}\mathrm{BC}\mathrm{concentration}\mathrm{at}\mathrm{respective}\mathrm{stations}}$. ³ win: monthly average value of December, January, and February. ⁴ pmn: monthly average value of March, April, and May.

Download Print Version | Download XLSX

Figure 2Figures showing spatio-seasonal variation of atmospheric mass BC concentration for pre-monsoon (filled symbols) and winter (unfilled symbols) for (a) higher altitude stations, (b) lower altitude stations, (c) modeled versus measured atmospheric BC concentration at lower altitudes using RA and PA, (d) stationwise BC concentration in snow (BC_c) and comparison of modeled and measured values for higher altitudes, (e) Glacierwise value of BC_c considering only dry deposition for winter and pre-monsoon season, and (f) glacierwise percentage snow albedo reduction (SAR) from both SNICAR and the empirical approach for pre-monsoon; and spatial distribution of (g) averaged snow density from ECMWF for pre-monsoon season (PMN), (h) average of the modeled BC_c for PMN, and (i) percentage SAR estimated using empirical approach (${\mathit{\alpha }}_{\mathrm{dec}}^{\mathrm{empirical}}$) for PMN.

Download

The two constrained simulations formulated using the ratio approach (RA) and profile approach (PA) are compared with measurements. On a fundamental basis the two simulations, namely RA and PA, use BC-AOD (${\mathit{\tau }}_{\mathrm{bc}}^m$) and total AOD (τ^m) from the freesimu of GCM-indemiss to obtain the ratio R_bc as follows:

where $\stackrel{\mathrm{‾}}{t}$ specifies the seasonal mean for winter and pre-monsoon. The i and j correspond to the grid positions on the spatial map with the resolution of $\mathrm{1}{}^{\circ }\times \mathrm{0.8}{}^{\circ }$ as in the GCM-indemiss simulation output. The total AOD from GCM-indemiss, τ^m is constrained with the ground-based observed AOD (τ^o), which is then scaled with R_bc to calculate the constrained BC-AOD (${\mathit{\tau }}_{\mathrm{bc}}^{\mathrm{c}}$) as given in Eq. (2).

In order to incorporate the ${\mathit{\tau }}_{\mathrm{bc}}^{\mathrm{c}}$ for calculating the RA constrained surface concentration of BC, namely ${\mathrm{SC}}_{\mathrm{bc}}^{\mathrm{cr}}$, a ratio (η_bc) between the GCM-indemiss BC concentration (${\mathrm{SC}}_{\mathrm{bc}}^m$) and ${\mathit{\tau }}_{\mathrm{bc}}^m$ is considered.

The spatial distribution of η_bc is confined to the same resolution as the GCM-indemiss. Using this η_bc the constrained surface concentration is calculated based on the theory that the sensitivity of ${\mathit{\tau }}_{\mathrm{bc}}^{\mathrm{c}}$ to ${\mathrm{SC}}_{\mathrm{bc}}^{\mathrm{cr}}$ is numerically equivalent to the sensitivity of ${\mathit{\tau }}_{\mathrm{bc}}^m$ to ${\mathrm{SC}}_{\mathrm{bc}}^m$.

The constrained surface concentration of BC as obtained above by RA is evaluated with measurements.

In the PA, station height specific aerosol extinction coefficients for the corresponding grid coordinates (i,j) are taken into account. This involves the calculation of the monthly mean of the aerosol extinction coefficient (σ) vertical profile derived from Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO). The extinction coefficients are available layerwise at a wavelength of 532 nm and vertical resolution dz=0.06 km. The previously obtained ${\mathit{\tau }}_{\mathrm{bc}}^{\mathrm{c}}$ is disaggregated vertically using the retrieved profile of the aerosol extinction coefficient (σ in km⁻¹) from CALIPSO, renormalized by the CALIPSO AOD (τ^calip) averaged during the seasonal periods (as indicated by $\stackrel{\mathrm{‾}}{t}$) at given HKH LA stations to obtain the BC extinction coefficient:

