Atmospheric pollutant concentration mapping method considering multi-dimensional geographic similarity

Abstract

The invention discloses an atmospheric pollutant concentration mapping method considering multi-dimensional geographic similarity. The method is used for solving the problems that in an existing atmospheric pollutant concentration mapping technology, interpretability is insufficient, and non-linear correlation degree measurement is difficult. The method comprises the following steps: acquiring long-term geographical environment statistical data, and measuring geographical scene similarity between a to-be-monitored pixel in a to-be-monitored area and a foundation observation station by using a similarity calculation method to obtain a similarity matrix; dividing according to the similarity matrix to obtain a non-Euclidean neighborhood monitoring station set of a pixel scale; acquiring historical observation data of the to-be-monitored area, and constructing a graph topology network structure and a training data set of the to-be-monitored area according to the non-Euclidean neighborhood monitoring station set and the historical observation data; and training by utilizing the graph topology network structure and the training data set to obtain a target drawing model, wherein the target drawing model is used for generating the space seamless atmospheric pollutant concentration graph with the all-day hour resolution according to the real-time observation data of the to-be-monitored area.

Description

Technical Field

[0001] The present application relates to the field of meteorological data processing, and in particular to a method for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities. Background Art

[0002] Air pollutants refer to harmful substances produced by natural processes and human activities that have a negative impact on the environment or humans, including ozone, PM 10 、PM 2.5 Etc. Air pollution has significant negative impacts on the natural environment homeostasis, social and economic development, people's production and life, and many other aspects. Air quality monitoring stations can provide real-time, high-quality observation data on air pollutant concentrations, but the monitoring stations are mainly distributed in urban areas, and the spatial distribution is sparse, and the overall spatial coverage is insufficient, which cannot meet the task requirements of real-time assessment of regional air quality. Satellite optical remote sensing is easily affected by weather factors such as clouds and fog, and there are spatial omissions in the estimated products. At the same time, there is a lack of satellite optical remote sensing observation data at night, which makes it impossible to estimate the concentration of atmospheric pollutants at night. The spatial resolution and data accuracy of numerical model data are difficult to meet the needs of near-real-time regional atmospheric pollutant concentration monitoring.

[0003] In the prior art, the mainstream method for producing seamless grid data of atmospheric pollutant concentrations throughout the day is to infer unknown points through spatial interpolation methods using known sparse station observation data. However, for areas with sparse observation stations such as western China, traditional spatial interpolation methods based on spatial proximity show low interpolation accuracy. At the same time, many improved algorithms of traditional spatial interpolation use variables with high correlation between known points and unknown points to measure the similarity of their attributes and use this similarity as interpolation weights to improve the accuracy of interpolation results to a certain extent. However, due to the complex similarity of geographical scenes, such methods cannot simulate the nonlinear and complex relationship between the interpolation points and the neighboring monitoring stations, and cannot further improve the data accuracy of the interpolation results and limit their interpretability, which ultimately leads to low mapping accuracy of atmospheric pollutant concentrations and poor interpretability of the results. Summary of the invention

[0004] The present application provides a method for mapping atmospheric pollutant concentrations that takes into account multi-dimensional geographical similarities, which is used to solve the problems of insufficient interpretability and difficulty in measuring nonlinear correlations in existing atmospheric pollutant concentration mapping technologies.

[0005] The first aspect of the present application provides a method for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities, comprising:

[0006] Obtain long-term geographic environmental statistical data, and use similarity calculation methods to measure the geographic scene similarity between the pixels to be monitored in the monitored area and the ground-based observation sites to obtain a similarity matrix;

[0007] A non-Euclidean neighborhood monitoring site set at a pixel scale is obtained according to the similarity matrix division;

[0008] Acquiring historical observation data of the area to be monitored, wherein the historical observation data includes short-term meteorological reanalysis data and time series of atmospheric pollutant concentration observation data;

[0009] Constructing a graph topology network structure and a training data set of the area to be monitored according to the non-Euclidean neighborhood monitoring site set and the historical observation data, wherein the training data set includes a neighborhood monitoring site feature data set at a pixel scale and an atmospheric pollutant concentration observation target data set at a site scale;

[0010] The target mapping model is obtained by training using the graph topology network structure and the training data set. The target mapping model is used to generate a spatial seamless atmospheric pollutant concentration map with a full-day hourly resolution based on the real-time observation data of the area to be monitored.

[0011] Optionally, the similarity calculation method is used to measure the geographical scene similarity between the pixels to be monitored in the area to be monitored and the ground-based observation site to obtain a similarity matrix, including:

[0012] Determine the geographic environment feature vectors of each pixel to be monitored and all ground-based observation sites in the area to be monitored, and calculate statistical indicators for evaluating feature differences based on the geographic environment feature vectors;

[0013] Inputting the sequence composed of the statistical indicators into a similarity calculation method to obtain a similarity index value for measuring the similarity of the geographical environment scenes;

[0014] A similarity matrix of each pixel to be monitored in the area to be monitored is generated according to the similarity index value.

[0015] Optionally, the calculating of a statistical index for evaluating feature differences according to the geographic environment feature vector includes:

[0016] For each pixel to be monitored, calculate the normalized statistical index of the geographic environment characteristic vector and the geographic environment characteristic matrix of all ground-based observation sites;

[0017] The step of inputting the sequence of statistical indicators into a similarity calculation method to obtain a similarity index value for measuring the similarity of geographical environment scenes includes:

[0018] Inputting the sequence composed of the normalized statistical indicators into a mathematical statistical model for evaluating geographical similarity;

[0019] The output value of the mathematical statistics model is determined as a similarity index value for measuring the similarity of geographical environment scenes.

