The southwest wind prevailed at the Yunnan University (YNU) Chenggong station from 1 March to 30 April 2019, with wind speed in a range of 2–5 m s⁻¹ (Figure 2A). The daily variations of PM, BC, aerosol scattering coefficient (525 nm), and absorbing coefficient (532 nm) were similar to each other (Figures 2B–D). The mean concentrations of PM_2.5 and PM₁ were 34.5 and 30.5 μg m⁻³, respectively, during the period. It was striking that the daily mean PM_2.5 and PM₁ concentrations increased remarkably from 20.2 μg m⁻³ to 17.2 μg m⁻³ on 30 March to 93.5 and 87.7 μg m⁻³ on 1 April, which exceeded the Grade II National Ambient Air Quality Standard (75 μg m⁻³) and indicated the occurrence of an air pollution event. At the same time, BC concentration, aerosol scattering coefficient, and absorbing coefficient also increased simultaneously within 48 h, with BC₃₇₀ and BC₈₈₀ concentrations, the scattering and absorbing coefficients increased by 12.1 μg m⁻³, 7.7 μg m⁻³, 471.6 Mm⁻¹, and 63.5 Mm⁻¹, respectively. During the pollution episode, the fractional contributions of long-range transport of BB pollutants from Southeast Asia to the YNU site were estimated to be approximately 80% for PM₁, 78% for PM_2.5, and 83% for BC₈₈₀ concentrations, respectively, which demonstrated a substantial influence of the BB on the air quality of KM in spring 2019. Previous studies found that BB from Southeast Asia could explain about 20–70% of PM_2.5 in Yunnan Province (Fu et al., 2012) and 57% of the total AOD in KM (Zhu et al., 2017). It is noted that our results were slightly higher compared to previous studies, which was due to the PM_2.5 concentration in a typical event was reasonably higher than that of the monthly mean.

For more details, Figure 3 shows the hourly mean values of PM and BC concentrations, scattering coefficient, absorbing coefficient, meteorological variables and the vertical aerosol profiles from 30 March to 2 April 2019, when the air pollution event occurred. PM and BC concentrations and scattering coefficient began to increase on 31 March, with the maximums of PM_2.5 and PM₁ concentrations reached 130.1 and 122.7 μg m⁻³ in the early morning of 2 April, respectively (Figures 3A–C). The southwest wind prevailed during the episode with a wind direction of 180°–270°, and the maximum wind speed in the daytime was larger than 5 m s⁻¹, suggesting the air mass was from the southwest. The westerly wind abruptly increased at 7:00 LST on 2 April, resulted in the dissipation of pollutants and the ending of the pollution episode (Figure 3D).

The maximum of BC concentrations, the scattering, and absorbing coefficients appeared around noon on 1 April (Figures 3B, C). Wang et al. (2011) indicated that BC₃₇₀ concentration was generally much larger than BC₈₈₀ concentration in the air mass of BB due to aerosols from BB could strongly absorb light at ultraviolet wavelengths. The concentration of BC₃₇₀ was larger than that of BC₈₈₀ during the pollution episode in this study, with the maximum differences between BC₃₇₀ and BC₈₈₀ up to 9 μg m⁻³ and 7 μg m⁻³ on 1 April and 2 April (Figure 3B), respectively, suggesting the considerable influence of BB from Southeast Asia through long-range transport. Previous studies indicated that wildfire was one of the primary sources of CO (Mafusire et al., 2016). Figure 3F shows the hourly variation of NOx, SO₂ and CO concentrations, demonstrating that CO rather than NOx and SO₂ increased simultaneously with PM and BC. The hourly CO concentration reached the maximum value of 1.1 ppm on 2 April, much higher than the mean CO concentration (0.5 ppm) in March and April 2019, indicating that the primary source of the pollutants observed in KM was BB over Southeast Asia.

