Prediction method for mixed pollution of pm2.5 and ozone under high spatiotemporal resolution

Abstract

This invention discloses a method for predicting mixed PM2.5 and ozone pollution at high spatiotemporal resolution, comprising the following steps: first, using historical air quality data as the dependent variable and the driving factor data corresponding to the historical air quality data as the independent variable, to obtain the optimal PM2.5 concentration. 2.5 A model for predicting ozone concentration was developed, and then significantly correlated driving factors were extracted using a hexagonal grid as the prediction unit, which were then input into the PM2.5 concentration prediction system. 2.5 In ozone concentration prediction models, PM2.5 concentrations are obtained at high spatiotemporal resolution. 2.5 The ozone concentration spatial distribution prediction map and the ozone concentration spatial distribution prediction map are then intersected spatially to obtain the PM2.5 concentration. 2.5 A prediction map of mixed pollution with ozone. The prediction method of this invention has advantages such as good applicability and high reliability, and can accurately predict PM2.5. 2.5 When mixed with ozone, the concentration of PM2.5 in the atmosphere... 2.5 It has important guiding significance for the prevention and control of ozone, has high application value, and has good application prospects.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of air pollution control, and relates to a PM 2.5 and ozone mixed pollution prediction method with high spatiotemporal resolution. BACKGROUND

[0002] PM 2.5 is one of the most common air pollutants, and is widely studied due to its significant impact on human health and the environment. In recent years, researchers have proposed many control measures, which have effectively reduced the concentration of PM 2.5 . However, although the reduction of PM 2.5 is very fast, the level of O3 is abnormally rising. Various studies have shown that the reduction of PM 2.5 leads to an increase in radiation flux, which exacerbates O3 pollution. It can be seen that the change in pollution source emission structure caused by the reduction measures focusing on controlling PM 2.5 may be the main reason for the continuous rise of O3 pollution. At present, although the detection equipment arranged for monitoring can obtain the pollution status of PM 2.5 and O3 in the region, there is a lack of air quality prediction research on the mixing of PM 2.5 and O3, which leads to the fact that people cannot intuitively judge the pollution status of PM 2.5 and O3 in a certain region. Therefore, it is necessary to obtain a method capable of accurately predicting the mixed pollution status of PM 2.5 and O3 with high spatiotemporal resolution, which is very necessary for studying the synergistic effect and trade-off between PM 2.5 and O3.

[0003] Machine learning modeling is an emerging technology for air prediction and management applications, among which the application of land use regression (LUR) model and random forest (RF) model is the most extensive. For example, the influence relationship between pollutants and driving factors can be obtained by using the LUR model. However, in previous studies, the number of monitoring points is less than 40, which cannot meet the requirement of independent variable data for constructing the LUR model, and the sampling period of the monitoring points is short, which leads to the lack of complete annual hourly data in the monitoring time, resulting in the non-standardization of the LUR model construction and the error, especially the single selection of driving factors, which does not fully include the characteristic factors (such as AOI, POI, landscape index, etc.) that can reflect urbanization, which will cause the model to be one-sided and single, lack of wide applicability, and difficult to comprehensively evaluate the influence of urbanization on air quality, so it is difficult to obtain a reliable LUR model with high spatio-temporal resolution. In addition, the RF model has black box effect, and the influence relationship between pollutants and driving factors cannot be accurately seen, so the single prediction model still has the problem of poor reliability. In addition, the prediction grid used in the existing prediction method is usually quadrilateral and circular, which may cause folding and cross-influence between the prediction grids, resulting in the defects of inconsistent neighborhood, isotropy, compactness, and low sampling rate, and finally it is difficult to construct a high-resolution spatial distribution map.

[0004] Therefore, how to construct a prediction model with high accuracy is of great significance for accurately predicting the mixed pollution status of PM 2.5 and O3 and carrying out in-depth research on the mixed pollution status of PM 2.5 and O3. SUMMARY

[0005] The technical problem to be solved by the present application is to overcome the shortcomings of the prior art and provide a prediction method for the mixed pollution of PM 2.5 and ozone with high spatio-temporal resolution, good applicability and good reliability.

