We use the Community Atmosphere Model version 5 (CAM5; Neale et al., 2012), which is the atmospheric component of the Community Earth System Model version 1 (CESM1) (Hurrell et al., 2013). It includes relatively comprehensive representations of aerosols and clouds, and mechanisms for their interactions with each other and with climate (Gettelman et al., 2010; Liu et al., 2012). CAM5 employs a modal aerosol module (MAM) to represent aerosols in multiple log-normally distributed modes, with internal mixing assumed for aerosol species within each individual mode, including a three-mode standard representation (MAM3) and a more complex seven-mode representation (MAM7). The major difference between MAM3 and MAM7 related to carbonaceous aerosols lies in the treatment of aging. In MAM3, BC and primary organic matter (POM) particles are emitted into the accumulation mode, which also contains highly hygroscopic species such as sulfate and sea salt, while in MAM7, BC and POM are emitted into a primary carbon mode, which contains no other species. BC is hydrophobic upon emission, and thus the hygroscopicity of the primary-carbon-mode particles depends on the assumed hygroscopicity for POM. As more hygroscopic species (e.g., H2SO4 and NH3) condense onto the primary-carbon-mode particles, the particles become more hygroscopic and are gradually transferred into the MAM7 accumulation mode. The rate of transfer is controlled by uncertain aging parameters and the availability of gas precursors (Liu et al., 2012). In the accumulation mode of both MAM3 and MAM7, BC is internally mixed with other more hygroscopic species and is thus subject to wet scavenging and removal processes. During the transport from sources to remote regions, aerosols are removed too efficiently in the default CAM5 (Liu et al., 2012). Recently, H. Wang et al. (2013) revised some key processes associated with aerosol wet removal and convective transport, which significantly improved the vertical distribution of aerosols and their transport to remote regions such as the Arctic.

To better characterize the sensitivity of BC spatial distributions to emission uncertainties, Wang et al. (2014) implemented a direct source-tagging method in CAM5, whereby BC emitted from a number of independent source regions and/or sectors can be tagged and explicitly tracked within a single model simulation. This approach provides the quantitative characterization of source–receptor relationships for BC in any receptor region without perturbing emissions from individual BC source regions or sectors. In this study, we apply the BC tagging technique to the accumulation-mode BC in the MAM3 treatment. BC particles emitted from 16 geographical BC source regions and two emissions sectors (i.e., biomass burning and biofuel emissions and fossil fuel emissions) in each of the regions are tagged and explicitly tracked. Instead of using the global emissions from all sectors for the original one BC mass mixing ratio variable, the 32 regional/sectoral emissions provide sources to the respective tagged BC mass mixing ratio variables that are all added to the accumulation mode, including both interstitial and cloud-borne states. All physical and dynamic tendencies (e.g., transport, dry and wet removal) are calculated explicitly for the tagged BC mass mixing ratio variables in the same way as the original single BC mass mixing ratio. Also, when aerosol optical properties are calculated, all of the tagged BC mass mixing ratios contribute to the volume-mean refractive index of the accumulation mode that is used in the radiation calculation.

In addition to the free-running mode, CAM5 can also be configured in an offline mode, in which temperature, wind, surface fluxes (heat, moisture, and momentum), and pressure are constrained to agree closely with observations, while clouds and aerosol are allowed to evolve freely (Rasch et al., 1997; Lamarque et al., 2012; Ma et al., 2013). In this study, we run the CAM5 model in the offline mode with the direct BC source-tagging capability, including the improved representation of convective transport and wet removal of aerosols. We use the NASA Modern Era Retrospective-Analysis for Research and Applications (MERRA) reanalysis data set (Rienecker et al., 2011), using a horizontal resolution of 1.9 ∘ × 2.5 ∘ and 56 vertical levels. The goal is to characterize the fate of BC emitted from various geographical regions and sectors, their transport pathways to the HTP, and their radiative forcing with seasonal variations. The simulation is performed for year 2001 with prescribed sea surface temperatures.

