Geographically weighted regression modeling method and device for spatial filter value of CO2 concentration

Abstract

The invention provides a CO2 concentration spatial filter value geographically weighted regression modeling method and device, and the method comprises the steps: constructing a spatial weight matrix based on the geographic position of a carbon dioxide concentration data monitoring point, extracting the spatial weight matrix to obtain a feature vector representing a global position relation, and taking the feature vector as a global spatial feature factor; on the basis of a preset model reflecting local heterogeneity, screening in the global spatial feature factors to obtain feature vectors representing spatial local heterogeneity, and taking the feature vectors as local spatial factors; and constructing a carbon dioxide concentration model based on the local space factors. According to the method, the feature vectors representing the local heterogeneity are screened from the global spatial feature factors representing the global spatial relationship and are used as the local spatial factors to construct the CO2 concentration model, so that the constructed model can give consideration to the global spatial features and the local spatial heterogeneity at the same time.

Description

Technical Field

[0001] The present invention relates to the technical field of spatial statistical analysis, and in particular to a method and device for modeling a CO2 concentration spatial filter value geographical weighted regression model. Background Art

[0002] Carbon dioxide (CO2), as a major greenhouse gas, has a profound impact on human daily production and life, and plays a key role in global climate change and greenhouse effect.

[0003] Carbon dioxide shows certain patterns in spatial distribution, which are affected by spatial effects, specifically spatial autocorrelation and spatial heterogeneity. The distribution of CO2 is continuous, which means that there is a certain correlation in space, that is, the CO2 concentration in one area may be similar to that in its neighboring areas. At the same time, the distribution of CO2 is also affected by many factors, such as topography, meteorological conditions, and carbon emission sources, which vary in different geographical locations, resulting in spatial heterogeneity of CO2 distribution.

[0004] Traditional CO2 concentration regression models usually only consider one aspect of global spatial autocorrelation or local spatial heterogeneity, but not both. Therefore, existing CO2 concentration regression models have certain limitations. Summary of the invention

[0005] The present invention provides a method and device for geographically weighted regression modeling of spatial filtering values ​​of CO2 concentration, so as to solve the defect that the CO2 concentration regression model in the prior art usually only considers the global or local spatial effects, and realizes a method and device for geographically weighted regression modeling of spatial filtering values ​​of CO2 concentration that takes into account both the global spatial autocorrelation and the local spatial heterogeneity.

[0006] The present invention provides a CO2 concentration spatial filtering value geographical weighted regression modeling method, comprising:

[0007] A spatial weight matrix is ​​constructed based on the geographical location of the carbon dioxide concentration data monitoring point, and a feature vector representing the global position relationship is extracted from the spatial weight matrix as a global spatial feature factor;

[0008] Based on a preset model reflecting local heterogeneity, a characteristic vector representing the local heterogeneity of space is screened out from the global spatial characteristic factor as a local spatial factor;

[0009] A carbon dioxide concentration model is constructed based on the local spatial factors.

[0010] According to a method for modeling CO2 concentration spatial filtering values ​​based on geographical weighted regression provided by the present invention, the preset model based on the response local heterogeneity, the step of screening the characteristic vector representing the spatial local heterogeneity in the global spatial characteristic factor as the local spatial factor, specifically includes:

[0011] The global spatial characteristic factor is used as an independent variable of the preset model, and the preset model is calculated in combination with the pre-acquired influencing factor data;

[0012] Calculate the corrected determination coefficient corresponding to each of the global spatial characteristic factors;

[0013] In the case where the correction determination coefficient calculated each time is greater than the current maximum correction determination coefficient, the global spatial characteristic factor corresponding to the current correction determination coefficient is retained, and the current correction determination coefficient is updated to the current maximum correction determination coefficient; otherwise, the global spatial characteristic factor corresponding to the current correction determination coefficient is removed, wherein the initial value of the current maximum correction determination coefficient is set to zero;

[0014] The global space characteristic factor retained after the cyclic calculation is used as the local space factor.

[0015] According to a CO2 concentration spatial filtering value geographical weighted regression modeling method provided by the present invention, the step of using the global spatial characteristic factor as the independent variable of the preset model and combining the pre-acquired influencing factor data to calculate the preset model specifically includes:

[0016] Randomly sorting all the global spatial feature factors and adding them into the spatial feature vector set;

[0017] The global spatial characteristic factors in the spatial characteristic vector set are sequentially added to the independent variable set of the preset model as the independent variables of the preset model and combined with the pre-acquired influencing factor data to perform calculations on the preset model.

[0018] According to a CO2 concentration spatial filtering value geographically weighted regression modeling method provided by the present invention, the preset model is a geographically weighted regression model, a spatial error model or a spatial lag model.

[0019] According to a CO2 concentration spatial filtering value geographically weighted regression modeling method provided by the present invention, after the step of constructing a carbon dioxide concentration model based on the local spatial factors, the method further includes:

[0020] Performing grid calculation on the carbon dioxide concentration model to obtain a carbon dioxide concentration distribution map;

[0021] The abnormal values ​​in the carbon dioxide concentration distribution map are processed.

[0022] According to a CO2 concentration spatial filtering value geographically weighted regression modeling method provided by the present invention, the step of constructing a spatial weight matrix based on the geographical location of the carbon dioxide concentration data monitoring point includes:

[0023] Based on adjacency, inverse distance weighting or a combination of adjacency and inverse distance weighting, a spatial weight matrix is ​​constructed in combination with the geographical locations of the carbon dioxide concentration data monitoring points.

