Regional Impact of Biomass Burning in Southeast Asia on Atmospheric Aerosols during the 2013 Seven South-East Asian Studies Project

A nested air quality prediction modeling system (NAQPMS) with an online tracer-tagged module was utilized to investigate the regional impact of biomass burning (BB) on aerosols and source–receptor relationships in Southeast Asia during March–April 2013. NAQPMS could reproduce the three-dimensional spatial distribution of aerosols. Both monthly and episodic analyses indicated that BB significantly contributed to surface and column aerosol concentrations in Southeast Asia along two long-range transport pathways. In the first pathway, aerosols from BB were blown northward from the Indochina peninsula to southwestern provinces in China. The mean contributions of BB decreased from 70%– 80% in the source regions to 10%–40% in southwestern China. Myanmar was the largest exporter. In the second pathway, PM2.5 emitted by BB was uplifted into the mid-altitudes (2000 m) in the Indochina peninsula and transported eastward to the western Pacific at altitudes of 2500–4000 m, passing the South China Sea, southern China and western Pacific. In downwind regions, BB contributed 30%–60% of aerosols at altitudes of 2000–4000 m and 10%–30% below 2000 m. A simple estimation based on source–receptor relationships showed that BB emissions were likely overestimated by 35%– 50% in the Fire Inventory from National Center for Atmospheric Research (v1.5).


INTRODUCTION
Recently, biomass burning (BB) from crop stubble, forest residues, and other vegetation has been receiving increased attention because of the associated releases of large amounts of aerosols (solid carbon combustion particles) and their gaseous precursors (Viana et al., 2008).Bio-aerosols mix with anthropogenic pollutants and change the physical and chemical (e.g., hygroscopic growth) properties of aerosols in downwind regions (Kim et al., 2006).Studies have confirmed that BB has substantial effects on global air quality and climate (IPCC, 2007).
Southeast Asia, one of the most intensive BB regions in the world, emits half and two-thirds of global elemental and organic carbon, respectively (Gustafsson et al., 2009).
Studies have revealed that BB contributes 70% of the fine particles (PM 2.5 ) in Southeast Asia (Fu et al., 2012).Compared with the mid-latitude regions, the intensive convection and vertical advection in the subtropics are more conducive to transportation of pollutants from burning regions to downwind regions that are thousands of kilometers away, with far-reaching effects on air quality.Liu et al. (2003) reported that contributions from BB in Southeast Asia to the western Pacific were comparable to those from Asian anthropogenic emissions.Using a regional chemical transport model, Huang et al. (2013) provided a higher estimation of the impact of BB in Southeast Asia on southern China, the South China Sea, and the Taiwan Strait (approximately 62%) in the BASE-ASIA campaign.Fu et al. (2012) employed two widely used BB emission inventories to analyze the impact of BB episodes in March-April, and determined that such episodes can affect the Pearl River Delta and the Yangtze River Delta in China.In addition, new long-range transport mechanisms of BB emissions (e.g., the mountain leeside trough) in Southeast Asia have been discovered (Lin et al., 2009).The pollutants from BB can persist for 2 months and cause an enhancement in the following months in the subtropical Pacific (Nam et al. 2010).The significant longrange transport of BB emissions likely offsets the policydriven reductions of domestic anthropogenic emissions in South China, where the central government is necessitating a 10%-20% decrease in annual PM 2.5 by 2017 compared with the 2013 levels.Therefore, understanding the impact of BB is urgent not only for local air quality management but also for downwind countries, and this is evident from the agreement on transboundary haze pollution made by the Association of Southeast Asian Nations and the Special Funds for Scientific Research on Public Welfare by the China Ministry of Environmental Protection.
To date, comprehensive estimates of the impact of BB in Southeast Asia on the downwind regions are still highly uncertain (Fu et al., 2012).Compared with ground and satellite observations, the simulated PM 2.5 is still significantly overestimated in source regions or is underestimated in the downwind regions (Huang et al., 2013).Fu et al. (2012) evaluated two BB emission inventories (Global Fire Emissions Database version [GFEDv] 2.1 and Fire Locating and Modeling of Burning Emissions [FLAMBE]) using the Models-3/Community Multiscale Air Quality (CMAQ) modeling system and identified great divergences (by a factor of 7.58) between the inventories, which were likely caused by modeling uncertainties regarding the impact of BB.Consequently, additional studies using different regional models and BB emission inventories are urgently required (Fu et al., 2012).Second, current approaches for evaluating the impact of BB in Southeast Asia on regional aerosols primarily use the "brute force" method (Fu et al., 2012;Huang et al., 2013).Recently, an "online tracer-tagged" method was developed and successfully used to calculate the contribution of each individual source to the regional pollutants in East Asia (Grewe, 2004;Li et al., 2014Li et al., , 2016)).This approach facilitated the assessment of the contributions of pollutants from different regions, by avoiding uncertainties in atmospheric chemistry.Thus, the application of the online tracer-tagged method in Southeast Asia is strongly favored for understanding the regional impact of BB.Finally, previous simulations examined Southeast Asia as a whole and did not assess the impact of BB in each country.Studies have proven that the intensity of BB significantly differs across the countries in Southeast Asia.Although the area of the two nations is similar, the black carbon emitted in Myanmar was three times that emitted in Vietnam in March 2010 (Wiedinmyer et al., 2011).For policy makers, estimating the impact of BB in each country is urgently necessary to effectively manage regional air quality.
In March-April 2013, the Seven South-East Asian Studies (7-SEAS) organization arranged a field campaign called the 2013 BASELnE (Biomass-burning Aerosols & Stratocumulus Environment: Lifecycles and Interactions Experiment) to study BB aerosols from the source to receptor regions in Southeast Asia.Three observation stations in the BB source regions were established (Lin et al., 2013).In addition, a special fund for scientific research was provided by the China Ministry of Environmental Protection to examine the impact of BB in Southeast Asia on air quality in China.This study aimed to simulate the regional impact of BB in Southeast Asia on aerosols in East Asia and the western Pacific in March-April 2013 using a nested air quality prediction modeling system (NAQPMS).Model simulations were compared with in situ ground measurements in Southeast Asia and China and with satellite and LIDAR remote-sensing observations (Section 3).The monthly and episodic impact of BB and anthropogenic emissions in Southeast Asia on the regional surface and column PM 2.5 was quantified by an online tracer-tagged module implemented in the NAQPMS (Section 4).Particularly, the vertical transport patterns of PM 2.5 were illustrated.Finally, the uncertainties of BB emissions were analyzed.

