Wildfire Smoke Transport and Air Quality Impacts in Different Regions of China

The air quality and human health impacts of wildfires depend on fire, meteorology, and demography. These properties vary substantially from one region to another in China. This study compared smoke from more than a dozen wildfires in Northeast, North, and Southwest China to understand the regional differences in smoke transport and the air quality and human health impacts. Smoke was simulated using the Hybrid Single Particle Lagrangian Integrated Trajectory Model (HYSPLIT) with fire emissions obtained from the Global Fire Emission Database (GFED). Although the simulated PM2.5 concentrations reached unhealthy or more severe levels at regional scale for some largest fires in Northeast China, smoke from only one fire was transported to densely populated areas (population density greater than 100 people/km2). In comparison, the PM2.5 concentrations reached unhealthy level in local densely populated areas for a few fires in North and Southwest China, though they were very low at regional scale. Thus, individual fires with very large sizes in Northeast China had a large amount of emissions but with a small chance to affect air quality in densely populated areas, while those in North and Southwest China had a small amount of emissions but with a certain chance to affect local densely populated areas. The results suggest that the fire and air quality management should focus on the regional air quality and human health impacts of very large fires under southward/southeastward winds toward densely populated areas in Northeast China and local air pollution near fire sites in North and Southwest China.


Introduction
Wildfires have increased in many regions of the world with a large number of devastating wildfires in recent years [1][2][3][4][5][6][7]. About two dozen extremely large fires occurred in the western United States in 2017 and 2018 with totally burned areas of about 22 thousand km 2 (2.3 million hectares (hm 2 )). The 2018 Camp Fire in northern California damaged nearly 19 thousand structures and led to 85 deaths. The 2019-20 Australia bushfires burned 160 thousand km 2 (16 million hm 2 ) lands, leading to losses of over 3000 houses and 33 lives. The 2019 Amazon fires burned about 1500 km 2 (150 thousand hm 2 ) rainforest. The 2017 Portugal fires burned 5000 km 2 (0.5 million hm 2 ) with a loss of 119 human lives and the 2018 Greece Mati fires caused a loss of 99 people. In Russia, 25 thousand km 2 (2.5 million hm 2 ) were burned in 2019. Analyses and simulations have been conducted to understand fire occurrence and spread, weather and climate conditions, and the impacts of these fires.

Study Area
The study area was in eastern China, which was divided into five regions (Figure 1). The fire cases investigated in this study were from the Northeast, North, and Southwest China regions. China has a three-step topography with generally increasing elevations from east to west (Figure 1a). The step 1 topography consists of the coastal plains and hills mostly below 500 m. The step 2 topography consists of the mountain ranges up to 3000 m (expect a lower area called the Sichuan Basin in northern Southwest China). The step 3 topography consists of the Tibet Plateau of 3000-6000 m and the northwestern deserts and mountains of 1000-3000 m. The eastern Northeast and North China are within the step 1 topography; the Southwest, western Northeast, and western North China are within the step 2 topography, where a majority of lands are covered by needle and broad leaf trees ( Figure  1b). The population density distributions described in the introduction section are illustrated in Figure 2a. The population density was classified as sparsely (<10 people/km 2 ), moderately (10-100 people/km 2 ), and densely (>100 people/km 2 ) populated in this study. The three classifications approximately accounted for 60, 30, and 10% areas of the Northeast region, one third each of the North region, and 20, 60, and 20% of the Southwest region.
Northeast, North, and Southwest China are under the control of the East Asian monsoon, which is extremely wet and warm during the summer phase and dry and cold during the winter phase. Southwest China is also affected by the South Asian monsoon, which includes wet and dry phases. Fires in the three regions occurred mainly during the East Asian winter/South Asia dry monsoon phase. The spring 500 hPa geopotential height (Figure 2b) shows the westerly zone north of about 25° N with a trough located east of China and a ridge over western China and Mongolia. The corresponding prevailing airflows over the Northeast, North, and Southwest China are mainly westerly (i.e., eastward), as indicated by the arrows. The ground airflows are more complex due to the impacts of topography and weather systems such as fronts.
(a)   . Population density (a) (from [28]) and atmospheric circulation in the 500 hPa geopotential height field; (b) (drawn using the data from the US National Centers for Environmental Prediction/Department of Energy Reanalysis-2 [34]).)