[0024] According to a method for modeling CO2 concentration spatial filtering values ​​based on geographical weighted regression provided by the present invention, the preset model based on the response local heterogeneity, before the step of screening the characteristic vector representing the spatial local heterogeneity in the global spatial characteristic factor as the local spatial factor, further comprises:

[0025] Obtain carbon dioxide concentration data for a preset area;

[0026] Taking surface factors, atmospheric factors, human activity factors and radiation factors as influencing factors, the influencing factor data of the preset area are obtained;

[0027] The carbon dioxide concentration data and the influencing factor data are preprocessed.

[0028] According to a method for modeling CO2 concentration spatial filtering values ​​based on geographical weighted regression provided by the present invention, the preset model based on the response local heterogeneity, before the step of screening the characteristic vector representing the spatial local heterogeneity in the global spatial characteristic factor as the local spatial factor, further comprises:

[0029] Calculate the variance inflation factor between each influencing factor data and the carbon dioxide concentration data. When the currently calculated variance inflation factor is greater than a preset threshold, remove the influencing factor corresponding to the current maximum variance inflation factor.

[0030] Repeat the above steps until the variance inflation factor corresponding to each influencing factor after screening is less than the preset threshold.

[0031] The present invention also provides a CO2 concentration spatial filtering value geographical weighted regression modeling method, comprising:

[0032] A global module is used to construct a spatial weight matrix based on the geographical location of the carbon dioxide concentration data monitoring point, and extract a feature vector representing the global position relationship from the spatial weight matrix as a global spatial feature factor;

[0033] A local module, for obtaining a characteristic vector representing the local heterogeneity of the space as a local spatial factor by screening the characteristic vector from the global spatial characteristic factor based on a preset model reflecting the local heterogeneity;

[0034] A building module is used to build a carbon dioxide concentration model based on the local spatial factors.

[0035] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for geographically weighted regression modeling of spatial filtered CO2 concentration values ​​as described above is implemented.

[0036] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for geographically weighted regression modeling of spatial filtered CO2 concentration.

[0037] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-mentioned methods for geographically weighted regression modeling of spatial filtering values ​​of CO2 concentration.

[0038] The method and device for modeling the spatial filtering value of CO2 concentration by geographical weighted regression provided by the present invention extract the global spatial characteristic factors by constructing a global spatial matrix, and extract the local spatial factors that take into account both global and local characteristics from the global spatial characteristic factors, and then perform geographical weighted regression modeling together with the factor data that affects the CO2 concentration to obtain a CO2 concentration estimation model, so that the constructed model can take into account both global spatial characteristics and local spatial heterogeneity. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 This is one of the flow diagrams of the CO2 concentration spatial filtering value geographically weighted regression modeling method provided by the present invention;

[0041] Figure 2 It is a flow chart mainly used to show the global spatial characteristic factor screening process in the CO2 concentration spatial filtering value geographical weighted regression modeling method provided by the present invention;

[0042] Figure 3 It is a flow chart mainly used to show the influencing factor screening process in the CO2 concentration spatial filtering value geographical weighted regression modeling method provided by the present invention;

[0043] Figure 4 This is the second flow chart of the CO2 concentration spatial filtering value geographical weighted regression modeling method provided by the present invention;

[0044] Figure 5 It is a structural schematic diagram of the CO2 spatial filtering value geographical weighted regression modeling device provided by the present invention;

[0045] Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] Combine the following Figures 1 to 4 The present invention introduces the CO2 concentration spatial filtering value geographical weighted regression modeling method, such as Figure 1 As shown, including:

[0048] Step 101, constructing a spatial weight matrix based on the geographical locations of the carbon dioxide concentration data monitoring points, and extracting a feature vector representing the global position relationship from the spatial weight matrix as a global spatial feature factor;

[0049] Traditional CO2 concentration regression models usually only consider one aspect of global spatial autocorrelation or local spatial heterogeneity, but not both. In other words, when processing spatial data, these models either consider the global correlation between data points or the specific heterogeneity of local areas, but do not combine the two, thus failing to fully capture the complex spatial distribution characteristics of CO2.

[0050] In order to take into account the impact of global spatial autocorrelation and local spatial heterogeneity on the CO2 concentration regression model, the present invention proposes a modeling method of the Eigenvector spatial filtering-based geographically weighted regression model (ESF-GWR), which believes that the impact of space on CO2 concentration can be divided into two parts: one is the spatial impact caused by the local environmental conditions themselves, and the other is the global spatial impact and the local manifestation of the global impact.

[0051] To this end, it is first necessary to extract the global spatial characteristic factors that characterize the global position relationship of CO2.

[0052] Specifically, a spatial weight matrix is ​​constructed based on the geographical location of the CO2 concentration data monitoring points, and then the global spatial characteristic factors are screened from the eigenvectors of the spatial weight matrix.

[0053] Among them, the geographical location of the CO2 concentration data monitoring point is determined according to the method of obtaining the CO2 concentration data.

[0054] Optionally, when the CO2 concentration data comes from a CO2 concentration monitoring station established on the ground, the coordinates of each station in the study area are used as the geographical location of the CO2 concentration data monitoring point.

