CN116698688A - Method for estimating concentration of atmospheric particulates based on double-star of cloud number 4 - Google Patents

Abstract

The invention discloses a method for estimating the concentration of atmospheric particulates based on the synergy of wind and cloud number 4 double stars, which comprises the steps of matching the top reflectance (TOAR) data of the atmosphere after cloud pixels are removed with relevant auxiliary data, establishing a TOAR data filling model based on a data set and an Extreme Tree (ET) model obtained by matching, filling TOAR data after cloud pixels are removed by FY-4A and FY-4B satellites through the TOAR data filling model, and obtaining TOAR filling data under the completely covered assumed clear sky condition; PM is observed by TOAR filling data and related auxiliary data and site ₁₀ And PM _2.5 And (3) matching the data, establishing a TOAR-particulate matter estimation model based on the matched data set and the ET model, assimilating the double-star data by using the model, and estimating the concentration of the particulate matters. The invention can realize the double-star cooperative estimation of the concentration of the particles, the built TOAR-particle estimation model has better performance, and the obtained particlesThe particle distribution is more accurate, and the method has high coverage rate and high space-time resolution.

Description

Method for estimating concentration of atmospheric particulates based on double-star of cloud number 4

Technical Field

The invention belongs to the technical field of atmospheric pollutant monitoring, and particularly relates to a method for cooperatively estimating atmospheric Particulate Matters (PM) based on observation data of China domestic wind cloud No. 4 double satellites (stationary satellites FY-4A and FY-4B) _2.5 and PM₁₀ ) Concentration method.

Background

Atmospheric Particulates (PM) having aerodynamic diameters of less than 2.5 μm and 10 μm _2.5 and PM₁₀ ) Has great influence on the ecological environment of the earth and the life health of human beings. Since 2013, china successively established thousands of environmental monitoring stations to acquire particulate matter concentration information. However, these ground sites are unevenly distributed, and the concentration of particulate matter in a large spatial scale cannot be obtained, limiting the research. In recent years, in order to obtain the concentration of the particulate matter with higher spatial resolution, a method for obtaining the concentration of the particulate matter based on satellite remote sensing data is widely used.

The satellite remote sensing product that is often used today to estimate particulate matter concentration is aerosol optical thickness (AOD). AOD is the integral of the extinction coefficient of an aerosol from the ground to the top of the atmosphere, which has a strong correlation with near-ground particulate. Particulate matter concentration can be effectively estimated using AOD, but the AOD inversion algorithm has uncertainty that can further affect the estimation of particulate matter.

AOD data used in earlier particulate matter estimation studies have been from a variety of sources, such as Moderate-resolution Imaging Spectroradiometer (MODIS), visible Infrared Imaging Radiometer Suite (VIIRS), multiangle Imaging Spectro Radiometer (MISR) and Advanced Himawari Imager (AHI). MODIS, VIRRS and MISR sensors are mounted on polar satellites and AHI is mounted on Himawari-8 stationary satellites. However, neither satellite can observe the chinese world with high space-time frequency. Polar satellites can provide remote sensing data covering the entire china area, but with a time resolution of only daily; himaware-8 stationary satellites increase the time resolution of remote sensing data to an hour, but cannot cover the world of China and cannot acquire valid data in the western region of China.

To avoid the effects of uncertainty in the AOD inversion algorithm, some students directly acquire the concentration of particulate matter using satellite atmospheric top-of-the-Atmosphere Reflectance. In addition, the new generation of geostationary meteorological satellite FY-4A in China emits and goes off in the year of 12 and 11 in 2016, the satellite point is 104.7 DEG E, and a Advanced Geosynchronous Radiation Imager (AGRI) imager carried by the satellite can provide a multi-band full-disc image covering the whole world of China, and the time resolution can reach 15 minutes; the second new generation of stationary orbit meteorological satellite FY-4B in China is launched and lifted off in 2021, the position of the satellite below the satellite is 133 DEG E, the carried improved Advanced Geosynchronous Radiation Imager (AGRI) imager can provide observation information (one more water vapor channel than FY-4A) in 15 wave bands, the time resolution is 15 minutes, and the spatial resolution is 1-4 km.