[0006] To solve the above technical problems, the following technical solutions are adopted in the present application:

[0007] A prediction method for the mixed pollution of PM 2.5 and ozone with high spatio-temporal resolution, comprising the following steps:

[0008] S1, collecting historical air quality data of a to-be-predicted area and driving factor data corresponding to the historical air quality data; the historical air quality data comprises the concentration data of PM 2.5 and ozone measured every hour by the monitoring station in the to-be-predicted area in the past; and the driving factor data comprises coordinates, meteorological factors, remote sensing factors, population density, land coverage, road density and landscape index;

[0009] S2, monitoring the PM 2.5 and ozone concentration data in the prediction area into annual average, quarterly average, weekend average, weekday average and special day average, and constructing the PM 2.5 and ozone LUR model corresponding to the different time periods respectively with the related factors in the driving factor data. 2.5 and ozone improved RF model.

[0010] S3, verifying the PM 2.5 and ozone LUR model, PM 2.5 and ozone improved RF model respectively, and screening out the LUR model or RF model with the largest determination coefficient R2 and the lowest root mean square error RMSE as the PM 2.5 and ozone concentration prediction model.

[0011] S4, dividing the prediction area into a plurality of hexagonal grids, and taking the hexagonal grid as a prediction unit.

[0012] S5, extracting the significant driving factors corresponding to the PM 2.5 and ozone concentration prediction model from each hexagonal grid area.

[0013] S6, inputting the driving factors significantly related to the PM 2.5 and ozone concentration in each hexagonal grid area into the PM 2.5 and ozone concentration prediction model respectively, to obtain the PM 2.5 and ozone concentration corresponding to each hexagonal grid area, to obtain the PM 2.5 concentration spatial distribution prediction map and ozone concentration spatial distribution prediction map under high spatio-temporal resolution.

[0014] S7, intersecting the PM 2.5 concentration spatial distribution prediction map and ozone concentration spatial distribution prediction map under high spatio-temporal resolution in space to obtain the PM 2.5 and ozone mixed pollution prediction map in the prediction area.

[0015] The above prediction method is further improved, in step S1, the related factors in the meteorological factors include temperature, humidity, wind direction and wind speed; the related factors in the remote sensing factors include point of interest data, interest area data, normalized vegetation index, rainfall grid data, light data, building proportion data.

[0016] The above prediction method is further improved, in step S2, the construction method of the PM 2.5 and ozone LUR model corresponding to the different time periods includes the following steps:

[0017] (1) constructing a PM 2.5 and ozone concentration LUR model in the corresponding time period by using the driving factor data and the PM 2.5 and ozone concentration data.

[0018] (2) constructing a mapping relationship between the PM 2.5 and ozone concentration and the driving factor by using the driving factor data and the PM 2.5 and ozone concentration data. 2.5

[0019] The improved RF model of the PM 2.5 and ozone concentration in step S2 comprises the following steps:

[0020] (a) constructing a random forest model by using the annual average, quarterly average, weekend average, weekday average and special day average of the PM 2.5 and ozone concentration and the driving factor data.

[0021] (b) constructing a mapping relationship between the PM 2.5 and ozone concentration and the driving factor by using the random forest model. 2.5 2.5

[0022] In step S4, the hexagonal grid is a regular hexagonal grid, and the side length of the regular hexagonal grid is 1 km.

[0023] In step S7, the predicted concentration of the PM 2.5 and O3 in the corresponding time period is divided into low, medium and high levels by using the natural breakpoint method, and the pollution conditions corresponding to the three levels are light pollution, medium pollution and high pollution.

[0024] In step S7, the mixed pollution prediction map of the PM 2.5 and ozone in the to-be-predicted area includes the following mixed pollution conditions:

[0025] First, the pollution condition of the PM 2.5 in the to-be-predicted area is low pollution, and the pollution condition of the ozone is low pollution.​​​

[0026] The second, the pollution situation of PM 2.5 in the to-be-predicted region is low pollution, and the pollution situation of ozone is medium pollution;

[0027] The third, the pollution situation of PM 2.5 in the to-be-predicted region is low pollution, and the pollution situation of ozone is high pollution;

[0028] The fourth, the pollution situation of PM 2.5 in the to-be-predicted region is medium pollution, and the pollution situation of ozone is low pollution;

[0029] The fifth, the pollution situation of PM 2.5 in the to-be-predicted region is medium pollution, and the pollution situation of ozone is medium pollution;

[0030] The sixth, the pollution situation of PM 2.5 in the to-be-predicted region is medium pollution, and the pollution situation of ozone is high pollution;

[0031] The seventh, the pollution situation of PM 2.5 in the to-be-predicted region is high pollution, and the pollution situation of ozone is low pollution;

[0032] The eighth, the pollution situation of PM 2.5 in the to-be-predicted region is high pollution, and the pollution situation of ozone is medium pollution;

[0033] The ninth, the pollution situation of PM 2.5 in the to-be-predicted region is high pollution, and the pollution situation of ozone is high pollution.

