Spatiotemporal Distribution Patterns and Exposure Risks of PM 2.5 Pollution in China

: The serious pollution of PM 2.5 caused by rapid urbanization in recent years has become an urgent problem to be solved in China. Annual and daily satellite-derived PM 2.5 datasets from 2001 to 2020 were used to analyze the temporal and spatial patterns of PM 2.5 in China. The regional and population exposure risks of the nation and of urban agglomerations were evaluated by exceedance frequency and population weight. The results indicated that the PM 2.5 concentrations of urban agglomerations decreased sharply from 2014 to 2020. The region with PM 2.5 concentrations less than 35 µ g · m − 3 accounted for 80.27% in China, and the average PM 2.5 concentrations in 8 urban agglomerations were less than 35 µ g · m − 3 in 2020. The spatial distribution pattern of PM 2.5 concentrations in China revealed higher concentrations to the east of the Hu Line and lower concentrations to the west. The annual regional exposure risk (RER) in China was at a high level, with a national average of 0.75, while the average of 14 urban agglomerations was as high as 0.86. Among the 14 urban agglomerations, the average annual RER was the highest in the Shandong Peninsula (0.99) and lowest in the Northern Tianshan Mountains (0.76). The RER in China has obvious seasonality; the most serious was in winter, and the least serious was in summer. The population exposure risk (PER) east of the Hu Line was signiﬁcantly higher than that west of the Hu Line. The average PER was the highest in Beijing-Tianjin-Hebei (4.09) and lowest in the Northern Tianshan Mountains (0.71). The analysis of air pollution patterns and exposure risks in China and urban agglomerations in this study could provide scientiﬁc guidance for cities seeking to alleviate air pollution and prevent residents’ exposure risks.


Introduction
In the decades since the reform and opening up, China's rapid economic development, urbanization and population have caused severe damage to air quality [1]. Meanwhile, the major fossil energy consumption of transportation, industry, and residents' living needs is a direct source of PM 2.5 pollution, which has further exacerbated the problem of air pollution characterized by atmospheric particulate matter [2,3]. PM 2.5 refers to fine particulate matter with a dynamic diameter less than 2.5 µm, which is composed of a variety of complex chemical substances discharged from various natural and anthropogenic sources [4]. PM 2.5 is an important air pollutant that poses serious risks to public health all over the world and has become an issue of widespread concern to scientists and residents [5]. Since 2000, the Chinese government has issued a series of air pollution control policies, such as the 12th five-year plan on prevention and control of air pollution in key regions, the air pollution prevention and control action plan, and the three-year action plan for the defense of the blue sky [6][7][8]. Although many policies and measures have been taken and

Data
The PM data of China from 2001 to 2020 were obtained from the ChinaHighPM 2.5 data set, with a spatial resolution of 1 km [10,35]. ChinaHighPM 2.5 is one of a series of long-term, full-coverage, high-resolution, and high-quality datasets of ground-level air pollutants for China. Considering the spatiotemporal heterogeneity of air pollution, the PM 2.5 concentration at 1 km resolution was estimated by using the spatiotemporal extra tree (STET) model and the atmospheric correction (MAIAC) algorithm. At the same time, auxiliary data that may affect PM 2.5 concentration, including meteorological variables, surface conditions, pollutant emissions and population distribution, were collected to improve data accuracy. In the evaluation process, several main meteorological variables were analyzed and extracted from ERA-Interim, including temperature, relative humidity, precipitation, evaporation, surface pressure, wind speed and wind direction. The results showed that the dataset captured variations in PM 2.5 concentrations at different spatiotemporal scales well, with higher accuracies (cross-validation coefficient of determination, CV-R 2 = 0.86-0.90) and stronger predictive powers (R 2 = 0.80-0.82) than previously reported. The annual average PM 2.5 concentration data were used to analyze the changes in the pollutant interannual distribution pattern and exposure risk, and the daily average PM 2.5 concentration data of 2020 were used to analyze the changes in the pollutant seasonal distribution pattern and exposure risk.

Population Density Data
The population density from 2001 to 2020 in China was obtained from the Worldpop gridded population dataset (https://www.worldpop.org/methods/populations (accessed on 1 March 2022). The resolution of this dataset is 30 arc seconds, which is about 1 km for longitude resolution and latitude resolution at the equator. The population density data were used to calculate the population exposure risks under the condition of different population distributions.