At present, research shows that the machine learning method and FY-4A satellite TOAR data can be used for estimating the concentration of the particles with high space-time resolution covering the whole Chinese environment, and the accuracy of inverting the concentration of the particles by satellite remote sensing data is effectively improved. PM (particulate matter) _2.5 Estimating hourly R of model ² 0.83-0.88, RMSE of 8.81 [ mu ] g/m 2.7 [ mu ] g/m; PM (particulate matter) ₁₀ Estimating hourly R of model ² 0.72 to 0.85, and an RMSE of 18.22 mu g/m of the total length of 33.67 mu g/m of the total length of the total. These studies all employ a machine learning model named deep forest that references the structure of the neural network and replaces neurons with decision tree models. By combining the advantages of the neural network and the tree model, the deep forest model can better fit data and provide feature importance, so that the model has interpretability.

However, either AOD data or stationary satellite TOAR data, only one type of satellite data is involved in performing the particulate matter estimation. Because the satellite has a fixed orbit and a fixed satellite point, poor observation results can be obtained in some observation blind areas under the influence of an observation range and an angle, and thus, the data cannot be represented in some places.

FY-4A and FY-4B two geostationary satellites form a first group of meteorological double-star observation systems in China. In view of the excellent performance of FY-4A in particulate matter estimation, through assimilating TOAR observation data of FY-4A and FY-4B, atmospheric particulate matters are estimated again by using satellite data after assimilation, and the double-star collaborative observation of the concentration of the atmospheric particulate matters can be realized. Estimating the concentration of particulate matter using satellite TOAR data may help one monitor the distribution of the concentration of particulate matter at a higher spatial-temporal resolution. However, satellite-observed TOAR data suffers from a low spatial coverage by the cloud, particularly at the hour level, and has a large number of missing values. This is very disadvantageous for obtaining a fully covered particulate concentration product. Therefore, there is a need to find an effective way to pad satellite observations. In earlier studies, AOD data products of MODIS were filled based on linear regression, kriging interpolation, and machine learning methods to obtain fully covered particulate products. However, these studies result in a low time resolution product of particulate concentration and do not allow continuous monitoring of the spatial distribution of particulate.

Disclosure of Invention

The invention aims to provide a method for synergistically estimating atmospheric Particulate Matters (PM) based on observation data of China domestic Fengyun No. 4 double satellites (stationary satellites FY-4A and FY-4B) _2.5 and PM₁₀ ) The concentration method aims at solving the problems that the acquisition of the current atmospheric particulate concentration data product is difficult, the coverage rate of the existing data space is insufficient and the space-time resolution is low.

The invention discloses a method for estimating the concentration of atmospheric particulates based on the cooperative estimation of the No. 4 double stars of wind and cloud, which comprises the following steps:

s1, matching the TOAR data of the top reflectivity of the atmosphere after cloud removal with related auxiliary data, establishing a TOAR data filling model based on a data set and an extreme tree model obtained by matching, filling TOAR observation data of FY-4A and FY-4B satellites after cloud removal by the TOAR data filling model, and obtaining TOAR filling data under the completely covered assumed clear sky condition;

s2, combining the TOAR filling data, the related auxiliary data and site observation PM ₁₀ and PM_2.5 And (3) matching the data, establishing a TOAR-particulate matter ET estimation model based on the matched data set and the extreme tree model, assimilating the double-star data based on the model, and estimating the concentration of the particulate matters.

Preferably, before step S1, the method further comprises the step of:

s0, data preparation: the spatial resolution of the associated assistance data is adjusted to 0.04 x 0.04 unified with the satellite observations using bilinear interpolation.