BC emission data sets have large uncertainties (e.g., Bond et al., 2013), and there are different inventories available for climate modeling. We use the present-day (i.e., year 2000) monthly mean emission inventories for BC provided by Lamarque et al. (2010). They were built for the climate model simulations in the Coupled Model Intercomparison Project Phase 5 (CMIP5) (Taylor et al., 2012) performed for the Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC). The AR5 BC emissions being used in our CAM5 simulation include monthly varying elevated open-fire emissions (injection altitude up to 6 km), and yearly constant surface emissions from shipping and from six sectors over land: agricultural waste burning, domestic, energy, industry, transportation, and waste treatment. These surface BC emissions sectors do not distinguish between biofuel and fossil fuel combustion. To prepare for the BC source sector tagging, we divide the total surface emissions into two broader sectors, biofuel and fossil fuel, by using the ratio of biofuel to biofuel plus fossil fuel at each model grid provided by Dentener et al. (2006). We then combine the biomass burning (open fire) emissions and surface biofuel emissions, hereafter referred to as the BB (biofuel and biomass) sector. The shipping emissions are combined with the fossil fuel emissions over land to form the FF (fossil fuel) sector. Note that emissions in the BB sector have seasonal variations (associated with the open-fire emissions) but the FF sector emissions used in this study have no seasonal variation at all.

The 16 geographical BC source regions (Fig. 1a) are defined using the definition of source/receptor regions by Work Plan (WP 2.1) of the Task Force on Hemispheric Transport of Air Pollution (http://iek8wikis.iek.fz-juelich.de/HTAPWiki/WP2.1). They are ARC (Arctic), NAM (North America), CAM (Central America), SAM (South America), EUR (Europe), NAF (North Africa), SAF (Sub-Saharan Africa), MDE (Middle East), CAS (Central Asia), SAS (South Asia), EAS (East Asia), SEA (South East Asia), PAN (Pacific, Australia and New Zealand), RBU (Russia, Belarus and Ukraine), HTP (Himalayas and Tibetan Plateau), and ROW (rest of world).

Figure 1b and Table S1 in the Supplement summarize the fractional contributions of BC emissions from the different source regions and sectors. The global annual mean BC emission rate is 7.78 Tg yr-1, with 56.2 % (sum of the red bars) from BB emissions (33.6 % from fires and 22.6 % from biofuel) and 43.8 % (sum of the blue bars) from FF emissions. The two largest contributors are BB emissions from SAF (about 20 %) and FF emissions from EAS (about 15 %), followed by BB emissions from SEA (7.7 %), EAS (6.4 %), SAS (6.2 %), and SAM (5.7 %), as well as EUR FF (6.4 %) emissions. The geographical distributions of BC annual mean emission fluxes from BB and FF sectors for year 2000 are shown in Fig. S1 (in the Supplement). The global annual and seasonal mean lifetimes of BC emitted from the tagged source regions and sectors are summarized in Table S2. On global average, BB BC has a longer lifetime than FF BC in all seasons, especially in boreal winter (6.9 vs. 3.1 days), due in part to higher open-fire emissions (in the BB sector) during local dry seasons. Another reason is that open-fire emissions have initial injection heights of up to 6 km, resulting in less removal below 6 km. The availability of co-emitted hygroscopic species that are internally mixed with BC in the accumulation mode of the MAM3 aerosol treatment also impacts the scavenging and wet removal rate of BC. This also in part explains the variability of BC lifetime among the different source regions and sectors. Regarding the seasonal cycle, BC emitted from the major source regions (e.g., SAF, EAS, SEA, SAS) has substantially lower lifetime in summer (JJA) than in the other seasons, likely due to relatively strong removal by the summer monsoon precipitation.

We use two metrics for quantifying source–receptor relationships and the sensitivity of BC in a receptor region to various sources following Wang et al. (2014), but we extend them to treat BB and FF sectors separately.

The fractional contribution of BB and FF emissions from source region i to a BC property in receptor region j (the entire HTP or a subset of it), CiBB or CiFF, is defined as CiBB=AiBBAj,CiFF=AiFFAj,

where AiBB and AiFF are a BC property (e.g., mass mixing ratio, column burden, or deposition flux) in/over receptor region j resulting from BB and FF emissions, respectively, in source region i, and Aj=∑i=1NAiBB+AiFF represents the total BC property in the receptor region from all source regions (N=16 in this study) and sectors (BB and FF). Note that for BC properties such as column burden, surface mixing ratio, and deposition flux, the tagging method in CAM5 explicitly calculates how much is due to emissions from each source region and sector.