[0034] Compared with the prior art, the present application has the advantages that:

[0035] The present application provides a PM 2.5 and ozone mixed pollution prediction method under high space-time resolution, which uses air quality data obtained by complete annual hourly monitoring of air quality micro-stations, combines traditional driving factors and emerging driving factors in seven categories of influencing factors including coordinates, meteorological factors, remote sensing factors, population density, land coverage rate, road density and landscape index, constructs a prediction model, screens out driving factors significantly related to PM 2.5 and ozone concentration, constructs a mapping relationship between PM 2.5 concentration, ozone concentration and significantly related driving factors, respectively obtains a LUR model of PM 2.5 and ozone and a RF model of PM 2.5 and ozone in different time periods, and further, verifies the above-mentioned models, screens out a LUR model or a RF model with the largest determination coefficient R2 and the lowest root mean square error RMSE as PM 2.5and the concentration prediction model of ozone, namely the regression model with high spatiotemporal resolution; on this basis, the to-be-predicted region is divided into a plurality of hexagonal grids, the hexagonal grid is taken as a prediction unit, and the driving factors related to PM 2.5 and the concentration prediction model of ozone are input into the PM 2.5 and the concentration prediction model of ozone, the corresponding significant driving factors of each hexagonal grid region in the PM 2.5 and ozone concentration are obtained, and the PM 2.5 and ozone concentration of different hexagonal grid regions are obtained under high spatiotemporal resolution. 2.5 and ozone concentration prediction map, and finally the PM 2.5 and ozone concentration of each hexagonal grid region in the corresponding time period is obtained. 2.5 and ozone concentration prediction map. In the present application, the air quality data obtained by continuous monitoring is taken as the dependent variable, which can effectively avoid the non-standard and error of model construction, and the traditional driving factors and emerging driving factors in the corresponding seven categories of influencing factors within the adaptive time are taken as the independent variables, which can completely include the characteristic factors of the to-be-predicted region, better integrity and better adaptability, thereby being conducive to constructing a prediction model with higher accuracy. At the same time, the to-be-predicted region is divided into a plurality of hexagonal grids, which is conducive to constructing a honeycomb-shaped prediction region, reducing the folding and poor between grids, improving the sampling rate, and obtaining high-resolution PM 2.5 and ozone concentration prediction map, which is conducive to analyzing the PM 2.5 and ozone mixed pollution situation under high resolution, accurately identifying the mixed pollution situation in different regions, and putting forward targeted solutions, especially for the regions suffering from PM 2.5 and O3 heavy pollution, which can be more timely and effective governance, and has important practical significance for effective governance of air pollution. 2.5 and ozone mixed pollution prediction method has the advantages of good applicability, good reliability, etc., can accurately predict the PM 2.5 and ozone mixed pollution situation, and has important guiding significance for the prevention and treatment of PM 2.5 and ozone in the atmosphere, high use value and good application prospect. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 for the prediction method of PM 2.5 and ozone mixed pollution under high spatiotemporal resolution in embodiment 1 of the present application.

[0037] Figure 2Figure 1 is a land use type and air quality monitoring station distribution map of the region to be predicted in Example 1 of the present application.

[0038] Figure 3 Figure 6 is a scatter plot of the actual and predicted concentrations of ozone and PM 2.5 in the annual mean value, quarterly mean value time in Example 1 of the present application.

[0039] Figure 4 Figure 7 is a scatter plot of the actual and predicted concentrations of ozone and PM 2.5 in the working day mean, weekend mean time in Example 1 of the present application.

[0040] Figure 5 Figure 8 is a scatter plot of the actual and predicted concentrations of ozone and PM 2.5 in the daily mean time in Example 1 of the present application.

[0041] Figure 6 Figure 9 is a mixed pollution prediction map of PM 2.5 annual mean and ozone annual mean under high spatio-temporal resolution in Example 1 of the present application.

[0042] Figure 7 Figure 10 is a mixed pollution prediction map of PM 2.5 working day / weekend mean and ozone working day / weekend mean under high spatio-temporal resolution in Example 1 of the present application. DETAILED DESCRIPTION

[0043] The present application will be further described below in conjunction with the accompanying drawings and specific preferred embodiments, but the scope of protection of the present application is not limited thereby. The materials and instruments used in the following examples are all commercially available.

