A method for estimating PM2.5 based on MCD19A2 AOD product

Abstract

The present invention discloses a method for estimating PM2.5 based on MCD19A2 AOD products, comprising the following steps: step 1, acquisition of MCD19A2 data and auxiliary data; step 2, selection of the main research band, and linear regression based on the main research band; step 3, extraction of overlapping pixels based on each band; step 4, obtaining the auxiliary band after time normalization based on the overlapping pixels; step 5, kriging interpolation based on the fused AOD data; step 6, predicting PM2.5 in the study area based on the least squares method. The method of the present invention is based on the main band (the band with better data quality in the study area), linear normalization is performed based on the overlapping pixels of each band, and a fitting equation is obtained, thereby time normalization is performed on the bands with different acquisition times.

Description

Technical Field

[0001] The invention belongs to the field of PM2.5 prediction, and in particular relates to a method for estimating PM2.5 based on an MCD19A2 AOD product in the field. Background Art

[0002] PM2.5 monitoring is the basis of pollution prevention and control. Currently, remote sensing is mainly used to monitor over a large area. The mainstream method is to predict PM2.5 based on remote sensing aerosol (AOD) data combined with other auxiliary data. However, the commonly used aerosol products are 3KM resolution (MOD / MYD04_3K AOD) and 5KM resolution (Himawari-8AOD). The spatial resolution is low and there are many null values ​​in the data, which cannot meet the research of fine scale. Although the MCD19A2 product has 1KM resolution AOD data and few null values, it has good application value, but there is little research on it. The MCD19A2 product contains data collected by the satellite passing through the area in one day, so the data contains different numbers of bands (each band represents an AOD data acquisition, and different bands represent data at different times). The number of bands depends on the number of times the satellite acquires data in a day. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a method for estimating PM2.5 based on the MCD19A2 AOD product.

[0004] The present invention adopts the following technical solution:

[0005] A method for estimating PM2.5 based on MCD19A2 AOD product, the improvement of which is that it includes the following steps:

[0006] Step 1, acquisition of MCD19A2 data and auxiliary data; the auxiliary data include meteorological factors and land use factors, meteorological factors include air pressure, temperature, relative humidity, rainfall, wind speed and wind direction; the normalized difference vegetation index NDVI is used to replace the land use factor;

[0007] Step 2, selection of the main research band. The main research band is a band in the MCD19A2 product. You can choose any one of them. The other bands are auxiliary bands. Linear regression is performed based on the main research band.

[0008] Step 3, extract overlapping pixels based on each band;

[0009] Step 4: Based on overlapping pixels, the least square method is used to obtain the linear relationship between the auxiliary band and the main research band, and linear regression is performed to obtain the time-normalized auxiliary band. The time-normalized auxiliary band is merged with the main research band to reduce the null value area of ​​the main research band.

[0010] Step 5, performing Kriging interpolation based on the fused AOD data;

[0011] Step 6: Combine the PM2.5 data of the site, the interpolated AOD data, meteorological factors, and NDVI data to predict PM2.5 in the study area based on the least squares method;

[0012] In the following formula, T is the temperature, P is the air pressure, RH is the relative humidity, WS is the wind speed, WD is the wind direction, and NDVI is the normalized difference vegetation index.

[0013] PM2.5=β0+β1*AOD+β2*T+β3*P+β4*RH+β5*WS+β6*WD+β7*NDVI

[0014] Each factor parameter β0-β7 is estimated by the least squares method, and finally the PM2.5 value in the study area is obtained.

[0015] The beneficial effects of the present invention are:

[0016] At present, there are few studies on estimating PM2.5 distribution based on MCD19A2 products, and only spatial interpolation is used for multi-band fusion without considering time fitting. Therefore, the method of the present invention proposes PM2.5 distribution prediction based on AOD band fusion according to the characteristics of MCD19A2 products. Based on the consideration of both space and time, the AOD data of a single moment is obtained. The obtained fused data has a clear time meaning and can be better fitted with the site data. At the same time, in the process of fitting AOD and PM2.5, meteorological factors and land use factors are added, and the prediction accuracy is increased based on the fitting of multiple factors, which overcomes the shortcomings of many data null values ​​and low spatial resolution in previous studies.

[0017] Since the acquisition time of each band is similar in the time-normalized AOD band fusion, only spatial interpolation was performed in the past without considering the acquisition time of different bands. However, in fine-scale research, local AOD changes are very sensitive to time. In order to obtain AOD fusion data of a single time, the method of the present invention is based on the main band (the band with better data quality in the study area), and linear normalization is performed based on the overlapping pixels of each band to obtain a fitting equation, thereby performing time normalization on bands with different acquisition times. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1This is a schematic diagram of PM2.5 concentration in Xinxiang, Henan Province on December 10, 2020 obtained using the method of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0020] Example 1: This example discloses a method for estimating PM2.5 based on the MCD19A2 AOD product, which uses the time-normalized AOD band fusion method to combine AOD, meteorological elements, and land use factors to predict PM2.5. Specifically, it includes the following steps:

[0021] Step 1, acquisition and processing of MCD19A2 data and auxiliary data;

[0022] MCD19A2 data comes from NASA website. MCD19A2 data is an aerosol product obtained by multi-angle atmospheric correction algorithm, derived from two data sources, Terra and Aqua. The 1KM AOD in this product is a daily product, which contains data collected by the satellite passing through the area in one day. Therefore, the daily data contains different numbers of bands (each band represents an AOD data acquisition, and different bands represent data acquisition at different times). The number of bands depends on the number of times the satellite acquires data in a day.