[0020] Optionally, when the atmospheric pollutant concentration map is a PM2.5 concentration map, the geographic environment feature vector includes at least one of historical meteorological variables Met, historical PM2.5 concentrations, historical AOD data, surface elevation data DEM, land use type data LULC, night light data NLT, and normalized vegetation index data NDVI;

[0021] The mathematical and statistical model for evaluating geographical similarity adopts:

[0022] P ij =||R(F ij ),(1-NRMSE(F ij ))||2

[0023] F ij ={AOD,PM 2.5 ,Met,DEM,LULC,NLT,NDVI}

[0024] Where i and j represent the pixel to be monitored and a specific ground-based observation station number, respectively. ij represents the similarity index value, ||·||2 represents the Euclidean norm, R is the relevant preset coefficient, NRMSE is the normalized root mean square error, F ij Represents the geographic environment feature vector.

[0025] Optionally, the non-Euclidean neighborhood monitoring site set at a pixel scale obtained by partitioning the similarity matrix includes:

[0026] Defining a neighborhood monitoring site of each pixel to be monitored according to the similarity matrix of each pixel to be monitored, wherein the neighborhood monitoring site is a plurality of sites arranged in descending order according to the similarity index value;

[0027] The neighborhood monitoring sites are integrated to obtain a non-Euclidean neighborhood monitoring site set for each pixel to be monitored.

[0028] Optionally, the non-Euclidean neighborhood monitoring site set is expressed as:

[0029]

[0030] Among them, S i represents the set of non-Euclidean neighborhood monitoring sites corresponding to the pixel i to be monitored; topK(·) represents the set of independent variable points with the largest K values, that is, a function that selects the K values ​​with the largest geographical environment similarity; K is the number of the neighborhood monitoring sites corresponding to each pixel to be monitored, represents the similarity matrix corresponding to the pixel i to be monitored, and N is the total number of ground-based observation sites.

[0031] Optionally, constructing the graph topology network structure and the training data set of the area to be monitored according to the non-Euclidean neighborhood monitoring site set and the historical observation data includes:

[0032] Dynamically calculating the geographical environment similarity between the pixel to be monitored and the neighborhood monitoring site according to the non-Euclidean neighborhood monitoring site set and the historical observation data;

[0033] Taking the geographical environment similarity as edge weight and the geographical environment characteristics of the pixels to be monitored and the neighboring monitoring sites as nodes, a graph topology network structure of the area to be monitored is constructed;

[0034] A neighborhood monitoring site characteristic data set at a pixel scale is formed according to the graph network topology structure and the observed atmospheric pollutant concentration data of the neighborhood monitoring sites, and an atmospheric pollutant concentration observation target data set at a site scale is formed according to the atmospheric pollutant concentration observation data of the sites to be monitored.

[0035] The second aspect of the present application provides a system for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities, comprising:

[0036] A calculation unit is used to obtain long-term geographic environment statistical data and use a similarity calculation method to measure the geographic scene similarity between the pixels to be monitored in the monitored area and the ground-based observation site to obtain a similarity matrix;

[0037] A partitioning unit, used for obtaining a non-Euclidean neighborhood monitoring site set at a pixel scale according to the similarity matrix partitioning;

[0038] An acquisition unit, used for acquiring historical observation data of the area to be monitored, wherein the historical observation data includes short-term meteorological reanalysis data and a time series of atmospheric pollutant concentration observation data;

[0039] A construction unit, used to construct a graph topology network structure and a training data set of the area to be monitored according to the non-Euclidean neighborhood monitoring site set and the historical observation data, wherein the training data set includes a neighborhood monitoring site feature data set at a pixel scale and an atmospheric pollutant concentration observation target data set at a site scale;

[0040] A training unit is used to train a target mapping model using the graph topology network structure and the training data set, and the target mapping model is used to generate a spatial seamless atmospheric pollutant concentration map with a full-day hourly resolution based on the real-time observation data of the area to be monitored.

[0041] A third aspect of the present application provides a device for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities, the device comprising:

[0042] Processor, memory, input-output unit, and bus;

[0043] The processor is connected to the memory, the input and output unit, and the bus;

[0044] The memory stores a program, and the processor calls the program to execute the first aspect and any optional method of mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities in the first aspect.

[0045] The fourth aspect of the present application provides a computer-readable storage medium having a program stored thereon, which, when executed on a computer, executes the method for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities in the first aspect and any optional method in the first aspect.

[0046] It can be seen from the above technical solutions that this application has the following advantages:

[0047] First, long-term geographic environmental statistical data is obtained, and similarity calculation methods are used to measure the geographic scene similarity between the pixels to be monitored and the ground-based observation sites in the monitored area, so as to construct a similarity matrix and divide the non-Euclidean neighborhood monitoring site set at the pixel scale. This process can effectively capture the similar characteristics between different regions, ensuring that more levels of geographical and meteorological environmental influences can be considered when selecting neighborhood monitoring sites, avoiding the single influence of simple spatial proximity on the results. After that, the historical short-term meteorological reanalysis data and atmospheric pollutant concentration observation data are combined to construct a graph topology network structure and training data set, and the target mapping model is trained based on this. The target mapping model uses graph attention network technology, which can consider the potential nonlinear associations between geographic objects and enhances the ability to handle complex relationships between geographic objects compared to existing spatial interpolation technologies.