[0044] Example 1:

[0045] A prediction method for mixed pollution of PM 2.5 and ozone under high spatio-temporal resolution, the flowchart is shown in Figure 1 , comprising the following steps:

[0046] S1, as shown in Figure 2 , the region to be predicted is the region where Changsha city is located. The PM 2.5 and O3 data (street level) of 144 air quality monitoring stations in Changsha city measured by the Environmental Protection Bureau of Changsha city in 2020 are collected, and the concentration data of PM 2.5 and ozone measured every hour are obtained, which are used as dependent variables. According to meteorology (northern hemisphere), spring: March to May; summer: June to August; autumn: September to November; winter: December to February. Changsha area represents longer time series level with annual mean, seasonal mean, daily mean and hourly concentration data representing instantaneous concentration level.

[0047] At the same time, the air quality monitoring station in the main urban area of Changsha in 2020 is taken as the center to establish a plurality of buffer zones, and the corresponding driving factor data of the main urban area of Changsha is collected, which can be specifically four buffer zones (i.e. 500-2000 meters, with an interval of 500 meters), and the corresponding driving factor data (independent variable data) of Changsha in 2020 is extracted, wherein the driving factor data includes coordinates, meteorological factors (such as temperature, humidity, wind direction, wind speed), remote sensing factors (such as area of interest data (AOI), point of interest data (POI), normalized difference vegetation index (NDVI), rainfall grid data, light data and building proportion data), population density, land coverage, road density and landscape index.

[0048] S2, converting the PM 2.5 and ozone concentration data measured by the monitoring station in the to-be-predicted area every hour into annual average, quarterly average, weekend average, weekday average and special day average, and constructing different time period corresponding PM 2.5 and ozone LUR model, and improved RF model of PM 2.5 and ozone, specifically:

[0049] The construction method of different time period corresponding PM 2.5 and ozone LUR model, which is specifically constructing PM 2.5 and O3 concentration and the stepwise multiple regression model of the related factors in the driving factor data by using R4.1.2, including the following steps:

[0050] (1) using R4.1.2 version software to perform bivariate correlation analysis on the annual average, quarterly average, weekend average, weekday average and special day average of PM 2.5 and ozone concentration and the related factors in the driving factor data, filtering out the insignificant correlation (P>0.05) between PM 2.5 and ozone concentration, and screening out the driving factors significantly correlated with PM 2.5 and ozone concentration.

[0051] (2) performing stepwise linear regression processing on the driving factors screened out and significantly correlated with PM 2.5 and ozone concentration, respectively, to construct the mapping relationship of PM 2.5 concentration, ozone concentration and the significantly correlated driving factors, and to obtain different time period corresponding PM 2.5 and ozone LUR model.

[0052] As shown in Table 1, the variance inflation factor VIF of the model is less than 10, which indicates that there is no multicollinearity in the model.

[0053] Table 1 corresponding PM 2.5 and ozone LUR model parameters

[0054]

[0055] Table 2 corresponding PM 2.5 and ozone LUR model parameters

[0056]

[0057]

[0058]

[0059] PM 2.5 The method for constructing the improved RF model of PM

[0060] (a) constructing a random forest model with annual average, quarterly average, weekend average, weekday average and special day average of PM 2.5 and ozone concentration and the driving factor data, specifically using Python 3.10 to construct a random forest model of PM 2.5 and O3 concentration and the corresponding independent variable data set.

[0061] (b) screening out driving factors significantly related to PM 2.5 and ozone concentration by twice screening of the random forest model, and constructing the mapping relationship between PM 2.5 and ozone concentration and the significantly related driving factors, respectively obtaining the improved RF model of PM 2.5 and ozone.

[0062] S3, performing ten-fold cross-validation test on the LUR model of PM 2.5 and ozone and the improved RF model of PM 2.5 and ozone respectively, and screening out the LUR model or RF model with the largest determination coefficient R2 and the lowest root mean square error RMSE as the concentration prediction model of PM 2.5 and ozone.