The vertical aerosol profiles detected by lidar at the YNU Chenggong station are shown in Figure 3G. Aerosols were mainly concentrated near surface on 30 March, and the aerosol extinction coefficient was less than 0.15 km⁻¹. The aerosol extinction coefficient slightly increased to 0.3 km⁻¹ on 31 March, while high aerosol concentration was still mainly concentrated below 1 km. It was also noticed that, below 1 km, the aerosol concentration gradually accumulated on the night of 31 March and the high extinction coefficient larger than 1 km⁻¹ abruptly appeared in the early morning of 1 April. In addition to the accumulation of near surface aerosols, there could be a downward transport of aerosols from the altitude of 1.5–3 km to the surface, suggesting that long-range transport coming from different paths occurred at upper and lower altitudes. There was another weak peak of the aerosol extinction coefficient on 2 April, and the aerosol extinction coefficient decreased at noon on 2 April.

Figure 4 shows the observed hourly concentrations of PM_2.5 at the four cities in Yunnan Province during the pollution episode and the 29-h backward trajectories ending at 0 m altitude and 05:00 UTC (13:00 LST) on 1 April at these cities, when the hourly PM_2.5 concentration in KM reached the maximum. It was noteworthy that the temporal variations of PM_2.5 concentration were alike among these cities (Figure 4A), the maximum PM_2.5 gradually decreased from south to north, negatively correlated with the distance to the fire regions in Southeast Asia, and the air mass arriving at the four cities came from near the borders of Myanmar, Thailand, and Laos (Figure 4B). It is noticed that the PM_2.5 concentration substantially increased (up to 173 μg m⁻³) at midnight on 1 April in JH, a city near the border between Myanmar and China (JH in Figure 1B). The PM_2.5 maximum up to 114 μg m⁻³ appeared around noon on 1 April in KM (the northernmost site among the four cities) (Figure 4A), about 12 h later than that in JH, reflecting an evident regional transport path of Southeast Asian BB regions to Yunnan Province.

To further identify the potential sources of this pollution event, the vertical aerosol profiles detected by the CALIPSO satellite and the backward trajectories were analyzed (shown in Figure 5). The CALIPSO satellite passed over Southeast Asia and western Yunnan Province from north to south at 19:22–19:26 UTC on 31 March. It shows that the air masses arriving at the locations on the satellite orbit track in Yunnan Province were mainly from Myanmar (Figure 5A). It was striking that there was an approximately 2 km vertically thick smoke layer at 1–4 km altitude extending from northern Thailand (98.4°E, 19°N), crossing eastern Myanmar (99.1°E, 22.2°N) to western Yunnan Province (99.8°E, 25°N) (Figures 5B, C, blank areas denote data unavailable due to cloud, etc.), with the maximum aerosol extinction coefficient >2 km⁻¹. Figure 5D exhibited negative vertical velocities (upward movement of air mass) from 19°N to 24°N along the track, uplifting most fire plumes at 3 km altitude. These satellite retrievals indicated the smoke plumes mainly originated in the fire emissions over eastern Myanmar and the long-range transport of fire aerosols towards Yunnan Province, consistent with the analysis of surface observations and backward trajectory. It should be mentioned that the elevation gradually increased from Southeast Asia to Yunnan Province (Figure 5C), and the height of the Chenggong station and KM was 1920 m above sea level. Although there was no CALIPSO track crossing KM (102.71°E, 25.01°N) during this episode, the proximity of KM to the track and the large-scale transport characteristic of the fire plumes suggested that the BB over Southeast Asia likely contributed to the pollution event in KM on 1 April.

The MODIS and VIIRS data provide more details on aerosol properties (Figure 6), although the data is unavailable over the central and northern parts of Yunnan Province. The regions with high AOD corresponded well to those with large fire spots, indicating the presence of BB particles. For example, the fires were widely distributed in Myanmar, Laos, and northern Thailand on 30 March 2019 (Figure 6A), and high AOD was concentrated over Laos and northern Thailand (Figure 6E), and smoke exactly existed on the border between Myanmar, Laos, Thailand, and China (Figure 6I). Furthermore, the fire spots were concentrated in Laos on 31 March (Figure 6B), and high AOD moved northward and westward to the border between Laos, Vietnam, and China (Figure 6F). The smoke on the border between Myanmar, Laos, Thailand, and China also moved northward, and new smoke appeared in northern Myanmar (Figure 6J). The fire spots increased in eastern Myanmar and Laos on 1 April (Figure 6C); it was impressive that the smoke plumes moved further northward and reached KM (Figure 6K), with AOD up to 1.8 (Figure 6G), consistent with the above surface observation analysis for the pollution event in KM. AOD values decreased in KM and most parts of Yunnan Province on 2 April (Figure 6H).