Trend Analysis
Ordinary least-squares regression is widely used in trend analysis of sequence data evolution. In this study, trend analysis was used to characterize the change trends of annual PM 2.5 concentrations from 2001 to 2020 and determine the degree of its increase or decrease. The calculation formula is: where T is the change trends of the PM 2.5 concentration of each grid, n is the time span, i is the number of years, and pi is the annual average concentration of PM 2.5 in the ith year. If T is greater than 0, it indicates that the PM 2.5 concentration has an increasing trend; if T is less than 0, it indicates that the PM 2.5 concentration has a decreasing trend.
where r is the correlation coefficient; t is the time series from 1 to 20; and p and t are the mean values of p and t, respectively. The value of r ranges from −1 to 1.
2.3.2. Regional Exposure Risk Analysis The regional exposure risk (RER) of air pollution refers to the time when the air pollutant concentration exceeds the limitation in a region, which reflects the length or proportion of time people in this region are exposed to air pollution. The exceedance frequency was used to estimate the annual and seasonal RER of PM 2.5 , and the calculation formula is: where RER j is the RER of PM 2.5 in grid j; C j is the PM 2.5 concentration in grid j; S is the interim target 1 (IT-1) of the WHO; the IT-1 for annual PM 2.5 concentration is 35 µg·m −3 , and that for daily PM 2.5 concentration is 75 µg·m −3 . The annual and seasonal RERs were calculated based on the annual PM 2.5 concentration and daily PM 2.5 concentration, respectively. n represents the time span, which is years for annual RER and days of each season for seasonal RER. The value of RER ranges from 0 to 1. A value of 0 indicates that there is no event exceeding the pollution standard of IT-1, while the larger the value of RER is, the greater the proportion of pollution events, and a value of 1 means that the PM 2.5 concentration exceeds the standard of IT-1 throughout the whole period.

Population Exposure Risk Analysis
High concentrations of PM 2.5 pose serious threats to human life and health. However, due to the uneven distribution of population density, the RER cannot fully reflect the exposure risk of PM 2.5 for residents. To solve this problem, it is necessary to quantify the risk under different population densities. In this study, population exposure risk (PER) was used to quantify the population exposure risk of PM 2.5 pollution on a spatial scale. The calculation formula is: where PER ij is the PER of PM 2.5 in grid j in the ith year, C ij is the average annual PM 2.5 concentration of grid j in the ith year, P ij is the population density of grid j in the ith year, and m is the total number of grids in the study area. PER was divided into five levels, including level 0 (0 ≤ PER ≤ 1), level 1 (1 < PER ≤ 2), level 2 (2 < PER ≤ 3), level 3 (3 < PER ≤ 5) and level 4 (PER > 5).

Distribution Patterns of PM 2.5 Concentrations
The temporal and spatial distribution patterns of annual average PM 2.5 from 2001 to 2020 were calculated by annual ground-level PM 2.5 data. The distribution patterns of PM 2.5 concentrations in 5 representative years are displayed in Figure 2. Over the past 20 years, the distribution patterns of PM 2.5 in China have been roughly divided by the Hu Line. The PM 2.5 concentration in the eastern region was higher, and the concentration in the western region was lower, except for the Tarim Basin in Western China. From the overall distribution of PM 2.5 concentrations in the five representative years, it can be clearly found that the PM 2.5 pollution in most regions of China was obviously reduced from 2001 to 2020. Remote Sens. 2022, 14, 3173 6 of 15 2020 were calculated by annual ground-level PM2.5 data. The distribution patterns of PM2.5 concentrations in 5 representative years are displayed in Figure 2. Over the past 20 years, the distribution patterns of PM2.5 in China have been roughly divided by the Hu Line. The PM2.5 concentration in the eastern region was higher, and the concentration in the western region was lower, except for the Tarim Basin in Western China. From the overall distribution of PM2.5 concentrations in the five representative years, it can be clearly found that the PM2.5 pollution in most regions of China was obviously reduced from 2001 to 2020.     trend. Among them, CDCQ (−6.14 µg·m −3 /year), MYZ (−5.65 µg·m −3 /year) and JH (−5.47 µg·m −3 /year) decreased the most, while NTM (−2.24 µg·m −3 /year) and HBCC (−3.30 µg·m −3 /year) decreased the least. It can be seen that the relevant policies adopted by China to control PM 2.5 in recent years have achieved remarkable results.

