-
In this work, we compared the simulated T2, as well as RH and wind field at 500 hPa, with the CRU and ERA-interim reanalysis datasets (Fig. 2). The WRF-Chem model represents the spatial distribution pattern of annual T2 well (Fig. 2a). Warmer regions appeared in South Asia (T2s>20°C), but colder regions appeared over the TP (T2s<0°C). Compared with the CRU (Fig. 2b), WRF-Chem captured the topographically induced variation of T2 over the TP in more detail, owing to its higher spatial resolution. Both WRF-Chem and ERA-interim results indicate that high 500-hPa RH occurred over the TP, but RH was relatively lower in South Asia (Figs. 2c and d). The model predicted slightly lower RH at 500 hPa over the northern TP compared with ERA-interim, which could be because of the simulation bias in temperature affecting the saturation pressure of water vapor (Yang et al., 2018b). The model also effectively reproduced the dynamics for the wind field at 500 hPa. There were prevailing westerly winds over the northern TP and southwesterly winds over the southern TP (Figs. 2e and f). Over South Asia, westerly winds from land to ocean were predominant. Overall, this simulation configuration captured the meteorological fields well, which is crucial to assure prediction accuracy of air pollutant concentrations.
Figure 2. Comparisons of annual T2, as well as RH and wind at 500 hPa between the WRF-Chem simulation and the reanalysis datasets.
Further, we quantitatively evaluated the model performance by using observations at stations in the study area (Fig. S1a in the Electronic Supplementary Material, ESM). The WRF-Chem model generally represents the correct annual variation trends of temperature and humidity at the stations. Compared with the observation, the model simulated higher T2 from June to September and slightly lower values in other months (Fig. S1b). On the other hand, the model simulated lower 2-m relative humidity than the observation from June to September (Fig. S1c). The corresponding statistics between WRF-Chem simulations and in situ observations are presented in Fig. S1.
-
We summarized surface O3 (Fig. 3a) and PM2.5 (Fig. 3b) concentrations from 12 sites to compare with the simulation results. The observed results show very high O3 concentrations at TP sites. As in Haizi, Yushu, and Ngari, surface O3 concentrations exceeded 100 µg m−3 in the pre-monsoon months. However, low O3 levels appeared over South Asia, such as in Delhi (below 30 µg m–3), at 1–2 orders of magnitude less than those over the TP sites. The simulation reflects the spatial variability of O3 concentrations from the TP to South Asia. When the monthly variation of O3 is considered, its concentration is found to have been lowest during the monsoon months over the TP. This might be because pollutants experience the washout effect in the rainy months (Yin et al., 2017). Moreover, the sky was always overcast during the monsoon season, thus, lower solar flux weakened the photochemical processes (Reddy et al., 2008). O3 concentration was also lowest during the summer months at sites over South Asia. The highest O3 concentration was found in May at Delhi and Mohali, partly owing to low humidity and high concentration of precursors (Lamaud et al., 2002). The model exhibited the annual trends of O3 concentrations well at all TP sites, except for Haixi. Overestimation of O3 concentration occurred in spring and underestimation occurred in autumn at Yushu, Ngari, Qamdo, Lhasa, and Nyingchi. The O3 concentration was uniformly underpredicted at other TP sites, such as Guolog, Haixi, and Nagqu. In South Asia, the seasonal trend of O3 concentration was reproduced well at all sites, despite a slight overestimation.
Figure 3. The seasonal variations of observed and simulated O3 (a) and PM2.5 (b) concentrations at sites.