There are seven remote sites that have surface measurements of seasonal BC aerosol concentrations available. The locations and elevations of the sites and the sampling time periods and observation methods are described in Table 1. Figure 2 shows the comparison of seasonal mean BC concentrations between observations and CAM5 results. Note that model results represent mean concentrations in the grid box that the sampling sites reside in and at the grid-mean elevation, which could deviate significantly from the sampling point near complex terrain. All sites have non-negligible amounts of BC in the near-surface air. The error bars indicate the intraseasonal and interannual variations if multi-year data were used for given season and site. However, the uncertainties of observed BC surface concentrations mainly originate from the large discrepancies between different measurement methods, the mixing of BC with other components (e.g., organic carbon and mineral dust) in the aerosol samples, and the sampling time and location (Bond et al., 2013; Petzold et al., 2013). BC surface concentrations over the various sites show strong seasonal variations, which are reasonably captured by the model. The modeled magnitude of BC concentrations has a good agreement with observations at some sites (e.g., Fig. 2b, d, g), but the model clearly overestimates BC at the Muztagh Ata site (Fig. 2a) and underestimates at the Lulang site (Fig. 2e). The large underestimation (about 1000 m; see Table 1) of the Muztagh Ata site elevation in the model, determined by the model grid resolution, could largely explain the overestimation of BC since BC concentrations have sharp decreases with height in this region. At the sites over the southern HTP (i.e., Hanle, Manora Peak, NCO-P, Lulang, and NCOS), the BC surface concentrations in the summer (JJA) are lower, mainly due to wet scavenging by more frequent precipitation and partly due to the minimal emissions from domestic heating and wildfires over the Himalaya foothills and Indo-Gangetic Plain (IGP) during the Indian summer monsoon season (Marinoni et al., 2010, 2013). Among all these sites, the largest BC surface concentrations occur at the Manora Peak site, which is closer to the major sources in South Asia, especially in the winter (DJF), when the model underestimates the concentrations by about 50 %. The high concentrations in winter at Manora Peak is mainly due to the dry winter monsoon conditions and increased transport of emissions from regional biomass burning, agricultural waste, and wood fuel burning from the IGP (e.g., Ram et al., 2010; Moorthy et al., 2013). The BC surface concentrations peak in the springtime (MAM) at Hanle, NCO-P, Lulang, and NCOS sites. This might be related to an increase in BB and/or FF emissions in the Indian Subcontinent, along with the higher regional boundary-layer top over the IGP during the springtime that may favor the transport of particles from the surface up to higher altitudes (e.g., Marinoni et al., 2010, 2013). Moreover, long-range transport of pollution emitted from distant regions like the Middle East, North Africa, or Europe (Marinoni et al., 2010) could further contribute to BC variability over the southern Himalayas, which will also be examined in this study. Part of the discrepancies between observations and model results can be attributed to the inherent difficulty in simulating the cloud/precipitation fields over the complex topography and subsequent wet removal of aerosols during the transport, but emission uncertainties (e.g., Bond et al., 2013) might play a primary role.

It is important to evaluate the performance of model in simulating seasonal snow over this region in order to assess the importance of BC-in-snow effect. Figure 3 shows the CAM5 simulated seasonal and annual mean snow cover fraction (SCF) during year 2001, in comparison to observed mean SCF, derived from the Moderate Resolution Imaging Spectrometer (MODIS) (Hall et al., 2006) monthly mean of daily products at 0.05 ∘ resolution. For a better comparison, the MODIS monthly mean SCFs are mapped to the CAM5 grid. The summer (JJA) season only includes July and August for both CAM5 and MODIS due to missing MODIS data in June 2001. To illustrate whether year 2001 can represent the average condition in terms of SCF, the MODIS SCF climatology (2000–2013) is also plotted. The overall SCF in HTP has very small difference between climatology and year 2001 in all seasons except for JJA, when SCF is notably higher over the northwest plateau for the climatology that included June SCF in the average. On average, SCF is about 5 % (absolute amount) higher in June than in July and August. Over the 52 HTP grid cells, the CAM5 SCF is highly correlated spatially with that of MODIS (for both 2001 and 2000–2013) with the statistical confidence level greater than 99 %, except for summer (JJA), when the linear correlation is significant only at 80 % level.