Fire Cases
A total of 16 fires were investigated, seven from the Northeast, six from the North, and three from the Southwest region ( Figure 1 and Table 1). Four fires each in the Northeast and North regions occurred in the step 1 topography and the rest of the fires occurred in the step 2 topography. Thirteen fires occurred in spring, two in fall, and one in summer. The burned areas in the Northeast region were over 1000 km 2 (100 k hm 2 ) for three fires, 100-1000 km 2 (10-100 k hm 2 ) for two fires, and 50-100 km 2 (5-10 k hm 2 ) for two fires. The durations were longer than a week for five fires and four days for two fires. The fires in the North and Southwest regions had burned areas of over 10 km 2 (1 k hm 2 ) for four fires and 1-10 km 2 (0.1-1 k hm 2 ) for five fires. The durations were longer than a week for two fires and 2-6 days for seven fires. The annual fire numbers during 1999-2017 were about 350, 230, 1850, and 7080 in the Northeast, North, and Southwest regions and in entire China based on the data from the China National Forestry and Grassland Administration's China National Forest Fire Statistical System [35]. The corresponding burned areas were about 1190, 210, 240, and 2110 km 2 (119, 21, 24, and 211 k hm 2 ). The averaging burned areas were about 3.5, 0.9, 0.13, and 0.3 km 2 (350, 90, 13, and 30 hm 2 ) each fire. The examples from the Northeast region were much larger than the averaging size. The examples from the North and Southwest regions were also larger than their averaging sizes.
Fire boundaries, which were not available from the fire information provided by the fire management agency of China, were obtained from the Fire Information for Resource Management System (FIRMS) [36]. The FIRMS products had a resolution of 0.25° (approximately 25 km), obtained based on the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Visible Infrared Imaging Radiometer Suite (VIIRS) satellite remote sensing detection. The latitude and longitude ranges of a fire were determined based on the latitude and longitude values of the fire and the FIRMS boundaries. Figure 2. Population density (a) (from [28]) and atmospheric circulation in the 500 hPa geopotential height field; (b) (drawn using the data from the US National Centers for Environmental Prediction/Department of Energy Reanalysis-2 [34]).
Northeast, North, and Southwest China are under the control of the East Asian monsoon, which is extremely wet and warm during the summer phase and dry and cold during the winter phase. Southwest China is also affected by the South Asian monsoon, which includes wet and dry phases. Fires in the three regions occurred mainly during the East Asian winter/South Asia dry monsoon phase. The spring 500 hPa geopotential height (Figure 2b) shows the westerly zone north of about 25 • N with a trough located east of China and a ridge over western China and Mongolia. The corresponding prevailing airflows over the Northeast, North, and Southwest China are mainly westerly (i.e., eastward), as indicated by the arrows. The ground airflows are more complex due to the impacts of topography and weather systems such as fronts.

Fire Cases
A total of 16 fires were investigated, seven from the Northeast, six from the North, and three from the Southwest region ( Figure 1 and Table 1). Four fires each in the Northeast and North regions occurred in the step 1 topography and the rest of the fires occurred in the step 2 topography. Thirteen fires occurred in spring, two in fall, and one in summer. The burned areas in the Northeast region were over 1000 km 2 (100 k hm 2 ) for three fires, 100-1000 km 2 (10-100 k hm 2 ) for two fires, and 50-100 km 2 (5-10 k hm 2 ) for two fires. The durations were longer than a week for five fires and four days for two fires. The fires in the North and Southwest regions had burned areas of over 10 km 2 (1 k hm 2 ) for four fires and 1-10 km 2 (0.1-1 k hm 2 ) for five fires. The durations were longer than a week for two fires and 2-6 days for seven fires. The annual fire numbers during 1999-2017 were about 350, 230, 1850, and 7080 in the Northeast, North, and Southwest regions and in entire China based on the data from the China National Forestry and Grassland Administration's China National Forest Fire Statistical System [35]. The corresponding burned areas were about 1190, 210, 240, and 2110 km 2 (119, 21, 24, and 211 k hm 2 ). The averaging burned areas were about 3.5, 0.9, 0.13, and 0.3 km 2 (350, 90, 13, and 30 hm 2 ) each fire. The examples from the Northeast region were much larger than the averaging size. The examples from the North and Southwest regions were also larger than their averaging sizes. Fire boundaries, which were not available from the fire information provided by the fire management agency of China, were obtained from the Fire Information for Resource Management System (FIRMS) [36]. The FIRMS products had a resolution of 0.25 • (approximately 25 km), obtained based on the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Visible Infrared Imaging Radiometer Suite (VIIRS) satellite remote sensing detection. The latitude and longitude ranges of a fire were determined based on the latitude and longitude values of the fire and the FIRMS boundaries.

