Black carbon (BC) deposited on snow and ice accelerates glacier melting and contributes to climate change of the Himalayas and Tibetan Plateau (HTP). Taking into account emissions, hydrophilic-to-hydrophobic conversion, and removal processes of BC, a novel back-trajectory approach is developed to study the origin of BC reaching the HTP. The results indicate that BC received by the HTP increased by 41% from 1996 to 2010, implying that the BC problem is accelerating in the HTP region. South Asia and East Asia are the main source regions, accounting for 67% and 17% of BC transported to the HTP on an annual basis, followed by Former USSR (∼8%), Middle East (∼4%), Europe (∼2%), and Northern Africa (∼1%). BC reaching the HTP is high in winter and low in summer, and the relative contributions of different source regions vary with seasons. We show the seasonal spatial distribution of BC sources directly on a 0.5° × 0.5° grid, which provides information to policymakers about the best target areas for mitigating the climate changes and other effects on the HTP.
 Black carbon (BC) in the atmosphere is mainly emitted from the incomplete combustion of fossil fuels and biomass burning. In addition to the adverse effects on public health and air quality, BC can alter the Earth's energy budget by strongly absorbing solar radiation (direct effect), acting as a cloud condensation nucleus (CCN) to modify the microphysical properties of clouds (indirect effect), and heating the troposphere that in turn affects the cloud formation and lifetime (semi-direct effect) [Ramanathan and Carmichael, 2008]. Furthermore, BC particles deposited on snow and ice can reduce the surface albedo and consequently accelerate the melting of glaciers [Flanner et al., 2009; Menon et al., 2010; Xu et al., 2009a]. These BC effects may be more pronounced for the Himalayas and Tibetan Plateau (HTP, also known as Earth's “third pole”) than for other regions, because it is surrounded by the world's two largest BC generating regions, South Asia and East Asia (∼40% of global emissions [Lamarque et al., 2010]). It is reported that the warming rate of the HTP is twice the global average [Xu et al., 2009a] and that the glaciers of the HTP have been melting at an accelerating rate since the mid-1990s [Lau et al., 2010]. BC transported to the HTP (both in the air and deposited at the surface) is identified as a major reason, together with the increasing global greenhouse gases, for the rapid climate change and glacier retreat of the HTP [Menon et al., 2010; Xu et al., 2009a] and has been shown to influence the weather, hydrological cycles, and ecosystems at regional and global scale [Menon et al., 2002, 2010]. Therefore, identifying the origin of BC on the HTP is of great importance.
 Due to scarce observing networks and limited long-term measurement data, the origin of BC on the HTP is insufficiently studied. The traditional back-trajectory approach only identifies the possible source regions by tracking air mass flow and cannot give any quantitative results [Ming et al., 2008, 2009]. Recently, using a GEOS-Chem (Goddard Earth Observing System-Chemistry) adjoint model,Kopacz et al. provided for the first time the spatial extent of BC source regions that impact the HTP seasonally. However, they only presented the results for four representative locations on the HTP in a single year (i.e., 2001), due to the heavy computational requirements of the complex chemical transport model, the absence of reliable and consistent emissions data for multiple years, and the fact that the origin of emissions is not accurately attributed to particular countries because of the relatively coarse model resolution (i.e., 2° × 2.5°). In this work, we develop a novel back-trajectory approach that takes into account emissions, hydrophilic-to-hydrophobic conversion, and removal processes of BC to study the origin of BC on the HTP. The annual trend and seasonality of BC transported to the HTP are shown for the first time from 1996 to 2010. Furthermore, the relative contribution of BC emissions from different areas to the BC on the HTP is quantified on a 0.5° × 0.5° grid by season.
2.1. Back-Trajectory Calculations
 The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model version 4.9 is used to investigate the transport pathways of BC to the HTP during 1996–2010 [Draxler and Hess, 1998]. Different from most previous studies, we applied the back-trajectory analysis not to a single receptor location, but to the whole area of the HTP.Figure 1bshows the boundary of the HTP on a 1° × 1° grid, and the centers of the grids are selected as receptors (367 points in total). Seven-day back-trajectories arriving at each receptor were initialized four times daily at 00:00, 06:00, 12:00, and 18:00 coordinated universal time (UTC). The trajectories were run for seven days, because the atmospheric lifetime of BC is about one week [Cooke et al., 1999; Reddy and Boucher, 2007]. The arrival height is 500 m above ground level which is within the typical height of the planetary boundary layer (PBL) over the HTP [Ram et al., 2010]. The model was driven by the 3-D meteorological fields of NCEP GDAS (3 h temporal resolution, 1° × 1°, 23 vertical levels) for the period of 2005–2010, and NCEP/NCAR reanalysis data (6 h temporal resolution, 2.5° × 2.5°, 18 vertical levels) for 1996–2004. Typically, the individual trajectory contains a position error of 20% [Stohl, 1998]. In this study, however, the statistical uncertainties can be reduced, to some extent, by using the highest resolution meteorological fields (NCEP GDAS) after 2004 [Stohl, 1998], computing trajectories for the whole of the HTP region, analyzing a large set of trajectories (∼535820 trajectories for a year) [Kahl, 1990], and using transport efficiency density maps (see Section 3) instead of individual trajectories to describe the overall transport characteristics of BC to the HTP.