The vertical distribution of the BC concentration, termed as the PA constrained concentration (${\mathrm{SC}}_{\mathrm{bc}}^{\mathrm{cp}}$), is then calculated based on the formula given below:

where α_bc is the mass extinction coefficient for BC, taken as 10–16 m² g⁻¹ (Ram et al., 2010). The uncertainty in the BC surface concentration from constrsimu using the RA, taking into account the uncertainty in observed AOD, α_bc, and interannual variability in AOD, is estimated to be within 30 % (Kumar et al., 2018). The uncertainty using the PA, through application of CALIPSO data, is estimated as within 45 %.

2.3 Estimation of BC concentration in snow and its impact on the annual snowmelt runoff for the glaciers under study

BC concentration in the snow (BC_c, µg kg⁻¹) during pre-monsoon and winter is calculated with the following equation:

where ${\mathrm{SC}}_{\mathrm{bc},\stackrel{\mathrm{‾}}{t}}^m$ is the seasonal mean for atmospheric BC mass concentration (µg m⁻³) simulated with GCM-indemiss, v_d is the dry deposition velocity (m s⁻¹), t is the season-specific time interval considered for snow deposition, ${\mathit{\rho }}_{\stackrel{\mathrm{‾}}{t}}$ is the seasonal mean snow density (kg m⁻³), and d is the snow depth (m). BC_c is calculated assuming a uniform distribution of BC in the top 2 cm of pure snow, as it is more contaminated and contributes to a larger albedo reduction than the deeper layers (Tanikawa et al., 2009). Deposition velocity of particles has been inferred to be higher than 0.010 cm s⁻¹ over land (Nho-Kim et al., 2004); the minimal v_d of 0.010 cm s⁻¹ is taken in the present study (similar to Yasunari et al., 2010, for the Himalayan glaciers), and a higher v_d than the above will lead to increase in estimated BC_c.

The uncertainty in the estimated BC_c is 69 % due to model uncertainty in ${\mathrm{SC}}_{\mathrm{bc}}^m$ and lack of information on v_d (model uncertainty for BC dry deposition velocity is 55 %, consistent with that mentioned in Yasunari et al. (2013) as 0.01–0.054 cm s⁻¹). The percentage 1σ variability in the pre-monsoon mean of ${\mathit{\rho }}_{ij\stackrel{\mathrm{‾}}{t}}$ and ${\mathrm{SC}}_{bc,\left(ij\stackrel{\mathrm{‾}}{t}\right)}^m$ is respectively 26 %–30 % and 42 %–55 %. This variability of BC_c is estimated as 50 %–70 %.

Deposition of atmospheric BC over the snow also takes place by wet deposition through below cloud scavenging during snowfall events. Calculation of this deposition therefore requires information about size-resolved snow and BC. There is a lack of this information for the Himalayan region from observations and is therefore estimated using numerical models through parameterization of snow scavenging (Ming et al., 2008). Measurements of snowfall from meteorological stations are not available at and around Hanle (Nair et al., 2013) or over the HKH stations under study. Using information from Nair et al. (2013) on the rate of wet scavenging based on atmospheric BC measurements and on snow accumulation depth at Hanle and snow density from ECMWF, our estimated value of BC_c from wet deposition is 36 µg kg⁻¹ for the pre-monsoon season (using the prescribed snow density of 195 kg m⁻³, the value is the same as that in Nair et al., 2013). This value is obtained in the range of 32–90 µg kg⁻¹ for the HKH glaciers, extrapolating the information at Hanle for the entire HKH region and glaciers under study. The total precipitation amount and the precipitation events have been inferred to be notably low during the pre-monsoon season over the HKH region (Bonasoni et al., 2010b; Yasunari et al., 2010). Hence, due to the lack of an adequate estimation of wet deposition, calculation of BC impacts on SAR during the pre-monsoon season is done by neglecting the wet deposition and considering the dry deposition as the governing mechanism for the removal of atmospheric BC during the period of study.