Distribution Patterns of PM 2.5 Trends
The change trends of annual PM 2.5 concentrations from 2001 to 2020 were explored by the ordinary least-squares regression method, and the spatial distribution of trends is presented in Figure 4. The T value was negative in most regions of China, indicating that the PM 2.5 concentration has gradually decreased in the past two decades (Figure 4a). The area with a significant decrease (p < 0.05) in the PM 2.5 concentration accounted for 86.34% of the total national area (Figure 4b). There were few areas with significant (p < 0.05) and nonsignificant (p ≥ 0.05) increases in PM 2.5 concentrations, accounting for 0.05% and 3.13% of the national area, respectively, and they were mainly concentrated in Western China. For urban agglomerations, except NTM, the PM 2.5 concentration in other urban agglomerations illustrated a complete decreasing trend (Figure 4c). The trends of PM 2.5 concentration in SGX, PRD and WTS showed a significant decrease, and other urban agglomerations revealed a significant and nonsignificant decrease from 2001 to 2020.  3.3. Regional Exposure Risks of PM2.5 The annual RERs of the PM2.5 concentration in China and 14 urban agglomerations from 2001 to 2020 are presented in Figure 5. Overall, the annual RER in China was at a high level, with a national average of 0.75, while the average of 14 urban agglomerations

Regional Exposure Risks of PM 2.5
The annual RERs of the PM 2.5 concentration in China and 14 urban agglomerations from 2001 to 2020 are presented in Figure 5. Overall, the annual RER in China was at a high level, with a national average of 0.75, while the average of 14 urban agglomerations was as high as 0.86. The annual RERs in Central, Eastern and Northwest China were the highest, and those in Northeast and Southwest China were the lowest (Figure 5a). Among the 14 urban agglomerations, the average annual RER was the highest in SDP (0.99), followed by CPL (0.98) and JH (0.97), and the lowest annual RERs were in WTS (0.79), HBCC (0.78) and NTM (0.76) (Figure 5b).

Population Exposure Risks of PM2.5
The PERs of PM2.5 concentrations in China and 14 urban agglomerations from 2001 to 2020 are presented in Figure 7. The PERs of the 14 urban agglomerations showed a fluctuating trend, with no significant change in 20 years. The average PER of urban agglomerations was higher than that of China as a whole. Among the 14 urban agglomerations, the PERs of BTH, YRD, SDP and CPL were higher, while the PERs of CSLN, HBCC and NTM were lower than the national average.
Since the PER from 2001 to 2020 was basically the same, the PER distribution pattern of China and urban agglomerations was analyzed by using the PER in 2020 (Figure 8). The PER east of the Hu Line was significantly higher than that west of the Hu Line, and the PER in the eastern coastal and central regions was higher (Figure 8a). Among the 14 urban agglomerations, the average PER was the highest in BTH (4.09), followed by YRD (3.75), SDP (2.91) and CPL (2.9), and the lowest in CSLN (0.85), HBCC (0.76) and NTM (0.71) (Figure 8b). The PER level east of the Hu Line was higher than that west of the Hu Line (Figure 8c). The PERs of urban agglomerations were higher than those of other re- To analyze the RER of PM 2.5 concentrations in each season, the 24 h average concentration limitation (75 µg·m −3 ) of WHO's IT-1 was set as the standard. The proportion of days exceeding the limitation in each season of 2020 was calculated, and the results of the seasonal RER are presented in Figure 6. The RER in China has obvious seasonality; the most serious was winter, and the least serious was summer. Among the different urban agglomerations, the RER of HBCC in spring was the highest (0.08). In summer, except for MYZ, for which the RER was 0.05, the RERs of the other urban agglomerations were very low. The high RER values in autumn were mainly concentrated in Central and Eastern China, where the RERs of BTH, SDP and CPL were relatively high. In winter, the RERs of all urban agglomerations except SGX and WTS were high, and the highest RER was in CPL (0.49). The RERs of SGX, WTS and PRD along the southern coast of China were low throughout the year. The RERs of HBCC, NTM and CSLN in northern China were higher in spring and winter.

Population Exposure Risks of PM 2.5
The PERs of PM 2.5 concentrations in China and 14 urban agglomerations from 2001 to 2020 are presented in Figure 7. The PERs of the 14 urban agglomerations showed a fluctuating trend, with no significant change in 20 years. The average PER of urban agglomerations was higher than that of China as a whole. Among the 14 urban agglomerations, the PERs of BTH, YRD, SDP and CPL were higher, while the PERs of CSLN, HBCC and NTM were lower than the national average.    Since the PER from 2001 to 2020 was basically the same, the PER distribution pattern of China and urban agglomerations was analyzed by using the PER in 2020 (Figure 8). The PER east of the Hu Line was significantly higher than that west of the Hu Line, and the PER in the eastern coastal and central regions was higher (Figure 8a). Among the 14 urban agglomerations, the average PER was the highest in BTH (4.