Preferably, the relevant auxiliary data includes meteorological elements, altitude, land use type and population density.

Preferably, the TOAR padding data and related auxiliary data are input values, and the site observes PM ₁₀ and PM_2.5 The data is a tag value.

Preferably, PM is observed for a site ₁₀ and PM_2.5 If only one site exists in one grid, setting the observed value of the site as a tag value; if there are multiple sites in a grid, the average of these site observations is set as the tag value.

Preferably, in step S1, the processing procedure of the TOAR data after cloud rejection is:

TOAR data of the FY-4A satellite and the FY-4B satellite are subjected to cloud removal and removal processing by using cloud processing products CLM corresponding to the TOAR data, and pixels which are determined and possibly are cloud in the cloud products are removed;

and projecting the TOAR data subjected to cloud rejection processing to a plane coordinate system to obtain data which can be directly used.

Preferably, in step S1, the TOAR data padding model is:

wherein ,frepresenting the ET model,ithe position of the grid is indicated,jthe time is represented by the time period of the day,nrepresents a channel number; the dependent variables are TOARs after cloud rejection to be filled, and the independent variables comprise: boundary layer heightBLHRelative humidity ofRHTwo meters of air temperature on the earthTWind speedWSWind directionWDLow vegetation indexLLHigh vegetation indexLHSurface air pressureSPSolar radiation on earth's surfaceFSType of earth's surfaceLUCCAltitude of seaHEIGHTYear (year)YEARAnd what day of the yearDOY。

Preferably, in step S2, the TOAR-particulate matter ET estimation model is:

wherein ,f representing a TOAR-particulate ET estimation model,ithe position of the grid is indicated,jthe time is represented by the time period of the day,nTOAR data representing culling clouds, retaining clouds and assuming clear sky conditions under different conditions, dependent variables are PM ₁₀ and PM_2.5 The independent variables include: TOAR, boundary layer heightBLH、Relative humidity ofRHTwo meters of air temperature on the earthTWind speedWSWind directionWDLow vegetation indexLLHigh vegetation indexLHSurface air pressureSPSolar radiation on earth's surfaceFSType of earth's surfaceLUCCAltitude of seaHEIGHTYear (year)YEARAnd what day of the yearDOY。

Preferably, after step S2, the method further comprises the steps of:

s3, combining and smoothing the estimated particulate matter concentration data results of the FY-4A and the FY-4B by using a linear regression model to obtain a particulate matter distribution result, wherein the calculation formulas are shown in (4) and (5):

wherein ,frepresenting a linear regression model, PM _2.5i,j,4A and PM_10i,j,4A PM representing FY-4A on 4km resolution grid in China _2.5 and PM₁₀ Concentration; PM (particulate matter) _2.5i,j,4B and PM_10i,j,4B PM representing FY-4B on 4km resolution grid in China _2.5 and PM₁₀ Concentration.

The invention overcomes the defects of the prior art and provides a method for estimating the concentration of atmospheric particulates based on the cooperative estimation of the wind cloud No. 4 double stars. According to the invention, the satellite TOAR data of FY-4A and FY-4B are filled by using a machine learning model, so that satellite observation TOAR data which is completely covered and has high space-time resolution under the assumed clear sky condition is obtained, the FY-4A and FY-4B satellite data are assimilated based on the satellite observation TOAR data, and finally the concentration of particles is estimated.

Specifically, in the present invention, the resolution of meteorological elements, altitude, land use type and population density is adjusted to 4km (spatial resolution of satellite data) using bilinear interpolation. For site data, if only one site exists in one grid, setting the observed value of the site as a tag value; if there are multiple sites within a grid, the average of these site observations is set to the tag value. Finally, all data (satellite TOAR data, meteorological elements, altitude, land use type and population density) were matched based on a 0.04 ° x 0.04 ° data grid for FY number four satellites. Based on the data obtained by matching and the extreme tree model, a TOAR data filling method is constructed, namely satellite TOAR data, meteorological elements, altitude, land utilization types, population density and the like are input into the extreme tree model, a TOAR data filling model is built, and based on the model, TOAR filling data under the full-coverage assumed clear sky condition is obtained.