The seasonal mean of spatial distribution of snow density obtained from ECMWF is used to calculate BC_c. This calculation is also done assuming the fixed snow density of 195 kg m⁻³ (for fresh snow) as per information from Yasunari et al. (2010) in order to facilitate the comparison between estimations of BC_c from the present study and previous studies over Himalayan stations (e.g., from Nair et al., 2013).

The percentage reduction in snow albedo (${\mathit{\alpha }}_{\mathrm{dec}}^{\mathrm{empirical}}$) (%) for the estimated BC concentration in snow cover (µ g kg⁻¹) is assessed using the following empirical relationship proposed by Ming et al. (2009).

The equation describing the relationship between BC concentration in snow and albedo reduction (Ming et al., 2009) was obtained based on snowpack observations and model calculations. As obtained from Table 4 of Yasunari et al. (2010), the BC-induced SAR over the NCO-P from this equation was within −17 % to 8 % of that estimated for an internal mixture of BC with new snow by Hansen and Nazarenko (2004).

BC-induced SAR is also calculated from the online radiative model, Snow, Ice, and Aerosol Radiation (SNICAR) (${\mathit{\alpha }}_{\mathrm{dec}}^{\mathrm{SNICAR}}$) (Flanner et al., 2007) to compare with the value obtained from Eq. (8). This calculation is done taking the difference of snow albedo between two sets of simulations with SNICAR, one with BC_c and the other without BC_c. SNICAR uses parameters such as snow grain size, solar zenith angle, species concentration, albedo of the underlying ground, snowpack thickness, and snowpack density (refer to Table 3). The effective radius of snow grain size is taken as 100 µm (Warren and Wiscombe, 1980) and snowpack density and snowpack thickness are referred from ECMWF gridded archive files corresponding to the period of present study.

The relative percentage of uncertainty in estimated SAR (%) due to uncertainty in BC_c is estimated as within 30 %. The percentage 1σ variability in SAR associated with the variability in ${\mathit{\rho }}_{ij\stackrel{\mathrm{‾}}{t}}$ is 9 %–14 %.

The BC-induced SAR (%) for each of the corresponding HKH stations is calculated. When compared, it is observed that calculated SAR values from SNICAR are lower than those using the empirical approach (refer to Fig. 2e).

Lower values from SNICAR than from the empirical approach is attributed to consideration of external mixing of snow with BC, and possibly a large snow grain size (of 100 µm) for addressing snow aging; a snow grain radius size of 100 to 1000 µm is considered for new and aged snow by Warren and Wiscombe (1980).

The consideration of old aged snow may lead to an enhanced BC accumulation onto the snow surface and a higher reduction of snow albedo than estimated in the present study, taking into account BC accumulation in the top 2 cm of pure snow (not contaminated from the preexisting debris cover). Also, there is a possibility of enrichment of BC in snow (not taken into account in the present study) due to the predominance of sublimation compared to precipitation during the period under study, likely including a higher v_d than in the present study, which justifies our assumption that we are using a lower bound estimate of SAR from what could be the actual scenario. Also, BC-induced SAR used in the sensitivity analysis of annual snowmelt runoff for a glacier (discussed in the next section) is considered for BC_c with dry deposition only (due to lack of adequate estimations of wet deposition). All the above would thereby lead to an enhanced BC_c and impact compared to that estimated in the present study.

The present study aims to evaluate the estimated impact of BC aerosols over the glaciers in the HKH region and identify the glacial region most vulnerable to BC-induced impacts by considering the lower bound estimates of their concentration in snow and the corresponding BC-induced SAR.