Smoke Modeling
The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model [37] was used to simulate smoke transport and dispersion. HYSPLIT is a complete system for computing simple air parcel trajectories and complex dispersion and deposition. This model uses a hybrid modeling approach of either puffs, particles, or a combination of the two. In the particle model, which was used in this study, a fixed number of initial particles are advected over the model domain by the combined mean and turbulent wind fields. The mean wind fields are the wind field values at the grid points of the input meteorological data. The turbulent wind fields are calculated by HYSPLIT based on the vertical shear of the mean wind fields, atmospheric stability calculated based on temperature, and ground surface roughness which is mainly determined by vegetation type. The plume rise (that is, the height of the smoke plume) is calculated by the model based on the PM 2.5 emissions and heat release. HYSPLIT uses a simple plume rise algorithm [38] originally developed for power plant stacks. Applications of more complex plume rise algorithms for wildfires [39,40] would possibly improve the smoke transport and the air quality modeling but were not used in this study. The heat release of fire missions was calculated using the scheme described in the Fire Emission Production Simulator (FEPS) [41]. HYSPLIT has been widely used for fire smoke modeling [42][43][44]. It was used in our recent smoke modeling study of the 2016 Rough Ridge Fire in northern Georgia, USA [45].
The location and size of the simulation domain varied with fire cases. The domain size in the zonal direction (west-east) or the meridional direction (south-north) ranged 10-40 degrees (approximately 1000-4000 km). A resolution of 0.25 • (approximately 25 km) and 23 vertical levels (6 in the atmospheric boundary layer up to about 1.5 km above ground level) were used. The option of varied integration time step automatically set each hour by HYSPLIT was selected with the stability ratio of 0.75. The major simulation inputs included fire emissions and meteorology. The fire emissions were obtained from Atmosphere 2020, 11, 941 7 of 24 the gridded Global Fire Emissions Database, version 4 (GFED4) [8], which included small fires [46]. The GFED database was developed based on MODIS products. The resolution was 0.25 • (approximately 25 km) with a daily time frequency. There were 32 species and products of fire emissions, including PM 2.5 , CO 2 , CO, NO x , and SO 2 . The simulations included no lateral chemical boundary transport. An evaluation study of burned areas in Daxing'anling showed good agreement in spring but poor in fall between the ground reported and the GFED data [47]. Calculations of fire emissions based on ground reported fire information and measured fuel conditions could improve smoke simulations from fires, especially those occurring in the fall. This approach was not used for this study.
The meteorological variables were from the US National Oceanic and Atmospheric Administration (NOAA) reanalysis with a resolution of 2.5 • (approximately 250 km, available until 2007) and the NOAA National Centers for Environmental Prediction (NCEP) Global Data Assimilation System (GDAS) with a resolution of 1 • (approximately 100 km, available from 2005). HYSPLIT includes algorithms to interpolate meteorological fields at modeling grids (including those at lateral boundaries). Meteorological modeling using mesoscale models for the smoke modeling domains would provide high-resolution variables and improve smoke modeling. This approach was not used for this study.
The US Environmental Protection Agency (EPA) Air Quality Index (AQI) color codes [48] were used to assess the human health impacts of smoke. The The air quality impacts of smoke were evaluated at local and regional scales based on simulated PM 2.5 spatial distributions. The size to separate local and regional impacts is indeterminate, but the size of a region usually incorporates one or more cities, and is on the order of 100 to 10,000 km 2 according to the American Meteorological Society (AMS) [49]. Because fire emissions are an elevated source, which is usually transported much longer in distance than the air pollutants emitted from the surface sources, we used 10,000 km 2 (100 km in distance or about 4 grid points) to separate local and regional scales. The PM 2.5 measurements from [50] were used for model evaluation.