[0055] Furthermore, due to the limited number of ground CO2 concentration monitoring stations and the uneven geographical distribution of the stations, with the development of remote sensing technology, data collected by carbon monitoring satellites and other remote sensing methods can also be used to estimate the ground CO2 concentration. Among them, carbon monitoring satellites are divided into two categories: active satellites and passive satellites.

[0056] Optionally, when the CO2 concentration data is derived from a passive satellite, the geographical location corresponding to the data received by the passive satellite in the study area is used as the geographical location of the CO2 concentration data monitoring point.

[0057] Optionally, when the CO2 concentration data originates from an active satellite, the scanning point coordinates of the active satellite are used as the geographical location of the CO2 concentration data monitoring point.

[0058] Compared with passive satellites, active satellites have obvious advantages in data accuracy and spatial coverage, and active satellite data are scanned point data with clear time and space coordinates. Therefore, CO2 concentration data is preferably data acquired by active satellites. In this embodiment, the CO2 concentration data acquired by the active satellite DQ-1 will be taken as an example for introduction.

[0059] It is understandable that in other feasible implementations, the above-mentioned methods can also be combined to obtain CO2 concentration data. For example, CO2 concentration data can be obtained by using active satellites and ground-based CO2 concentration monitoring stations at the same time. In this case, the scanning point positions of the active satellites and the coordinates of the CO2 concentration monitoring stations in the study area are jointly used as the geographical locations of the CO2 concentration data monitoring points.

[0060] On this basis, in this embodiment, the spatial weight matrix W1 is constructed according to the projection coordinates of the scanning point data of the active satellite, and is centralized into a symmetric matrix W:

[0061]

[0062] Where I represents the n-dimensional unit matrix, 11 Trepresents an n×n matrix, all elements in the matrix are equal to 1, and n represents the number of CO2 concentration data monitoring points in the study area, which in this embodiment is the number of active satellite scanning points in the study area.

[0063] Furthermore, according to the eigenvalues ​​and eigenvectors of the symmetric matrix W, several global spatial characteristic factors that enable the best OLS (Ordinary Least Squares) model fitting effect are screened from the eigenvectors of the symmetric matrix W.

[0064] In a specific implementation, the eigenvalues ​​and eigenvectors of the symmetric matrix W are first calculated and preliminarily screened. The eigenvalues ​​are solved to obtain the eigenvalue λ i and the eigenvector E i There are two conditions for the initial screening: one is its eigenvalue λ i >0; second, according to the empirical model, the following conditions are required:

[0065]

[0066] In the formula, λ i represents the feature value to be screened, λ max Indicates the maximum eigenvalue, and 0.25 is the empirical value set. The eigenvectors that meet the conditions are initially screened out as spatial eigenvectors.

[0067] On this basis, the spatial feature vectors are further screened, and the spatial feature vectors that make the evaluation result optimal are screened out as the global spatial feature factors through the stepwise regression method.

[0068] Specifically, Figure 2 As shown in FIG. 1 , the spatial eigenvectors are arranged from small to large according to the eigenvalue size to obtain the spatial eigenvector set E. The eigenvectors in the spatial eigenvector set E are then added to the independent variable set of the OLS model in sequence, and the least squares regression is performed on the residuals of the OLS model to calculate the AIC value (Akaike Information Criterion) of the regression equation.

[0069] Then perform the screening step: select the regression equation Y with the smallest AIC value K , and its corresponding spatial eigenvector is e k , and then remove e from the spatial eigenvector set E k The eigenvectors outside are added to the regression equation Y K , and calculate the AIC value of the regression equation.

[0070] Determine whether the AIC value has decreased significantly. If so, continue to perform the screening step; if the AIC value has hardly decreased, stop adding the eigenvectors to the regression equation and record the eigenvectors that have been added to the regression equation. These eigenvectors are the eigenvector sets that are screened out to optimize the evaluation criteria, that is, the global spatial characteristic factors that are screened out.

[0071] Step 102, based on a preset model reflecting local heterogeneity, a feature vector representing spatial local heterogeneity is screened out from the global spatial feature factor as a local spatial factor;

[0072] Furthermore, in order to comprehensively consider the impact of local spatial heterogeneity on CO2 concentration, the characteristic vector representing the local spatial heterogeneity is screened out from the global spatial characteristic factors as the local spatial factor, which serves as the basis for the final construction of CO2 concentration.

[0073] On this basis, in order to find suitable local spatial factors among the global spatial characteristic factors, the global spatial characteristic factors are screened based on the preset model.

[0074] Optionally, the preset model may be a geographically weighted regression model, a spatial error model, a spatial lag model or other models that can reflect local heterogeneity.

[0075] In this implementation, the geographically weighted regression model is used as an example to explain spatial heterogeneity with spatially varying coefficients, and further screen appropriate local spatial factors.

[0076] Step 103: construct a carbon dioxide concentration model based on the local spatial factors.

[0077] After screening the local spatial factors, the CO2 concentration estimation model (ESF-GWR) that takes into account global spatial autocorrelation and local heterogeneity can be constructed as follows:

[0078]

[0079] In the formula, y i represents the CO2 concentration corresponding to the i-th sampling point, (u i ,v i ) is the coordinate of the i-th sampling point, β0 is a constant, β k (u i ,v i ) is the kth regression parameter at the i-th sampling point, x ik is the value of the independent variable at the i-th sampling point, Eα ik is the kth spatial feature weight value at the i-th sampling point, and its value is calculated based on the local spatial factor, ε iis the random error of the i-th sampling point, satisfying the basic assumptions such as zero mean, homoscedasticity, and mutual independence. p represents the number of regression parameters, and q represents the sum of the number of regression parameters and local spatial factors.