Model Description
The NAQPMS utilized in this study is a fully modularized three-dimensional (3D) regional Eulerian chemical transport model driven by the Weather Research and Forecasting (WRF)-ARW 3.5 system.This model has been widely used in simulating and predicting the long-range transport of aerosols and their precursors in East Asia (Li et al., 2011(Li et al., , 2012(Li et al., , 2014;;Tang et al., 2015).Since 2006, the NAQPMS has been successfully applied as a routine air quality forecast model to predict air quality during the Beijing Olympic Games, Shanghai Expo, and Guangzhou Asian Games (Chen et al., 2010;Wu et al., 2011;Wu et al., 2015).
The NAQPMS reproduces the physical and chemical evolution of reactive air pollutants (gaseous species, inorganic and organic anthropogenic and biogenic aerosols, sea salts, and mineral dust) by solving the mass balance equation in terrain-following coordinates.An accurate mass-conservative, peak-preserving algorithm was used to simulate the advection by applying a time-splitting technique (Walcek and Aleksic, 1998).Vertical eddy diffusivity was parameterized using a scheme by Byun and Dennis (1995).Dry and wet depositions were parameterized according to Wesely (1989) and the Regional Acid Deposition Model mechanism in the CMAQ system, version 4.6 (CMAS, 2016), respectively.The gasphase chemical reaction scheme used in the present study was the Carbon-Bond Mechanism Z (CBM-Z), which is composed of 133 reactions for 53 species (Zaveri and Peters, 1999).The composition and phase state of an ammoniasulfate-nitrate-chloride-sodium-water inorganic aerosol were calculated using an aerosol thermodynamic model (ISORROPIAI1.7)(Nenes et al., 1998).A bulk yield scheme was used to model the formation of secondary organic aerosols (SOAs) (Pandis et al., 1992;Odum et al., 1997).Naturally produced aerosol emissions (dust and sea salt) were parameterized with a size-segregated dust deflation and sea salt module (Athanasopoulou et al., 2008;Wang et al., 2000).The interaction between gaseous pollutants and aerosols was parameterized using a heterogeneous chemistry module with 28 reactions and an accurate radiative transfer model (TUV, version 4.5) (Li et al., 2011(Li et al., , 2012)).Aerosol optical depths were converted from mass concentrations through a "reconstructed extinction coefficient" method proposed by Malm (2000) as a part of the Interagency Monitoring of Projected Visual Environment (IMPROVE) program.Model validation using 1-year LIDAR and satellite remote-sensing observations indicated that this method could reproduce the general aerosol optical properties over East Asia, although a few discrepancies existed because of the empirical relationship derived from the United States (Li et al., 2014).
An online tracer-tagged module, similar to the Particulate Matter Source Apportionment Technology (PSAT), was embedded into the NAQPMS (Wagstrom et al., 2008).During each step, the module attributes pollutant concentrations to different types of emissions (i.e., BB and anthropogenic) and geographical locations, without disturbing the original calculations.All secondary aerosol components are transformed into specific precursor gas species (e.g., sulfate to sulfur dioxide, nitrate to gaseous nitric acid (HNO 3 ), ammonium to ammonia (NH 3 ), SOA to semivolatile gases).These modules provide a similar chemical transformation of air pollutants and eliminate the effects of chemical nonlinear errors.The module has been widely applied to quantitatively evaluate the source of pollutants and has been validated by the Ministry of Environmental Protection of China (CMEP, 2013).In this study, we tagged two types of emissions (BB and anthropogenic emissions, including fossil fuel, biofuel, industrial, transportation, and power plant emissions) and 18 regions (Fig. 1), five of which are located within the Indochina Peninsula.Detailed descriptions of the tagged regions are provided in Table 1.