[0048] Through multi-dimensional geographic similarity analysis, the division of non-Euclidean neighborhood monitoring site sets and the application of graph attention network technology, the problems of insufficient interpretability and difficulty in measuring nonlinear correlations in existing atmospheric pollutant concentration mapping technologies have been solved, the accuracy and flexibility of mapping results have been improved, and all-day, hourly resolution, and spatially seamless atmospheric pollutant concentration monitoring has been successfully achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0050] Figure 1 A schematic flow chart of an embodiment of a method for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities provided in this application;

[0051] Figure 2 A schematic diagram of the method architecture for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities provided in this application;

[0052] Figure 3 The method for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities provided in this application is aimed at PM 2.5 A schematic flow diagram of an embodiment of a concentration diagram;

[0053] Figure 4 A scatter plot of the accuracy of mapping atmospheric pollutant concentrations in the method for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities provided in this application;

[0054] Figure 5 A schematic diagram of the structure of an embodiment of a system for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities provided by the present application;

[0055] Figure 6 A schematic diagram of the structure of an embodiment of a device for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities provided in the present application. DETAILED DESCRIPTION

[0056] The present application provides a method for mapping atmospheric pollutant concentrations that takes into account multi-dimensional geographical similarities, which is used to solve the problems of insufficient interpretability and difficulty in measuring nonlinear correlations in existing atmospheric pollutant concentration mapping technologies.

[0057] It should be noted that the method for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities provided in this application can be applied to terminals or servers. For example, the terminal can be a smart phone or computer, tablet computer, smart TV, smart watch, portable computer terminal or a fixed terminal such as a desktop computer. For the convenience of explanation, this application uses the terminal as an example for illustration.

[0058] See also Figure 1 , Figure 1 An embodiment of a method for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities provided in the present application comprises:

[0059] 101. Obtain long-term geographic environmental statistical data, and use similarity calculation methods to measure the geographic scene similarity between the pixels to be monitored in the monitored area and the ground-based observation sites to obtain a similarity matrix;

[0060] In the mapping of atmospheric pollutant concentrations, the traditional spatial interpolation method mainly selects reference observation sites based on Euclidean distance or spatial proximity, but this method ignores the differences in geographical, meteorological and environmental characteristics between different regions, resulting in insufficient interpretability and accuracy of the interpolation results. Based on this, in this embodiment, it is first necessary to collect long-term geographic environmental statistical data of the area to be monitored, that is, data reflecting long-term changes in geography, climate, pollutant concentrations and land use in the area to be monitored. The geographic environmental statistical data include but are not limited to atmospheric pollutant concentrations, atmospheric parameters that are highly correlated with atmospheric pollutant concentrations, meteorological variables, surface elevation data, etc., and all data are pre-processed by time and space matching. It should be noted that the long-term time scale includes but is not limited to the monthly scale or the annual scale, etc., and the pre-processing such as time and space matching includes: matching all input data according to the production time; unifying the spatial range and spatial resolution of the input data; normalizing the numerical range of the input data to the [0,1] interval.

[0061] After that, the collected geographic environment statistical data is combined with the similarity calculation method to measure the similarity between each pixel to be monitored and all ground-based observation sites in the monitored area, and the output value of the similarity calculation method is used as an indicator to measure the similarity of the geographic environment scene. For each pixel to be monitored in the monitored area, it is necessary to calculate the geographic similarity matrix between the pixel and the set of ground-based observation sites, and finally obtain the pixel-scale similarity matrix used to characterize the geographic environment in the monitored area.

[0062] 102. Obtain a non-Euclidean neighborhood monitoring site set at the pixel scale based on the similarity matrix partition;

[0063] Each element in the similarity matrix represents the similarity value between a pixel to be monitored and a ground-based observation site. For each pixel to be monitored, a certain number of neighborhood observation sites can be selected for it based on the data in its similarity matrix, so that the selection of neighborhood monitoring sites does not only depend on spatial distance, but is sorted according to the similarity between the pixel and each ground-based observation site. By selecting appropriate neighborhood monitoring sites for each pixel to be monitored, a non-Euclidean neighborhood monitoring site set containing all pixels to be monitored and their corresponding neighborhood monitoring sites can be finally obtained.

[0064] 103. Obtain historical observation data of the area to be monitored, including short-term meteorological reanalysis data and time series of atmospheric pollutant concentration observation data;

[0065] The distribution of atmospheric pollutant concentrations is greatly affected by meteorological conditions. Meteorological factors such as temperature, humidity, and wind speed can directly affect the diffusion and concentration changes of pollutants. In order to accurately map the atmospheric pollutant concentrations in the monitored area, in addition to considering the spatial geographical environment characteristics, it is also necessary to fully consider time series data, specifically the continuous historical meteorological reanalysis data and atmospheric pollutant concentration observation data time series on a short time scale. These historical observation data can reveal the law of pollutant concentration changes over time.

[0066] Specifically, meteorological reanalysis data is usually obtained by combining meteorological satellites, numerical forecast models and ground meteorological station data, which can provide accurate meteorological data within a certain time range. The atmospheric pollutant concentration observation data comes from ground monitoring stations or satellite remote sensing data, mainly including PM 2.5 、PM 10 , NO2, SO2, CO, O3 and other pollutant concentration values. The time scale of the historical observation data can be selected as hourly scale or daily scale, which can be determined according to the monitoring needs.

[0067] 104. Construct a graph topology network structure and a training data set of the area to be monitored based on the non-Euclidean neighborhood monitoring site set and historical observation data. The training data set includes a pixel-scale neighborhood monitoring site feature data set and a site-scale atmospheric pollutant concentration observation target data set.

[0068] The non-Euclidean neighborhood monitoring site set and historical observation data obtained in the previous calculation are used to construct the graph topology network structure of the monitored area, and generate feature data sets and target data sets for training the all-day atmospheric pollutant concentration mapping model.