2.2. BC Emissions
 A global monthly BC emission inventory at a resolution 0.5° × 0.5° is built for the period 1996–2010 to support trajectory analysis. The emission sources are categorized into nine sectors: power generation, industry, residential, land transport, international shipping, aviation, agricultural waste burning, open forests burning, and open savanna/grasslands burning. For China and India, the world's two largest BC generating countries, monthly anthropogenic emissions are directly taken from our recent inventory, which was developed specifically for these two countries using a detailed technology-based methodology [Lu et al., 2011]. For the other regions, annual anthropogenic gridded emissions are from the historical and RCP 4.5 emissions of the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5) [Lamarque et al., 2010]. Due to the intensive use of heating stoves in winter, we consider the residential emissions in winter (or summer) over northern Europe (latitude > 45° N) to be 30% higher (or lower) than the annual average value based on the recommendation of the Global Emissions Initiative (GEIA) [Chin et al., 2009]. For the other anthropogenic sectors in other regions, no seasonality is presumed due to the lack of information. For the open biomass burning of forests and grasslands, we use emissions from the Global Fire Emissions Database (GFED) version 3.1, which provides a global assessment of the different types of fire emissions with a monthly time step [van der Werf et al., 2010].
2.3. Removal Processes
 The removal of the emitted BC from the atmosphere is largely dependent on the dry and wet deposition, as well as the hydrophilic-to-hydrophobic aging of BC aerosols. In this study, we follow the emission and transformation scheme developed byCooke et al. , assuming that: 80% of fresh emitted BC aerosols are hydrophobic and the rest are hydrophilic; the hydrophobic aerosols become hydrophilic with an e-folding time of 1.15 days (i.e., first-order conversion rate,kc, of 1.01 × 10−5 s−1); and wet deposition is applied only to the hydrophilic part of the aerosol. Comparing to wet deposition, dry deposition is small [Reddy and Boucher, 2007]. Therefore, to simplify the calculation, we assume that the dry deposition is subject to exponential decay with a first-order rate (kd) of 4.25 × 10−7 s−1, corresponding to an e-lifetime of about 27.2 days [Reddy and Boucher, 2007]. Based on the scheme used by Liu et al. , the wet deposition related parameters can be calculated from the hydrological data in the GEOS meteorological archive. In this work, monthly 3-D global wet deposition rate (kw, s−1) fields were generated from a 15-year run of the GEOS-Chem model (version v9-01-01, http://acmg.seas.harvard.edu/geos/) with default emissions at a horizontal resolution of 2° × 2.5°. Wet deposition includes contributions from rainout and washout of both convective and large-scale precipitation. The model was driven by the 3-D assimilated meteorological fields from the GEOS-5 (47 vertical levels) for the period of 2005–2010, and GEOS-4 (30 vertical levels) for 1996–2004. It should be noted that the indirect and semi-direct effects of the aerosols on cloud formation, properties, and lifetime, and consequently wet scavenging are beyond the scope of this work, and are not considered.
2.4. Effective Emission Intensity (EEI)
 Imagine that, following a trajectory l, freshly emitted BC aerosols in a spatial grid box b will pass through a series of boxes in sequence (i.e., b1, b2, …) to a receptor at the HTP (i.e., bn). Then the change of the fraction of hydrophobic and hydrophilic BC (FHPO and FHPI) in box bi with time (t) can be expressed as
where kc, kd, and kw,iare the first-order conversion, dry deposition, and wet deposition rates in boxbi, respectively (see Section 2.3). Equation (1) is integrated to give
where FHPO,i and FHPI,i are the remaining fractions of hydrophobic and hydrophilic BC after passing through box bi, respectively, and ti (in s) is the residence time of BC aerosols in the box bi. The initial conditions of equation (2) are FHPO,0 = 0.8 and FHPI,0 = 0.2, derived from the scheme developed by Cooke et al. . Meanwhile, the fraction of BC deposited to the ground surface below box bi (Fdep,i) can be determined by
 Combining equations (2) and (3), the fraction of BC transported to the HTP can be calculated, and we define it as the BC transport efficiency (TE) of trajectory l at box b to the HTP, i.e.,
where j is the index of boxes that are above the area of the HTP. The first term on the right side of equation (4) represents the proportion of BC transported to the receptor site, while the second term represents the deposition proportion at the HTP surface during transportation.