The model also represented the spatiotemporal variation of PM2.5 concentrations well at all sites (Fig. 3b). Both the observation and simulation showed lower PM2.5 over the TP (e.g., in Yushu and Lhasa) and higher PM2.5 at sites over South Asia (e.g., in Agra and Kosmarra), which is opposite to the spatial variability of O3 concentration from the TP to South Asia. Simulated PM2.5 concentrations showed underestimations compared to in situ observations over the TP. However, it should be noted that there is uncertainty in the emissions inventory. For example, in Lhasa, the underestimation of residential emissions was considered as a crucial factor causing the underestimation of PM2.5 (Li et al., 2019). The temporal variability of PM2.5 concentration was reproduced well over the TP, with higher concentration values in winter months but lower values in summer. However, at Haixi and Guolog, PM2.5 concentration reached its peak in May (42 µg m–3) and June (43.5 µg m–3), respectively, possibly due to westerly winds bringing dust to those locations during this period (Jia et al., 2015). High PM2.5 concentrations were observed at sites over South Asia, which is one of the most densely populated regions in the world and has high local emission sources including household, vehicular, mining, and urbanization (Nair et al., 2007). Both observed and simulated PM2.5 concentrations reached their maximum values during winter months and reached their minimums during summer months over South Asia, except at Punjab. The model accurately captures the highest PM2.5 concentration in April (101.1 µg m–3) at Punjab, which could be caused by agricultural waste burning emissions (Kharol et al., 2012). April in Punjab, India happens to be the start of one growing season and end of another, so crop residue burning was prevalent there. As seen in Fig. S2 in the ESM, the pre-monsoon PM2.5 emissions from deforestation and wildfires increased significantly in western India, compared with other seasons.
Table 1 summarizes the model statistics of O3 and PM2.5 concentrations at the observed sites, including mean observation (OBS), mean simulation (SIM), normalized mean bias (NMB), and normalized mean error (NME). The performance criteria (NMB <±30%, NME <50%) for PM2.5 and the performance criteria (NMB <±15%, NME <25%) for O3 are suggested by Emery et al. (2001). Overall, the model statistics for O3 and PM2.5 in the region meet the model performance criteria. There are large underestimations of PM2.5 in Guolog and Nagqu, partly because the model grid represents a regional average, but also, the observation site is intensely affected by local anthropogenic emissions such as industrial facilities and traffic. In addition, the model generally reflects the O3 and PM2.5 concentrations at these sites during the pre-monsoon season, with R2 values of 0.89 for PM2.5 and 0.81 for O3 (Fig. S3 in the ESM), respectively.
Obs
SimO3
NMB
NME
R
Obs
SimPM2.5
NMB
NME
RGuolog 85.3 66.7 −14.4 23.8 0.86 29.1 13.7 −36.9 36.9 0.71 Haixi 82.4 64.1 −12.9 21.1 0.74 24.0 16.8 −28.1 43.7 0.76 Yushu 55.9 52 −6.9 22.6 0.93 18.4 13.1 −29.1 29.3 0.97 Mgaro 54.9 54.2 −1.4 15.7 0.94 13.9 12.5 −10.2 21.9 0.77 Qamdo 62.1 58.5 −5.8 10.7 0.89 20.9 14.5 −28.5 34.1 0.86 Lhasa 65.9 59.5 −9.7 15.4 0.80 23.1 17.3 −25.4 40.4 0.73 Nylingchi 66.7 59.9 −10.1 15.5 0.92 13.4 16.6 24.2 30.6 0.89 Nagqu 38.6 44.2 12.5 14.5 0.91 48.7 32.9 −32.9 32.9 0.72 Pune 30.9 37.1 13.8 22.5 0.94 − − − − − Delhi 16.4 21.5 12.3 21.4 0.76 − − − − − Anantapur 31.0 34.5 11.3 11.3 0.98 − − − − − Mohali 40.4 37.6 4.4 7.6 0.95 − − − − − Agra − − − − − 112.0 79.8 −28.7 30.8 0.95 Mumbai − − − − − 27.7 25.8 −6.9 14.7 0.92 Kosmarra − − − − − 73.6 72.5 −1.8 8.6 0.96 Punjab − − − − − 65.4 69.7 6.4 13.7 0.88 Table 1. Model performance for O3 and PM2.5 concentrations at sites. OBS is mean observation; SIM is mean simulation; NMB is Normalized mean bias; NME is Normalized mean error. The performance criteria (NMB <±30%, NME <50%) for PM2.5 and the performance criteria (NMB <±15%, NME <25%) for O3 are suggested by Emery et al. (2001). The values that do not meet the criteria are marked in bold.