Biomass fires occurred in the northern Indochina Peninsula almost over the entire period of March-April 2019, but BB pollutants were not always transported into Yunnan Province during this period. There was a northwest wind in Myanmar before 29 March, and the fire plumes were transported to Thailand and downwind regions (Figures 7A, B). A shallow trough (the India-Myanmar trough) appeared around 90°E just to the west of Myanmar on 29 March and was strengthened and moved eastward on 30–31 March (Figures 7C, D). Although the trough line disappeared on 1 April, southwestern wind was strengthened under the compression of the South China Sea high pressure (Figure 7E). The occurrence of the trough changed the wind direction from northwest to the southwest over the fire regions and drove the pollutants to southwest China, which led to a remarkable enhancement of PM levels over most areas of Yunnan Province and the pollution event in KM on 1–2 April. The high pressure disappeared after 2 April, accompanied by reducing fire spots and weakening southwestern wind (Figure 7F).

The PM_2.5 gradually moved toward the northeast of Yunnan Province, accompanied by the development of the India-Myanmar trough. The daily mean PM_2.5 concentrations exceeded 70 μg m⁻³ in southern Yunnan Province on 28–30 March due to the transport of BB plumes, whereas those in KM were less than 30 μg m⁻³ (Figures 8A–C). A southerly convergence zone formed in the vicinity of KM on 31 March, accompanied by the strengthening of the India-Myanmar trough (Figure 8D), which favored the accumulation of pollutants from BB and contributed to the PM increase, thus, the daily mean PM_2.5 concentrations increased in the whole of Yunnan Province. The PM_2.5 concentration exceeded 110 μg m⁻³ in southern Yunnan Province, whereas that in KM was larger than 50 μg m⁻³. The southerly further increased on 1 April, which contributed to the PM increase, with the daily PM_2.5 concentration exceeding 80 μg m⁻³ in KM (Figure 8E). The southerly gradually decreased after the India-Myanmar trough disappeared on 2 April (Figure 8F), leading to an abrupt decrease in the PM level from 07:00 LST (Figure 3A).

Along the southwest wind direction in front of the trough, the vertical profiles of daily mean meridional wind and the model simulated vertical profiles of daily mean PM_2.5 concentrations produced by BB in Southeast Asia across JH and KM from 28 March to 2 April 2019 are shown in Figure 9. The deep southerly wind produced by the India-Myanmar trough favored the transmission of pollutants. When the India-Myanmar trough appeared just to the west of Myanmar on 29 March, the weak southerly layer appeared near the ground north of JH (Figure 9B1), and the PM_2.5 produced by BB in Southeast Asia remained to the south of JH (Figure 9B2). When the India-Myanmar trough strengthened on 30–31 March, the southerly also strengthened (Figures 9C1, D1), and the PM_2.5 had been transported from the border between Myanmar and Thailand to KM (Figures 9C2, D2). The southerly was most robust on 1 April, the maximum wind speed exceeded 8 m s⁻¹ (Figure 9E1), and PM_2.5 produced by BB in Southeast Asia spread furthest (Figure 9E2). The southerly decreased on 2 April (Figure 9F1), accompanied by the reduction of the transmission of PM_2.5 (Figure 9F2).