[0080] The present invention extracts global spatial characteristic factors by constructing a global spatial matrix, and extracts local spatial factors that take into account both global and local characteristics from the global spatial characteristic factors, and then performs geographically weighted regression modeling together with the factor data that affects CO2 concentration to obtain a CO2 concentration estimation model, so that the constructed model can take into account both global spatial characteristics and local spatial heterogeneity.

[0081] In the method for modeling CO2 concentration spatial filtering values ​​by geographical weighted regression of the present invention, the preset model based on the reaction local heterogeneity, the step of screening the characteristic vector representing the local heterogeneity of the space from the global spatial characteristic factor as the local spatial factor, specifically includes:

[0082] The global spatial characteristic factor is used as an independent variable of the preset model, and the preset model is calculated in combination with the pre-acquired influencing factor data;

[0083] Calculate the corrected determination coefficient corresponding to each of the global spatial characteristic factors;

[0084] In this embodiment, the geographically weighted regression model is used as a preset model for explanation. The global spatial characteristic factors obtained by screening are put into a data set in the order of screening, and a characteristic vector set E1 is constructed. The characteristic vectors in E1 are sequentially added to the independent variables of the geographically weighted regression model, and the geographically weighted regression calculation is performed together with the pre-acquired influencing factor data.

[0085] For each global spatial characteristic factor, calculate the corrected coefficient of determination (Adj.R 2 ), as the correction determination coefficient corresponding to the global spatial characteristic factor.

[0086] Among them, the influencing factors represent the relevant factors that affect the CO2 concentration, which may include surface factors such as temperature, geothermal flux, and surface cover; atmospheric factors such as air pressure and wind speed; human activity factors and radiation factors; and can be selected and determined based on the geographical location and surrounding environment of the study area.

[0087] The data of the selected influencing factors are collected in advance and preprocessed together with the acquired CO2 concentration data to obtain the influencing factor data for the calculation of the geographical weighted regression model.

[0088] It is understandable that the selection of influencing factors and the model calculation method can be set accordingly depending on the preset model used.

[0089] In the case where the correction determination coefficient calculated each time is greater than the current maximum correction determination coefficient, the global spatial characteristic factor corresponding to the current correction determination coefficient is retained, and the current correction determination coefficient is updated to the current maximum correction determination coefficient; otherwise, the global spatial characteristic factor corresponding to the current correction determination coefficient is removed, wherein the initial value of the current maximum correction determination coefficient is set to zero;

[0090] The global space characteristic factor retained after the cyclic calculation is used as the local space factor.

[0091] Furthermore, the optimal global spatial feature factor is found through cyclic comparison as the local spatial factor.

[0092] Specifically, we first initialize the evaluation index, that is, The value is set to 0.

[0093] When the global spatial characteristic factor is used as an independent variable for geographically weighted regression calculation, its corresponding Adj.R 2 The maximum corrected coefficient of determination The value of is compared.

[0094] If the current global spatial characteristic factor Adj.R 2 Greater than , the current global space characteristic factor is saved and updated The value of .

[0095] Otherwise, remove the current global spatial feature factor. The above steps are executed repeatedly until all feature vectors in E1 are traversed, and finally, the saved feature vector is used as the local spatial factor that can reflect the local heterogeneity of the space.

[0096] In the CO2 concentration spatial filtering value geographical weighted regression modeling method of the present invention, the step of using the global spatial characteristic factor as the independent variable of the preset model and combining the pre-acquired influencing factor data to calculate the preset model specifically includes:

[0097] Randomly sorting all the global spatial feature factors and adding them into the spatial feature vector set;

[0098] The global spatial characteristic factors in the spatial characteristic vector set are sequentially added to the independent variable set of the preset model as the independent variables of the preset model and combined with the pre-acquired influencing factor data to perform calculations on the preset model.

[0099] Due to the cyclic comparison process, The value of the current global space characteristic factor Adj.R 2The value of is updated in real time, so the order of the global spatial feature factors will have a certain impact on the screening results.

[0100] Therefore, before the cyclic comparison step, the global spatial feature factors are first randomly sorted, and then added to a data set in sequence according to the randomly sorted order to construct a spatial feature vector set E2. When executing the cyclic comparison step, one feature vector is selected in E2 for calculation.

[0101] In the method for geographically weighted regression modeling of spatial filtering values ​​of CO2 concentration of the present invention, after the step of constructing a carbon dioxide concentration model based on the local spatial factors, the method further includes:

[0102] Performing grid calculation on the carbon dioxide concentration model to obtain a carbon dioxide concentration distribution map;

[0103] When the CO2 concentration estimation model was obtained, the correlation coefficients of various factors at various locations in the entire study area were also obtained.

[0104] Each location point in the study area is converted into a raster image, and the raster values ​​are set as the coefficient values ​​of each factor. The extracted spatial feature vector is interpolated into a raster image with the same resolution as the variable covering the entire study area, and then raster calculation is performed, the factor correlation coefficient raster image is multiplied by the factor raster image and then added, and finally the intercept term is added to obtain the CO2 concentration estimate, and finally the carbon dioxide concentration distribution map is obtained.

[0105] The abnormal values ​​in the carbon dioxide concentration distribution map are processed.