Then respectively collecting TOAR data of the removed cloud, TOAR data of the reserved cloud and TOAR data of the full coverage under the assumed clear sky condition, and observing PM with meteorological elements, altitude, land utilization type, population density and sites ₁₀ and PM_2.5 The data are matched to obtain three training data sets (TOAR data of the cloud are removed, TOAR data of the cloud are reserved, and full coverage TOAR data under the condition of clear sky is assumed, and the full coverage TOAR data are respectively matched with other data), an extreme tree model is input, a TOAR-particulate matter ET estimation model is built, and model performance is compared.

The results show that compared with the defects and shortcomings of the prior art, the invention has the following beneficial effects: compared with the TOAR data of the removed cloud and the TOAR data of the reserved cloud, the TOAR-particle ET estimation model performance obtained by the TOAR data establishment under the full-coverage assumed clear sky condition obtained by using the TOAR data filling model is the best. When the data is used, the sample size is improved by 2 times, and PM ₁₀ and PM_2.5 Model R of (2) ² About 0.72 to 0.78, and the RMSE is 6.7 to 15.23 mug/m. The TOAR filling method has the advantages that the TOAR under the assumed clear sky condition is obtainedR between data and raw TOAR data ² Above 0.8, the method is superior to the past AOD filling method, and the time resolution can be improved to the minute level. When the TOAR data of the two satellites FY-4A and FY-4B are used for estimating the concentration of the particulate matters, the same model performance can be achieved; and (3) carrying out linear fusion on the estimation results of the two satellites so as to realize double-satellite collaborative observation. The particulate matter concentration product obtained by the full-coverage TOAR data under the condition of clear sky is applied, and has complete coverage and high space-time resolution.

Drawings

FIG. 1 is a schematic flow diagram of the process of the present invention;

FIG. 2 is a schematic diagram of a machine learning model;

FIG. 3 is a particulate matter filling result;

fig. 4 is a distribution result of the double star cooperative observation of the particulate matter.

Detailed Description

The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

The embodiment of the invention discloses a method for estimating the concentration of atmospheric particulates based on the cooperative estimation of a wind cloud No. 4 double star, which comprises the following steps:

s0, data preparation: the spatial resolution of the associated aiding data was adjusted to be uniform with the satellite observations (0.04. Times.0.04. Using bilinear interpolation)

As shown in part A of FIG. 1, TOAR data for FY-4A and FY-4B are used to estimate atmospheric particulates, which includes 4 bands: 0.45-0.49 [ mu ] m (blue light band), 0.55-0.75 [ mu ] m (green light band), 0.75-0.90 [ mu ] m (red light band) and 2.1-2.35 [ mu ] m (near infrared band). Cloud detection products (CLMs) were used to perform the cloud removal process with a spatial resolution of 4km of data.

As shown in part B of FIG. 1, the meteorological parameters include boundary layer height [ ]BLH) Temperature of 2mTM) Relative humidity%RH) U, V component of 10m windU10、V10) Surface air pressureSP). Some geographic information is also used, including: high and low vegetation index [ ]LH，LL) To represent the earth's surface coverage type; altitude of the seaHEIGHT) SRTM-3 elevation data, spatial resolution 90m, measured by a combination of the United states space agency (NASA) and the national surveying and mapping agency (NIMA); population density data was adjusted to 0.04 ° by 0.04 ° spatial resolution using NASA socioeconomic data and 2015 united nations provided by the application center (secac). Time variable [ ]TIME) And the spatial resolution of all the data is adjusted to 4km by bilinear interpolation and is unified with satellite observation data, wherein the time variable is the same as that in ERA-5 analysis data.