[0063] As shown in Figure 3 , Figure 3 (a) and Figure 3 (f) are scatter plots of actual and predicted concentrations of ozone and PM 2.5 in the annual average time period. Figure 3 (b) and Figure 3(g) represents the ozone and PM2.5 concentrations corresponding to the average spring time. 2.5 A scatter plot of the actual and predicted concentrations; Figure 3 (c) and Figure 3 (h) represents the ozone and PM2.5 concentrations during the average summer time period. 2.5 A scatter plot of the actual and predicted concentrations; Figure 3 (d) and Figure 3 (i) represents the ozone and PM2.5 concentrations corresponding to the average autumn time period. 2.5 A scatter plot of the actual and predicted concentrations; Figure 3 (e) and Figure 3 (j) is a scatter plot of the actual and predicted concentrations within the winter mean time period.

[0064] like Figure 4 As shown, Figure 4 (a) and Figure 4 (c) Ozone and PM2.5 concentrations during the average working day. 2.5 A scatter plot of the actual and predicted concentrations; Figure 4 (b) and Figure 4 (d) represents the ozone and PM2.5 concentrations corresponding to the average time of the weekend. 2.5 A scatter plot of the actual and predicted concentrations.

[0065] like Figure 5 As shown, Figure 5 (a) is a scatter plot of the actual and predicted concentrations of ozone over a daily average period; Figure 5 (b) PM2.5 during the daily average time period 2.5 A scatter plot of the actual and predicted concentrations.

[0066] S4. Using appropriate geoprocessing software, the area to be predicted (Changsha City) is divided into several regular hexagonal grids, with the regular hexagonal grids serving as prediction units. The side length of each regular hexagonal grid is 1 km.

[0067] S5. Extract PM from each hexagonal grid region. 2.5 The driving factors that are significantly correlated with ozone concentration prediction models.

[0068] S6. Connect the PM within each hexagonal grid region. 2.5 The driving factors that are significantly correlated with ozone concentration are respectively input into PM 2.5 In the ozone concentration prediction model, the PM2.5 concentration corresponding to each hexagonal grid region is obtained. 2.5 And ozone concentration, to obtain PM2.5 concentrations with high spatiotemporal resolution. 2.5 The predicted spatial distribution maps of annual average concentration and ozone annual average concentration are shown below. Figure 6 As shown.

[0069] S7、the PM 2.5 The concentration spatial distribution prediction map and the ozone concentration spatial distribution prediction map intersect in space to obtain the PM 2.5 and ozone mixed pollution prediction map in the to-be-predicted region.

[0070] In step S7, the predicted concentrations of PM 2.5 and O3 in the corresponding time period are divided into three levels of low, medium and high, corresponding to light pollution, medium pollution and high pollution, and the degree of PM 2.5 and O3 pollution in the same region is identified by means of geographic processing software and spatial connection, and the synergistic pollution of PM 2.5 and O3 is further analyzed at high spatiotemporal resolution.

[0071] Figure 6 The PM 2.5 annual mean and ozone annual mean mixed pollution prediction map in Example 1 of the present application at high spatiotemporal resolution. Figure 6 As shown, the black color belongs to the high concentration region of the divided grades, that is, the pollution prevention and control region to be found in the to-be-predicted region. The white region belongs to a non-polluted or lower polluted region. If we only predict the street-level pollution region from the 144 sites, it will result in low resolution of the prediction result and inaccurate prediction result. However, it is obviously unrealistic to arrange as many sample points as possible to cover the air quality situation of Changsha. Because arranging too many sample points requires more financial and material resources, and the operation cost is too high. Therefore, the present application proposes to obtain the PM 2.5 and O3 pollution concentration distribution map of Changsha at high spatiotemporal resolution by constructing a high spatiotemporal resolution model. The high concentration of PM 2.5 and O3 concentration is overlapped in space and time by means of geographic processing software, so as to obtain the PM 2.5 and O3 concentration distribution region at high spatiotemporal resolution, as shown in Figure 7 The black hexagon of the 1Km hexagonal grid map is the result we want to obtain. Air management personnel can more accurately identify the fine region of high PM 2.5 and high O3 simultaneous action according to the black block, which is more conducive to the fine regional prevention and control requirements of air pollution management. When exploring the PM 2.5 and O3 mixed pollution exposure situation, time accuracy is also particularly important, so the present method explores the PM 2.5 and O3 mixed pollution situation from multiple time scales. Figure 7 The PM 2.5 working day / weekend mean and ozone working day / weekend mean mixed pollution prediction map at high spatiotemporal resolution in Example 1 of the present application.