Fire Cases in Northeast China
The seven fire cases were classified into three types according to their smoke transport direction and the air quality impacts.   (Figure 4). Smoke was transported eastward, traveling more than 1000 km into the Russia territory on May 24th (Figure 4a). The PM2.5 concentrations reached 55-150 µg/m 3 (unhealthy level) in a sparsely populated area a few hundred kilometers from the fire site. Thus, the human health impacts should have been minimal. In addition to eastward direction, the smoke was transported northward. It was also dispersed southward over the entire area of Northeast China during the rest of the fire period (Figure 4b-d). Despite densely populated with many cities, the human health should have been minimal due to the small PM2.5 concentrations of less than 12 µg/m 3 (good air quality level).   (Figure 6a). It appeared that the fire intensified and spread toward the south on the next day leading to a line of dense smoke longer than 500 km (Figure 6b) with sparsely and moderately populated areas. The PM2.5 concentrations were over 250 µg/m 3 (hazardous). The smoke was transported further south into Russia where the PM2.5 concentrations were 55-150 µg/m 3 in the following two days (Figure 6c,d). Thus, the high PM2.5 concentrations should have had a moderate impact on human health in some areas.
For three other fires of the third type (illustrated in Appendix A), the Huma fire occurred near the Yichun fire site also during the fall (24-31 October 2005) ( Figure A1). The smoke was transported to east and southeast. The PM2.5 concentrations of more than 250 µg/m 3 affected some moderately populated areas. The Songling fire occurred in Daxing'anling, northwest of the Yichun fire site, during 22 May to 2 June 2006 ( Figure A2). The smoke was transported to east, northeast, and north. The PM2.5 concentrations of more than 250 µg/m 3 affected sparsely populated areas. The Yimuhe fire occurred further west during 1-4 June 2006 ( Figure A3). The smoke was transported north with minimal health impacts.  (Figure 6a). It appeared that the fire intensified and spread toward the south on the next day leading to a line of dense smoke longer than 500 km (Figure 6b) with sparsely and moderately populated areas. The PM 2.5 concentrations were over 250 µg/m 3 (hazardous). The smoke was transported further south into Russia where the PM 2.5 concentrations were 55-150 µg/m 3 in the following two days (Figure 6c,d). Thus, the high PM 2.5 concentrations should have had a moderate impact on human health in some areas.
For three other fires of the third type (illustrated in Appendix A), the Huma fire occurred near the Yichun fire site also during the fall (24-31 October 2005) ( Figure A1). The smoke was transported to east and southeast. The

Fire Cases in North China
The fires and smoke transport were classified into two types.

Fire Cases in North China
The fires and smoke transport were classified into two types.  The Funing fire occurred from 12 to 18 April 2011 in Hebei Province, less than 500 km east of Beijing (Figure 9). On 14 and 15 April, smoke was transported northeastward to Northeast China and eastward to the sea area (Figure 9a,b). The smoke continued to move this way the following two days, but in the meantime, smoke was also transported southward to South China (Figure 9c,d). The PM2.5 concentrations were smaller than 12 µg/m 3 during the entire fire period. Thus, smoke from this fire spread over densely populated areas but should not have had much impact on human health. The smoke from either the Laiwu fire occurring during 17-19 April 2011 ( Figure A6    The Funing fire occurred from 12 to 18 April 2011 in Hebei Province, less than 500 km east of Beijing (Figure 9). On 14 and 15 April, smoke was transported northeastward to Northeast China and eastward to the sea area (Figure 9a,b). The smoke continued to move this way the following two days, but in the meantime, smoke was also transported southward to South China (Figure 9c,d). The PM2.5 concentrations were smaller than 12 µg/m 3 during the entire fire period. Thus, smoke from this fire spread over densely populated areas but should not have had much impact on human health. The smoke from either the Laiwu fire occurring during 17-19 April 2011 ( Figure A6  The Funing fire occurred from 12 to 18 April 2011 in Hebei Province, less than 500 km east of Beijing (Figure 9). On 14 and 15 April, smoke was transported northeastward to Northeast China and eastward to the sea area (Figure 9a,b). The smoke continued to move this way the following two days, but in the meantime, smoke was also transported southward to South China (Figure 9c,d). The PM 2.5 concentrations were smaller than 12 µg/m 3 during the entire fire period. Thus, smoke from this fire spread over densely populated areas but should not have had much impact on human health. The smoke from either the Laiwu fire occurring during 17-19 April 2011 ( Figure A6) or the Jinan fire occurring during 18-20 April 2011 ( Figure A7) in Shandong Province was similar to that from the Funing fire, but spread in two directions of northeast and south only over the land areas. Atmosphere 2020, 11, x FOR PEER REVIEW 13 of 23