Model comparison against various observations demonstrated that the model could reasonably reproduce meteorological variables, and the model simulated air pollutants concentration generally agreed with observations. Table 2 presents the statistics between model simulations and observations in the four cities of Yunnan Province. The air temperature and relative humidity at 2 m were well reproduced, with correlation coefficients (COR) of 0.85–0.94 and 0.82–0.86, the normalized mean biases (NMB) of −11.7% ∼-1.3% and −34.9%–0.2%, respectively. The underestimation of RH2 in the JH could be due to more active and complex land-atmosphere interactions in the tropics. The model simulated wind speed and direction reasonably well in KM, with COR of 0.78 and a slight bias (<5%), but the model tended to predict higher wind speed in JH and lower wind speed in YX and PE, with NMBs of 54.8% and approximately −20%, respectively. Statistics for comparison between observations and the S1 simulations for air pollutants at these cities are also presented in Table 2. The performance statistics for PM_2.5 were generally good, with the COR of 0.32–0.51 and the NMB of −36%∼−67.3%, respectively. The performance for trace gases (O₃, NO₂, and CO) was generally poorer than that for PM_2.5, with COR of 0.13–0.58. In order to evaluate the simulation performance of PM_2.5, the simulation results of 10 provinces in southern China other than Yunnan Province were also compared with the observations (not shown in Table 2), with the mean COR of 0.36 and the NMB of 11.7%, respectively, indicating that the simulation results of PM_2.5 were approved.

The underestimation of PM_2.5 in Yunnan province could be associated with potential uncertainties in emission inventories, meteorological fields, heterogeneous chemistry (Zheng et al., 2015), the representatives of ground-based measurements, or the relatively coarser grid resolution. The possible underestimation of BB emissions in Southeast Asia and anthropogenic emissions in Yunnan province was one of the critical reasons for the underestimation of PM_2.5. The simulation bias of PM_2.5 would be significantly improved when the anthropogenic emission of Yunnan Province was doubled, with the NMB in a range of −50%–16.4%. The NMB and COR of the simulated mean concentration of PM_2.5 in JH decreased from −38.6% to 1.6%, and from 0.32 to 0.23, respectively, when BB emissions from FINN were applied, which was due to the higher emission amount (about 3–5 times of that in GFED) and courser temporal resolution of FINN. While the simulation results of the cities farther from the border (KM, YX, and PE) had not been improved obviously by using BB emission from FINN, with main underestimation in peak concentrations during 31 March to 1 April when the transport pollution occurred, indicating that there could be possible simulated bias in the meteorological fields. It is worth noting that the WS10 of YX and PE was significantly underestimated, with NMB of −19.9% and −20.7%, respectively, thus the BB pollutants that should be transported here might not arrive totally, which also reasonably explained the occurrence of main underestimation during transport pollution episode. The underestimation of simulated PM_2.5 could also be affected by meteorological fields through the heterogeneous chemistry. For example, the underestimation of PM_2.5 in KM and JH might be due to the underprediction of RH2, with had a significant impact on the secondary formation of PM_2.5 by heterogeneous reaction. In addition, ground-based measurements were mainly conducted in urban aeras and the observation results might be affected by the surrounding highways or catering, and the coarser resolution of the model could not reproduce this situation well.

The simulation results derived from GFED with finer temporal resolution were used for further analysis considering that a typical pollution event with short duration was mainly focused on in this study, and the simulation results derived from FINN were used to analyze the potential uncertainty of simulation results caused by emission differences among two BB inventories. Figure 10 presents the simulated daily mean near-surface PM_2.5 concentrations (Figures 10A1–F1) and PM_2.5 column concentrations in the troposphere (Figures 10A2–F2) produced by BB in Southeast Asia on 28–31 March and 1–2 April during the severe BB episode, which are derived from the simulation of Base run and S1 (S1 minus Base) by using GFED. It is noteworthy that the contributions of BB in Southeast Asia to the PM_2.5 concentration in Yunnan Province were calculated using the difference between the two scenarios (the last paragraph of Section 2.2 for details), which effectively reduced the bias caused by the underprediction of aerosols. High concentrations persistently occurred over western Myanmar, the border between Myanmar and Thailand, and northern Laos, consistent with the location of BB emissions (Figure 1B), with the maximum daily mean PM_2.5 concentration reaching 250 μg m⁻³. From 28 March to 31 March, the daily mean near-surface PM_2.5 concentration in western Myanmar gradually decreased from >250 μg m⁻³ to <50 μg m⁻³, while the PM_2.5 concentration in the Myanmar-Thailand border and northern Laos showed a trend of first increasing and then decreasing (Figures 10A1–D1). Afterward, another small peak appeared on 1 April and gradually dissipated on 2 April (Figures 10E1, F1). It should be noted that the near-surface PM_2.5 concentrations in most areas of the low-latitude plateau in China (e.g., Yunnan Province) were slightly affected by the transboundary process on 29 March, when there was a northerly wind; the contribution of the transboundary process to the near-surface PM_2.5 concentration in KM (red star located) was less than 1.0 μg m⁻³.