[0023] Auxiliary data include meteorological factors and land use factors. Meteorological factors include air pressure (hPa), temperature (℃), relative humidity (%), rainfall (mm), wind speed (km / h) and wind direction (°); the normalized difference vegetation index NDVI is used to replace the land use factor; the meteorological factor data comes from the Meteorological Bureau, and the NDVI uses MOD13A2 data (16-day resolution data) from the NASA website.

[0024] Step 2: Select the main research band based on the study area;

[0025] The MCD19A2 1KM AOD product has multiple bands at different times. Generally, the band with better data quality (few null values ​​and high coverage) in the study area is selected as the main research band, and other bands are used as auxiliary bands to supplement the main research band and reduce the null value area in the study area.

[0026] Step 3, extract overlapping pixels based on each band;

[0027] Step 4: Based on overlapping pixels, the least square method is used to obtain the linear relationship between the auxiliary band and the main research band, and linear regression is performed to obtain the time-normalized auxiliary band. The time-normalized auxiliary band is merged with the main research band to reduce the null value area of ​​the main research band.

[0028] Band fusion based on time normalization: Since different bands have different acquisition times, time normalization is required when the auxiliary band is used as a supplement to the main band. The method of this embodiment performs time normalization based on the linear relationship between the auxiliary band and the main band, extracts the overlapping pixels of the auxiliary band and the main band, linearly fits the auxiliary band and the main band based on the overlapping pixels, obtains the linear relationship between the main band AOD and the auxiliary band AOD, and finally normalizes the auxiliary band AOD to the main band data acquisition time based on the obtained model;

[0029] Step 5: Perform Kriging interpolation based on the fused AOD data; when the fused data still has null values, use the Kriging interpolation method to interpolate.

[0030] Step 6: Combine the PM2.5 data of the site, the interpolated AOD data, meteorological factors, and NDVI data to predict PM2.5 in the study area based on the least squares method;

[0031] Implementation of the least squares model OLS model: Based on the interpolated AOD data combined with meteorological factor data and land use factors, a least squares model (OLS) was constructed according to the existing PM2.5 station data to estimate the PM2.5 concentration in the study area.

[0032] In the following formula, T is the temperature, P is the air pressure, RH is the relative humidity, WS is the wind speed, WD is the wind direction, and NDVI is the normalized difference vegetation index.

[0033] PM2.5=β0+β1*AOD+β2*T+β3*P+β4*RH+β5*WS+β6*WD+β7*NDVI

[0034] Each factor parameter β0-β7 is estimated by the least squares method, and finally the PM2.5 value in the study area is obtained.

[0035] With the continuous development of my country's economy, the problem of ambient air pollution has become increasingly prominent. Using satellite remote sensing to monitor the temporal and spatial distribution and changing trends of ambient air pollution can solve the problem that existing ground environmental monitoring cannot conduct large-scale regional ambient air quality monitoring. The method disclosed in this embodiment uses the MCD19A2 1KM AOD product to estimate PM2.5. On the basis of solving the problem of large-scale monitoring of regional ambient air quality, it further improves the refinement of remote sensing monitoring of air quality. Figure 1This is a schematic diagram of the PM2.5 concentration in Xinxiang, Henan Province on December 10, 2020, obtained using the method disclosed in this embodiment.

Claims

1. A method for estimating PM2.5 based on MCD19A2 AOD product, characterized in that: The steps include: Step 1, acquisition of MCD19A2 data and auxiliary data; the auxiliary data include meteorological factors and land use factors, meteorological factors include air pressure, temperature, relative humidity, rainfall, wind speed and wind direction; the normalized difference vegetation index NDVI is used to replace the land use factor; Step 2, selection of the main research band. The main research band is a band in the MCD19A2 product. You can choose any one of them. The other bands are auxiliary bands. Time normalization is performed based on the linear relationship between the auxiliary band and the main band. Step 3, extract overlapping pixels based on each band; Step 4: Based on overlapping pixels, the least square method is used to obtain the linear relationship between the auxiliary band and the main research band, and linear regression is performed to obtain the time-normalized auxiliary band. The time-normalized auxiliary band is merged with the main research band to reduce the null value area of ​​the main research band. Step 5, performing Kriging interpolation based on the fused AOD data; Step 6: Combine the PM2.5 data of the site, the interpolated AOD data, meteorological factors, and NDVI data to predict PM2.5 in the study area based on the least squares method; In the following formula, T is the temperature, P is the air pressure, RH is the relative humidity, WS is the wind speed, WD is the wind direction, and NDVI is the normalized difference vegetation index. PM2.5=β0+β1*AOD+β2*T+β3*P+β4*RH+β5*WS+β6*WD+β7*NDVI Each factor parameter β0-β7 is estimated by the least squares method, and finally the PM2.5 value in the study area is obtained.

Images