2.4 Impact of BC-induced SAR on increase in annual snowmelt runoff for glaciers

SAR is considered one of the precursors of excess snow melting. To quantify the impact of BC concentration in the snowpack on annual snowmelt runoff for the glacier (also termed as annual glacier runoff in the text), we calculated the mass balance and runoff from nine selected glaciers with an energy and mass balance model for glaciers (Fujita and Ageta, 2000) (Table 1 and Fig. 1b). As mentioned earlier, the selected glaciers are widely distributed over the HKH region and located near the zone of high BC_c values (discussed in Sect. 3.2). We have used a calculation scheme in which the albedo of the glacier surface was reduced from the albedo value corresponding to the null value of BC_c during the control run to an albedo value of 0.5, by which we assumed the glacier surface darkened by BC at a given date, and then snowmelt runoff in the following summer season was calculated (Fujita, 2007). Surface albedo of the control run was not a fixed value, but calculated from snow density so that it changed temporally and spatially. In the calculation, other meteorological parameters were not changed so that we evaluate the impact of surface darkening by BC only. The lower portions of the selected glaciers are covered by debris mantle. The sub-debris ice melt is highly altered with the debris thickness, and the albedo lowering by BC should not affect the ice melting under the debris layer. We, therefore, excluded the debris-covered area (Fig. 1c), which is delineated manually by following previous studies in the eastern Himalayas (Nagai et al., 2016; Ojha et al., 2016, 2017). The calculation was performed with the ERA-Interim reanalysis data (Dee et al., 2011), but precipitation was calibrated to set an initial climatic condition for each glacier, which surely affects glacier response to climate change (Sakai et al., 2015; Sakai and Fujita, 2017). In the calculation, Fujita (2007) has pointed out that the timing of surface darkening affected the consequence of annual runoff significantly. Darkened surfaces during winter could be covered by succeeding snowfall while those in the early melting season significantly enhanced glacier melting and thus runoff. On the other hand, the impact of surface darkening in the late melting season was limited because of the remaining short period for melting. Therefore, in the calculation of this study, the date of albedo lowering was changed at 5-day intervals from October to April. Summer mean albedo and annual runoff were calculated, and then averaged for 35 years (1979–2014). The entire process is carried out for each of the glaciers. The output data set is then further analyzed statistically and discussed in Sect. 3.2.1 and 3.2.2. The results of annual glacier runoff depth and summer mean albedo obtained are then used to generate the data set for values of SAR and the corresponding annual runoff increase (ARI), estimating the respective values with respect to the control run for each of the nine glaciers. The BC-induced SAR values estimated for each of the glaciers under study is then interpolated with the above data set to estimate the corresponding range for ARI, which is analyzed in Sect. 3.2.1.

3.1 Comparison between model estimated and measured BC concentration

Model estimates (free-running simulations and constrained simulations as given in Tables 2 and 3) are compared to measurements. This comparison is presented as a scatter plot in Fig. 2a–d for HA and LA stations. The dashed lines on both the sides of the scatter plot represent a deviation factor of 2 and 0.5, respectively. A detailed performance of the model outputs with respect to the measurements is also provided in Table 4.

3.2 Analysis of spatial distribution of BC concentration in snow: estimates of impact on SAR and annual glacier runoff

The comparison of BC concentration in snow (BC_c) between GCM-indemiss-derived values and those obtained from available studies (e.g., Nair et al., 2013; Yasunari et al., 2010) at HKH HA stations is presented in Fig. 2d. As mentioned earlier, snow density used to estimate BC_c is extracted from the ECMWF ERA-Interim daily reanalysis data set. In order to compare our BC_c estimates with those from available studies, we also estimate BC_c using the fixed snow density of 195 kg m⁻³ as done in the available studies. These estimates using fixed snow density and those using ECMWF data at HA stations are presented in Fig. 2d.