There are strong spatial and seasonal variations in SCF due to the complex terrain and seasonal variation in snowfall and melting. The SCF over the entire HTP reaches the maximum in the winter (DJF), while it decreases to almost none (less than 5 %) in July and August. Snow covers the western and southeastern plateau during the transition seasons (MAM and SON). The CAM5 simulation shows a good agreement with MODIS in the annual mean (ANN) SCF and the strong seasonality. The most persistent snow cover at the southern and western edges of the HTP and the relatively less persistent in the HTP interior are captured by the CAM5 model. The performance of the CAM5 has been improved, in comparison to its earlier version (CAM3) that remarkably overestimated the SCF especially over the HTP interior (Qian et al., 2011), although the CAM5 still significantly overestimates the SCF in the western plateau in DJF and MAM and underestimates it in JJA. The CAM3 model used by Qian et al. (2011) overestimates SCF by up to a factor of 2 during the cold season (November to April). The CAM3 spring (MAM) mean SCF is greater than 35 %, while the CAM5 spring mean (21 %) in the present study is in good agreement with the MODIS spring SCF (18 ± 5 %).

Although we believe that the CAM5 SCF biases are qualitatively robust, it is worth noting that the MODIS products have uncertainties as well. Pu et al. (2007) evaluated the MODIS SCF products over the HTP against ground-based snow observations and showed that total error in MODIS SCF products over the HTP is about 10 %. However, their analysis based on MODIS 8-day snow cover composite gave a significantly higher SCF (more than 10 %) than the one we show here using daily products, especially in winter and early spring. Interestingly, based on a different source of observation, Qin et al. (2006) found that snow covers about 59 % of the Tibetan Plateau in winter, which is comparable to the mean SCF (50 %) in our CAM5 simulation. Nonetheless, we keep this discrepancy in mind when interpreting the wintertime BC-in-snow radiative forcing that suffers the most from such potential SCF bias.

The direct source-tagging method can clearly characterize the three-dimensional transport pathways of BC emitted from various source regions and sectors to the HTP receptor region. General circulation patterns over the HTP and surroundings are typically affected by midlatitude westerlies in the winter and Asian monsoon in the summer, including the South Asian summer monsoon and East Asian summer monsoon (Xu et al., 2009; Yao et al., 2012; Wu et al., 2012; also see Fig. S2).

Figure 4 illustrates circulation patterns over HTP and BC transport pathways from six major source regions to the HTP in the winter (DJF) and summer (JJA). (See similar plots in Figs. S3–S5 for other tagged source regions.) In the winter, the strong surface cooling over the HTP leads to subsidence/divergence and the formation of an enhanced local circulation cell, while in the summer air converges toward the HTP from the surroundings, particularly from South Asia, due to the ascending of strongly heated air over the HTP (e.g., Wu et al., 2012), as also indicated by the arrows in the vertical cross sections in Fig. 4. In the winter, the subtropical westerlies extend to about 10∘ N in middle/upper troposphere and 20∘ N near the surface, and the tropical easterlies are weak (see the white contours of latitude–height cross-section panels in Fig. 4). The circulation patterns near the HTP change dramatically during the summer monsoon season. The reversal of surface wind regime in the tropics (e.g., Arabian Sea, Bay of Bengal, and South China Sea) is characteristic of the Asian summer monsoon climate (see Fig. S2e, g). The subtropical westerlies recede to north of 30∘ N and the center of the westerly jet shifts to about 40∘ N in JJA (from about 30∘ N in DJF). The strong easterlies characterize the upper troposphere of tropical region (south of HTP), while the southwesterly flow prevails in the lower troposphere (white contours of latitude–height cross-section panels in Fig. 4). The prevailing winds during the transition seasons (MAM and SON) between DJF and JJA are still westerlies (Fig. S2b, d).