Fire Cases in Southwest China
The Chuxiong fire (case 14) occurred during 23-28 April 2013 in Yunnan Province ( Figure 10). Smoke was transported eastward to form a smoke line of over 1000 km long from central Yunnan Province to eastern Guizhou Province on 24 April (Figure 10a). The PM2.5 concentration was 35-55 µg/m 3 near the fire site. It affected Kunming, the province capital with a population of 6 million people. The concentrations were very small in other areas. The smoke started to disperse southward on 26 April reaching the coastal area of Guangxi (Figure 10b). In the following days, the smoke further spread toward the east and north. The PM2.5 concentrations remained small (Figure 10c,d). The transport of smoke from two other fires (cases [15][16] in Southwest China was similar to that of the Chuxioang fire except that the PM2.5 concentrations from the Dali fire were smaller than 12 µg/m 3 (Figures A8 and A9).

Fire Cases in Southwest China
The Chuxiong fire (case 14) occurred during 23-28 April 2013 in Yunnan Province ( Figure 10). Smoke was transported eastward to form a smoke line of over 1000 km long from central Yunnan Province to eastern Guizhou Province on 24 April (Figure 10a). The PM 2.5 concentration was 35-55 µg/m 3 near the fire site. It affected Kunming, the province capital with a population of 6 million people. The concentrations were very small in other areas. The smoke started to disperse southward on 26 April reaching the coastal area of Guangxi (Figure 10b). In the following days, the smoke further spread toward the east and north. The PM 2.5 concentrations remained small (Figure 10c,d). The transport of smoke from two other fires (cases [15][16] in Southwest China was similar to that of the Chuxioang fire except that the PM 2.5 concentrations from the Dali fire were smaller than 12 µg/m 3 (Figures A8 and A9).

Discussion
The simulation results of the16 fire cases summarized in Table 2 indicate that the investigated fires in the Northeast region occurred in the remote mountains north of the densely populated areas. The PM 2.5 concentrations reached the hazardous level for most cases. However, smoke was transported to the sparsely populated areas for a majority of fire cases. Smoke affected densely populated areas only for one fire case and moderately populated areas for two fire cases. The locations and impacts of the fires in this region were similar to those in the Rocky Mountains of the United States.