Model Configuration
Fig. 1 shows our model domain, which covers most of Southeast Asia and all of East Asia, with a 45-km grid resolution on a Lambert conformal map projection.The NAQPMS is configured with 20 layers extending from the surface to 20 km above sea level; 10 of the layers are within the lowest 2 km above the surface.In WRF, the final analyses dataset (ds083.2) from the National Centers for Environmental Prediction (NCEP), with 1° × 1° resolution and a temporal resolution of 6 h, was used for the initial and boundary conditions.A four-dimensional data assimilation nudging toward the NCEP dataset was performed to increase the accuracy of WRF.For the NAQPMS, the initial and boundary conditions were taken from a global chemical transport model (MOZART-V2.4) with 2.8° resolution.
The simulation was performed from February 15, 2013 to April 30, 2013.The first 15 days were regarded as spinup time to reduce the influence of initial conditions.

Emissions
In this study, anthropogenic emissions (fossil fuel, biofuel, industrial, transportation, and power plant emissions) were determined from an inventory by harmonizing different local emission inventories through the mosaic approach (Li et al., 2015).This inventory was prepared for the Model Inter-Comparison Study for Asia (MICS-Asia) and Task Force on Hemispheric Transport of Air Pollution projects.The inventory was established by four countries: China, at Tsinghua University and Peking University; Japan, at the National Institute of Environmental Sciences and Asia Center for Air Pollution Research; the United States, at the Argonne National Laboratory; and Korea, at Konkuk University.The original resolution and initial year of the MIX were 0.25° and 2010, respectively.
The biogenic emission inventory used in this study was obtained from an emission model (Meganv2, with a 0.5°  Studies have reported that carbon emissions reported for FLAMBE were 7.58 times more than those for GFEDv2.1 in Southeast Asia (Fu et al., 2012).CMAQ modeling studies by Fu et al. (2012) and Huang et al. (2013) have demonstrated that simulated results obtained using FLAMBE overestimated observations in source regions such as Thailand, whereas simulations using GFEDv2.1 underestimated these values.In this study, we used the Fire Inventory from National Center for Atmospheric Research (FINN) with a daily, 1-km resolution (Wiedinmyer et al., 2011).The tracer and particle emissions from wildfires, agricultural fires, and prescribed burning were included in FINNv1.5.Nonmethane volatile organic compounds were categorized into 18 subspecies to match the CBM-Z mechanism used in the NAQPMS.Comparisons among these three inventories revealed that BB emissions in Southeast Asia in FINNv1.5 were between those in FLAMBE and GFEDv2.1.For example, in FINNv1.5, black carbon emissions in Southeast Asia were 514 Gg yr -1 , which was 4.7 times that in GFEDv2.1.The injection heights of anthropogenic and BB emissions were obtained from Simpson et al. (2012).Fig. 1 and Table 1 show the spatial distribution of black carbon emission rates in this study.BB emissions were clearly much higher than the anthropogenic and BB emissions in south China.