Influencing Factors of the Evolution of PM 2.5 Distribution
The distribution patterns of PM 2.5 concentrations in China illustrated obvious regional differences. The general trend of PM 2.5 distributions in China is high in the East and low in the West. PM 2.5 concentration is closely related to urbanization level, energy use, production mode, population density, etc. [36][37][38]. From 2001 to 2010, PM 2.5 in China rose rapidly, especially in eastern China [39]. Due to large population, energy consumption, frequent economic activities and rapid urbanization process, the PM 2.5 concentration in eastern China maintained at a high level. Subsequently, the Chinese government realized the serious harm caused by PM 2.5 and began to adopt a series of policies to reduce PM 2.5 concentration to balance the relationship between economic development and pollution emissions [7].Therefore, PM 2.5 concentrations in China began to decrease significantly, especially after 2013.Due to the dust transported by the Taklimakan Desert and the Kunlun Mountains block the northeast wind, the PM 2.5 concentration in Taklimakan Desert in Xinjiang (yellow area in the Figure 4a) is unstable from 2001 to 2020 [40,41], therefore, the results showed an upward trend (p > 0.05).
Meanwhile, there were seasonal differences of PM 2.5 concentrations in China. The PM 2.5 concentrations in densely populated areas of North China were significantly higher than those in other areas. Due to the cold winter weather in northern China, coal-fired heating is an important pollution emission source, emitting a large amount of fine particulates, sulfides and nitrogen oxides [42]. In addition, straw burning in agricultural planting areas is also an important factor in seasonal PM 2.5 pollution [43].
The outbreak of COVID-19 pandemic in 2020 has brought abnormal impact on the analysis of air quality. In different cities and regions in China, PM 2.5 is significantly affected by COVID-19 [44]. In response to the epidemic, China has adopted the city closure policy, reducing human activities and some economic production, resulting in a decrease of PM 2.5 concentration in some areas. Wuhan, the most severely affected city, was blocked for 76 days, and the PM 2.5 concentration decreased by 35% compared with the same period in previous years [45]. However, the closure of the city does not strictly mean the reduction of pollution, which may cause a significant impact on the short-term air pollution [46]. For example, during the COVID-19 blockade, severe haze still occurred in northern China due to high relative humidity and poor air [47]. Other factors such as transportation and secondary pollution offset the reduction of PM 2.5 concentration [48].
Meteorology is an important factor affecting PM 2.5 pollution. PM 2.5 is negatively correlated with daily temperature, relative humidity, precipitation and wind speed, and positively correlated with atmospheric pressure [49]. However, atmospheric circulation is conducive to the dissipation of PM 2.5 [50]. At the same time, PM 2.5 concentration is affected by traffic. The most important local factor causing air pollution is the exhaust gas emitted by road vehicles, especially heavy diesel powered vehicles [51]. In addition, tire wear, road dust resuspension, traffic facilities area, road width and length also affect PM 2.5 concentration [52][53][54].

Exposure Risk Characteristics and Evaluation Methods
Two methods based on exceedance frequency and population weight were used to evaluate the exposure risk in China. For RER, all of China was at a high-risk level, with an annual average of 0.75, while the average of 14 urban agglomerations was as high as 0.86. The RER in China has obvious seasonality; the most serious occurs in winter and the least serious in summer. The PERs of PM 2.5 east of the Hu Line were obviously higher than those west of the Hu Line. The PER of urban agglomeration was higher, and that of other regions was at a lower level. The great differences in PER distribution patterns are mainly due to the dense population in the east and sparse population in the west in China [55].
RER and PER have both advantages and disadvantages in assessing the exposure risks of air pollution. RER provides a better understanding and a better basis for the spatial management of pollution prevention. Due to the vulnerability of air pollutants to climate, PM 2.5 in a certain area may vary greatly in different years [56]. This study uses 20 years of data to calculate the multiyear cumulative RER to provide a long-term mechanism for formulating management strategies to improve air quality as additional information beyond the widely accepted global disease burden research [57]. However, RER ignores the uneven spatial distribution of the population, which makes it difficult to reflect the difference in exposure risk in areas with different population densities under the same PM 2.5 concentration [58]. And many previous studies revealed that the severity of PM 2.5 pollution is closely related to population density [59]. PER considers the spatial distribution of the population and compares the exposure risks in different regions [60]. To assess the level of PM 2.5 exposure risk, especially in specific regions, it is more appropriate to use PER which considering the population distribution [61]. Therefore, due to the uneven distribution of China's population, PER seems to be a more appropriate method for assessing PM 2.5 exposure risks in China.