The estimated value of BC_c from the present study compares well with that obtained from previous work and has a variation of 10 % to 27 % from the earlier studies. These estimates compare relatively well at Hanle and NCO-P; however, at Satopanth during winter they are overestimated compared to the respective measured value. The percentage difference of BC_c estimates using fixed snow density with respect to that using ECMWF retrieved values is found to be 34 %–38 %. The estimated BC_c (with fresh snow consideration) over HKH glaciers (refer to Fig. 2e) during the pre-monsoon (20–68 µg kg⁻¹) and that during the month of June (84 µg kg⁻¹) in the present study is inferred to be comparable to that measured over the southeastern Tibetan Plateau e.g., for fresh snow and ice samples (four numbers during June) at Demula Glacier (29.26^∘ N, 97.02^∘ E; 56.6±26.1 µg kg⁻¹) (Zhang et al., 2017). This inference is consistent with the observational studies indicating the BC concentration in snow over the southeastern Tibetan Plateau is the most comparable to those over the Himalayan regions (Zhang et al., 2017). We also compare the estimated BC_c from our study with the observed values at a few other stations over the Himalayas (e.g., East Rongbuk Glacier, Kangwure, Qiangyong). This comparison over the mentioned stations has also been carried out in a study by Kopacz et al. (2011) using estimates from Geos-Chem simulations. The estimated BC_c from the present study at East Rongbuk Glacier (28^∘ N, 88^∘ E) for the month of October (Ming et al., 2009), at Kangwure (28.5^∘ N, 85.8^∘ E), and Qiangyong (28.3^∘ N, 90.3^∘ E) for the month of July (Xu et al., 2006) are respectively 13, 17, and 24 µg kg⁻¹, which are found to be lower than the respective observed values by 27 %, 23 %, and 44 %. Some of the above discrepancies are expected due to the comparison of the model estimates, which are monthly averaged (corresponding to the month of observation), with the respective measured values at stations which are mostly based on a single sample observation (Ming et al., 2009; Xu et al., 2006). The above bias is, however, within the range of uncertainty in BC_c as estimated in the present study. It is also noted that the bias in the estimated BC_c for the above stations as obtained from the Geos-Chem simulations (Kopacz et al., 2011) is respectively −155 %, −22 % (negative bias indicates the model value is higher than the observed), and 54 %, which is found to be higher than that from the present study. The lower estimated values of BC_c than the measured, specifically at Qiangyong, are, although, as expected in our study due to non-consideration of wet deposition in the estimation of BC_c. Nevertheless, the comparable values of the model estimates (within the range of uncertainty in BC_c) with the respective observation indicate the conformity with our consideration of the dry deposition as a governing removal process of atmospheric BC during the period of study over the region.

Figure 3(a) Glacierwise decadal trend (1961–2010, percent rate of increase with respect to 1961) of BC_c based on atmospheric BC concentration (µg m⁻³) modeled over the glacier using SPRINTARS, (b) estimated increase in annual glacier runoff (mm w.e. y⁻¹) shown as box-and-whisker plot and average yearly recession (m y⁻¹) for the nine glaciers, (c) simulated summer mean albedo and annual runoff depth of glaciers, and those of sensitivity analysis for (d) a single glacier hypsometry at different ERA-Interim grids and (e) different glacier hypsometries at a single ERA-Interim grid.

Download

We further examine the estimates of BC_c and the relative percentage SAR (%) for the HKH glaciers under study as presented in Fig. 2e and 2f, respectively. The pre-monsoon mean of BC_c and BC-induced SAR for the HKH glaciers is higher by 39 %–76 % than the winter mean. Among the glaciers, Pindari, Poting, Chorabari, and Gangotri, located over the northern HKH region near the Manora Peak, all exhibit specifically high values of BC_c (63–68 µg kg⁻¹) and BC-induced SAR (≈5 %).

We also evaluate the percentage rate of increase in BC_c estimates over the HKH glaciers from 1961 to 2010 (with respect to 1961) (Fig. 3a) estimated from the available long-term simulation of atmospheric BC concentration in SPRINTARS. While there is a relative increase of BC_c from 1961 to 2010 for all glaciers under study, the percentage increase for Pindari Glacier is 26 % compared to that for Zemu, Milam, and Bara Shigri being 130 %, 55 % and 50 %, respectively.