[0106] For the obtained carbon dioxide concentration distribution map, due to the existence of errors, the CO2 concentration values ​​in a very small number of areas may be negative, and there may also be a small number of abnormally high CO2 concentration values, which need to be processed. Negative concentration values ​​indicate that the CO2 concentration at this point is low, so the CO2 concentration value is set to 0, and a threshold is set for the CO2 concentration value. All CO2 concentration values ​​exceeding the threshold are modified to the threshold for visualization, and finally a continuous ground CO2 concentration distribution map is obtained.

[0107] In the CO2 concentration spatial filtering value geographically weighted regression modeling method of the present invention, the step of constructing a spatial weight matrix based on the geographical location of the carbon dioxide concentration data monitoring point includes:

[0108] Based on adjacency, inverse distance weighting or a combination of adjacency and inverse distance weighting, a spatial weight matrix is ​​constructed in combination with the geographical locations of the carbon dioxide concentration data monitoring points.

[0109] Optionally, a spatial weight matrix may be constructed based on adjacency, inverse distance weighting, or a combination of adjacency and inverse distance weighting.

[0110] In a feasible implementation, a spatial weight matrix W1 is constructed based on adjacency relationships. First, a global spatial weight matrix is ​​constructed based on Thiessen polygons of multiple CO2 concentration data monitoring points in the study area. Further, a global spatial weight matrix can be constructed based on Bishop adjacency, Rook adjacency or Queen adjacency.

[0111] In another feasible implementation, the spatial weight matrix W1 is constructed based on inverse distance weighting. First, the adjacency relationship is constructed according to the distance of multiple CO2 concentration data monitoring points in the study area, and the distance is converted into a weight value through the kernel function to obtain the global spatial weight matrix. The kernel function can be divided into exponential, Gaussian, and spherical models. The specific formulas are as follows:

[0112]

[0113] Where i and j represent position i and position j respectively, W i,j represents the adjacency (weight) between location points i and j, and r represents the maximum distance of the edges in the minimum spanning tree of all monitoring points.

[0114] In another feasible implementation, a composite method such as a K-nearest neighbor matrix and a spatial weight matrix W1 based on adjacency relationship and distance is used. First, a threshold K is determined, and the K nearest neighbor monitoring points of each monitoring point are searched to construct an adjacency matrix, W i,j Represents the adjacency (weight) between location points i and j; constructs a weight matrix based on adjacency relationship and distance. First, a threshold d is determined, and other monitoring points whose distance from each monitoring point does not exceed the threshold d are searched. The two are considered to be adjacent. i,j =1; otherwise, it is not adjacent, W i,j =0, and the spatial weight matrix W1 is constructed in this way.

[0115] In the method for modeling CO2 concentration spatial filtering values ​​by geographical weighted regression of the present invention, the preset model based on the reaction local heterogeneity, before the step of screening the characteristic vector representing the spatial local heterogeneity in the global spatial characteristic factor as the local spatial factor, further comprises:

[0116] Obtain carbon dioxide concentration data for a preset area;

[0117] The preset area is the research area where the CO2 concentration estimation model needs to be constructed.

[0118] Optionally, the CO2 concentration data of the preset area during the study period can be obtained at least through a CO2 concentration monitoring station, a passive satellite, an active satellite or a combination thereof.

[0119] In this embodiment, the CO2 concentration data is remote sensing data from the atmospheric environment monitoring satellite DQ-1.

[0120] Taking surface factors, atmospheric factors, human activity factors and radiation factors as influencing factors, the influencing factor data of the preset area are obtained;

[0121] Optionally, the surface factors may include at least air temperature, normalized difference vegetation index, geothermal flux, surface cover, etc.

[0122] Optionally, the atmospheric factors may include at least air pressure and wind speed.

[0123] Optionally, the human activity factor may include at least population density, CO2 emissions from fossil fuel combustion, etc.

[0124] Optionally, the radiation factor may include at least long-wave radiation, short-wave radiation, solar surface radiation, etc.

[0125] The above factors all reflect the spatial distribution of ground CO2 concentration to a certain extent. The factor data can be selected as influencing factors based on the climate characteristics and data quality of the study area.

[0126] Collect and obtain the data of the selected influencing factors in the study area and during the study period to obtain the influencing factor data.

[0127] The carbon dioxide concentration data and the influencing factor data are preprocessed.

[0128] The preprocessing of CO2 concentration data and influencing factor data includes quality inspection of active satellite scanning data and elimination of obvious outliers.

[0129] Preprocessing includes removing null values ​​and outliers, projection transformation, and unification of temporal and spatial resolutions.

[0130] Optionally, the original CO2 data is a concentration with a time field, and the daily, monthly, seasonal and annual average concentration values ​​can be calculated in turn. The remaining influencing factors also unify the time resolution to the same time resolution as the CO2 data.

[0131] Optionally, in terms of spatial resolution, other influencing factors need to be converted into a consistent spatial resolution, and the spatial resolution can be adjusted by methods such as resampling or spatial interpolation.

[0132] Both CO2 emissions from fossil fuel combustion and DQ-1 satellite CO2 data are point data, which can be converted into raster data covering the entire study area through the fishing net statistical method.

[0133] Data with different spatial reference systems need to be projected and transformed to unify the spatial reference system.

[0134] After the preprocessing is completed, the data of various factors at the CO2 satellite scanning points are extracted to screen the local spatial factors and build a CO2 concentration estimation model.