[0072] From the above results, the high spatial and temporal resolution PM 2.5 And ozone mixed pollution prediction method, using the air quality data obtained by the complete annual hourly monitoring of the air quality micro station, combining the traditional driving factors and emerging driving factors in the seven categories of influencing factors of coordinates, meteorological factors, remote sensing factors, population density, land coverage, road density and landscape index, constructing a prediction model, by screening the driving factors significantly related to PM 2.5 And ozone concentration, constructing the mapping relationship of PM 2.5 And ozone concentration and significantly related driving factors, respectively obtaining the corresponding PM 2.5 And ozone LUR model in different time periods, PM 2.5 And ozone RF model, further, by verifying the above model, screening out the LUR model or RF model with the largest determination coefficient R2 and the lowest root mean square error RMSE as the concentration prediction model of PM 2.5 And ozone, that is, the regression model with high spatial and temporal resolution; on this basis, the to-be-predicted region is divided into several hexagonal grids, the hexagonal grid is taken as the prediction unit, and the significantly related driving factors corresponding to the concentration prediction model of PM 2.5 And ozone are extracted from each hexagonal grid region and input into the concentration prediction model of PM 2.5 And ozone, respectively obtaining the corresponding PM 2.5 And ozone concentration of each hexagonal grid region, obtaining the PM 2.5 Concentration spatial distribution prediction map and ozone concentration spatial distribution prediction map of different hexagonal grid regions under high spatial and temporal resolution, finally, the PM 2.5 Concentration spatial distribution map and ozone concentration spatial distribution map of each hexagonal grid region in the corresponding time period are intersected in space, and the mixed pollution prediction map of PM 2.5 And ozone in the to-be-predicted region is obtained. In the application, the air quality data obtained by continuous monitoring is taken as the dependent variable, which can effectively avoid the non-standard and error of model construction, and the traditional driving factors and emerging driving factors in the corresponding seven categories of influencing factors in the corresponding time period are taken as the independent variables, which can completely include the characteristic factors of the to-be-predicted region, better integrity and better adaptability, thereby being conducive to constructing a prediction model with higher accuracy, at the same time, the to-be-predicted region is divided into several hexagonal grids, the hexagonal grid is conducive to constructing a honeycomb-shaped prediction region, conducive to reducing the folding and poor between grids, conducive to improving the sampling rate, obtaining the PM 2.5 And ozone concentration spatial distribution prediction map under high resolution, and conducive to analyzing the mixed pollution of PM 2.5 And ozone in different regions, and putting forward targeted solutions, especially for the mixed pollution of PM2.5 The present application can effectively manage the PM 2.5 The present application can accurately predict the PM 2.5 and ozone mixed pollution, and has important guiding significance for the prevention and treatment of PM 2.5 and ozone in the atmosphere, and has high use value and good application prospect.

[0073] The above is only the preferred embodiment of the present application, and does not limit the present application in any form. Although the present application has been disclosed as above with the preferred embodiment, it is not intended to limit the present application. Any skilled person in the art can make many possible changes and modifications to the technical solutions of the present application, or modify equivalent embodiments, without departing from the spirit and technical solutions of the present application, by using the disclosed methods and technical contents. Therefore, any simple modification, equivalent replacement, equivalent change and modification made to the above embodiments according to the technical essence of the present application, without departing from the technical solutions of the present application, still falls within the scope of protection of the technical solutions of the present application.

Claims

1. A method for predicting mixed PM2.5 and ozone pollution at high spatiotemporal resolution, characterized in that, Includes the following steps: S1. Collect historical air quality data for the area to be predicted and driving factor data corresponding to the historical air quality data; the historical air quality data includes PM2.5 measured hourly at monitoring stations in the area to be predicted throughout history. 2.5 The data includes ozone concentration data; the driving factor data includes coordinates, meteorological factors, remote sensing factors, population density, land cover, road density, and landscape index. S2, The PM2.5 levels measured historically hourly at monitoring stations in the area to be predicted. 2.5 The ozone concentration data were converted into annual averages, quarterly averages, weekend averages, weekday averages, and special day averages, and then compared with the relevant factors in the driving factor data to construct PM values ​​for different time periods. 2.5 LUR model of ozone, PM 2.5 An improved RF model for ozone; S3. For the PM corresponding to different time periods obtained in step S2 2.5 LUR model of ozone, PM 2.5 The improved RF model for ozone was validated separately, and the LUR model or RF model with the largest coefficient of determination R² and the lowest root mean square error (RMSE) was selected as the PM model. 2.5 And ozone concentration prediction models; S4. Divide the area to be predicted into several hexagonal grids, and use the hexagonal grids as prediction units; S5. Extract the PM from each hexagonal grid region. 2.5 The driving factors that are significantly correlated with ozone concentration prediction models; S6. Connect the PM within each hexagonal grid region. 2.5 The driving factors that are significantly correlated with ozone concentration are respectively input into the PM 2.5 In the ozone concentration prediction model, the PM2.5 concentration corresponding to each hexagonal grid region is obtained. 2.5 And ozone concentration, to obtain PM2.5 concentrations with high spatiotemporal resolution. 2.5 Predicted spatial distribution maps of concentration and ozone concentration; S7, PM under high spatiotemporal resolution 2.5 The spatial intersection of the PM2.5 concentration prediction map and the ozone concentration prediction map yields the PM2.5 concentration in the region to be predicted. 2.5 A prediction diagram of mixed pollution with ozone.