-
Figures 4a–d show the seasonal and spatial variations of simulated surface O3 concentration. O3 concentration was found to be higher over the TP but lower in South Asia, because of more intense photochemical reactions and downward transport of stratospheric air mass over the TP (Yin et al., 2017). Seasonally, higher O3 concentrations occurred in the pre-monsoon season over the TP (Fig. 4a), likely owing to stronger stratosphere–troposphere exchange during this pre-monsoon season compared with other seasons (Yin et al., 2017). O3 concentration declined to its minimum during the monsoon season over the TP (Fig. 4b), consistent with previous in situ observations (Lin et al., 2015; Yin et al., 2019). Since the greatest precipitation occurs during the monsoon season (Fig. S4 in the ESM), O3 and its precursors could be removed, thus reducing photochemistry and O3 concentration over the TP (Ma et al., 2014). As for South Asia, surface O3 concentrations are higher over northeastern India during all seasons except for the monsoon season, which may be attributed to higher O3 precursor concentrations in this region, such as CO (Fig. S5 in the ESM).
Figures 4e–h illustrate the seasonal and spatial variation of surface PM2.5 concentrations. PM2.5 concentration is observed to be high in eastern and northeastern regions of South Asia during the pre-monsoon season (Fig. 4e), possibly due to the local forest fires. As seen in Fig. S2, PM2.5 emissions from deforestation and wildfires in eastern South Asia were significantly greater during the pre-monsoon season than during other seasons. Meanwhile, higher PM2.5 concentrations appeared over the TP during the pre-monsoon season compared with other seasons; this was accompanied with a decreasing trend from the southwest TP to the northeast TP. There are very little local emissions over the TP, and westerly winds prevail during the pre-monsoon season. Thus, high PM2.5 mass over the TP during the pre-monsoon season probably was caused by the cross-border transmission from western India. The lowest PM2.5 concentrations (below 60 μg m–3) are observed during the monsoon season (Fig. 4f) because the large amount of precipitation (Fig. S4) in this season leads to intense wet scavenging of particulate matter. PM2.5 concentration is found to be high over the southern Indian subcontinent during the post-monsoon season (Fig. 4g) and winter (Fig. 4h) because low wind speeds at the surface (Fig. S6 in the ESM) as well as little precipitation (Fig. S4) are unfavorable to pollutant diffusion and dispersion. Higher PM2.5 concentrations occurred over the TP during the post-monsoon season and winter compared with the monsoon season, which is partly because of stagnant meteorological conditions during the post-monsoon season and winter, i.e., less precipitation (Fig. S4d) and lower wind speeds (Fig. S6d); it is also partly because of the cross-border transmission by the large-scale air circulation and regional mountain–valley wind.
The spatial distributions of PM2.5 components had significant seasonal variations (Fig. 5). Secondary inorganic components (SO4, NO3, and NH4) showed higher concentration values during the post-monsoon season and winter over South Asia, whereas the concentrations were lower during the monsoon season. Increasing residential emissions of SO2 and NOx during winter (Fig. S7 in the ESM) resulted in an increase of surface heterogeneous reactions for the formation of SO4 and NO3 in this period. The NO3 to SO4 mass ratio refers to the relative dominance of stationary versus mobile emission sources (Wan et al., 2016). The NO3/SO4 mass ratios below 1 in India during the pre-monsoon season indicate that the mobile emissions are not the main source there. Besides, colder weather in winter is more preferential to ammonium nitrate being partitioned into the particle phase (Aw and Kleeman, 2003). Primary PM2.5 component (BC and OC) concentrations over the TP reached their highest values during the pre-monsoon season (Figs. 5m and q) due to the cross-border transmission. High primary component concentrations occurred over the Indian subcontinent during winter (Figs. 5p and t), driven by the significant increase of residential emissions during winter, as well as stagnant weather conditions as discussed above.
Winter SO4, NO3, and NH4 concentrations were 50%–250% higher than the annual average concentrations in southern India and the Bay of Bengal (Fig. S8 in the ESM). However, winter SO4 concentration is lower than the annual mean over the TP, where NO3 and NH4 do not evidently deviate from the average. During the pre-monsoon season, secondary inorganic components are 50%–100% higher than the annual mean over the western TP (Figs. S8a, c, and e). As for primary components over the TP and its southern slope, BC concentration is 50%–200% higher during the pre-monsoon season than its annual mean, while OC concentration during the pre-monsoon season is approximately 100%-250% higher than its annual mean (Fig. S9 in the ESM). Higher BC and OC concentrations during the pre-monsoon season indicate that primary components were the more vital contributor to the maximum PM2.5 concentrations over the TP during the pre-monsoon season. Moreover, maximum PM2.5 concentrations in the northern Indo-Gangetic plain and the Himalayan foothills could be attributable to the primary components which are approximately 100-150% of the annual mean, whereas secondary inorganic components are lower than their annual means during the pre-monsoon season (Fig. S8).
