Changes in source contributions of particulate matter during COVID-19 pandemic in the Yangtze River Delta, China

Abstract. Within a short time after the outbreak of coronavirus disease 2019 (COVID-19) in Wuhan, Hubei, the Chinese government took a nationwide lockdown to prevent the spread of the pandemic. The quarantine measures have significantly decreased the anthropogenic activities, and thus improving the air quality. To study the impacts caused by the lockdown on specific source sectors and regions in the Yangtze River Delta (YRD), the Community Multiscale Air Quality (CMAQ) model was used to investigate the changes in source contributions to fine particulate matter (PM2.5) from January 23 to February 28, 2020, based on different emission control cases. Compared to Case 1 (without emission reductions), the total PM2.5 mass for Case 2 (with emission reductions) decreased by larger than 20 % over the entire YRD and the reduction ratios of its components were 15 %, 16 %, 20 %, 43 %, 34 %, and 35 % in primary organic aerosol (POA), elemental carbon (EC), sulfate, nitrate, ammonium, and secondary organic aerosol (SOA), respectively. The source appointment results showed that PM2.5 concentrations from transportation decreased by 40 % while from residential and power decreased by less than 10 % due to the lockdown. Although all sources decreased, the relative contribution changed differently. Contributions of residential increased by more than 10 % to 35 %, while that of the industry decreased to 33 %. Considering the regional transport, the total PM2.5 mass of all regions decreased 20–30 % in the YRD with the largest decreased value of 5.0 μg m−3 in Henan, Hebei, Beijing, and Tianjin (Ha-BTH). In Shanghai, the lower contributions from local emissions and regional transmission (mainly Shandong and Ha-BTH) led to the reduced PM2.5. This study suggested adjustments of control measures for various sources and regions. Also, it was necessary to strengthen the regional collaboration at a large scale to improve the air quality in the YRD.


2018; Song et al., 2017;Song et al., 2016;Yan et al., 2018;Du and Li, 2016;Liu et al., 2016;Shen et al., 2020a). To improve air quality, China has promulgated stringent emission control plans such as the Air Pollution Prevention and Control Action Plan and PM2.5 concentrations have been reduced significantly in different regions (Zheng et al., 2018;Cai et al., 2017;Zhang et al., 2016;Zheng et al., 2017). The Yangtze River Delta (YRD), one of the largest economic centers, PM2.5 concentrations were 35 reduced by 34.3% from 2013 to 2017 due to significant efforts (China, 2018). However, PM2.5 concentrations are still much higher than the recommended annual mean criteria of 10 μg m -3 by the World Health Organization (WHO). The significant reductions in emissions lead to changes in the local and regional transport contributions of key pollutants. Consequently, the air quality strategies need further improvement according to the source apportionment results.
PM2.5 is a complete mixture of primary PM components (PPM) and secondary formed components, and its source 40 apportionment based on quantifying the contributions of different sources to all the components. Statistical methods using observed PM2.5 composition information can only resolve the contributions of source sectors to PPM. Source-oriented chemical transport models (CTMs) are capable of investigating the contributions of both source sectors and regional transports to both PPM and secondary components (Wang et al., 2014;Ying et al., 2014;Wang et al., 2013;. For instance, Hu et al. (2015) reported that local emissions account for the highest fraction of PPM compared to the regional transport in 45 Shanghai. Zhang et al. (2012) showed that the power sector (~30%) was the predominant contributor to sulfate, a component of secondary inorganic aerosol (SIA), and remaining contributions are from industrial and residential sectors in Shanghai. Liu et al. (2020) reported that the industry sector was the major secondary organic aerosol (SOA) emissions source, and additionally both regional transport and local emissions were critical to Shanghai. With source contributions changed, the information provided by these studies is not suitable for the further reduction of PM2.5 in the YRD. Therefore, the updated 50 source appointment information is needed to support the further reduction policy.
To prevent the spread of the COVID-19 pandemic, the unprecedented nationwide lockdown has been implemented to limit anthropogenic activities from January 2020. As a result, anthropogenic emissions decreased drastically, especially in the transportation and industry sectors . As a natural experiment with high research values, this provides a valuable opportunity to understand pollution changes with extremely strict measures. Studies have reported significant 55 decreases of PM2.5 in the YRD based on absolute concentrations (Chen et al., 2020;Li et al., 2020;Chauhan and Singh, 2020;Yuan et al., 2020). However, it is not clear how the contributions of local sources and regional transport changed, and cannot be used to design control strategies. Thus, it is critical to investigate changes in source sectors and regions during the COVID-19 pandemic.
In this study, a source-oriented version of the Community Multiscale Air Quality (CMAQ) model was used to determine the 60 contributions of source sectors and regional transport to PM2.5 in the YRD from January 23 to February 28. The impacts of quarantine measures were estimated by comparing the contributions before and after January 23, the start point of the lockdown.