2. The prediction method according to claim 1, characterized in that, In step S1, the relevant factors among the meteorological factors include temperature, humidity, wind direction, and wind speed; the relevant factors among the remote sensing factors include point of interest data, region of interest data, normalized vegetation index, rainfall raster data, light data, and building proportion data.

3. The prediction method according to claim 2, characterized in that, In step S2, the PM corresponding to the different time periods 2.5 The method for constructing a LUR model for ozone includes the following steps: (1) PM 2.5 Bivariate correlation analysis was performed between the annual average, quarterly average, weekend average, weekday average, and special day average of ozone concentration and the relevant factors in the driving factor data to screen out factors related to PM2.

5. 2.5 Driving factors that are significantly correlated with ozone concentration; (2) Select the ones that match PM 2.5 The driving factors significantly correlated with ozone concentration were subjected to stepwise linear regression to construct PM2.5 concentration. 2.5 The mapping relationship between PM2.5 concentration, ozone concentration, and significantly related driving factors was obtained, yielding the corresponding PM2.5 concentrations for different time periods. 2.5 LUR model for ozone.

4. The prediction method according to claim 2, characterized in that, In step S2, the PM 2.5 The method for constructing an improved RF model for ozone includes the following steps: (a) PM 2.5 The annual average, quarterly average, weekend average, weekday average, and special day average of ozone concentrations were used to construct a random forest model with the driving factor data. (b) Two rounds of screening were performed using a random forest model to select those that match PM. 2.5 Driving factors significantly correlated with ozone concentration were used to construct PM 2.5 The mapping relationship between PM2.5 concentration, ozone concentration and significantly related driving factors was obtained. 2.5 An improved RF model for ozone.

5. The prediction method according to any one of claims 1 to 4, characterized in that, In step S4, the hexagonal grid is a regular hexagonal grid; the side length of the regular hexagonal grid is 1 km.

6. The prediction method according to any one of claims 1 to 4, characterized in that, In step S7, the natural breakpoint method is used to divide PM in the corresponding time period. 2.5 The predicted concentrations of O3 are divided into three levels: low, medium, and high, corresponding to light pollution, medium pollution, and high pollution levels, respectively.

7. The prediction method according to claim 6, characterized in that, In step S7, the PM in the area to be predicted 2.5 The mixed pollution prediction map for ozone and other pollutants shows the following mixed pollution scenarios: The first type is PM2.5 in the area to be predicted. 2.5 The pollution level is low, and the ozone pollution level is low. The second type is PM in the area to be predicted. 2.5 The pollution level is low, while the ozone pollution level is moderate. The third type is PM in the area to be predicted. 2.5 The pollution level is low, while the ozone pollution level is high. The fourth type is PM2.5 in the area to be predicted. 2.5 The pollution level is moderate, while the ozone pollution level is low. The fifth type is PM2.5 in the area to be predicted. 2.5 The pollution level is moderate, and the ozone pollution level is also moderate. The sixth type is PM2.5 in the area to be predicted. 2.5 The pollution level is moderate, and the ozone pollution level is high. The seventh type is PM2.5 in the area to be predicted. 2.5 The pollution level is high, while the ozone pollution level is low. The eighth type is PM2.5 in the area to be predicted. 2.5 The pollution level is high, and the ozone pollution level is medium. The ninth type is PM2.5 in the area to be predicted. 2.5 The pollution level is high, and the ozone pollution level is also high.

Images