 Theoretically, TE represents the transport ability of BC from the source region to the HTP. Therefore, in conjunction with gridded emissions, TE can be treated as a weighting factor to identify the major BC source regions that impact the HTP. For a surface grid h in month g, we define the effective emission intensity (EEI, same units as emissions) as
where E represents the BC emissions; mh,g is the total number of trajectories passing through the atmospheric column above the grid h in month g; and TNTgis the total number of back-trajectories originating at the HTP in monthg (i.e., 44,040 for a month with 30 days). The second term on the right side of equation (5) can be considered as the monthly TE density of grid h (Figure 2, middle), which ensures that EEIh,g < Eh,g. EEI is a quantity that links BC source regions with receptor regions, taking into account the magnitude of emissions, transformation, transport, and removal processes of BC in the atmosphere. A high value of EEI for a surface grid implies that this grid has high BC emissions and/or has many high TE trajectories passing above it, and consequently has a big effect on the HTP. Hence, EEI values represent transport weighted BC emissions, and the relative values among regions, sectors, years, months, etc., will provide valuable information on the sources, trends, and seasonal variations of BC transported to the HTP. These will be discussed in detail in the next section.
3. Results and Discussion
 The average BC emissions by season during 1996–2010 are shown in Figure 2 (left). Globally, anthropogenic BC emissions increased by 17% from 5.66 Tg in 1996 to 6.61 Tg in 2010. Emissions from open biomass burning are estimated in the range of 1.55–2.88 Tg/yr, depending on the year. East and South Asia are the major emitting regions, accounting for 49 ± 3% of the global anthropogenic emissions. China and India are the dominant emitting countries in these two regions. Driven by the rapid growth of population, energy consumption, and industrial production, BC emissions in China and India increased by 21% and 41% to 1.85 Tg and 1.02 Tg, respectively, from 1996 to 2010 [Lu et al., 2011]. For the other regions which have a potential effect on the HTP, BC emissions were either relatively constant (Former USSR, ∼0.51 Tg/yr; Middle East, ∼0.14 Tg/yr; Northern Africa, ∼0.10 Tg/yr) or showed a decreasing trend (Europe, 0.51 Tg/yr to 0.39 Tg/yr) during the study period.
 The average BC TE density maps shown in Figure 2(middle) give the mean BC transport characteristics from the meteorological point of view. Typically, the climate of the HTP is characterized by the influence of the mid-latitude westerlies in winter and South Asian monsoon (SAM) in summer. Due to the dry weather conditions and high speed of westerlies in winter, BC aerosols can be transported long distances from upwind areas (e.g., Europe, Northern Africa, Former USSR, Middle East, and north part of South Asia) to the HTP. The atmospheric circulation is changed during the SAM summer season. The Arabian Sea Branch and the Bay of Bengal Branch – two sub-systems of the SAM – make the air converge toward the northern and eastern HTP, and bring BC emissions from South Asia, the western Indo-China peninsula, and southwestern China to the HTP. Meanwhile, the SAM also brings abundant precipitation, which removes BC effectively from the atmosphere. For this reason, the TE density map has a smaller affected area in summer than in winter. Spring and autumn are transition seasons between winter and summer, and their prevailing winds are still westerlies.
 Showing significant monthly variations, the amount of BC transported to the HTP (hereinafter quantified in terms of EEI) is high in winter and low in summer (Figures 3b and 3d). This is consistent with the seasonal variability of BC content observed in the atmosphere [Ram et al., 2010], surface snow [Ming et al., 2009], and ice cores [Xu et al., 2009a, 2009b] at different locations of the HTP. Except for the effective wet scavenging of BC during the summer monsoon season, the EEI seasonal variation can also be attributed to the seasonality of BC emissions. For example, due to the increased residential heating needs and the carryover effect of industrial production at the end of the year, BC emissions in China are higher in winter and lower in summer [Lu et al., 2011].