Model description 65
The SAPRC-11 photochemical mechanism and AERO6 aerosol module were applied in the CMAQ v5.0.2 to separately quantify source contributions to PPM and SIA (Carter and Heo, 2013;Zhang et al., 2015). The CMAQ model used in this study was modified to add additional non-reactive tracers of PPM from various source sectors and regions . The emission rates of these tracers only account for 0.001% of total PPM emission rates in each grid cell so that they will not have an impact on the atmospheric process. The concentrations of tracers from a given source or region were then estimated by 70 multiplying 10 5 to represent the concentrations of PPM from that source or region. The concentrations of components in PPM were calculated based on the ratio of each component to total PPM from sources or regions. Details are discussed in Hu et al. (2015).
The contributions of source sectors and regions to SIA were quantified by tagging reactive tracers. Precisely, both the components of SIA and their precursors from diverse source types and regions were tracked separately by adding labels on 75 NOx, SO2, and NH3 through the atmospheric process (Shi et al., 2017). In this study, contributions from different emission sectors including residential, industry, transportation, power, and agriculture, and those from source regions including Jiangsu, Shanghai, Zhejiang, Anhui, Ha-BTH (Henan, Hebei, Beijing, and Tianjin), Shandong, HnHb (Hunan and Hubei) and other provinces were tracked ( Fig. S1 and Table S1). It should be noted that considerable uncertainties due to emission of precursors and lack of formation pathways, contributions of SOA to total PM2.5 was small (less than 10%) and SOA sources were not 80 tracked in this study. More information of SOA source apportionment was discussed in Wang et al. (2018).

Model application
Two nested domains were used to simulate pollution changes during the COVID-19 pandemic from January 5 to February 28, 2020. As shown in Fig Two cases were simulated in this study (Table 1). The base case (Case 1) used the original inventory. In Case 2, the emissions 95 of carbon monoxide (CO), nitric oxide (NOx), sulfur dioxide (SO2), volatile organic compounds (VOC), and PM were decreased during the COVID-19 period since January 23 with provincial specific factors from Huang et al. (2020). The differences between the cases represent the changes in sources and regions.

WRF evaluation
Since air quality simulations would be influenced by meteorology difference, it is critical to validate WRF performance before simulating source appointment . The model performance of meteorological parameters temperature at 2 m above the ground surface (T2), wind speed (WSPD), wind direction (WD) and relative humidity (RH) in the COVID-19 period could be found in Table S2. The statistical values of mean prediction (PRE); mean observation (OBS); mean bias (MB); gross 105 error (GE); and root mean square error (RMSE) have been calculated and the calculation formulas were listed in Table S4.
The WRF model predicted slightly higher T2 than observations in the two periods. The MB values of T2 before and after the lockdown were both 1.6, while the GE value of T2 before the lockdown period was slightly larger than the recommended criterion based on Emery et al. (2001). Except for the MB values of WSPD, both GE (1.3 and 1.6) and RMSE (1.7 and 2.0) were met the benchmarks during the two periods. The MB (1.8) and GE (29.2) values of WD were all within the benchmarks 110 after the lockdown, but the GE value of WD before the lockdown was slightly higher than the benchmark. The simulated RH was underestimated with the MB values of -2.4 and -5.6 during the two periods. The hourly comparisons of T2, WSPD and RH were shown in Fig. S9 based on Wang et al. (2020b, under review) also indicated good model performance. Compared to previous studies (Chen et al., 2019;Liu et al., 2020), the meteorology predictions in this study were robust to drive air quality simulation. Generally, the simulation of meteorological conditions in this study was acceptable. 115

