Mining area drought index inversion method and system based on multi-source data

Abstract

The invention discloses a mining area drought index inversion method and system based on multi-source data. The method comprises the steps that S1, a multi-source data sample set containing the multi-source data and drought index labels is constructed; s2, a drought index inversion model is constructed, the drought index inversion model obtains a temperature vegetation drought index, a vegetation condition index, a temperature condition index and a rainfall condition index through multi-source data inversion of the multi-source data sample set, and the temperature vegetation drought index, the vegetation condition index, the temperature condition index and the rainfall condition index serve as prediction factors; s3, collecting a multi-source data set of the research area, inputting the multi-source data set into the drought index inversion model, and obtaining predicted drought indexes corresponding to all positions of the research area. According to the method, the prediction factors are obtained through inversion of the multi-source data, then the trained drought index inversion model is obtained through model inversion training and used for prediction processing of the drought indexes, and the long-time-sequence drought time-space change characteristics of mining area vegetation can be accurately inverted.

Description

Technical Field

[0001] The present invention relates to the field of mining area drought assessment, and in particular to a mining area drought index inversion method and system based on multi-source data. Background Art

[0002] Drought refers to a long-term lack of rain or high temperature and little rain in a certain area, which causes a lack of moisture in the air and soil. Drought can be divided into two types: climate drought and soil drought. Climate drought refers to a long-term lack of precipitation, while soil drought refers to a significant lack of moisture in the soil. Both types of drought will have an impact on vegetation and cause vegetation drought, which is more significant in mining areas because human mining activities in mining areas have damaged the soil structure and reduced the soil's ability to retain moisture. China is rich in mineral resources, especially coal resources, but most coal resources are distributed in arid and semi-arid areas in the northwest. These areas have scarce precipitation, large water evaporation, large temperature differences and fragile ecology. The mining of coal resources has caused the destruction of water resources, aggravating the drought situation in the region, thereby affecting the growth of vegetation. Vegetation drought is closely related to soil moisture. In order to quantitatively evaluate the impact of drought on vegetation, relevant scholars have used vegetation condition variables from multiple angles to create a drought index to effectively monitor vegetation drought. However, in the existing remote sensing drought monitoring, most studies currently use a single remote sensing index or a combination of fewer remote sensing indices. These result in low accuracy of the inverted drought results and cannot effectively reflect the drought conditions in mining areas. Therefore, it is necessary to develop a method that uses multi-source data to extract multiple indices to invert the mining area drought index. Summary of the invention

[0003] The purpose of the present invention is to provide a mining area drought index inversion method based on multi-source data, which uses multi-source data including remote sensing image data, DEM data, meteorological data, surface albedo data, surface solar radiation data, and land use data to respectively invert the temperature vegetation drought index, vegetation condition index, temperature condition index, and precipitation condition index as prediction factors, and then obtains a trained drought index inversion model through model inversion training and uses it for drought index prediction processing, so as to accurately invert the spatiotemporal variation characteristics of drought in the long time series of vegetation in the mining area.

[0004] The purpose of the present invention is achieved through the following technical solutions:

[0005] A mining area drought index inversion method based on multi-source data, the method comprising:

[0006] S1. Construct a multi-source data sample set containing multi-source data and drought index labels;

[0007] S2. Construct a drought index inversion model. The drought index inversion model uses the multi-source data of the multi-source data sample set to invert the temperature vegetation drought index, vegetation condition index, temperature condition index, and precipitation condition index as prediction factors. The drought index inversion model uses the multi-source data sample set to perform model training according to the following expression:

[0008] SDI i =α×VCI i +β×TCI i +γ×PCI i +δ×TVDI i +b, where SDI i Represents the drought index of pixel i, VCI i Represents the vegetation condition index of pixel i, TCI i Represents the temperature condition index of pixel i, PCI i Represents the precipitation condition index of pixel i, TVDI i Represents the temperature vegetation drought index of pixel i; α, β, γ, δ are coefficients respectively, and b is a constant term;

[0009] S3. Collect multi-source data sets in the study area and input them into the drought index inversion model to obtain the predicted drought index corresponding to each location in the study area.

[0010] In order to better implement the mining area drought index inversion method of the present invention, the present invention also includes the following method:

[0011] S4. Extract data source factors of multi-source data from the drought index inversion model; use structural equation modeling to analyze the drought driving relationship between all data source factors, all prediction factors and drought index, and obtain the direct driving relationship data of the data source factors directly driving the drought index and the indirect driving relationship data of the data source factors driving the drought index through the prediction factors.

[0012] Preferably, in step S4, before drought-driven analysis, structural equation modeling uses the Mantel test method, geographic detector method and Pearson correlation analysis method to screen out data source factors that are closely related to the drought index from all data source factors.