[0135] In the method for modeling CO2 concentration spatial filtering values ​​by geographical weighted regression of the present invention, the preset model based on the reaction local heterogeneity, before the step of screening the characteristic vector representing the spatial local heterogeneity in the global spatial characteristic factor as the local spatial factor, further comprises:

[0136] Calculate the variance inflation factor between each influencing factor data and the carbon dioxide concentration data. When the currently calculated variance inflation factor is greater than a preset threshold, remove the influencing factor corresponding to the current maximum variance inflation factor.

[0137] Repeat the above steps until the variance inflation factor corresponding to each influencing factor after screening is less than the preset threshold.

[0138] Furthermore, in order to simplify the steps of screening local spatial factors and simplify the constructed CO2 concentration estimation model, the extracted influencing factors were first tested for collinearity to screen out some influencing factors.

[0139] Specifically, Figure 3 As shown, first, each influencing factor data is constructed to obtain an alternative set of influencing factors, from which data and CO2 concentration data are selected in turn for collinearity test, and the variance inflation factor (VIF) corresponding to each influencing factor is calculated. If there is an influencing factor with a VIF greater than a preset threshold, the influencing factor corresponding to the maximum VIF needs to be removed, and the above steps are repeated until the VIF corresponding to each influencing factor is less than the preset threshold.

[0140] In this implementation, the preset threshold is determined to be 10 based on experience.

[0141] On this basis, the construction process of a complete carbon dioxide concentration estimation model ESF-SWR is as follows: Figure 4 shown.

[0142] In addition, for the CO2 concentration estimation model that has been built, the R 2 , adjusted R 2 , root mean square error (RMSE) and Akaike information criterion (AIC) were used as evaluation indicators to verify the accuracy of the proposed CO2 concentration estimation model taking into account global spatial autocorrelation and local spatial heterogeneity.

[0143]

[0144] In the formula, y iis the CO2 concentration observation value at CO2 concentration monitoring point i, y is the average value of the observation data, is the CO2 concentration at i predicted by the model, and n is the number of monitoring points.

[0145]

[0146] In the formula, p is the number of independent variables, R 2 and Adj.R 2 The value range is [0,1]. The larger the value, the higher the model accuracy.

[0147]

[0148] In the formula, the smaller the RMSE, the higher the model accuracy.

[0149]

[0150] In the formula, the smaller the AIC is, the higher the model accuracy is.

[0151] In addition, the model can be cross-validated. Optionally, a 10-fold cross-validation method can be used to evaluate the prediction accuracy of the model for CO2 concentration in non-scanning areas.

[0152] Specifically, the obtained CO2 concentration data and influencing factor data are randomly divided into 10 parts, 9 parts are selected as training sets each time to model according to the above method, and the remaining 1 part is used as a test set. The CO2 concentration data and influencing factor data of the test set are brought into the model obtained from the training set as independent variables, and the root mean square error of the test set is calculated; after each data has been tested once, the mean of the 10 root mean square errors RMSE is calculated, which is the result of cross-validation. The smaller the RMSE, the higher the prediction accuracy of the model, the stronger its robustness, and the stronger its applicability.

[0153] The CO2 concentration spatial filtering value geographical weighted regression modeling device provided by the present invention is described below. The CO2 concentration spatial filtering value geographical weighted regression modeling device described below and the CO2 concentration spatial filtering value geographical weighted regression modeling method described above can be referenced to each other.

[0154] like Figure 5 As shown, the CO2 concentration spatial filtering value geographical weighted regression modeling device includes a global module 501, a local module 502 and a construction module 503:

[0155] The global module 501 is used to construct a spatial weight matrix based on the geographical location of the carbon dioxide concentration data monitoring point, and extract a feature vector representing the global position relationship from the spatial weight matrix as a global spatial feature factor;

[0156] Traditional CO2 concentration regression models usually only consider one aspect of global spatial autocorrelation or local spatial heterogeneity, but not both. In other words, when processing spatial data, these models either consider the global correlation between data points or the specific heterogeneity of local areas, but do not combine the two, thus failing to fully capture the complex spatial distribution characteristics of CO2.

[0157] In order to take into account the impact of global spatial autocorrelation and local spatial heterogeneity on the CO2 concentration regression model, the present invention proposes a modeling method of the Eigenvector spatial filtering-based geographically weighted regression model (ESF-GWR), which believes that the impact of space on CO2 concentration can be divided into two parts: one is the spatial impact caused by the local environmental conditions themselves, and the other is the global spatial impact and the local manifestation of the global impact.

[0158] To this end, it is first necessary to extract the global spatial characteristic factors that characterize the global position relationship of CO2.

[0159] Specifically, a spatial weight matrix is ​​constructed based on the geographical location of the CO2 concentration data monitoring points, and then the global spatial characteristic factors are screened from the eigenvectors of the spatial weight matrix.

[0160] Among them, the geographical location of the CO2 concentration data monitoring point is determined according to the method of obtaining the CO2 concentration data.

[0161] Optionally, when the CO2 concentration data comes from a CO2 concentration monitoring station established on the ground, the coordinates of each station in the study area are used as the geographical location of the CO2 concentration data monitoring point.