The circulation patterns determine the transport of BC around the HTP region. However, the variations in spatial distributions of BC emitted from the different source regions and in different seasons could be due to the differences in source location and strength, wet removal rate, and lifting. Note that although we combined BC emitted from BB and FF sections to characterize transport pathways in Fig. 4, only BC emissions from the BB sector have seasonal variations in the emission inventory we use.

The HTP region is surrounded by two major BC source regions, SAS and EAS (Fig. 1a), which potentially have great impact on BC in the HTP (e.g., Menon et al., 2010; Bond et al., 2007; Ohara et al., 2007; Xu et al., 2009; Kopacz et al., 2011; Lu et al., 2012). BC emissions from SAS are dominated by the BB sector, and by FF sector from EAS (Fig. 1b). As shown in Fig. 4, in the winter, a significant amount of BC from SAS can be transported to the eastern plateau by the strong westerlies under the dry winter monsoon conditions. During the South Asian summer monsoon BC from SAS is effectively removed by the local abundant precipitation, as indicated by the low lifetime in summer (Table S2), but can still affect large area in the southwest of the HTP. However, BC from EAS can be uplifted higher and transported more to the northeast plateau in the summer monsoon season than in the winter. Along the wintertime westerlies, BC from upwind source regions (e.g., EUR, NAF, SAF, MDE, and CAS; see Fig. S3) can easily move to the HTP, while the HTP local emissions are transported far away (Fig. 4). BC originating from the distant sources such as SAF and MDE reaches up high (to 300hPa) in the HTP. In the summer, continental deep convection can loft BC into higher altitudes, where it can be transported to the HTP along the relatively weaker westerlies from upwind source regions (e.g., EUR, RBU, MDE, and CAS; see Figs. 4 and S3). However, BC from distant low-latitude source regions such as SAF barely reaches the HTP region due to weak emissions but strong removal along the transport pathways to the HTP during the summer monsoon season.

BC concentrations in the HTP have strong dependence on season and location. Figure 5 shows the annual mean and seasonal variations in BC column burden and deposition rate over the HTP and five subregions. The seasonal variation in the ratio of wet to total BC deposition is superimposed. The central plateau is the cleanest region during all seasons, compared to other subregions in the HTP (Fig. 5e). Both BC column burden and deposition rate from the BB sector peak in MAM over the HTP, mostly in the Himalayas and southeast plateau region. The FF BC burden in the HTP peaks in the summer mainly due to the seasonal maximum over the northwest and northeast plateaus. However, BC wet removal rate over the northwest plateau is at minimum in the summer, as opposed to the summer maximum in other subregions and the entire HTP region. For the Himalayas and the southeast and central plateau, the seasonal variation (i.e., maximum in MAM followed by a sharp decrease to JJA) in BB and FF column burden (Fig. 5c, d and e) is similar to the variation in observed surface concentrations at sites located in these subregions (Fig. 2b, d, e and f). In the Himalayas and the southeast plateau, the ratio of regional mean BC column burden to deposition rate (Fig. 5c and d), an indicator of removal timescale or lifetime, is the smallest (less than 1 day) during the Asian summer monsoon (JJA) due to the efficient wet scavenging of BC by abundant precipitation. In northwest and northeast plateau, the BC column burden increases from DJF to JJA and reaches the maximum in JJA, and then decreases in SON (Fig. 5b and f), partly due to the peak contribution of EAS and CAS emissions in JJA. This trend is also similar to that in the observed surface concentrations (Fig. 2a and g). The deposition rate follows the same seasonal variation in column burden over the northeast plateau, while the deposition has a minimum in JJA over the northwest plateau when the column burden is at maximum, likely due to the less efficient wet removal in this region (Fig. 5b and f).

The annual mean BC column burden over the HTP has almost the same contributions from BB and FF emission origins, with BB dominating in DJF and MAM and FF in JJA. In the Himalayas, BC is predominantly from the BB sector for all seasons (Fig. 5c). In the southeast and central plateaus, the dominant source sector is BB in DJF and MAM, but FF dominates in JJA. The dominant source sector over the northwest and northeast plateaus is always FF, especially in the summer. We need to analyze the source–receptor relationships in order to quantify the roles of BB and FF emissions from the various source regions in determining BC over the HTP and the subregions.