In Situ and Remote-Sensing Measurements
As shown in Fig. 1, three remote rural monitoring stations in Thailand and Vietnam were used in the 7-SEAS project.Doi Ang Khang (99.05°E, 19.94°N, approximately 1.5 km above sea level) is close to the border with Myanmar and approximately 125 km north of Chiang Mai city in Thailand; it primarily monitors forest and agricultural burning (Sayer et al., 2016).Surface particulate matter mass loadings were measured by the Beta Attenuation Monitor 9600 (MetOne instruments, USA).Another station in Bangkok (100.53°E,13.77°N) monitors emissions in southern Thailand.The Sonla site (103.90°E,21.33°N, 675 m above sea level) is located at the Sonla Atmospheric Observatory Station, northern Vietnam, which is 140 km south of the border with China and 110 km east of the border with Laos (Lee et al., 2016).PM 2.5 was collected using collocated R&P ChemComb Model 3500 speciation sampling cartridges (Thermo Fisher Scientific Co., Inc., Waltham, MA, USA).
Numerous monitoring stations operated by the Chinese National Environmental Monitoring Center routinely measure PM 2.5 in more than 300 cities across China.Six cities were selected for this study (Fig. 1).
Two LIDAR remote-sensing stations in central Thailand and southern Japan were used to evaluate the vertical profile of simulated results.The extinction coefficients at Phimai (102.57°E,15.83°N) and Cape Hedo (128.25°E,26.87°E) were observed using dual-wavelength (1064 nm, 532 nm), depolarization LIDAR under the NIES LIDAR network over Asia (NIES, 2016).The aerosol optical depth at 550 nm (AOD550), retrieved from a MODerate Resolution Imaging Spectrometer (MODIS) instrument aboard the Aqua satellite with 1°× 1° resolution, was used in this study.We obtained MODIS Collection 5.1 (051) from the MODIS Science Team at NASA (NASA, 2016).the NAQPMS successfully captured the temporal variation at all sites throughout the study period.Factor 2 analysis indicated that modeled PM 2.5 captured 75%-98% of the fraction between 0.5-to 2.0-fold of the measured data (Table 2).The normalized mean bias (NMB) and error ranged from -0.11 to 0.56 and 0.29 to 0.61, respectively.At Doi Ang Khang and Sonla, two sites close to intensive BB areas (Fig. 1), the simulated PM 2.5 overestimated observations by 42% and 56%, respectively; this was mainly caused by uncertainties in BB emissions.The simulation by Huang et al. (2013) overestimated observed carbon monoxide (CO) by 130%-150% in northern Thailand.This indicated that emissions in FINNv1.5 were still overestimated, albeit less than those in FLAMBE.In other regions, including southern Thailand and China, the model performance was much higher than that in Doi Ang Khang and Sonla, with NMBs of -0.11 to 0.36.