[0162] Furthermore, due to the limited number of ground CO2 concentration monitoring stations and the uneven geographical distribution of the stations, with the development of remote sensing technology, data collected by carbon monitoring satellites and other remote sensing methods can also be used to estimate the ground CO2 concentration. Among them, carbon monitoring satellites are divided into two categories: active satellites and passive satellites.

[0163] Optionally, when the CO2 concentration data is derived from a passive satellite, the geographical location corresponding to the data received by the passive satellite in the study area is used as the geographical location of the CO2 concentration data monitoring point.

[0164] Optionally, when the CO2 concentration data originates from an active satellite, the scanning point coordinates of the active satellite are used as the geographical location of the CO2 concentration data monitoring point.

[0165] Compared with passive satellites, active satellites have obvious advantages in data accuracy and spatial coverage methods, and active satellite data are scanned point data with clear time and space coordinates. Therefore, CO2 concentration data is preferably data acquired by active satellites. In this embodiment, the CO2 concentration data acquired by the active satellite DQ-1 will be taken as an example for introduction.

[0166] It is understandable that in other feasible implementations, the above-mentioned methods can also be combined to obtain CO2 concentration data. For example, CO2 concentration data can be obtained by using active satellites and ground-based CO2 concentration monitoring stations at the same time. In this case, the scanning point positions of the active satellites and the coordinates of the CO2 concentration monitoring stations in the study area are jointly used as the geographical locations of the CO2 concentration data monitoring points.

[0167] On this basis, in this embodiment, the spatial weight matrix W1 is constructed according to the projection coordinates of the scanning point data of the active satellite, and is centralized into a symmetric matrix W:

[0168]

[0169] Where I represents the n-dimensional unit matrix, 11 T represents an n×n matrix, all elements in the matrix are equal to 1, and n represents the number of CO2 concentration data monitoring points in the study area, which in this embodiment is the number of active satellite scanning points in the study area.

[0170] Furthermore, according to the eigenvalues ​​and eigenvectors of the symmetric matrix W, several global spatial characteristic factors that enable the best OLS (Ordinary Least Squares) model fitting effect are screened from the eigenvectors of the symmetric matrix W.

[0171] A local module 502 is used for selecting a characteristic vector representing the local heterogeneity of the space from the global spatial characteristic factors based on a preset model reflecting the local heterogeneity as a local spatial factor;

[0172] Furthermore, in order to comprehensively consider the impact of local spatial heterogeneity on CO2 concentration, the characteristic vector representing the local spatial heterogeneity is screened out from the global spatial characteristic factors as the local spatial factor, which serves as the basis for the final construction of CO2 concentration.

[0173] On this basis, in order to find suitable local spatial factors among the global spatial characteristic factors, the global spatial characteristic factors are screened based on the preset model.

[0174] Optionally, the preset model may be a geographically weighted regression model, a spatial error model, a spatial lag model or other models that can reflect local heterogeneity.

[0175] In this implementation, the geographically weighted regression model is used as an example to explain spatial heterogeneity with spatially varying coefficients, and further screen appropriate local spatial factors.

[0176] The construction module 503 is used to construct a carbon dioxide concentration model based on the local spatial factors.

[0177] After screening the local spatial factors, the CO2 concentration estimation model (ESF-GWR) that takes into account global spatial autocorrelation and local heterogeneity can be constructed as follows:

[0178]

[0179] In the formula, y i represents the CO2 concentration corresponding to the i-th sampling point, (u i ,v i ) is the coordinate of the i-th sampling point, β0 is a constant, β k (u i ,v i ) is the kth regression parameter at the i-th sampling point, x ik is the value of the independent variable at the i-th sampling point, Eα ik is the kth spatial feature weight value at the i-th sampling point, and its value is calculated based on the local spatial factor, ε i is the random error of the i-th sampling point, satisfying the basic assumptions such as zero mean, homoscedasticity, and mutual independence. p represents the number of regression parameters, and q represents the sum of the number of regression parameters and local spatial factors.

[0180] The present invention extracts global spatial characteristic factors by constructing a global spatial matrix, and extracts local spatial factors that take into account both global and local characteristics from the global spatial characteristic factors, and then performs geographically weighted regression modeling together with the factor data that affects CO2 concentration to obtain a CO2 concentration estimation model, so that the constructed model can take into account both global spatial characteristics and local spatial heterogeneity.

[0181] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the CO2 concentration spatial filtering value geographical weighted regression modeling method, the method comprising: constructing a spatial weight matrix based on the geographical location of the carbon dioxide concentration data monitoring point, and extracting the characteristic vector representing the global position relationship from the spatial weight matrix as the global spatial characteristic factor; based on the preset model reflecting the local heterogeneity, screening the characteristic vector representing the spatial local heterogeneity from the global spatial characteristic factor as the local spatial factor; and constructing the carbon dioxide concentration model based on the local spatial factor.

[0182] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the 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 for 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 invention. 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 codes.

[0183] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the CO2 concentration spatial filtering value geographical weighted regression modeling method provided by the above methods. The method includes: constructing a spatial weight matrix based on the geographical location of the carbon dioxide concentration data monitoring point, and extracting a characteristic vector representing the global position relationship from the spatial weight matrix as a global spatial characteristic factor; based on a preset model that reflects local heterogeneity, screening out a characteristic vector representing spatial local heterogeneity from the global spatial characteristic factor as a local spatial factor; and constructing a carbon dioxide concentration model based on the local spatial factor.