 The spatial origin of BC by season is shown in Figure 2(right). Northern South Asia (e.g., the Indo-Gangetic Plain) and southwestern China (e.g., the Sichuan Basin) are identified as two hotspots for EEI in nearly all seasons, reflecting the intensive BC emissions and high transport efficiencies for these regions. BC sources that influence the HTP vary with seasons (Figure 3b). In January, South Asia is the dominant source of BC transported to the HTP (70%), followed by East Asia (9%), Former USSR (8%), Middle East (6%), Europe (5%), and Northern Africa (2%); whereas in July, only South Asia (56%), East Asia (36%), and Former USSR (6%) are major EEI sources. India and China are the two largest contributing countries, accounting for 50% and 9% (or 40% and 34%) of BC transported to the HTP in winter (or summer), respectively. Although annual BC emission in China is at least 80% higher than that in India [Lu et al., 2011], its contribution to BC on the HTP is smaller. The main reason is that the high BC flux areas (eastern and central China) are mostly downwind of the HTP and only the SAM in summer can effectively bring BC emissions from China to the HTP. However, emissions from India can be transported to the HTP throughout the year (via SAM during summer and westerlies during other times). The conclusion that most of BC on the HTP is from South Asia is also consistent with recent studies. For example, using a back-trajectory method,Ming et al. [2008, 2009] pointed out that BC emissions from South Asia dominate the BC deposition on the southern slope of the HTP. Based on sunphotometer and satellite observations, Xia et al.  found that background aerosols in the central HTP are mainly from South Asia. Menon et al.  estimated that Indian BC emissions alone contribute ∼36% of the snow/ice cover decrease since 1990.
 The interannual variation of EEI during 1996–2010 is shown in Figures 3a and 3c. BC transported to the HTP increased by 41% from 1996 to 2010 with a linear annual growth rate of 2.5%. This is in line with the increasing BC deposition observed in ice cores [Ming et al., 2008; Xu et al., 2009a, 2009b] and accelerating warming and glacier melting over the HTP [Menon et al., 2010] since the mid-1990s. On average, South Asia and East Asia are responsible for about 67% and 17% of the BC transported to the HTP, respectively, followed by Former USSR (∼8%), Middle East (∼4%), Europe (∼2%), and Northern Africa (∼1%). The increasing BC emissions in India and China contribute 47% of the EEI increase. Examining the sectoral distributions, the contributions of residential, industry, land transportation, and agricultural waste burning to the EEI are 60 ± 5%, 17 ± 3%, 15 ± 5%, and 6 ± 3%, respectively (Figures 3c and 3d). Open forest burning has significant interannual and seasonal variability. It accounted for 11% of the BC transported to the HTP in 1999 when extensive forest burning occurred.
Figure 4 shows how our results compare with a recent study conducted by Kopacz et al. , who employed a GEOS-Chem adjoint model to estimate the amount of BC reaching the atmospheric column above four representative sites on the HTP in 2001. This was the first and so far the only study to identify the magnitude of contributions from different BC source regions to the HTP. We extract our EEI values that contribute to the BC at the four sites, and convert the units to Mg/day for 2° × 2.5° grids, corresponding to the unit used byKopacz et al. . As shown in Figure 4, good agreement is found for each month, as well as for the annual average and all data points (R > 0.79). It is not surprising that EEI values are about two orders of magnitude lower than Kopacz et al.'s estimates, because EEI is correlated with TNT, as indicated in equation (5). This further confirms that EEI is a valid term to represent the amount of BC reaching the receptor to some extent, and the relative EEI values reflect the relative importance of source regions. It is also worth noting that our estimates are higher in January and lower in July than the GEOS-Chem results. This is probably because we consider seasonal variability in BC emissions for India, China, and northern Europe, whereasKopacz et al. did not.
 In this work, we use an improved back-trajectory approach to study the origin of BC on the HTP. Because we study a 15-year time period, we are able to reliably quantify the increasing trend of BC reaching the HTP. The results also provide an improved scientific understanding of the seasonality and interannual variation of BC transported to the HTP during 1996–2010. More importantly, the major source regions of BC are directly identified, and this information should benefit policymakers when planning future mitigation strategies to slow glacier melting and mitigate other climate effects in the region. Furthermore, the method described here provides a novel way of trajectory analysis to quantify the relative contribution of sources to a certain receptor region and it could be readily applied to the source apportionment of other species in other regions.
 This work was funded in support of the Ganges Valley Aerosol Experiment (GVAX) by the Office of Biological and Environmental Research in the U.S. Department of Energy, Office of Science. The 15-year BC emission trends were developed with the support of the Modeling, Analysis and Predictability (MAP) program of the National Aeronautics and Space Administration (NASA) under proposal 08-MAP-0143. The work at Tsinghua University was supported by China's National Basic Research Program (2010CB951803). Argonne National Laboratory is operated by UChicago Argonne, LLC, under contract DE-AC02-06CH11357 with the U.S. Department of Energy.
 The Editor would like to thank Gregory Carmichael and an anonymous reviewer for their assistance in evaluating this paper.