The decadal increase indicates the possibility of the largest relative impact due to BC_c on BC-induced glacial mass balance for the Zemu Glacier (over eastern Himalayas) under conditions of similar glacier hypsometry and climatic setting. Nevertheless, the BC_c for Pindari Glacier is about 7 times that for the Zemu Glacier and hence would be more vulnerable than Zemu to these impacts. A sensitivity analysis of the change in the glacial mass balance due to relative change in hypsometry and climatic setting is discussed in Sect. 3.2.2.

In order to identify the location of the hot spots over the HKH region, we also analyze the spatial distribution of pre-monsoon mean of BC_c (Fig. 2h) and SAR (Fig. 2i). The spatial distribution of pre-monsoon mean of snow density from the ECMWF is presented in Fig. 2g. The pre-monsoon mean of snow density is seen in the range 190–290 kg m⁻³ over the HKH glaciers. Please refer to Fig. 1b for the locations of the glaciers in the study region. High BC_c (40 to 125 µg kg⁻¹) and the BC-induced SAR (2 % to 10 %) over the HKH region are located between coordinates 26 to 33^∘ N and 70 to 90^∘ E, specifically over the grids around Manora Peak. A gradually increasing trend westwards from NCO-P and moving north towards Hanle over the snow-covered (Fig. 2) HKH region is observed in the spatial distribution of BC_c and SAR. The extent of the glaciers studied under this work (as listed in Fig. 1b) over the region of 78 to 27 to 34^∘ N and 88^∘ E thus allows us to locate the glaciers which are the most affected due to BC_c; this is examined in the next section. The ablation zone of Gangotri Glacier (located from the field measurements), which lies at approximately 30^∘93^′ N and 79^∘ E, is also found to be within the zone of predicted high BC_c values. High BC_c in the range of 55 to 100 µg kg⁻¹ in the ablation zone of Gangotri near the snout, Gomukh (source of the Bhagirathi River, one of the primary headstreams of the Ganges River), is estimated in the present study.

Table 5Sensitivity of annual runoff and runoff increase (ARI) to albedo change for the glaciers considered in this study, along with sensitivity of annual runoff to albedo due to change in climatic condition and glacier hypsometry.

¹ Slope of annual runoff (mm w.e. y⁻¹) to albedo from Fig. 3c. ² Slope of increase in annual runoff (mm w.e.) to BC-induced percentage SAR. ³ Slope of annual runoff (mm w.e. y⁻¹) to albedo from Fig. 3d, where Gangotri hypsometry implies that the calculation with the same hypsometry of Gangotri Glacier as a sensitivity test of different climatic conditions by using the single hypsometry. ⁴ Slope of annual runoff (mm w.e. y⁻¹) to albedo from Fig. 3e, where Gangotri grid implies that the calculation with the same ERA-Interim grid including Gangotri Glacier as a sensitivity test of different hypsometry by using the single ERA-Interim grid.