[0069] The pixels and neighboring monitoring sites in the area to be monitored are used as nodes, and a graph topology network structure is constructed based on the similarity relationship between these nodes. Each node in the graph represents a pixel or an observation site, and the edges between the nodes represent the spatial-temporal association or similarity between these nodes. This graph topology network structure can effectively describe the geographical and meteorological similarities between the nodes, as well as the potential dependence of pollutant concentrations, providing a framework support for subsequent pollutant concentration prediction and interpolation.

[0070] The generated training data set specifically includes a pixel-scale neighborhood monitoring site feature data set and a site-scale atmospheric pollutant concentration observation target data set. The neighborhood monitoring site feature data set will be used as the input feature in the graph network, representing the geographical environment and pollutant concentration characteristics of the pixel to be monitored and its neighborhood monitoring site. The site-scale atmospheric pollutant concentration observation target data set contains the pollutant concentration data of each observation site at different time points, representing the pollutant concentration observation value at the site scale. This data set will be used as the target data set for model training and prediction, that is, for fitting the pollutant concentration value of the site as the target variable of the mapping model. Specifically, when constructing the neighborhood monitoring site feature data set, the feature data of the neighborhood monitoring site related to the pixel to be monitored can be extracted from the historical observation data based on the graph topology network structure. These feature data include information such as meteorological conditions and pollutant concentrations, which are the observation data of each neighborhood monitoring site at the current moment. These feature data are combined into a pixel-scale neighborhood monitoring site feature data set.

[0071] 105. The target mapping model is trained using the graph topology network structure and the training data set. The target mapping model is used to generate a spatial seamless atmospheric pollutant concentration map with all-day hourly resolution based on the real-time observation data of the area to be monitored.

[0072] Based on the constructed graph topology network structure and training data sets, namely the neighborhood monitoring station feature data set and the atmospheric pollutant concentration observation target data set, the initial mapping model is trained. Specifically, the initial mapping model adopts:

[0073] P = f r (G;s)

[0074] Where P represents the estimated concentration of atmospheric pollutants at the target time; f r (·) represents an atmospheric pollutant concentration estimation model with fitting characteristics to achieve the purpose of regression, including but not limited to graph convolutional neural network, random forest regression, support vector machine, etc.; G represents the graph topology network structure data; s represents the measured concentration of atmospheric pollutants at the current moment. 10-20% of the training data is randomly reserved as a validation set for model estimation cross-validation to test the generalization ability of the model.

[0075] The neighborhood monitoring site feature dataset in the training dataset provides the model with the geographical environment, meteorological characteristics and pollutant concentration information of the pixels to be monitored and their neighborhood monitoring sites. The target dataset provides the model with target pollutant concentration data to help the model learn the relationship between features and pollutant concentrations. The training dataset is input into the graph attention model for model parameter training, and modeling variable sensitivity analysis is carried out. By calculating the relative importance of the input variables, the variables that do not contribute significantly to the modeling accuracy are gradually eliminated, and the model structure is optimized until the model converges to obtain the target mapping model.

[0076] In actual mapping, the real-time meteorological reanalysis data and ground-based observation data on atmospheric pollutant concentrations are used, combined with the long-term geographic environmental statistical data and similarity matrix of the monitored area, to construct a pixel-scale neighborhood monitoring site characteristic data set for the monitored area. This data set is used as the input of the target mapping model, and the target mapping model is used to output an atmospheric pollutant concentration map of the monitored area with a full-day and hourly resolution. The map has high timeliness and spatial accuracy, and each area in the map can reflect the precise pollutant concentration.

[0077] In this embodiment, long-term geographic environmental statistical data is first obtained, and a similarity calculation method is used to measure the geographic scene similarity between the pixels to be monitored and the ground-based observation sites in the monitored area, so as to construct a similarity matrix and divide the non-Euclidean neighborhood monitoring site set at the pixel scale. This process can effectively capture the similar characteristics between different regions, ensuring that more levels of geographical and meteorological environmental influences can be considered when selecting neighborhood monitoring sites, avoiding the single influence of simple spatial proximity relationships on the results. After that, the historical short-term meteorological reanalysis data and atmospheric pollutant concentration observation data are combined, and these data are used to construct a graph topology network structure and a training data set, and the target mapping model is trained accordingly. The target mapping model uses graph attention network technology, which can consider the potential nonlinear associations between geographic objects, and enhances the ability to handle complex relationships between geographic objects compared to existing spatial interpolation technologies.

[0078] Through multi-dimensional geographic similarity analysis, the division of non-Euclidean neighborhood monitoring site sets and the application of graph attention network technology, the problems of insufficient interpretability and difficulty in measuring nonlinear correlations in existing atmospheric pollutant concentration mapping technologies have been solved, the accuracy and flexibility of mapping results have been improved, and all-day, hourly resolution, and spatially seamless atmospheric pollutant concentration monitoring has been successfully achieved.