CMAQ evaluation
The model performance of O3, NO2, SO2, PM2.5 and PM10 mass in the YRD during the COVID-19 pandemic has been described in Table S2 of a previous study (Wang et al., 2020b, under review). During the whole simulated period, the predicted PM2.5 and O3 were slightly higher than observations, but the model performance was within the criteria for PM2.5 (mean fractional bias (MFB)≤±60% and mean fractional error (MFE)≤75%, suggested by Boylan and Russell (2006)) and for O3 (MFB≤±15% 120 and MFE≤30%, suggested by U.S.EPA (2007)). Figure 1 showed predicted and observed daily PM2.5 and its components averaged over the YRD and at three major cities based on Case 2. Generally, the temporal trends of PM2.5 mass before and during the lockdown were successfully captured by the model simulations. The MFB and MFE values of PM2.5 mass were 0.14-0.41 and 0.38-0.57, which were all within the criteria. In Shanghai, the simulations missed the PM2.5 episodes from January 11 to 13, but the overall performance was good. In addition, observed SIA (including sulfate, nitrate and ammonium) 125 https://doi.org/10.5194/acp-2020-953 Preprint. Discussion started: 5 January 2021 c Author(s) 2021. CC BY 4.0 License.
from January 08 to February 10 2020 in Shanghai reported by Chen et al. (2020) were used to evaluate the model performance, as shown in Fig. S3. The daily simulated trends of SIA generally agreed with the observations, although the model slightly overpredicted SIA concentrations with the MFB values of 0.19-0.37 and the MFE values of 0.41-0.68 (Table S3). The overestimation of nitrate has been reported in the previous studies (Chang et al., 2018;Shen et al., 2020b;Choi et al., 2019) and the possible reason was the lack of chlorine heterogeneous chemistry in the model (Qiu et al., 2019). Compared to the 130 predictions before the lockdown, the MFB and MFE values of PM2.5 and its components during the lockdown were slightly lower than the values before the lockdown (Table S4), indicating the model performance during the lockdown was better. This discrepancy could be caused by the uncertainties in the emissions and meteorology (Ying et al., 2014). The model performance of the WRF model was similar during the two periods (Table S2). The 2016 MEIC emission was used for the year 2020, which may overestimate the anthropogenic emissions and thus the PM2.5 concentration in the before lockdown period. However, the 135 emission adjustments based on Huang et al. (2020) during lockdown may be closer to the real condition, leading to better model performance. Despite these uncertainties, the model results were acceptable for source apportionment studies. Figure 2 shows the predicted total PM2.5 and its components in the YRD during the COVID-19 lockdown. In both cases, PM2.5 and its components showed similar spatial distributions with the highest concentrations in the northwest and lower 140 concentrations in the southeast. Substantial PM2.5 was observed in north Anhui, similar patterns were found in elemental carbon (EC) and primary organic aerosol (POA), indicating similar sources and large contributions. For Case 2, averaged PM2.5 concentrations mainly decreased in south and northeast YRD due to the lockdown and all major components decreased in varying degrees. For EC and POA, similar decreases of 15% were observed in Anhui compared to Case 1. More important decreases were found in other regions especially in Zhejiang (up to 25%). SIA had the maximum decrease in Anhui (30-40%), 145 which was related to sharp drops of concentrations in nitrate and ammonium with decreasing ratios of 40-50% and 30-40% ( Fig. S4), respectively. On the contrary, the reductions of sulfate in Shanghai were higher than other regions in the YRD, mainly due to reduced more SO2 from industries during the lockdown based on Huang et al. (2020). Except for central and northwest YRD, SOA decreased significantly (35-40%) also due to the reductions of industrial activities, which was an important contributor to SOA . 150 Figure 3 shows the contributions of components to PM2.5 in the YRD and three major cities during the lockdown. For Case 2, over the entire YRD, the reductions in POA, EC, sulfate, nitrate, ammonium, and SOA was 2.4, 0.8, 2.1, 7.8, 2.9, and 0.9 μg m -3 with a total of 17.0 μg m -3 decrease in PM2.5. The most significant decrease was found in nitrate with the highest decrease rate of over 40%. In selected cities, PM2.5 was decreased by 15.1, 14.8 and 16.8 μg m -3 in Shanghai, Hangzhou and Nanjing, respectively, with the largest decrease ratio of 27% in Hangzhou. Secondary components (SIA + SOA) dropped more 155 significantly than primary components, especially for nitrate (35-45%) due to the severe decrease of NOx from transportation.