COMPARISON BETWEEN MODEL RESULTS AND MEASUREMENTS
The simulated vertical profiles of PM 2.5 in the source and downwind regions were also evaluated (Fig. 3).At Phimai, two intensive episodes were observed during March 10-15 and April 5-9, 2013, when high extinction coefficients of 0.3-0.5 extended from the surface to 3 km, suggesting strong BB activities and intensive vertical mixing in Southeast Asia.The model reproduced this vertical pattern reasonably well, although extinction coefficients in the second episode were slightly underestimated.At the downwind site (Cape Hedo), high extinction coefficients were mostly constrained to below 1 km.An exception was April 7-8, when higher extinction coefficients of 0.2, compared with those of the surface, appeared 2-4 km above sea level.the mean observed and modeled PM 2.5 (µg m -3 ) during March-April 2013.MB, RMSE, and R represent the mean bias (µg m -3 ), root mean square error (µg m -3 ), and correlation coefficients, respectively.NMB and NME are normalized mean bias and error.NUM is the number of paired samples.FAC2 is the percentage of the ratios between 0.5 and 2.  (Lin et al., 2013).These clouds obscured the MODIS AOD observations.As shown in Fig. 4, only 10-15 valid days for measurements were observed in April for the transport pathway.
Fig. 4 also shows the monthly impact of BB on AOD550.In March, the impact was concentrated in source regions and inner southern China, where it reached 0.5-0.9(50%-70% of the total AOD550).In the South China Sea, Taiwan Strait, and western Pacific, BB contributed 25% of the total AOD550.In April, more BB emissions were transported to the western Pacific.The AOD550 from BB in the Taiwan Strait and western Pacific reached 0.4-0.8(50% of the total AOD550).In China, the impact of BB was mostly noted in the southwestern Chinese borderland provinces.Figs.5(e) and 5(f)).The regions most influenced by longrange transport of BB were the border provinces between China and the Indochina peninsula (Yunnan, Guangxi, and Guizhou provinces), where it reached 5-20 µg m -3 (10%-40%).Compared with March, the transport of BB emissions spread over farther regions in April, including the South China Sea and Taiwan Strait, where it was 1-5 µg m -3 (3%-10%).In March-April, contributions of organic matter (OM), secondary inorganic aerosols (SIA) and black carbon (BC) from BB were 20-30%, 5-10% and 1-3% of PM 2.5 concentrations in Indochina peninsula and on the border with China (figure not shown).In the western Pacific, BB contributions are negligible.Wind patterns affected the spatial distribution.Between 15°N and 25°N, a convergence between the eastern winds in the South China Sea and the western winds in the Bay of Bengal appeared in the Indochina peninsula (Fig. 5(a)).As a result, the southern winds in the peninsula predominantly blew pollutants from the peninsula to the border provinces in China.The horizontal convergence of the mass flux of PM 2.5 in the boundary layer also caused upward transport to the free troposphere (Oshima et al., 2013) Table 3 shows the mean source-receptor relationships of five countries in the Indochina peninsula and China during the study period.In the BB source regions, local BB dominated PM 2.5 in Laos, Myanmar, and Cambodia, which accounted for 60%-85% (19-65 µg m -3 ) of the total concentrations.This was higher than their local anthropogenic emissions, which ranged from 5% to 20%, by almost a factor of 10-15.Compared with BB emissions, the long-range transport of Chinese anthropogenic emissions was negligible (0%-6%) in these three countries.Laos was the most intensive BB country, contributing 66% of PM 2.5 , but this was mostly restricted to its boundaries.In Thailand and Vietnam, the contributions from BB were lower, at 45%-56%, whereas those from local anthropogenic emissions were higher, at 18%-28%.This reflects the heterogeneity of BB in the source regions.Among the downwind regions, Yunnan province (YN in Fig. 1) was the most influenced by BB in Southeast Asia.The BB contribution to PM 2.5 reached 23.8 µg m -3 , 43.3% of the total concentration.Myanmar was the greatest contributor, with 18.1 µg m -3 PM 2.5 .This transport of BB emissions in Southeast Asia was even higher than that of Chinese anthropogenic emissions, which accounted for only 15.1 µg m -3 PM 2.5 (27.4%), suggesting that policy makers should consider the impact of BB when setting air quality goals in Yunnan.In the nearby provinces of China, such as Guangxi, Guizhou, Sichuan, and Hainan (GX, GZ-CQ, SC, and HN in Fig. 1), the contributions from long-range BB emissions in Southeast Asia markedly decreased to 4%-6% of the total concentrations.This was caused by the topography of the Indochina peninsula.Intensive fire points in Laos, Myanmar, and Thailand appeared on the plains, but these areas are surrounded by high mountains (Annam Mountains) and plateaus (Tibet and Yungui Plateau), and this blocked the transport of BB emissions at the surface.Compared with the findings of Fu et al. (2012), our estimated impact area of BB was smaller.In their 2006 study, BB was transported to the surface of the western Pacific.This was likely caused by the meteorology and emission inventories.
As shown in Fig. 6, the relationship between daily surface PM 2.5 and BB contributions showed different patterns, with an increase of the distance to fire spots.In the source regions (Doi Ang Khang and Sonla), PM 2.5 from BB was linearly correlated with total PM 2.5 concentrations, which indicated the dominant role of BB in the whole PM 2.5 range.At Kunming, the closest downwind site in China to fire spots, BB contributions seemed to be more significant during the high-pollution period (PM 2.5 > 150 µg m -3 ) than on clear days.
As shown in Figs. 4 and 5, the impact of BB on surface PM 2.5 and AOD550 seems to be in conflict.The column PM 2.5 was investigated (Figs.5(g)-5(i)), and contrary to the surface PM 2.5 , a high-column PM 2.5 belt extended to the northeast from Myanmar, northern Laos, and Vietnam in the Indochina peninsula (80-110°E, 20-25°N) to the western Pacific (140-150°E, 25-45°N), crossing southern China, the South China Sea, and the Taiwan Strait (Figs. 5(g) and 5(j)).BB majorly contributed to this spatial distribution pattern (Figs.5(h) and 5(k)).In the northern peninsula, the contribution of BB reached 80-100 mg m -2 (over 50% of the total column PM 2.5 ).OM and SIA contributed 20-30% and 10% to total column PM 2.5 , respectively.The downwind areas were strongly influenced by the long-range transport of BB plumes.This transport spread over the southeastern parts of the Chinese mainland, Taiwan Strait, and western Pacific (dark lines in Fig. 5(h)), with contributions of 30%-50% and 10%-30% in the Chinese mainland and western Pacific.Contributions of OM and SIA were 10-20% and 5-10%, respectively.In April, BB plumes from the Indochina peninsula extended farther into the western Pacific than did those in March.This result was different from that of Huang et al. (2013), who reported that the transport of BB emissions was stronger in March.In conclusion, the impact of BB in Southeast Asia on column PM 2.5 was spread over a broader region than was the impact on surface PM 2.5 , which was close to that on AOD550.Compared with BB emissions, the impact of anthropogenic emissions in Southeast Asia appears to be much weaker (Figs. 5(i) and 5(l)), at only 5%-10% and < 5% in the peninsula and downwind regions, respectively.