[0079] See also Figure 2 and Figure 3 , Figure 2 A schematic diagram of the method architecture for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities provided in this application, Figure 3 The method for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities provided in this application is aimed at PM 2.5 A schematic diagram of an embodiment of concentration map drawing process, the following is combined with Figure 2 and Figure 3 The method for mapping the concentration of atmospheric pollutants taking into account multi-dimensional geographical similarities provided in this application is described in detail. The method is also applicable to other atmospheric pollutant concentration mappings. The method specifically includes:

[0080] 301. Obtain long-term geographic environmental statistics;

[0081] In some specific embodiments, it is necessary to obtain the average monthly geographical environment statistical data for the past three years, and the data types specifically include: historical meteorological variables Met, historical PM 2.5 concentration, historical AOD data, surface elevation data DEM, land use type data LULC, night light data NLT, normalized vegetation index data NDVI. All statistical data are normalized and pre-processed, resampled to the range of the area to be monitored, and the corresponding geographic environment statistical data are assigned according to the latitude and longitude of the ground-based observation station. For example, the historical meteorological variable Met is obtained from the fifth generation reanalysis dataset of the European Center for Medium-Range Forecasts; the historical PM 2.5 The concentration data and historical AOD data are obtained from the LGHAP v2 dataset; the surface elevation data DEM is obtained from the SRTM project; the land use type data LULC is obtained from the CLCD v01 dataset; the night light data NLT is obtained from the NPP-VIIRS satellite data; and the normalized difference vegetation index data NDVI is obtained from the MOD13A3 product.

[0082] 302. Determine the geographic environment feature vectors of each pixel to be monitored and all ground-based observation sites in the area to be monitored, and calculate the statistical index for evaluating the feature difference based on the geographic environment feature vectors;

[0083] Each pixel to be monitored and the ground-based observation site can be defined by its geographical environment characteristics. The geographical environment characteristic vector contains various parameters that reflect the geographical environment differences. The geographical environment characteristics include but are not limited to historical meteorological variables, historical PM 2.5 concentration, historical AOD data, surface elevation data, land use type data, night light data, and normalized difference vegetation index data.

[0084] For each pixel to be monitored and the geographic environment feature vector of all ground-based observation sites, their differences can be measured by calculating different statistical indicators, so it is necessary to calculate a variety of statistical indicators to evaluate the feature differences between the two. The statistical indicators include but are not limited to root mean square error, correlation coefficient, average deviation, etc.

[0085] It should be noted that for each pixel to be monitored and the geographic environment characteristic vector of each ground-based observation site, it is recommended to calculate at least two statistical indicators, and complement each other through multiple statistical indicators to more comprehensively and comprehensively evaluate the similarity between the two.

[0086] 303. Inputting the sequence composed of statistical indicators into the similarity calculation method to obtain a similarity index value for measuring the similarity of geographical environment scenes;

[0087] Each statistical indicator calculated in step 302 represents an evaluation of the differences in the geographical environment from different angles. These indicators are normalized and the normalized statistical indicators are combined into a feature sequence. The feature sequence contains information in multiple dimensions and is used to describe the differences between the pixels to be monitored and the ground-based observation sites. The feature sequence is input into the similarity calculation method to integrate the various statistical indicators to obtain a single similarity index value between each pixel to be monitored and the ground-based observation site. This index value reflects the degree of similarity between the two geographical environment scenes and can effectively measure the geographical environment similarity between the pixels to be monitored and the ground-based observation sites. Specifically, the similarity calculation method can be expressed as:

[0088] P ij =f p ({R1(Geo i ,Geo j ),R2(Geo i ,Geo j ),...,R n (Geo i ,Geo j )})

[0089] Where i and j represent the pixel to be monitored and a specific ground-based observation station number respectively; P ij represents the similarity index value of the geographical environment scene between the pixel i to be monitored and the ground-based observation site j; f p (·) represents a similarity algorithm that measures the similarity of vector features and can achieve the purpose of measuring the similarity of geographical environment scenes, including but not limited to arithmetic mean, Euclidean norm, cosine similarity, etc.; R k (·), k∈[1,n] represents various statistical index calculation methods; Geo i With Geo j They represent the geographical environment characteristic variables corresponding to the pixel i to be monitored and the ground-based observation site j respectively.

[0090] In some specific embodiments, for each pixel to be monitored, the normalized statistical indicators of the geographic environment feature vector and the geographic environment feature matrix of all ground-based observation sites can be calculated; the sequence composed of the normalized statistical indicators is input into the mathematical statistical model for evaluating geographic similarity; the output value of the mathematical statistical model is determined as the similarity index value for measuring the similarity of the geographic environment scenes. Among them, taking the normalized statistical indicators as the correlation coefficient and the normalized root mean square error, and the mathematical statistical model as the Euclidean norm as an example, the mathematical statistical model for evaluating geographic similarity can be expressed as:

[0091] P ij =||R(F ij ),(1-NRMSE(Fij ))||2

[0092] F ij ={AOD,PM 2.5 ,Met,DEM,LULC,NLT,NDVI}

[0093] Among them, i and j represent the pixel to be monitored and a specific ground-based observation station number, respectively. ij represents the similarity index value, ||·||2 represents the Euclidean norm, R is the relevant preset coefficient, RMSE is the root mean square error, F ij Represents the geographic environment feature vector.

[0094] 304. Generate a similarity matrix of each pixel to be monitored in the area to be monitored according to the similarity index value;

[0095] According to the similarity index values ​​calculated above, a similarity matrix can be generated for each pixel to be monitored in the monitored area. This similarity matrix reflects the geographical similarity between each pixel to be monitored and all ground-based observation sites, providing support for the subsequent selection of neighborhood monitoring sites, pollutant concentration interpolation and mapping models. Assuming that the monitored area contains M pixels to be monitored and the total number of ground-based observation sites is N, the similarity matrix will be an M×N matrix, and each element P in the matrix ij Represents the similarity index value between the pixel i to be monitored and the jth ground-based observation station.