As shown in part C of FIG. 1, the data used includes site observation PM ₁₀ and PM_2.5 Data.

S1, matching TOAR data after cloud removal with related auxiliary data, establishing a TOAR data filling model based on a data set and an extreme tree model obtained by matching, filling TOAR observation data after cloud removal of FY-4A and FY-4B satellites by the TOAR data filling model, and obtaining TOAR filling data under completely covered assumed clear sky conditions

In step S1, cloud rejection processing is performed on TOAR data of FY-4A and FY-4B, respectively. Because the positions of the points under the satellites of the FY-4A satellite and the FY-4B satellite are different, cloud removal processing is needed by using respective corresponding cloud processing products, and pixels which are determined and possibly are cloud in the cloud products are removed. The satellite data is then projected onto a planar coordinate system based on satellite parameters (long half axis, earth to satellite centroid distance, offset, scale factor, etc.).

And matching the TOAR data after cloud removal with meteorological elements, geographic information, population density and the like. As shown in stage1 in fig. 1, a TOAR data padding model under the condition of assumption of clear sky is constructed by using the matched data set and an ET model (an extreme tree model, a model structure is shown in fig. 2), and the expression of the TOAR data padding model is shown in formula (1):

wherein ,frepresenting the ET model,ithe position of the grid is indicated,jthe time is represented by the time period of the day,nrepresents a channel number; the dependent variables are TOARs after cloud rejection to be filled, and the independent variables comprise: boundary layer heightBLHRelative humidity ofRHTwo meters of air temperature on the earthTWind speedWSWind directionWDLow vegetation indexLLHigh vegetation indexLHSurface air pressureSPSolar radiation on earth's surfaceFSType of earth's surfaceLUCCAltitude of seaHEIGHTYear (year)YEARAnd what day of the yearDOY。

Through the TOAR data filling model, the radiation quantity reaching the satellite observation instrument from the earth surface under the condition of the clear sky is researched and obtained, namely, the TOAR filling data of complete coverage under the condition of the clear sky is assumed. The coverage rate of TOAR data is greatly improved by the model, and the influence of cloud is eliminated.

S2, combining the TOAR filling data, the related auxiliary data and site observation PM ₁₀ and PM_2.5 Matching the data, establishing a TOAR-particulate matter ET estimation model based on a data set obtained by matching and an extreme tree model, assimilating double-star data based on the model, and estimating the concentration of particulate matters

In step S2, in order to facilitate the technical effect obtained by the present invention, in the embodiment of the present invention, using the data shown in fig. 1D, TOAR data of the reject cloud, TOAR data of the reserve cloud, and a TOAR-particulate matter ET estimation model (i.e., stage2 in fig. 1) for estimating the particulate matter concentration assuming completely covered TOAR filling data under clear sky conditions are respectively constructed, and the obtained model is shown in fig. 1E. For TOAR data sets under different conditions, the meteorological data and geographic information used are the same, and then the matched site data is observed for PM ₁₀ and PM_2.5 Modeling is performed as a tag value. The expression of the TOAR-particulate ET estimation model is shown in the formula (2) and the formula (3):

wherein ,f representing a TOAR-particulate ET estimation model,ithe position of the grid is indicated,jthe time is represented by the time period of the day,nTOAR data representing TOAR under different conditions (reject cloud, preserve cloud and assume TOAR under clear sky conditions). Dependent variable is PM ₁₀ and PM_2.5 The independent variables include: TOAR, boundary layer heightBLH、Relative humidity ofRHTwo meters of air temperature on the earthTWind speedWSWind directionWDLow vegetation indexLLHigh vegetation indexLHSurface air pressureSPSolar radiation on earth's surfaceFSType of earth's surfaceLUCCAltitude of seaHEIGHTYear (year)YEARAnd what day of the yearDOY。