Episodic Impact Case I
Case I (March 13-19) was selected to assess the longrange transport of BB emissions in Southeast Asia at ground level.Fig. 7 illustrates the AOD550, surface PM 2.5 , and column PM 2.5 concentrations in the model domain during this episode.The regional contributions of BB and anthropogenic emissions in Southeast Asia to AOD550 and PM 2.5 are also shown.Both the MODIS and the model show high AOD in the Indochina peninsula and its nearby regions along the border between the peninsula and China.This indicated that BB during this episode could play a crucial role in the air quality of these regions.For surface PM 2.5 , a high concentration belt extended to the northeast from Myanmar and Laos to Yunnan (YN in Fig. 1) and Guizhou (GZ-CQ in Fig. 1) and eventually linked with the high anthropogenic PM 2.5 region in eastern China.The mass flux of PM 2.5 reached 200-300 µg m -2 s -1 along this transport pathway (Fig. 7(d)).As shown in Fig. 7(e), BB significantly contributed to this pattern.In the source regions, the contribution of BB reached 80%.With the increase in transport distance, the contribution of BB decreased from 50% in Yunnan to 10% in Guizhou because of the rapid wet/dry scavenges.
Fig. 9 presents the vertical structure of PM 2.5 and BB contributions along the transport pathway (dark line in Fig. 7(e)).Clearly, two air masses with opposite characteristics dominated the two sides of the mountains (101-103°E) at 2000-2500 m above sea level (Fig. 9(a)).On the southwestern side (the Indochina peninsula), a dry air mass with < 50% relative humidity (RH) ascended the mountains, with a 5-10 m s -1 wind velocity.The air masses brought high PM 2.5 concentrations from BB to the summit along the mountain, where the contribution of BB reached 25-50 µg m -3 (30-60%; Fig. 9(e)).These PM 2.5 from BB was transported to the free atmosphere over central China in the prevailing western winds (Fig. 9(e)), and largely contributed the column densities in central China (Fig. 7(h)).Surface air masses on the northeastern side were much wetter than those on the southwestern side (RH > 60%; Fig. 9(a)).High PM 2.5 concentrations (60-80 µg m -3 ) came mainly from Chinese anthropogenic emissions, extending from the surface to the whole boundary layer (0-2500 m layer) (Fig. 9(c)).PM 2.5 from BB in the Indochina peninsula was 5-15 µg m -3 (< 30%).

Case II
Case II (April 6-8) was an ideal episode for understanding the long-range transport of BB emissions along the second pathway.In this episode, both the satellite and models demonstrated a high AOD belt (> 0.7) that extended from the Indochina peninsula to the western Pacific, crossing the South China Sea, southern parts of China, and the Taiwan Strait (Figs. 8(a) and 8(b)).Compared with the monthly mean values, the plume extended farther into the Pacific, and BB had a higher contribution in the downwind regions (Fig. 8(c); 50% over the western Pacific).shows that the long-range transport of BB emissions at ground level was not responsible for this BB plume from the peninsula to the Pacific.At ground level, most PM 2.5 from BB was constrained to the Indochina peninsula and its surrounding regions.Column PM 2.5 was consistent with AOD550 (Figs. 8(g)-8(i)).In the western Pacific, the transported column PM 2.5 from the BB plume was 30-80 mg m -2 , which accounted for 30%-50% of the total column concentrations.In southern China, including Yunnan, Guangxi, and Guangdong, contributions from the BB plume reached 100-180 mg m -2 (approximately 50%).As shown in Fig. 8(g), a strong mass flux (500 mg m -1 s -1 ) of PM 2.5 predominantly appeared from west to east at latitudes of 15-25°N (dark line in Fig. 8(h)), which transported the PM 2.5 emitted by intensive BB more northeastward.Compared with the 2006 episode (Fu et al., 2012), the mass flux was stronger and covered a larger area.
The vertical structures of total and BB PM 2.5 along the transport pathway (dark line in Fig. 8(h)) clearly illustrate that the main body of BB PM 2.5 appeared in the free atmosphere (2500-4000 m above sea level) of the downwind regions (Figs. 9(d) and 9(f)).PM 2.5 emitted by BB in the Indochina peninsula (100-105°E) was uplifted into the mid-altitudes (2000 m above sea level; Fig. 9(b)) and was quickly transported eastward to the western Pacific at a level of 2500-4000 m.In the source regions, 70% of PM 2.5 from the surface to 4500 m came from BB.In the downwind regions, contributions from BB were 30%-60% and 10%-30% above and below 2 km, respectively.Different from the slow ascent to the free atmosphere along the orography in Case I, BB PM 2.5 experienced more intensive vertical mixing and was directly transported to higher levels (4500 m) in source regions in this case.This may be because in this episode, the regions with BB (100-105°E) were controlled by wet air masses (Fig. 9(b)).The major transport body at the 2500-4000 m level was also observed at the downwind sites.During this episode, high extinction coefficients appeared at the 3000-4000 m level at Cape Hedo (Fig. 3).Ozonesounding data at Banqiao station in Taiwan (121.48°E,24.99°N) revealed that the long-range transport of BB emissions caused an ozone peak at 2-4 km (Fig. S1).