Advantages and Limitations
Due to the development of remote sensing technology, high time resolution remote sensing products promote the application of environmental issue analysis. In this study, daily average PM 2.5 concentration data were used to assess the seasonal variation in exposure risks. Moreover, the RER and PER were proposed to comprehensively evaluate the exposure risk of air pollution, and this assessment and analysis framework can avoid the deviation of risk estimation by a single index. Urban agglomerations are the areas with the most intense human activities, so the analysis of air pollution patterns and exposure risks of urban agglomerations in this study could provide scientific guidance to help cities alleviate air pollution and prevent residents' exposure risks. The air pollution data obtained from fixed meteorological stations and mobile monitoring can better reflect the pollution distribution pattern and exposure risk [62,63]. However, due to the limited number of monitoring stations, the monitoring data based on stations cannot meet the spatial resolution requirements [64,65]. Due to the wide space and time coverage of satellite observations, remote sensing estimation of PM 2.5 concentration is a powerful supplement to the limited ground observations [66], but may be different from the real pollutant concentration near the ground. Therefore, a model considering both monitoring station and satellite remote sensing data was a good choice [27]. The ChinaHighPM 2.5 dataset used in this paper used the STET model, which contained auxiliary information about meteorological variables, land use, pollutant emissions and population [10,35]. Therefore, compared with previous studies, the data used in this paper were more accurate and the research results were much more reliable.
In addition to the advantages of the methods and results of air pollution exposure risk assessment described above, there are some shortcomings. Population mobility, such as the population distribution during the day and at night, annual population migration, and Spring Festival transportation, will bring considerable uncertainty about the exposure risks of air pollution [67]. However, it is very difficult to obtain real-time dynamic population data, and with the development of science and technology, big data could help in this regard, such as mobile phone signaling data and software location data [68,69].
In this study, only the influence of PM 2.5 pollutant was considered, but there are other air pollutants that will cause adverse effects. China has a high incidence rate of asthma due to NO 2 exposure [70]. SO 2 will lead to the formation of haze, and increase the mortality and incidence rate of cardiopulmonary diseases [71]. In addition, there are atmospheric pollutants such as CO, CO 2 and PM 10 , which also have a great impact on the environment and human health.

Conclusions
In this study, the PM 2.5 remote sensing inversion data and population density data of China from 2001 to 2020 were used to explore the PM 2.5 concentration distribution patterns and exposure risks. Regional and population exposure risks of PM 2.5 based on exceedance frequency and population weight were proposed to evaluate the exposure risks in China and 14 urban agglomerations. The results indicated that the PM 2.5 concentrations of urban agglomerations decreased sharply from 2014 to 2020. The region with PM 2.5 concentrations less than 35 µg·m −3 accounted for 80.27% of the area of China, and the average PM 2.5 concentrations in 8 urban agglomerations were less than 35 µg·m −3 in 2020. The spatial distribution patterns of PM 2.5 concentrations in China were higher to the east of the Hu Line and lower to the west. The annual RER in China was at a high level, with a national average of 0.75, while the average of 14 urban agglomerations was as high as 0.86. Among the 14 urban agglomerations, the average annual RER was the highest in SDP (0.99), followed by CPL (0.98) and JH (0.97), and the lowest annual RERs were in WTS (0.79), HBCC (0.78) and NTM (0.76). The RER in China has obvious seasonality; the most serious was in winter, and the least serious was in summer. The PER east of the Hu Line was significantly higher than that west of the Hu Line, and the PER in the eastern coastal and central regions was higher. The average PER was the highest in BTH (4.09), followed by YRD (3.75), SDP (2.91) and CPL (2.9), and the lowest in CSLN (0.85), HBCC (0.76) and NTM (0.71). The analysis of air pollution patterns and exposure risks in China and urban agglomerations in this study could better provide scientific guidance to help cities alleviate air pollution, and reduce residents' exposure risks.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.