[0096] 305. According to the similarity matrix of each pixel to be monitored, a neighborhood monitoring site of each pixel to be monitored is defined, and the neighborhood monitoring site is a plurality of sites arranged in descending order according to similarity index values;

[0097] Based on the generated similarity matrix, the neighborhood monitoring site of each pixel to be monitored can be determined. By sorting the ground-based observation sites from high to low according to the similarity index value, the most similar sites can be selected as neighborhood monitoring sites to ensure that the pollutant concentration estimate of each pixel to be monitored depends on the most representative observation site of a similar geographical environment, thereby improving the accuracy and interpretability of the mapping results. The number of selected neighborhood monitoring sites can be adjusted according to actual needs, and is not specifically limited here. For example, the 32 sites with the largest similarity index values ​​can be defined as the neighborhood monitoring sites of the pixel to be monitored.

[0098] 306. Integrate the neighborhood monitoring sites to obtain a non-Euclidean neighborhood monitoring site set for each pixel to be monitored;

[0099] These neighborhood monitoring sites are integrated to obtain a set of non-Euclidean neighborhood monitoring sites for each pixel to be monitored. The relationship between these neighborhood monitoring sites no longer relies solely on the traditional Euclidean spatial distance. Through this integration, the similarity between the pixel to be monitored and the neighborhood monitoring site can be measured under multi-dimensional geographic environmental characteristics, and interpolation weights can be calculated based on this similarity. This method breaks through the traditional Euclidean neighborhood method and introduces more diverse geographic scene features into the interpolation process, thereby improving the accuracy and interpretability of atmospheric pollutant concentration mapping that takes into account multi-dimensional geographic similarities.

[0100] The non-Euclidean neighborhood monitoring site set is specifically expressed as:

[0101]

[0102] Among them, S i represents the set of non-Euclidean neighborhood monitoring sites corresponding to the pixel i to be monitored; topK(·) represents the set of independent variable points with the largest K values, that is, the function of selecting the K values ​​with the largest geographical environment similarity; K is the number of neighborhood monitoring sites corresponding to each pixel to be monitored, represents the similarity matrix corresponding to the pixel i to be monitored, and N is the total number of ground-based observation sites.

[0103] 307. Obtain historical observation data of the area to be monitored, including short-term meteorological reanalysis data and atmospheric pollutant concentration observation data;

[0104] In this embodiment, step 307 is similar to step 103 in the above embodiment and will not be described again here.

[0105] 308. Dynamically calculate the geographical environment similarity between the pixel to be monitored and the neighboring monitoring site based on the non-Euclidean neighborhood monitoring site set and historical observation data;

[0106] According to the non-Euclidean neighborhood monitoring site set, the geographical environment similarity between the pixel to be monitored and its neighborhood monitoring site is dynamically calculated, and the short-term meteorological reanalysis data and the atmospheric pollutant concentration observation data time series are combined to realize the real-time update of the similarity distance. For each pixel to be monitored, the geographical environment similarity between the pixel to be monitored and the neighborhood monitoring site is first calculated through short-term meteorological reanalysis data and atmospheric pollutant concentration observation data. In this step, it is necessary to replace the long-term meteorological data time series when calculating the geographical environment similarity in step 303 with the short-term meteorological reanalysis data in the historical observation data, so as to realize the dynamic update of the similarity calculation, so that the calculation can reflect the impact of the current moment and recent historical observation data on the monitored area.

[0107] Taking the historical time step of 5 hours as an example, the geographical environment similarity distance between the pixel to be monitored and the neighboring monitoring station is calculated. The dynamic similarity distance calculation method adopts:

[0108] W=||R(F′ ij ),(1-NRMSE(F′ ij ))||2

[0109]

[0110] where F′ ij A sequence of dynamic geographic features representing the pixels to be monitored and the neighboring monitoring sites; Represents the historical 5-hour meteorological reanalysis variable time series; the meanings of other symbols are consistent with those in step 303.

[0111] 309. Taking geographical environment similarity as edge weight and geographical environment characteristics of the pixels to be monitored and the neighboring monitoring sites as nodes, a graph topology network structure of the area to be monitored is constructed;

[0112] The geographical environment similarity calculated above is used as the edge weight in the graph, and the geographical environment characteristics of the pixels to be monitored and the neighboring monitoring sites are used as nodes to construct a graph topology network structure of the area to be monitored. This graph structure can effectively express the geographical environment relationship between each pixel in the area to be monitored and its neighboring monitoring sites, and provide a basis for subsequent interpolation and atmospheric pollutant concentration mapping models.

[0113] The network topology structure of this diagram can be constructed in the following ways:

[0114] G = {υ|E,W};

[0115] Among them, G represents the network topology of the neighborhood monitoring site graph; υ represents the geographic environment feature matrix of pixels and neighborhood monitoring sites. The geographic environment features include short-term meteorological reanalysis data and other long-term geographic environment statistical data; E represents whether there is a mutual influence relationship between pixels and neighborhood monitoring sites. Here, it is assumed that there is mutual influence between the observed pixels and the neighborhood monitoring sites, that is, a fully connected graph network topology is constructed; W represents the association strength between pixels and neighborhood monitoring sites, that is, the geographic environment similarity distance.

[0116] Furthermore, the calculation method of the association strength between the pixel and the neighboring monitoring station is as follows:

[0117] W i ={f p ({R1(Geo′ i ,Geo′ j ),R2(Geo′ i ,Geo′ j ),...,Rn (Geo′ i ,Geo′ j )})|j

[0118] ∈[1,k]}

[0119] Where j represents a specific station number in the set of neighborhood monitoring stations corresponding to pixel i; k represents the total number of neighborhood monitoring stations; f p (·) is the mathematical statistical model for measuring the similarity of geographical environment in step 303; R k (·), k∈[1,n] is a method for calculating various statistical indicators used in steps 301-303; Geo′ is the long-term geographic environment characteristic data used in steps 301-303, in which the long-term meteorological variable data used is replaced by short-term meteorological reanalysis data.