Discussion
The simulation results of the16 fire cases summarized in Table 2 indicate that the investigated fires in the Northeast region occurred in the remote mountains north of the densely populated areas. The PM2.5 concentrations reached the hazardous level for most cases. However, smoke was transported to the sparsely populated areas for a majority of fire cases. Smoke affected densely populated areas only for one fire case and moderately populated areas for two fire cases. The locations and impacts of the fires in this region were similar to those in the Rocky Mountains of the United States.
In contrast, smoke from all nine fires in North and Southwest China was transported to densely populated areas. However, the PM2.5 concentrations reached the level of unhealthy to sensitive groups only in local areas for three fire cases. Thus, the regional air quality and human health impacts of these fires were minimal. One reason for the minimal regional impacts was that the Chinese government had implemented a strict forest fire suppression policy since the catastrophic 1987 Black Dragon fires in Daxing'anling [51]. A fire would be suppressed as soon as possible. Once a fire was observed, the local government was notified immediately, and it determined the amount of human resources to deploy to fight the fire [52]. As a result, most fires would grow slowly and diminish in a period of a few days. Thus, burned areas and emissions of these fires were mostly small with only minimal impacts on air quality downwind.
The evaluation of the modeling results suggested a difference in the contributions of fire smoke to regional air quality and human health between the China regions and some other fire regions in  In contrast, smoke from all nine fires in North and Southwest China was transported to densely populated areas. However, the PM 2.5 concentrations reached the level of unhealthy to sensitive groups only in local areas for three fire cases. Thus, the regional air quality and human health impacts of these fires were minimal. One reason for the minimal regional impacts was that the Chinese government had implemented a strict forest fire suppression policy since the catastrophic 1987 Black Dragon fires in Daxing'anling [51]. A fire would be suppressed as soon as possible. Once a fire was observed, the local government was notified immediately, and it determined the amount of human resources to deploy to fight the fire [52]. As a result, most fires would grow slowly and diminish in a period of a few days. Thus, burned areas and emissions of these fires were mostly small with only minimal impacts on air quality downwind.
The evaluation of the modeling results suggested a difference in the contributions of fire smoke to regional air quality and human health between the China regions and some other fire regions in the world such as the United States. The PM 2.5 concentrations from non-fire sources in Figure 8 was about 40 µg/m 3 . This value was comparable to the average values for the eastern China from a global analysis [19]. The concentrations in the continental US from the analysis were less than half of the values in China. The PM 2.5 concentrations on 8 April 2014 from the Baoding fire were at the unhealthy level. With the background value, however, the total PM 2.5 (measured) reached the very unhealthy level. Thus, although the PM 2.5 concentrations from fires in the North or Southwest region of China usually were small, air pollutions could be still heavy during a smoke period because of the large background PM 2.5 level.
The impact of smoke from fires in Northeast China on air quality in Russia is a concern for the fire and air quality management in both China and Russia. A related issue is smoke transport into China from fires in Russia. The Euro-Asia boreal, mainly in Siberia, is one of the major wildfire regions in the world. The fire number, size, intensity, and duration in this region usually are much more than those in Northeast China. Smoke from the fires in Siberia can be transported a long distance to affect air quality in eastern Asia including China. For example, the smoke from the 2003 spring Siberia fires was transported to Northeast and North China with PM 2.5 concentrations exceeding 55 µg/m 3 (unhealthy) [53]. Further research is needed to simulate the across-boundary smoke transport to improve the assessment of the contributions of wildfires to the air pollutions in China.
One of the limitations with this study is that only a limited number of fire cases were simulated. Because the spatial extend and magnitude of exposures and resultant health outcomes depend on the location, timing and duration of wildfires as well the prevailing meteorological conditions, the results from this study only included certain spatial patterns and magnitude of smoke transport and impacts. The fire cases in the Northeast region simulated in this study included major largest fires in this region during recent two decades, including the fires during 2003 and 2006 when nearly 9000 km 2 and 4700 km 2 were burned, respectively, much more than other years during the two-decade period ( Figure A10). Fires in the North and Southwest regions were usually much smaller than these in the Northeast region because of the moister conditions (especially in the Southwest region) and better accessibility for fire suppression due to closer fire sites to densely populated areas (especially in the North region). Also, fires usually occurred either in the mountains west of the downwind densely populated areas or within densely populated areas. Thus, it is expected that the simulation results of the fire cases investigated in this study, despite a small number, reflected some general spatial patterns and magnitude of smoke transport and impacts in the two regions.

Conclusions
Smoke has been simulated for more than a dozen wildfires in the Northeast, North, and Southwest China. The regional differences in smoke transport and the air quality and human health impacts obtained from the modeling results provided evidence for the hypothesis for this study, that is, individual fires with very large sizes in Northeast China had a large amount of emissions but a small chance to affect air quality in densely populated areas, while fires in North and Southwest China usually had small emissions with large local impacts in some cases. This finding suggests that the fire and air quality management should focus on fires with very large sizes under wind directions toward the south in Northeast China and local air pollutions from fires in the North and Southwest China. Figure A1. Simulated smoke plume from the Huma fire. Each square box is a model grid point (approximately 25 km × 25 km). The names in black are provinces. The green, yellow, orange, red, brown, and purple colors represent good, moderate, unhealthy for sensitive groups, unhealthy, very unhealthy, and hazardous air quality levels, respectively. Panels