IMPLICATIONS FOR BB EMISSION ESTIMATIONS
An urgent issue in studying the regional impact of BB is the reduction of the uncertainties in BB emission inventories.In the current BB inventories, BB emissions in Southeast Asia differed by a factor of 10 (Fu et al., 2012).Retrieving useful information from the simulations is one of the most effective approaches to improving emission inventories.In this study, we conducted a simplified approach to evaluate the uncertainties of the FINNv1.5, according to the differences in the simulations, observations, and calculated source-receptor relationships between the different regions.This approach can offer assistance for improving the inventory.In this study, in situ observations were obtained from remote sites in the BB source regions (Doi Ang Khang and Sonla).
Comparisons between different BB emission inventories (e.g., GFEDv2.1,FLAMBE, and FINNv1.5)showed similar spatial and seasonable distributions in Southeast Asia (Wiedinmyer et al., 2011;Fu et al., 2012).In this simplified approach, we therefore did not alter the spatial distribution of fire spots in FINNv1.5.According to the calculated source-receptor relationship, surface PM 2.5 at each site was divided into contributions from BB in Southeast Asia (Cbb) and others (Coth).An adjustment factor (F j ) for BB emissions was set, and the mean error (ME) at the i th site and j th F j was calculated as follows: where Obs i represents the observations.We continually changed the adjustment factor (F j ) until the mean of ME i at all stations reached the minimum.We assumed that the contributions of BB to PM 2.5 were linearly correlated with BB emissions observed in this study.This likely introduces some uncertainties.However, Lin et al. (2008) discovered that errors resulting from the nonlinear responses of PM 2.5 to emissions over Southeast and East Asia were within 5%.Fig. 10 shows the mean errors at nine sites (Fig. 1), with a different adjustment factor (F j ).The mean errors exhibited a quadratic function with F j and reached the minimum when F j ranged from 0.5 to 0.65.This indicated that current BB emissions in Southeast Asia in FINNv1.5 likely overestimated observations by 35%-50%.Additional studies on emission factors and fire activities in this region are necessary.

DISCUSSIONS ON UNCERTAINTIES
Besides the impact of cloud on retrievals of remote sensing measurements, uncertainties of simulated meteorological fields partly caused the discrepancies of column PM 2.5 in downwind sites between satellite and lidar and simulation.Koffi et al. (2012) evaluated the performance of 12 models and found almost half of them overestimated the aerosol extinctions in 2-6 km in burning seasons over Southeast Asia (Koffi et al., 2012).Tost et al. (2010) found that simulated concentrations among five parameterisations in the same model even differed more than 100%.In particular, the convection scheme in this study easily resulted in higher outflow height in tropics (Emmanuel et al., 1999;Tost et al., 2010).
Overestimation of anthropogenic PM 2.5 plume may also largely result in discrepancies between simulation with satellite and lidar.In this study, anthropogenic PM 2.5 plume extended from the continent to open ocean and contributed more to column PM 2.5 than biomass burning.Previous studies that most models overestimated column PM 2.5 in East China and northwestern pacific in spring among 12 models (Koffi et al., 2012) by comparing with CALIOP Lidar satellite.So compared with absolute concentrations, relative contributions (in percentage) of BB were more affected by the overestimation of anthropogenic PM 2.5 plume.This indicated the relative contributions from BB were actually more than current estimation.