[0120] 310. According to the graph network topology and the observed atmospheric pollutant concentration data of the neighboring monitoring stations, a characteristic dataset of the neighboring monitoring stations at the pixel scale is formed, and according to the atmospheric pollutant concentration observation data of the stations to be monitored, an atmospheric pollutant concentration observation target dataset at the station scale is formed;

[0121] Based on the graph network topology constructed above and the observed atmospheric pollutant concentration data of the neighborhood monitoring stations, two data sets are formed, namely:

[0122] Pixel-scale neighborhood monitoring site feature data set: This set of data will contain the relevant information of the pixels to be monitored and their neighborhood monitoring sites, specifically including the atmospheric pollutant concentration observation data (such as PM 2.5 , NO2, SO2 and other concentration values), as well as other meteorological characteristic data of neighboring monitoring stations (such as wind speed, temperature, humidity, etc.). The observation data of these neighboring monitoring stations will be weighted according to the geographical similarity between them and the pixels to be monitored. The more similar the stations are, the greater the weight of the given observation data in the calculation.

[0123] Site-scale atmospheric pollutant concentration observation target dataset: This set of data contains the actual observation data of each ground-based observation site, that is, the atmospheric pollutant concentration of the site at a certain moment (for example, PM 2.5 concentration, NO2 concentration, etc.) as the target value when training the model and used to predict the pollutant concentration of the pixel to be monitored.

[0124] 311. The target mapping model is trained using the graph topology network structure and the training data set. The target mapping model is used to generate a spatial seamless atmospheric pollutant concentration map with all-day hourly resolution based on the real-time observation data of the monitored area.

[0125] In this embodiment, step 311 is similar to step 105 in the aforementioned embodiment and will not be described again here.

[0126] See also Figure 4 , Figure 4 The scatter plot of the accuracy of atmospheric pollutant concentration mapping in the method for mapping atmospheric pollutant concentration taking into account multi-dimensional geographical similarity provided in this application, where the correlation coefficient R=0.91 indicates that the PM estimated by the mapping method 2.5 There is a high positive correlation between the concentration and the actual observation data, indicating that it can accurately reflect the actual pollutant concentration distribution to a certain extent. The root mean square error RMSE is 17.75μg / m 3 , indicating PM 2.5 There is a certain error between the estimated value and the observed value, but the overall error is relatively small, indicating that there is a certain reliability in accuracy. Overall, it shows that the estimation effect of the atmospheric pollutant concentration mapping method provided in this application is relatively ideal.

[0127] See also Figure 5 , Figure 5 An embodiment of a system for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities provided by the present application comprises:

[0128] The calculation unit 501 is used to obtain long-term geographic environment statistical data, and use a similarity calculation method to measure the geographic scene similarity between the pixels to be monitored in the monitored area and the ground-based observation site to obtain a similarity matrix;

[0129] A division unit 502 is used to obtain a non-Euclidean neighborhood monitoring site set at a pixel scale according to the similarity matrix division;

[0130] An acquisition unit 503 is used to acquire historical observation data of the area to be monitored, wherein the historical observation data includes short-term meteorological reanalysis data and a time series of atmospheric pollutant concentration observation data;

[0131] A construction unit 504 is used to construct a graph topology network structure and a training data set of the area to be monitored according to the non-Euclidean neighborhood monitoring site set and the historical observation data, wherein the training data set includes a neighborhood monitoring site feature data set at a pixel scale and an atmospheric pollutant concentration observation target data set at a site scale;

[0132] The training unit 505 is used to train a target mapping model using the graph topology network structure and the training data set, and the target mapping model is used to generate a spatial seamless atmospheric pollutant concentration map with a full-day hourly resolution based on the real-time observation data of the area to be monitored.

[0133] In this embodiment, the functions of each unit are the same as those described above. Figure 1 or Figure 3 The steps in the method embodiment shown correspond to each other and will not be repeated here.

[0134] This application also provides a device for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities, see Figure 6 , Figure 6 An embodiment of a device for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities provided in the present application comprises:

[0135] Processor 601, memory 602, input and output unit 603, bus 604;

[0136] The processor 601 is connected to the memory 602, the input and output unit 603 and the bus 604;

[0137] The memory 602 stores a program, and the processor 601 calls the program to execute any of the above methods for mapping the concentration of atmospheric pollutants taking into account multi-dimensional geographical similarities.

[0138] The present application also relates to a computer-readable storage medium having a program stored thereon. When the program is executed on a computer, the computer is enabled to execute any of the above methods for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographic similarities.

[0139] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0140] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0141] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0142] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0143] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), disk or optical disk and other media that can store program code.

Claims

1. A method for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities, characterized in that: The method comprises: Obtain long-term geographic environmental statistical data, and use similarity calculation methods to measure the geographic scene similarity between the pixels to be monitored in the monitored area and the ground-based observation sites to obtain a similarity matrix; A non-Euclidean neighborhood monitoring site set at a pixel scale is obtained according to the similarity matrix division; Acquiring historical observation data of the area to be monitored, wherein the historical observation data includes short-term meteorological reanalysis data and time series of atmospheric pollutant concentration observation data; Constructing a graph topology network structure and a training data set of the area to be monitored according to the non-Euclidean neighborhood monitoring site set and the historical observation data, wherein the training data set includes a neighborhood monitoring site feature data set at a pixel scale and an atmospheric pollutant concentration observation target data set at a site scale; The target mapping model is obtained by training using the graph topology network structure and the training data set. The target mapping model is used to generate a spatial seamless atmospheric pollutant concentration map with a full-day hourly resolution based on the real-time observation data of the area to be monitored.