Previous studies (e.g., Xu et al., 2009; Kopacz et al., 2011) have shown that BC and its source–receptor relationships vary significantly with season and location in the HTP. We intend to quantify source contributions to BC at different locations of the HTP and in different seasons. Our analysis also shows that the relative contributions to BC from different source regions and sectors depend on season and location in the HTP. As shown in Fig. 6, the largest contribution to the annual mean BC burden and surface deposition for the entire HTP region is from BB emissions from SAS, followed by FF emissions from SAS and then the FF from EAS. The same roles hold for all the seasonal means except for the summer (JJA), when the EAS FF becomes more important for BC column burden in the HTP and, to a lesser extent, for deposition.

The SAS emissions account for 50 % of the annual mean burden over the HTP, including 33 % from BB and 17 % from FF. The other 50 % is mostly from the EAS (5 % BB and 14 % FF), HTP (6 % BB and 6 % FF), CAS FF (4 %), MDE FF (4 %), and SAF BB (3 %). The source attribution for annual mean BC deposition for the entire HTP is similar, but SAS contributes even more to BC deposition than to the column burden. Although RBU has a lower contribution to the annual mean BC in the HTP than the six regions shown in Fig. 6, its contribution to the JJA mean, especially at some locations, is quite substantial (included in the black bar) and even more important than some of the six regions in Fig. 6, as discussed in detail below.

BC annual mean burden over the HTP has nearly equal contributions from BB and FF emissions. However, the contribution by BB emissions, mainly from SAS, is larger than from the FF sector in DJF and MAM. In the summer (JJA), the largest contribution (about 29 %) to HTP BC is from EAS FF emissions. This is partly due to the change in circulation patterns (Fig. 4) and effective wet removal of SAS BB emissions. Note that EAS FF emissions are much larger than FF emissions from any other of the source regions, and are more than twice the EAS BB emissions.

For BC in the five finer receptor regions of interest (as defined in Fig. 5g), SAS BB and FF have the largest contribution to BC in the Himalayas and central plateau, while EAS FF and BB contribute the most to the northeast plateau in all seasons and southeast plateau in the summer. Central Asia and Middle East FF emissions have relatively more important contribution to BC reaching the northwest plateau, especially in the summer.

For the northwest plateau, the prevailing winds in this subregion are westerly throughout the year (Fig. S2; Cao et al., 2009; Xu et al., 2009), so the important source regions should be located at the west of the HTP (e.g., MDE, EUR, and parts of SAS and RBU in Fig. 1a). As shown in Fig. 6, SAS emissions are still the dominant source for the annual BC burden in this subregion (17 % from BB and 14 % from FF), followed by HTP local emissions (9 % from BB and 13 % from FF), CAS (2 % from BB and 14 % from FF), MDE FF (8 %), and EAS FF (7 %). BC emissions from SAS are the dominant source in DJF, MAM, and SON. CAS becomes the dominant source region (5 % from BB and 26 % from FF) in JJA, even though CAS is not a significant emission source region on a global basis (Fig. 1b). BC emitted from MDE is predominantly in the FF sector throughout the year. Emissions from the rest of the tagged sources (in addition to the six sources listed; black bar in Fig. 6) become more significant in this subregion, mostly from EUR and RBU through long-range transport (Fig. S3). The source attribution for BC deposition in this subregion is similar to that of the column burden, but BC emitted from SAS and MDE appears to be more efficient in deposition, except for the JJA season, when BC from EAS contributes more to deposition than to column burden.

The Himalayas subregion located along the southern edge of the HTP is in close proximity to the SAS. Thus emissions from SAS are absolutely the dominant source for BC in the Himalayas throughout the year. This subregion receives more BC from the BB sector than FF because BC emissions in SAS are mainly in the BB sector, especially in the MAM season (Figs. 1b and S6). For the annual mean burden, BC from SAS contributes 81 % (54 % from BB and 27 % from FF), followed by HTP local emissions (6 % from BB and 3 % from FF). It is worth noting that SAF BB emissions contribute about 10 % to burden in DJF through long-range transport. BC deposition in this subregion also predominantly originates from SAS, which is consistent with previous studies by Ming et al. (2008) and Kopacz et al. (2011).