Changes of PM2.5 and components during the lockdown
This also indicates that atmospheric reactions and meteorological conditions were important during the pandemic period. In addition to nitrate, a sharp decrease was observed in ammonium due to the decrease of both nitrate and sulfate (Erisman and https://doi.org/10.5194/acp-2020-953 Preprint. Discussion started: 5 January 2021 c Author(s) 2021. CC BY 4.0 License. Schaap, 2004). SIA concentrations contributed the most to PM2.5 in selected cities with the highest values of 26.5 μg m -3 in Nanjing. Furthermore, the largest contributor to SIA was nitrate in the YRD, Hangzhou, and Nanjing during the lockdown, 160 while sulfate became the dominant contributor in Shanghai and accounted for 22% of total PM2.5, similar to the result in Chen et al. (2020). This shows the changes in PM2.5, but sector and region contributions are unclear to help design regional control policies. Below shows the source appointment and regional transport of PM2.5 and changes caused by lockdown measures. Figure 4 shows the contributions of different source sectors to PM2.5 in the YRD during the lockdown. Source appointments 165 of SIA and PPM in two cases are illustrated in Fig. S5 and Fig. S7, respectively. The agricultural source of PPM is not shown due to minor contributions. Generally, residential activities were the most significant contributor to PM2.5 with the highest value of 45.0 μg m -3 , mainly due to the large contribution to PPM (Fig. S7). The contribution in Shanghai was ~20.0 μg m -3 and decreased to 15.0 μg m -3 during the lockdown. The overall decrease was less than 10% in the middle YRD and less than 15% in the rest regions. Contributions from transportation decreased the most due to the lockdown from larger than 10.0 μg 170 m -3 in Case 1 to less than 7.5 μg m -3 in most areas. This was shown in SIA as well (Fig. S5), over 40% decreases were found in the YRD except for southeast, with the maximum decrease value of ~7.0 μg m -3 . The industry contributed the most to PM2.5 values in industrial cities such as Suzhou and Hefei, positions as shown in Fig. S1, which decreased significantly by ~10.0 μg m -3 from >30.0 μg m -3 to ~20.0 μg m -3 in Case 2. PM2.5 from the power sector was decreased by less than 5% to less than 6 μg m -3 in most areas due to reduced emissions of SO2 and associated sulfate (Fig. S6). PM2.5 from agriculture also decreased by 175 the lockdown with the largest decrease of 5.0 μg m -3 in northwest YRD. Figure 5 shows the changes in contributions of sources to PM2.5 in the YRD, Shanghai, Hangzhou, and Nanjing caused by the lockdown. Overall in the YRD, residential and industrial sources were major sources with contributions of 35% and 33% with decreases of less than 20%. Transportation, power and agriculture sources contributed similar to PM2.5 but with different changing ratios of 40%, 6%, and 17%, respectively. Although all sources decreased, the relative contribution did not remain 180 unchanged. The contribution ratio of transportation decreased by 27% due to the decrease in both primary emission and secondary formation, as shown in Fig. S8 and Fig. S9. The contribution ratios of residential and power increased by more than 10% while industry and agriculture showed slight changes. In large cities, industrial sources were leading with 5.0-10.0 μg m -3 higher contribution than residential sources, while other sources were similar to the YRD averages. In Shanghai, the contributions of power and agriculture showed insignificant changes, that of the industry changed by ~20% and transportation 185 decreased by more than 30%. The relative contribution of transportation decreased by more than 15%, while that of power and agriculture increased by 14% and 9%, respectively. In Hangzhou and Nanjing, the trends were similar except contributions and changes of all sources were larger in Nanjing. Due to the lockdown measures, contributions of different sources decreased but their relative contribution changed differently, implying that adjustment of control measures for various sources is needed.