Because the undersea points of the two satellites FY-4A and FY-4B are far apart, there is some difference in the earth observation effect of the two satellites. Wherein, when estimating the atmospheric particulates by using the FY-4A or FY-4B satellite data alone, PM constructed by TOAR data of FY-4A and FY-4B reject clouds ₁₀ and PM_2.5 Model Performance is generally, R ² 0.7,0.7,0.67 and 0.66, respectively, RMSE was 13.86. Mu.g/m, 6.66. Mu.g/m, and 6.74. Mu.g/m, respectively; PM constructed from TOAR of FY-4A and FY-4B retention clouds ₁₀ and PM_2.5 The model performance is greatly improved, R ² To 0.78, 0.77, 0.73 and 0.72, respectively. The model performance when estimating the concentration of the particulate matter by using the fully covered TOAR filling data under the condition of supposing clear sky is similar to the model performance constructed by the TOAR data under the condition of no cloud processing. Wherein PM of FY-4A ₁₀ and PM_2.5 Model R ² 0.78 and 0.72, respectively, and RMSE of 15.18 μg/m < mu > and 6.7 μg/m < mu > respectively; PM of FY-4B ₁₀ and PM_2.5 Model R ² Also 0.78 and 0.72, respectively, and RMSE is 15.23. Mu.g/m, and 6.71. Mu.g/m, respectively. Although the estimated performance of the TOAR preserving cloud and the total coverage TOAR padding assuming clear sky conditions is similar, the TOAR model preserving cloud may lack representation (cloud impact) for the ground.

Therefore, the present invention considers that the estimated particulate matter concentration of the fully covered TOAR filling data under clear sky conditions is more representative of near-ground pollution.

S3, combining and smoothing the estimated particle concentration data results of the FY-4A and the FY-4B by using a linear regression model to obtain a particle distribution result

As shown in stage3 of FIG. 1, after estimating the particulate concentration data for FY-4A and FY-4B, respectively, the results were combined using a linear regression model. The calculation mode is that on grid points with data of FY-4A or FY-4B, the particle concentration estimated by using the two satellites is averaged; at other grid points with cloud coverage, which satellite has observation data uses the estimation result of which satellite, and then averages the estimation result with the total coverage TOAR data of the other satellite. That is, at each grid point, observations from two satellites are used. The calculation formulas are shown as (4) and (5):

wherein ,frepresenting a linear regression model, PM _2.5i,j,4A and PM_10i,j,4A PM representing FY-4A on 4km resolution grid in China _2.5 and PM₁₀ Concentration; PM (particulate matter) _2.5i,j,4B and PM_10i,j,4B PM representing FY-4B on 4km resolution grid in China _2.5 and PM₁₀ Concentration.

As shown in fig. 3, coverage rate of particulate matter concentration data estimated by the TOAR-particulate matter ET estimation model of the reject cloud is obviously lower; the coverage rate of the data of the particle concentration obtained by the TOAR-particle ET estimation model established by the TOAR data of the reserved cloud and the TOAR data under the assumed clear sky condition is 100%, but the particle concentration of the data of the reserved cloud is slightly higher. The method comprises the steps of establishing TOAR-particle ET estimation models by TOAR data under a clear sky condition, wherein the particle concentration distribution obtained by the TOAR-particle ET estimation models is assumed to be the most consistent with site distribution. Therefore, it is feasible to fill the TOAR data, obtain TOAR data under the condition of a hypothetical clear sky, and then establish a TOAR-particulate ET estimation model based on the data to obtain a full coverage and high time resolution particulate product.

Finally, because the data obtained by direct calculation are formed by isolated points, 2 points or 3 points (8-12 km) are used for smoothing the data obtained by calculation, and finally, a usable particle distribution result is obtained. The final particulate concentration distribution product is shown in fig. 4.

The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.