[0184] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the CO2 concentration spatial filtering value geographical weighted regression modeling method provided by the above-mentioned methods, the method comprising: constructing a spatial weight matrix based on the geographical location of the carbon dioxide concentration data monitoring point, and extracting a characteristic vector representing the global position relationship from the spatial weight matrix as a global spatial characteristic factor; based on a preset model reflecting local heterogeneity, screening out a characteristic vector representing spatial local heterogeneity from the global spatial characteristic factor as a local spatial factor; and constructing a carbon dioxide concentration model based on the local spatial factor.

[0185] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0186] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0187] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A CO2 concentration spatial filtering value geographically weighted regression modeling method, characterized in that: include: A spatial weight matrix is ​​constructed based on the geographical location of the carbon dioxide concentration data monitoring point, and a feature vector representing the global position relationship is extracted from the spatial weight matrix as a global spatial feature factor; Based on a preset model reflecting local heterogeneity, a characteristic vector representing the local heterogeneity of space is screened out from the global spatial characteristic factor as a local spatial factor; A carbon dioxide concentration model is constructed based on the local spatial factors.

2. The CO2 concentration spatial filtering value geographically weighted regression modeling method according to claim 1 is characterized in that: The step of selecting a characteristic vector representing the local heterogeneity of space from the global spatial characteristic factor based on the preset model reflecting the local heterogeneity as the local spatial factor specifically includes: The global spatial characteristic factor is used as an independent variable of the preset model, and the preset model is calculated in combination with the pre-acquired influencing factor data; Calculate the corrected determination coefficient corresponding to each of the global spatial characteristic factors; In the case where the correction determination coefficient calculated each time is greater than the current maximum correction determination coefficient, the global spatial characteristic factor corresponding to the current correction determination coefficient is retained, and the current correction determination coefficient is updated to the current maximum correction determination coefficient; otherwise, the global spatial characteristic factor corresponding to the current correction determination coefficient is removed, wherein the initial value of the current maximum correction determination coefficient is set to zero; The global space characteristic factor retained after the cyclic calculation is used as the local space factor.

3. The CO2 concentration spatial filtering value geographically weighted regression modeling method according to claim 2 is characterized in that: The step of using the global spatial characteristic factor as the independent variable of the preset model and calculating the preset model in combination with the pre-acquired influencing factor data specifically includes: Randomly sorting all the global spatial feature factors and adding them into the spatial feature vector set; The global spatial characteristic factors in the spatial characteristic vector set are sequentially added to the independent variable set of the preset model as the independent variables of the preset model and combined with the pre-acquired influencing factor data to perform calculations on the preset model.

4. The CO2 concentration spatial filtering value geographically weighted regression modeling method according to any one of claims 1 to 3, characterized in that: The preset model is a geographically weighted regression model, a spatial error model or a spatial lag model.

5. The CO2 concentration spatial filtering value geographically weighted regression modeling method according to any one of claims 1 to 3, characterized in that: After the step of constructing a carbon dioxide concentration model based on the local spatial factor, the method further includes: Performing grid calculation on the carbon dioxide concentration model to obtain a carbon dioxide concentration distribution map; The abnormal values ​​in the carbon dioxide concentration distribution map are processed.

6. The CO2 concentration spatial filtering value geographically weighted regression modeling method according to any one of claims 1 to 3, characterized in that: The step of constructing a spatial weight matrix based on the geographical location of the carbon dioxide concentration data monitoring point includes: Based on adjacency, inverse distance weighting or a combination of adjacency and inverse distance weighting, a spatial weight matrix is ​​constructed in combination with the geographical locations of the carbon dioxide concentration data monitoring points.

7. The CO2 concentration spatial filtering value geographically weighted regression modeling method according to claim 2 or 3, characterized in that: The preset model based on the reaction local heterogeneity, before the step of selecting the characteristic vector representing the local heterogeneity of the space from the global spatial characteristic factor as the local spatial factor, further comprises: Obtain carbon dioxide concentration data for a preset area; Taking surface factors, atmospheric factors, human activity factors and radiation factors as influencing factors, the influencing factor data of the preset area are obtained; The carbon dioxide concentration data and the influencing factor data are preprocessed.

8. The CO2 concentration spatial filtering value geographically weighted regression modeling method according to claim 7 is characterized in that: The preset model based on the reaction local heterogeneity, before the step of selecting the characteristic vector representing the local heterogeneity of the space from the global spatial characteristic factor as the local spatial factor, further comprises: Calculate the variance inflation factor between each influencing factor data and the carbon dioxide concentration data. When the currently calculated variance inflation factor is greater than a preset threshold, remove the influencing factor corresponding to the current maximum variance inflation factor. Repeat the above steps until the variance inflation factor corresponding to each influencing factor after screening is less than the preset threshold.

9. A CO2 concentration spatial filtering value geographically weighted regression modeling device, characterized in that: include: A global module, used to construct a spatial weight matrix based on the geographical location of the carbon dioxide concentration data monitoring point, and extract a feature vector representing the global position relationship from the spatial weight matrix as a global spatial feature factor; A local module, for obtaining a characteristic vector representing the local heterogeneity of the space as a local spatial factor by screening the characteristic vector from the global spatial characteristic factor based on a preset model reflecting the local heterogeneity; A building module is used to build a carbon dioxide concentration model based on the local spatial factors.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the CO2 concentration spatial filtering value geographical weighted regression modeling method as described in any one of claims 1 to 8 is implemented.

Images