CONCLUSIONS
The objective of this study was to quantitatively investigate the regional impact of BB in Southeast Asia on PM 2.5 in east Asia and the western Pacific during March-April 2013.We used a regional chemical transport model (NAQPMS) coupled with an online tracer-tagged module based on a high-resolution BB emission inventory.Comparisons of in situ measurements at the source and downwind sites with LIDAR and satellite remote-sensing retrievals demonstrated the capability of the NAQPMS in reproducing the spatial and vertical distribution and evolution features of aerosols.
In the study period, BB emissions played a significant role in aerosol levels in both Southeast Asia and its downwind regions, such as southern China, the South China Sea, the Taiwan Strait, and the western Pacific.Two transport pathways were identified: The first was a pathway from the Indochina peninsula northward to the southwestern provinces in China (Yunnan) at ground level.In this pathway, BB contributed 70%-80% of PM 2.5 in northern Myanmar,Laos,.With the increase in transport distance, the contributions of BB decreased to 10%-40% in southwest China (25-30°N).Myanmar was the source region for the downwind provinces in China.In particular, BB significantly contributed to high-pollution episodes at the downwind sites.Analyses of the vertical profiles revealed that under weak southern winds, PM 2.5 from BB was uplifted over the mountains (2500 m) on the boundary between the Indochina peninsula and China.These PM 2.5 from BB was transported to the free atmosphere over central China in the prevailing western winds, and largely contributed to the column densities in central China.The second pathway extended from the Indochina peninsula northwestward to the western Pacific, passing over the southern parts of the Chinese mainland and Taiwan Strait.In this pathway, PM 2.5 emitted by BB in the Indochina peninsula (100-105°E) was uplifted into the mid-altitudes (2000 m above sea level) in the source regions and was quickly transported eastward to the western Pacific at a level of 2500-4000 m.The major transport body appeared at a level of 2500-4000 m above sea level, which was validated by LIDAR and sounding observations in the downwind regions.In the source regions, 70% of PM 2.5 from the surface to 4500 m came from BB.In the downwind regions, such as the western Pacific, BB contributed 30%-60% of PM 2.5 at the 2000-4000 m level and 10%-30% at levels below 2000 m.Regarding AOD550 and column PM 2.5 , both monthly and episodic mean contributions of BB reached 30%-50% in the southern parts of China and the western Pacific.
Finally, a simple approach was used to estimate the uncertainties of BB emissions over Southeast Asia in FINNv1.5, according to the differences in the simulations, observations, and calculated source-receptor relationships between regions.BB emissions in FINNv1.5 likely overestimated observations by 35%-50%.Additional studies on emission factors and fire activity in this region are necessary.

Fig. 1 .
Fig. 1.Model domain for NAQPMS with black carbon emission rates (shaded) and tagged regions (sky blue).The solid blue circles represent the locations of the observation sites.

Fig. 2
Fig.2presents the simulated and observed daily PM 2.5 at nine ground stations during March-April 2013.In general,

Fig. 4 .Fig. 5 .
Fig. 4. Comparison between monthly mean satellite measured AOD550 (unitless) from MODIS (a and b) and calculations from NAQPMS (c and d) for March (left panels) and April (right panels) 2013.The monthly mean contributions from BB (e and f; shaded: unitless, dark contours: percentage) and the valid days for measurements (g and h) during these 2 months are also shown.
2.5 contributions (µg m -3 and mg m -2 ) and percentages (in parentheses) from the source regions to receptor regions over East Asia during March-April 2013.

Fig. 6 .
Fig. 6.Scatter plots of simulated daily PM 2.5 from BB with simulated total PM 2.5 concentrations at nine in situ stations.

Fig. 7 .
Fig. 7. Observed (a) and simulated (b) mean AOD550 and simulated contributions of BB (c) during March 13-18, 2013 (Case I).Simulated surface PM 2.5 and its mass flux (d), simulated contributions from BB (e), and anthropogenic emissions (f) in Southeast Asia.Column PM 2.5 and its mass flux (g) and contributions from BB (h) and anthropogenic emissions (i) in Southeast Asia.

Fig. 9 .
Fig. 9. Simulated vertical profiles of wind vectors, RH (a and b), PM 2.5 concentrations (c and d), and contributions from BB in Southeast Asia (e and f), along the transport pathway in Case I (dark line in Fig. 7) and Case II (dark line in Fig. 8).The vertical wind speeds were multiplied by 50 in (a) and (b).

Fig. 10 .
Fig. 10.Responses of the PM 2.5 mean errors (ME) to the adjusting factor of BB emissions in Southeast Asia at nine sites (S1) and 3 sites (Bangkok, Doi Ang Khang and Sonla) in Southeast Asia (S2).Also shown are responses of the mean ME (S3) and RMSE (S4) at three sites in Southeast Asia when BB contributions from Southeast Asia were more than 50%.

Table 1 .
Tagged regions in the model domain and their biomass burning and anthropogenic emissions of black carbon in the study period.
a Amthro and BB represent the anthropogenic and biomass burning emissions observed in this study.resolution produced by the National Center for Atmospheric Research, within the framework of the ACCENT-GEIA data portal (ACCENT-GEIA, 2016).

Table 2 .
Statistical parameters for model evaluation at in-situ measurement sites.

Table 3 .
Mean surface and column PM Anthro and BB represent the anthropogenic and biomass burning emissions observed in this study. a