-
In sections 3.1 and 3.2, we validated the WRF-Chem performance for meteorological fields, O3, and aerosol, using in situ observations and gridded data. The results suggest that this model framework is capable to further quantify the contribution proportion of South Asian biomass burning to O3 and aerosol concentrations. The contribution proportion is calculated by (A–W)/A, where A and W are pollutant concentration in the sensitivity and control simulations, respectively.
-
Biomass burning emissions affect O3 concentration mainly by influencing the emissions of O3 precursors (Andreae and Merlet, 2001). Figure 6 illustrates the contribution ratios of South Asian biomass burning to O3 concentrations during different seasons. South Asian biomass burning contributes to O3 primarily in South Asia, especially during the pre-monsoon season (Fig. 6a) and winter (Fig. 6d). This is because South Asian biomass burning significantly influences O3 precursors in South Asia, particularly CO (Fig. S10 in the ESM). During the pre-monsoon season, higher contribution ratios (up to 20%) of South Asian biomass burning to O3 concentration appeared in the Indo-Gangetic Plain and central India (Fig. 6a), which experience a large number of fire counts in this period (Xu et al., 2018). However, South Asian biomass burning contributed less than 1% of the O3 concentration over the TP during the pre-monsoon season, although the biomass burning from western India reached its maximum and westerly winds prevailed during the pre-monsoon season (Fig. S6a). Therefore, it can be inferred that high O3 concentration over the TP during the pre-monsoon season (Fig. 4a) is mainly a result of local weather and environmental factors such as strong stratosphere–troposphere exchange and high background concentration of O3. South Asian biomass burning has very little influence on O3 concentration during the monsoon season (Fig. 6b). During the post-monsoon season, high contribution proportion of South Asian biomass burning to O3 concentration was mainly concentrated in western India (Fig. 6c). Compared with that during the post-monsoon season, the fires from South Asia showed a larger contribution to O3 concentration in south India during winter (Fig. 6d), consistent with more fire counts in this region during winter.
-
Figure 7 shows the contribution ratios of South Asian biomass burning to PM2.5 and its components. During the pre-monsoon season, high contribution proportions (more than 60%) were observed over the TP and its southern slope (Fig .7a), coincident with the regions with highest PM2.5 concentration during the pre-monsoon season (Fig. 7e). This is primarily because South Asian biomass burning contributed significantly to primary PM2.5 components [BC (Fig. 7q) and OC (Fig. 7u)] over the TP and northern Indo-Gangetic plains during the pre-monsoon season. Moreover, the contribution ratios to primary components showed a gradual decreasing trend from the western TP to eastern TP. Considering that there are very few local emissions over the TP, the primary components over the TP must be mainly transported from western India. As seen in Fig. S11, South Asian biomass burning caused a clear increase of BC (Fig. S11a in the ESM) and OC (Fig. S11b in the ESM) concentrations along the foothills of the Himalayas and the eastern part of the Indian subcontinent, and these pollutants can be transported there from western India by northwesterly and southwesterly winds. We then analyze the cross-border transmission of South Asian biomass burning emissions into BC (Fig. S11c) and OC (Fig. S11d) concentrations over the TP along 30°N, which shows BC and OC in western India can reach 500 hPa and then be transported onto the TP by the westerly winds.
Figure 7. The contribution ratios of South Asian biomass burning to PM2.5 and its components during different seasons.
Meanwhile, the South Asian biomass burning contribution to secondary inorganic components was less than 10% over the TP and northern Indo-Gangetic plains (Figs. 7e, i, and m). The insignificant contributions of South Asian biomass burning to secondary inorganic components of PM2.5 could be attributed to its low contribution to the gaseous precursors of these secondary PM2.5 components (Fig. S12 in the ESM). The only exception is the relatively higher contribution ratios of South Asian biomass burning to NO3 concentrations over the Indian subcontinent in the pre-monsoon season (Fig. 6i), which is due to the larger contribution of South Asian biomass burning to local NO2 concentrations (Fig. S13 in the ESM).