The temporal variation of the daily mean PM_2.5 column concentrations during this episode was similar to that of the near-surface PM_2.5 concentrations, reaching a maximum value on 30 March and 31 March (Figure 10C2, D2) and then dissipating on 2 April (Figure 10F2). However, the spatial distribution of daily mean PM_2.5 column concentrations significantly differed from near-surface PM_2.5 concentrations. Before 29 March, PM_2.5 was transported toward the east in westerly. The distribution of daily mean PM_2.5 column concentrations was characterized by a southwest-northeast belt extending from the areas where BB occurred in Southeast Asia to southern China after 30 March (Figures 10C2–F2), caused by the deep southerly layer produced by the India-Myanmar trough, with the maximum concentration exceeding 250 mg m⁻² on 30 March at the southern border of Yunnan Province. Although the near-surface PM_2.5 concentrations decreased after 31 March, the high daily mean PM_2.5 column concentrations remarkably expanded toward the northeast of Yunnan Province and even reached Guizhou Province (Figure 10D2). The PM_2.5 column concentration in the troposphere in Yunnan Province was affected by the transboundary process of BB in Southeast Asia, with a maximum contribution of 150 mg m⁻² during this episode, which indicated that a large part of air pollutants had been uplifted during the burning and then transported and diffused into China. It is noticed that the near surface concentration and tropospheric column concentration derived from FINN showed the same spatial distribution and temporal variation characteristics as those from GFED, except that the daily mean concentrations in source regions and its nearby regions (e.g., the southern border of Yunnan Province) were higher, with the maximum reaching 300 μg m⁻³ near the surface and 300 mg m⁻² in the troposphere, respectively. This indicated that larger impacts on the air quality of Yunnan province would be produced by much higher BB emissions in Southeast Asia.

Figure 11 shows the simulated average contribution of BB in various regions of southeast Asia on mean near-surface PM_2.5 concentrations and PM_2.5 column concentrations in the troposphere during the severe BB episode from 28 March to 2 April 2019, which are derived from the differences between the two scenarios (the last paragraph of Section 2.2 for details) by using GFED. The aerosols in Regions 1 and 2 might be introduced from the western border and the southwestern border of Yunnan Province under the transmission of west and southwest winds. In contrast, the aerosols in Region 3 mainly affected aerosol concentrations in the southeastern border of Yunnan Province. The near-surface PM_2.5 concentration of KM was rarely transferred from Region 1 (only 1–10 μg m⁻³). However, Region 1 contributed 15–20 mg m⁻² of the PM_2.5 column concentration in the troposphere in KM, which demonstrated that the transfer from Region 1 to Yunnan Province might have occurred at a higher altitude (Figure 11A). BB in Region 2 (Myanmar-Thailand border) had the most dominant impact on PM_2.5 column concentration in the whole of Yunnan Province, contributing more than 100 mg m⁻² to the PM_2.5 column concentration in the southern border of Yunnan Province and it could be transported to Guizhou Province and Guangxi Province. However, Region 2 only contributed 1–10 μg m⁻³ to near-surface PM_2.5 concentration in KM (Figure 11B). BB in Region 3 (northern Laos) had a dominant impact on PM_2.5 column concentration in the southeastern border of Yunnan Province, and was further transported to Guangxi Province (Figure 11C). The contribution of other regions (OR) to near-surface and tropospheric column PM_2.5 concentrations in Yunnan Province was almost negligible (Figure 11D). There was a negligible discrepancy in the mean PM_2.5 concentration and spatial distribution near the surface over Yunnan Province between the simulation results from GFED and those from FINN except Region 3. The near surface PM_2.5 concentration in Region 3 derived from FINN was much higher than that from GFED, with a maximum up to 300 μg m⁻³. Moreover, the BB in Region 1 contributed 20–25 mg m⁻² to the tropospheric column concentration of KM, exceeding the contribution of Region 2, with almost constant contributions in Region 2, Region 3 and other regions to tropospheric column PM_2.5 concentration in KM.