2. The method according to claim 1, characterized in that The similarity calculation method is used to measure the geographical scene similarity between the pixels to be monitored in the monitored area and the ground-based observation site to obtain a similarity matrix, including: Determine the geographic environment feature vectors of each pixel to be monitored and all ground-based observation sites in the area to be monitored, and calculate statistical indicators for evaluating feature differences based on the geographic environment feature vectors; Inputting the sequence composed of the statistical indicators into a similarity calculation method to obtain a similarity index value for measuring the similarity of the geographical environment scenes; A similarity matrix of each pixel to be monitored in the area to be monitored is generated according to the similarity index value.

3. The method according to claim 2, characterized in that The calculating of the statistical index for evaluating the feature difference according to the geographic environment feature vector includes: For each pixel to be monitored, calculate the normalized statistical index of the geographic environment characteristic vector and the geographic environment characteristic matrix of all ground-based observation sites; The step of inputting the sequence of statistical indicators into a similarity calculation method to obtain a similarity index value for measuring the similarity of geographical environment scenes includes: Inputting the sequence composed of the normalized statistical indicators into a mathematical statistical model for evaluating geographical similarity; The output value of the mathematical statistics model is determined as a similarity index value for measuring the similarity of geographical environment scenes.

4. The method according to claim 3, characterized in that When the atmospheric pollutant concentration graph is PM 2.5 When plotting the concentration map, the geographic environment feature vector includes historical meteorological variables Met, historical PM 2.5 At least one of concentration, historical AOD data, surface elevation data DEM, land use type data LULC, night light data NLT, and normalized vegetation index data NDVI; The mathematical and statistical model for evaluating geographical similarity adopts: P ij =||R(F ij ),(1-NRMSE(F ij ))||2 F ij ={AOD,PM 2.5 ,Met,DEM,LULC,NLT,NDVI} Where i and j represent the pixel to be monitored and a specific ground-based observation station number, respectively. ij represents the similarity index value, ||·||2 represents the Euclidean norm, R is the correlation coefficient, NRMSE is the normalized root mean square error, F ij Represents the geographic environment feature vector.

5. The method according to claim 2, characterized in that: The non-Euclidean neighborhood monitoring site set at a pixel scale obtained by dividing the similarity matrix includes: Defining a neighborhood monitoring site of each pixel to be monitored according to the similarity matrix of each pixel to be monitored, wherein the neighborhood monitoring site is a plurality of sites arranged in descending order according to the similarity index value; The neighborhood monitoring sites are integrated to obtain a non-Euclidean neighborhood monitoring site set for each pixel to be monitored.

6. The method according to claim 5, characterized in that The non-Euclidean neighborhood monitoring site set is expressed as: Among them, S i represents the set of non-Euclidean neighborhood monitoring sites corresponding to the pixel i to be monitored; topK(·) represents the set of independent variable points with the largest K values, that is, a function that selects the K values ​​with the largest geographical environment similarity; K is the number of the neighborhood monitoring sites corresponding to each pixel to be monitored, represents the similarity matrix corresponding to the pixel i to be monitored, and N is the total number of ground-based observation sites.

7. The method according to any one of claims 1 to 6, characterized in that The step of constructing the graph topology network structure and the training data set of the area to be monitored according to the non-Euclidean neighborhood monitoring site set and the historical observation data includes: Dynamically calculating the geographical environment similarity between the pixel to be monitored and the neighborhood monitoring site according to the non-Euclidean neighborhood monitoring site set and the historical observation data; Taking the geographical environment similarity as edge weight and the geographical environment characteristics of the pixels to be monitored and the neighboring monitoring sites as nodes, a graph topology network structure of the area to be monitored is constructed; A neighborhood monitoring site characteristic data set at a pixel scale is formed according to the graph network topology structure and the observed atmospheric pollutant concentration data of the neighborhood monitoring sites, and an atmospheric pollutant concentration observation target data set at a site scale is formed according to the atmospheric pollutant concentration observation data of the sites to be monitored.

8. A system for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities, characterized in that: The system comprises: A calculation unit is used to obtain long-term geographic environment statistical data and use a similarity calculation method to measure the geographic scene similarity between the pixels to be monitored in the monitored area and the ground-based observation site to obtain a similarity matrix; A partitioning unit, used for obtaining a non-Euclidean neighborhood monitoring site set at a pixel scale according to the similarity matrix partitioning; An acquisition unit, used for acquiring historical observation data of the area to be monitored, wherein the historical observation data includes short-term meteorological reanalysis data and a time series of atmospheric pollutant concentration observation data; A construction unit, used to construct a graph topology network structure and a training data set of the area to be monitored according to the non-Euclidean neighborhood monitoring site set and the historical observation data, wherein the training data set includes a neighborhood monitoring site feature data set at a pixel scale and an atmospheric pollutant concentration observation target data set at a site scale; A training unit is used to train a target mapping model using the graph topology network structure and the training data set, and the target mapping model is used to generate a spatial seamless atmospheric pollutant concentration map with a full-day hourly resolution based on the real-time observation data of the area to be monitored.

9. A device for mapping atmospheric pollutant concentrations taking into account multi-dimensional geographical similarities, characterized in that: The device comprises: Processor, memory, input-output unit, and bus; The processor is connected to the memory, the input and output unit, and the bus; The memory stores a program, and the processor calls the program to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a program stored thereon, wherein the program, when executed on a computer, performs the method according to any one of claims 1 to 7.

Images