For the southeast plateau (Fig. 6), the BC source contribution profile is similar to that of Himalayas during DJF and MAM season, in which SAS is still the dominant source, especially in MAM (74 % contribution to column burden, including 53 % from BB and 21 % from FF), although the contribution from EAS is larger here than for the Himalayas. As also pointed by Ramanathan et al. (2007), BC over the SAS can be transported to the southeast plateau by the southern branch of the westerlies during the winter and spring. However, the BC source contribution profile changes dramatically during the summer, when emissions in EAS become the dominant source to this subregion (68 % to column burden, including 23 % from BB and 45 % from FF). Kopacz et al. (2011) also found that the BC from southeastern China is the dominant contributor to the southeast plateau in July. For the annual mean burden in this subregion, SAS is still the dominant contributor (40 % from BB and 17 % from FF), followed by EAS (8 % from BB and 15 % from FF), and HTP (6 % from BB and 6 % from FF). BC originating from EAS contributes more to deposition than to burden in this subregion.

For the central plateau (Fig. 6), source attribution profiles for annual and seasonal BC are very similar to those of the entire HTP region, with SAS being the dominant source region throughout the year, except that EAS has comparable contributions in JJA. Ming et al. (2010) pointed out that pollutants from the Indo-Gangetic Basin could be transported to the central plateau by both the summer monsoon and the westerlies. Xia et al. (2011) also found that the substantial regional atmospheric brown haze from the nearby regions of SAS is the main source for the background aerosols in the central plateau based on sun photometer and satellite observations.

Compared to the other subregions, the northeast plateau receives the largest contribution of BC from EAS throughout the year (50 % to annual mean burden, including 12 % from BB and 38 % from FF; see Fig. 6), especially in JJA (17 % from BB and 49 % from FF). The EAS FF sector contribution and the magnitude of burden (Fig. 5f) over the northeast plateau have strong seasonal variations, mostly due to variations in meteorology, because the FF emissions in our simulation do not vary seasonally. Kopacz et al. (2011) indicate that the primary contribution to BC over the northeast plateau is from western China during January and April (transported by midtropospheric westerlies), and from central-eastern China during July and October (transported by boundary-layer flow). Other main contributions to BC burden over the northeast plateau include 13 % from SAS and 10 % from HTP local emissions. Similar to the northwest plateau, some other upwind source regions (e.g., CAS, MDE, RBU, and EUR; see Figs. 4 and S3) have a significant contribution to the northeast subregion as well.

Following Wang et al. (2014), we defined the “efficiency” metric in Sect. 2.2 to quantify the sensitivity of BC response to absolute change (e.g., per unit perturbation) of emissions in different source regions. This metric has the value of 1 if the entire globe is treated as a single source region, so we may assume the global mean efficiency of 1 as a reference to measure the sensitivity to perturbation from different source regions/sectors.

Figure 7 shows efficiencies of tagged sources in affecting the BC seasonal and annual mean column burden and deposition in the HTP and five subregions (as defined in Fig. 5g). BC in the same receptor regions is generally most sensitive to change in local emissions, regardless of seasons, emission sectors, and locations of receptor regions. Among all the source regions, although the HTP local (FF + BB) emissions only contribute about 10 %, BC in the HTP is extremely sensitive to changes in the emissions within HTP (not shown in the figure), mainly because the emission rate is very low. In addition to the local emissions, BC in the HTP is also sensitive to emissions in neighboring source regions (e.g., SAS and CAS) and emissions from distant sources such as MDE. The SAS has a large contribution to BC burden and deposition over the HTP, as well as most of the subregions except for the northeast plateau (receptor V); the efficiencies for SAS emissions are also high for almost all of the subregions especially the Himalayas (receptor II). BC in the northeast plateau (receptor V) is quite sensitive to EAS emissions in all seasons, while BC in the southeast plateau is sensitive to EAS emissions in JJA and SON. Although BC emissions from MDE and CAS are weak (Fig. 1b) and their contributions to the HTP are relatively low, their efficiencies are high. BC over the northwest plateau (receptor I) and central plateau (receptor IV) is extremely sensitive to emissions from CAS in JJA. These source–receptor relationships of sensitivity will provide useful information for policymakers to improve the effective mitigation road map in order to potentially slow down the glacier retreat in the HTP region.