Claims (9)

1. The method for estimating the concentration of the atmospheric particulates based on the cooperative double-star of the wind cloud No. 4 is characterized by comprising the following steps of:

s1, matching the TOAR data of the top reflectivity of the atmosphere after cloud removal with related auxiliary data, establishing a TOAR data filling model based on a data set and an extreme tree model obtained by matching, filling TOAR observation data of FY-4A and FY-4B satellites after cloud removal by the TOAR data filling model, and obtaining TOAR filling data under the completely covered assumed clear sky condition;

s2, combining the TOAR filling data, the related auxiliary data and site observation PM ₁₀ and PM_2.5 And (3) matching the data, establishing a TOAR-particulate matter ET estimation model based on the matched data set and the extreme tree model, assimilating the double-star data based on the model, and estimating the concentration of the particulate matters.

2. The method of claim 1, further comprising the step of, prior to step S1:

s0, data preparation: the spatial resolution of the associated assistance data is adjusted to 0.04 x 0.04 unified with the satellite observations using bilinear interpolation.

3. The method of claim 2, wherein the relevant auxiliary data includes meteorological elements, altitude, land use type, and population density.

4. The method of claim 2 wherein the TOAR padding data and associated auxiliary data are input values and the site observes PM ₁₀ and PM_2.5 The data is a tag value.

5. The method of claim 4, wherein PM is observed for a site ₁₀ and PM_2.5 If only one site exists in one grid, setting the observed value of the site as a tag value; if there are multiple sites in a grid, the average of these site observations is set as the tag value.

6. The method of claim 1, wherein in step S1, the processing procedure of the TOAR data after cloud rejection is:

TOAR data of the FY-4A satellite and the FY-4B satellite are subjected to cloud removal and removal processing by using cloud processing products CLM corresponding to the TOAR data, and pixels which are determined and possibly are cloud in the cloud products are removed;

and projecting the TOAR data subjected to cloud rejection processing to a plane coordinate system to obtain data which can be directly used.

7. The method of claim 1, wherein in step S1, the TOAR data padding model is:

wherein ,frepresenting the ET model,ithe position of the grid is indicated,jthe time is represented by the time period of the day,nrepresents a channel number; the dependent variables are TOARs after cloud rejection to be filled, and the independent variables comprise: boundary layer heightBLHRelative humidity ofRHTwo meters of air temperature on the earthTWind speedWSWind directionWDLow vegetation indexLLHigh vegetation indexLHSurface air pressureSPSolar radiation on earth surfaceRadiationFSType of earth's surfaceLUCCAltitude of seaHEIGHTYear (year)YEARAnd what day of the yearDOY。

8. The method of claim 1, wherein in step S2, the TOAR-particulate ET estimation model is:

wherein ,f representing a TOAR-particulate ET estimation model,ithe position of the grid is indicated,jthe time is represented by the time period of the day,nTOAR data representing culling clouds, retaining clouds and assuming clear sky conditions under different conditions, dependent variables are PM ₁₀ and PM_2.5 The independent variables include: TOAR, boundary layer heightBLH、Relative humidity ofRHTwo meters of air temperature on the earthTWind speedWSWind directionWDLow vegetation indexLLHigh vegetation indexLHSurface air pressureSPSolar radiation on earth's surfaceFSType of earth's surfaceLUCCAltitude of seaHEIGHTYear (year)YEARAnd what day of the yearDOY。

9. The method of claim 1, further comprising the step, after step S2, of:

s3, combining and smoothing the estimated particulate matter concentration data results of the FY-4A and the FY-4B by using a linear regression model to obtain a particulate matter distribution result, wherein the calculation formulas are shown in (4) and (5):

wherein ,frepresenting a linear regression model, PM _2.5i,j,4A and PM_10i,j,4A PM representing FY-4A on 4km resolution grid in China _2.5 and PM₁₀ Concentration; PM (particulate matter) _2.5i,j,4B and PM_10i,j,4B PM representing FY-4B on 4km resolution grid in China _2.5 and PM₁₀ Concentration.