Download Print Version | Download XLSX

3.3 Source of origin of BC aerosols over HKH region

3.3.1 Contribution of BC emissions from nearby and far-off regions to BC aerosols

The relative distribution (%) of BC surface concentration (Fig. 4I) and BC-AOD (Fig. 4II) due to BC emissions originating from the nearby region, IGP (BC concentration or BC-AOD due to emissions from the IGP to the total BC concentration or BC-AOD) and the far-off region, AFWA during winter and pre-monsoon seasons over the HKH region is estimated. These estimations are done from region-tagged simulations in the GCM over the HKH region (please refer to Sect. 2.1). The spatial distribution (%) of winter and pre-monsoon mean of surface BC concentration (BC-AOD) due to emissions from the IGP is 70 % to 90 % (45 % to 80 %) of the total surface BC-AOD over the part of the HKH region south of 30^∘ N. This distribution decreases as we move towards the north (north of 30^∘ N), at higher latitudes in the Himalayan region. The spatial distribution (%) of surface BC-AOD due to emissions from AFWA increases towards the northern grids (north of 30^∘ N), with these being 25 % to 40 % (50 % to 60 %) in winter and 15 % to 25 % (40 % to 55 %) in pre-monsoon. It is found that while BC concentration and BC-AOD due to emissions from the IGP have a slightly higher value during pre-monsoon than winter, it is vice versa for those due to emissions from AFWA. These features are inferred likely due to emissions of biomass burning (crop waste and forest fires), which are more prominent during pre-monsoon over the IGP but are more prominent during winter over AFWA, as obtained from the information of fire count data from satellite-based measurements (Pohl et al., 2014; Verma et al., 2014). Estimates of pre-monsoon and winter means of surface BC concentration due to relative contribution (%) of emissions from OB, BF, and FF combustion from source-tagged (refer to Sect. 2.1) BC simulations in LMD-ZT GCM are also presented respectively in Fig. 5a to c and 5d to f. This figure indicates that while BC surface concentration during winter and pre-monsoon are influenced mostly due to biofuel emissions over part of the HKH region south of 30^∘ N (where there is a relatively higher contribution from emissions over the IGP), these are due to biomass burning over the north of 30^∘ N, thus corroborating the above inference. The high BC_c and BC-induced SAR over glaciers near the Manora Peak, over northern Himalayan region, are thus found to be mainly due to contribution of BF emissions from the IGP region.

Figure 4Spatial distribution of mean of (I) surface BC concentration and (II) BC-AOD for winter (a, c) and pre-monsoon (b, d) due to relative percentage contribution (%) of emissions from the IGP (a, b) and AFWA (c, d).

Download

Figure 5Spatial distribution of (a–c) pre-monsoon and (d–f) winter mean of surface BC concentration due to relative percentage contribution (%) of BC emissions from open burning (OB), biofuel (BF), and fossil fuel (FF) combustion.

Download

The present study is in agreement with the inference based on the source-tagging technique implemented in CAM5 (as also mentioned in Sect. 2.1) and that using an adjoint of the Geos-Chem model (Zhang et al., 2015; Kopacz et al., 2011) on the predominant contribution from biomass burning (BB) emissions to BC aerosols over the HTP. Although, unlike the previous studies (e.g., Kopacz et al., 2011; Zhang et al., 2015), the present study provides a more specific domain over the HKH region, e.g., south of 30^∘ N (covering the region of analysis of BC aerosols for Himalayas in Zhang et al., 2015), where a predominant contribution of BC emissions originating in IGP is evinced, corroborating that originating in SAS by Zhang et al. (2015). An enhanced contribution of emissions from the sub-Saharan Africa region over the Himalayas and central plateau during winter compared to that during pre-monsoon as inferred in Zhang et al. (2015) is also consistent with that in the present study. Further, for the region north of 30^∘ N within the HKH region under study (covering the region of analysis of BC aerosols for the central plateau in Zhang et al., 2015), the present study indicates that a predominating contribution to BC aerosols from OB emissions and those originating mainly in AFWA is in disagreement with Zhang et al. (2015), which once again indicates that BC aerosols from BB emissions originate mainly in SAS. This disagreement is expected due to (i) the tagged region of “SAS” in Zhang et al. (2015) also including parts of west Asia, unlike that of a more specific domain over the Indian subcontinent “IGP” in the present study and (ii) the fact that the source of BC is considered from the combined source sector of BF and OB as BB in Zhang et al. (2015), unlike that in the present study. While BC emissions from the BF combustion are predominant over the IGP region (part of SAS from Zhang et al., 2015), these from OB are over the African region specifically during the winter season. It is, however, noted that a considerable contribution of BC emissions from the African region over the Himalayas (Mount Everest), with this being larger than that from the Indian region during the winter month (January) compared to that during the pre-monsoon month (April), has been inferred based on the source analysis using an adjoint of the Geos-Chem model (Kopacz et al., 2011); this inference is in corroboration with the present study.