The lowest contribution proportions of South Asian biomass burning to PM2.5 mass and its components were found during the monsoon season, mainly because of the minimal biomass burning emissions in this season. During the post-monsoon season, South Asian biomass burning contributed up to 50% of the PM2.5 mass in western India (Fig. 7c), owing to its significant contribution to BC (Fig. 6s) and OC (Fig. 7w) in this season. South Asian biomass burning also evidently affected the concentrations of primary PM2.5 components over the TP during the post-monsoon season (Figs. 7s and w), with contribution ratios of more than 10%. However, South Asian biomass burning contributed less than 10% of the PM2.5 over the TP in the post-monsoon season (Fig. 7c). It reflects that primary PM2.5 components accounted for less PM2.5 mass over the TP compared with secondary PM2.5 components (SO4, NO3, and NH4), which is consistent with higher concentrations of secondary components being present during the post-monsoon season (Figs. 5c, g, and k).
Although PM2.5 concentration was higher over the southern Indian subcontinent during winter (Fig. 4h) than during the pre-monsoon season (Fig. 4e), South Asian biomass burning contributed less toward winter PM2.5 there (Fig. 7d). This can be attributed to the same reason that South Asian biomass burning contributed very little to primary (Fig. 7h) and secondary inorganic (Figs. 7h, l, and p) components over the Indian subcontinent during winter. It implies that the highest PM2.5 concentration over the southern Indian subcontinent during winter (Fig. 4h) might be caused by other emission sources and stagnant weather conditions (less precipitation (Fig. S2d in the ESM) and low wind speed (Fig. S4d)). As shown in Fig. S14, residential and industrial emissions of PM2.5 were significantly greater during winter in southern India.
Obs | Sim | O3 NMB | NME | R | Obs | Sim | PM2.5 NMB | NME | R | |
Guolog | 85.3 | 66.7 | −14.4 | 23.8 | 0.86 | 29.1 | 13.7 | −36.9 | 36.9 | 0.71 |
Haixi | 82.4 | 64.1 | −12.9 | 21.1 | 0.74 | 24.0 | 16.8 | −28.1 | 43.7 | 0.76 |
Yushu | 55.9 | 52 | −6.9 | 22.6 | 0.93 | 18.4 | 13.1 | −29.1 | 29.3 | 0.97 |
Mgaro | 54.9 | 54.2 | −1.4 | 15.7 | 0.94 | 13.9 | 12.5 | −10.2 | 21.9 | 0.77 |
Qamdo | 62.1 | 58.5 | −5.8 | 10.7 | 0.89 | 20.9 | 14.5 | −28.5 | 34.1 | 0.86 |
Lhasa | 65.9 | 59.5 | −9.7 | 15.4 | 0.80 | 23.1 | 17.3 | −25.4 | 40.4 | 0.73 |
Nylingchi | 66.7 | 59.9 | −10.1 | 15.5 | 0.92 | 13.4 | 16.6 | 24.2 | 30.6 | 0.89 |
Nagqu | 38.6 | 44.2 | 12.5 | 14.5 | 0.91 | 48.7 | 32.9 | −32.9 | 32.9 | 0.72 |
Pune | 30.9 | 37.1 | 13.8 | 22.5 | 0.94 | − | − | − | − | − |
Delhi | 16.4 | 21.5 | 12.3 | 21.4 | 0.76 | − | − | − | − | − |
Anantapur | 31.0 | 34.5 | 11.3 | 11.3 | 0.98 | − | − | − | − | − |
Mohali | 40.4 | 37.6 | 4.4 | 7.6 | 0.95 | − | − | − | − | − |
Agra | − | − | − | − | − | 112.0 | 79.8 | −28.7 | 30.8 | 0.95 |
Mumbai | − | − | − | − | − | 27.7 | 25.8 | −6.9 | 14.7 | 0.92 |
Kosmarra | − | − | − | − | − | 73.6 | 72.5 | −1.8 | 8.6 | 0.96 |
Punjab | − | − | − | − | − | 65.4 | 69.7 | 6.4 | 13.7 | 0.88 |