The vertical distribution of PM_2.5 from BB during the severe BB episode from 28 March to 2 April 2019 averaged over Yunnan province is exhibited in Figure 12, which are derived from the differences between the two scenarios (the last paragraph of Section 2.2 for details) by using GFED. The PM_2.5 mass per unit height in Yunnan Province produced by BB in Southeast Asia gradually decreased from ∼5,000 kg to 0 kg from near surface to 10 km, in which the larger transmission mass was concentrated in the boundary layer below 4.1 km. The maximum transmission occurred in the layer of 0.9–2.1 km, contributing 30% of the total transmission mass in the troposphere due to the larger transmission mass per unit height (∼4,000 kg) and the larger thickness (1,200 m) of this layer. Consistently, The PM_2.5 mass per unit height in Yunnan Province derived from FINN also showed a decreasing trend from near surface to 10 km, with the larger transmissi on mass concentrated in the boundary layer below 4.1 km. The PM_2.5 mass per unit height reached ∼8,000 kg near surface, higher than that from GFED.

The PM_2.5 mass per unit height of each vertical layer in Yunnan Province from different regions and pathways are further explored and identified. Figure 12A shows the relative contributions of individual regions to PM_2.5 mass. The average contribution of Region 1, Region 2, Region 3, and other regions to the mean mass of PM_2.5 in the entire 12 layers were 39.8%, 50.7%, 5.7%, and 3.8%, respectively. It is striking that the changes in the contributions of Region 1 from near surface to 10 km were opposite to those of Region 2. The larger contributions of Region 1 were concentrated in higher layers, and the larger contributions of Region 2 focused on lower layers. For example, the mean percentage contributions of Region 1 to PM_2.5 mass were from 21.4% at the near-surface to 75.8% at the layer of 7.9–10 km, increasing with height. However, the contributions of Region 2 to PM_2.5 mass decreased from 60.5% at the near-surface to 15.6% at the layer of 7.9–10 km, decreasing with height. Below 4.9 km, the PM_2.5 in Yunnan Province was mainly sourced from Region 2, while above 4.9 km, Region 1 contributed the most of PM_2.5 in Yunnan Province. Figure 12B shows the relative contributions of PM_2.5 mass produced by primary emission and secondary formation of BB. The average contributions of primary emission and secondary formation to the mean mass of PM_2.5 in the entire 12 layers were comparable, contributing 48% and 52%, respectively. Interestingly, the changes in the contribution of primary emission from near surface to 7.9 km were opposite to secondary formation. For example, the mean percentage contributions of PM_2.5 mass produced by primary emission were larger than those of PM_2.5 mass produced by secondary formation below 0.9 km. In comparison, the contributions of secondary formation exceeded those of primary emission in the layers from 0.9 to 7.9 km. It is also noticed that the contributions of Region 1 to PM_2.5 from BB averaged over Yunnan province increased with the height, being opposite to those of Region 2 by suing FINN, which was consistent with the simulation results from GFED. The average contribution of Region 1, Region 2, Region 3, and other regions to the mean mass of PM_2.5 in vertical layers were 35.2%, 44.0%, 18.5%, and 2.4%, respectively by using FINN. The contribution of Region 3 had increased significantly due to higher concentration near surface, while Region 2 was still the dominant contributor to the BB aerosols over Yunnan province. In addition, the simulation results presented a relatively higher contribution of PM_2.5 mass produced from secondary formation by using FINN, with an average over Yunnan province of 68%.