Regional contributions to PM2.5 190
https://doi.org/10.5194/acp-2020-953 Preprint. Discussion started: 5 January 2021 c Author(s) 2021. CC BY 4.0 License. Figure 6 illustrates the distribution of PM2.5 contributed by emissions from different regions for two cases in the YRD during the lockdown. Regional transmission of SIA and PPM was shown in Fig. S10 and Fig. S11, respectively. It is clear that the regional distributions of each source were the same in both cases and Case 2 had lower values and narrower distributions.
Contributions of local emissions from Jiangsu, Shanghai, Zhejiang, and Anhui generally peaked near the source regions with less than 5.0 μg m -3 transported to other areas. Emissions from HnHb were barely transported to the central YRD area. 195 Shandong and Ha-BTH emissions could be transported further due to north winds, as shown in Fig. S2, with ~10.0 μg m -3 and ~5.0 μg m -3 contributions to north YRD, respectively. It indicates that the regional transports among provinces were notable, which is consistent with Du et al. (2017). Other regions also had small contributions to YRD, but the contributions decreased significantly during the lockdown. The limitation of commercial activities and traffics caused by the pandemic lockdown significantly suppressed the emission and dispersion of PM2.5. Compared to Case 1, contributions from Jiangsu, Anhui, 200 Shandong, and Ha-BTH in Case 2 decreased by 20-30%. More significant decreases of 30-40% were found in Shanghai, Zhejiang, and HnHb. The largest decrease of ~18.0 μg m -3 was observed in Hubei, the center of the COVID-19 pandemic in China due to more strict lockdown measures. Figure S9 shows that after the implementation of quarantine measures, the SIA contributions decreased by more than 30% among each region and HnHb decreased by 51% to less than 10.0 μg m -3 . Figure   S10 shows the narrower distributions and smaller decreases of PPM in Case 2 compared with SIA, with a decrease of less than 205 30% in all selected regions. Figure 7 illustrates PM2.5 contributed by eight regions averagely in the YRD and Shanghai. In the YRD, averaged contributions due to local emissions from Jiangsu, Shanghai, Zhejiang and Anhui were 6.8, 0.8, 1.5, and 6.3 μg m -3 during the lockdown period, while the contribution of outside YRD areas from HnHb, Shandong, Ha-BTH and Other were 5.0, 9.1, 14.4, and 8.2 μg m -3 , respectively. The contributions of all regions decreased due to the COVID-19 lockdown with the averaged decreasing 210 rate of 20-30% with the largest decrease rate of 33% in HnHb and the least decrease of 21% in Jiangsu. In addition to the absolute contributions, Figure 7(b) also shows the relative contribution of different regions. Ha-BTH had the largest contribution of ~30%, followed by Shandong and Other. Jiangsu and Anhui were the largest local contributors with ~12% each. It is clear that long-range transport played an important role in PM2.5 pollution in the YRD with more than 70% contribution. Due to the COVID-19, although the absolute contributions decreased universally, their relative contributions did 215 not. The importance of Jiangsu and Shandong increased by ~5%, while that of Shanghai, Zhejiang and HnHb decreased with the largest rate of 12% in HnHb. The results show that although all regions reduced their concentrations to the YRD, the relative contribution changed. In the future, regional cooperative control is needed for the YRD and strategies should be adjusted according to changes in contributions.
At the city level, local emission was the major contributor with contributions of 10.0 μg m -3 within the YRD to Shanghai, the 220 largest city in the YRD (Fig. 7(c)). Jiangsu contributed 16% to Shanghai while Zhejiang and Anhui had few effects. Outside the YRD, Shandong had the largest contribution (11.5 μg m -3 ), followed by Ha-BTH and Other areas. In total, contributions from neighbor provinces (<10.0 μg m -3 ) were much smaller than long-range transport from outside YRD (23.7 μg m -3 ).
Prevailing northerly winds were a key factor in this (Fig. S2). The lockdown decreased the contributions from all regions by https://doi.org/10.5194/acp-2020-953 Preprint. Discussion started: 5 January 2021 c Author(s) 2021. CC BY 4.0 License. 20-45%, with the largest decrease from HnHb. The contribution order of different regions was unchanged but their relative 225 contributions changed. The relative contributions of local emissions from Shanghai decreased by ~10%, while that of Shandong and Jiangsu increased by ~10%. The relative contribution of HnHb decreased by more than 20%, although the absolute changes were small.
The quarantine measures during COVID-19 lockdown reduced emissions from transportation and industrials, and the total emissions for different areas changed differently. Although PM2.5 concentrations decreased in the whole YRD, the 230 contributions of source sectors and regions changed differently. It highlights the need for regional cooperative emission reduction and adjusting control strategies when significant reductions are achieved.

Conclusion
A source-oriented CMAQ model investigated the changes in contributions of source sectors and regions to PM2.5 during the COVID-19 lockdown in the YRD. Total PM2.5 mass decreased by larger than 20% across the YRD due to decreases of 30-40% 235 and 10-20% in secondary and primary components, respectively. The results of the source appointment showed that the residential and industrial sources were the major sources with contributions of 35% (18.0 μg m -3 ) and 33% (17.1 μg m -3 ) and decreased by less than 20% due to the lockdown. Contributions from transportation decreased by 40%, which was the most significant decrease, while the decrease in power was less than 10%. The relative contribution of sources changed due to differences in source decreases. The relative contribution of transportation decreased by more than 25%, while that of 240 residential and power increased by more than 10%, suggesting that further abatement policy should adjust control measures for various sources. Contributions from regional transport of emission outside YRD were the dominant contributor (more than 70%) to the YRD, and contributions from all regions decreased due to the lockdown. The relative contribution of each region also changed with increases in Jiangsu and Shandong (~10%) but decreases in all other regions. This implies that strengthening the regional joint preventions and control of transported pollution from heavily polluted regions could effectively mitigate 245 PM2.5 pollution in the YRD.
Author contributions. JM conducted the modelling and led the writing. JS assisted in writing papers. PW, SZ, YW and PW collected data and provided technical supports. GW assisted with data analysis. JC and ZH designed the study, discussed the results, and edited the paper.
Competing interests. The authors declare that they have no conflict of interest. 250 Financial support. This project was funded by Institute of Eco-Chongming (ECNU-IEC-202001).    and (d)) of Case 2 corresponding to left Y-axis and the relative change (corresponding to right Y-axis) from January 23 to February