[0013] Preferably, the method for inverting the temperature vegetation drought index by the drought index inversion model is as follows:

[0014] A10. Extract the reflectance NIR of the near infrared band and the reflectance R of the red light band from the multi-source data, and calculate the normalized vegetation index NDVI according to the following method:

[0015]

[0016] A11. Extracting the surface emissivity T from multi-source datas , the transmittance T of the atmosphere in the thermal infrared band, the downward radiation brightness L1 of the atmosphere, and the upward radiation brightness L2 of the atmosphere. The blackbody radiation brightness is inverted using the atmospheric correction method. The inversion expression is as follows:

[0017] B(T s ) represents the black body radiation brightness, ε represents;

[0018] The real surface temperature T is calculated based on the inverse function of Planck's formula s :

[0019] Where K1 and K2 are both constants;

[0020] A12. The temperature vegetation drought index is calculated by inversion as follows:

[0021]

[0022] T max =a 1 +b 1 ×NDVT

[0023] T min =a 2 +b 2 ×NDVT, where T max Indicates the highest true surface temperature; T min Indicates the lowest true surface temperature; a 1 、a 2 , b 1 , b 2 is the regression coefficient.

[0024] Preferably, the drought index inversion model inverts and calculates the temperature vegetation drought index according to the following formula:

[0025] VCI i Represents the temperature vegetation drought index of pixel i, NDVI i Represents the NDVI value of pixel i; NDVI max and NDVI min They represent the maximum and minimum NDVI values ​​during the study period, respectively;

[0026] The drought index inversion model is used to invert and calculate the temperature condition index according to the following formula:

[0027] TCI i Represents the temperature condition index of pixel i, LST i Indicates the true surface temperature of pixel i; LST maxand LST min They represent the maximum and minimum values ​​of the true surface temperature during the study period, respectively.

[0028] Preferably, the drought index inversion model extracts precipitation data from multi-source data and inverts and calculates the precipitation condition index according to the following formula:

[0029] PCI i represents the precipitation condition index of pixel i, PRE i represents the precipitation in pixel i, PRE max 、PRE min They represent the maximum and minimum precipitation during the study period, respectively.

[0030] Preferably, structural equation modeling is used to obtain the relationship between the observed variable and the latent variable, and the relationship between the two variables is determined according to the following expression:

[0031] X=λ x ω+μY=λ y ω+v

[0032] Where X represents the observed variable of the exogenous variable; Y represents the observed variable of the endogenous variable; η represents the endogenous latent variable; ω represents the exogenous latent variable. μ, υ and Represent the errors of X, Y and η respectively; λ x is the coefficient matrix between X and ω, λ y is the coefficient matrix between Y and ω, β is the coefficient matrix of η, and Ψ is the coefficient matrix between ω and η.

[0033] Preferably, the direct driving relationship data and the indirect driving relationship data both include influence degree data, and the influence degree data is calculated using the Pearson correlation coefficient according to the following formula:

[0034] Where n represents the total number of time series of the research time, j represents the i-th node in the time series of the research time, and x j and j represents the two variables of the ith node in the time series of the research time, and Represent the mean respectively; By setting the Pearson correlation coefficient threshold between the drought index SDI and the data source factor, the data source factors with significant correlation with the drought index SDI are screened out.

[0035] A mining area drought index inversion system based on multi-source data includes a data acquisition module, a multi-source data sample set and a drought index inversion model. The data acquisition module is used to collect multi-source data and mark drought index labels to form a multi-source data sample set. The multi-source data sample set is stored in association with the multi-source data and the drought index label; the drought index inversion model uses the multi-source data of the multi-source data sample set to invert the temperature vegetation drought index, the vegetation condition index, the temperature condition index, and the precipitation condition index as prediction factors. The drought index inversion model uses the multi-source data sample set to perform model training according to the following expression: SDI i =α×VCI i +β×TCI i +γ×PCI i +δ×TVDI i +b, where SDI i Represents the drought index of pixel i, VCI i Represents the vegetation condition index of pixel i, TCI i Represents the temperature condition index of pixel i, PCI i Represents the precipitation condition index of pixel i, TVDI i Represents the temperature vegetation drought index of pixel i; α, β, γ, δ are coefficients respectively, and b is a constant term; the data acquisition module also collects multi-source data sets of the study area and inputs them into the drought index inversion model to obtain the predicted drought index corresponding to each location in the study area.

[0036] The mining area drought index inversion system of the present invention also includes structural equation modeling, which extracts data source factors of multi-source data; uses structural equation modeling to analyze the drought driving of all data source factors, all prediction factors and drought index, and obtains direct driving relationship data of the data source factors directly driving the drought index and indirect driving relationship data of the data source factors driving the drought index through the prediction factors.

[0037] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0038] (1) The present invention utilizes multi-source data including remote sensing image data, DEM data, meteorological data, surface albedo data, surface solar radiation data, and land use data to invert the temperature vegetation drought index, vegetation condition index, temperature condition index, and precipitation condition index respectively and use them as prediction factors. Then, a trained drought index inversion model is obtained through model inversion training and used for drought index prediction processing, which can accurately invert the spatiotemporal variation characteristics of drought in the long time series of vegetation in the mining area.

[0039] (2) The present invention uses structural equation SEM to accurately quantify the direct and indirect effects of various driving factors on vegetation drought in mining areas, and obtains direct driving relationship data of the data source factor directly driving the drought index, and indirect driving relationship data of the data source factor driving the drought index through the prediction factor, which is conducive to the extraction of data source factors of multi-source data, as well as the drought driving relationship and quantitative data with all prediction factors and drought indices. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A method flow chart of the mining area drought index inversion method of the present invention;

[0041] Figure 2 It is a schematic diagram of the principle of the mining area drought index inversion method of the present invention;

[0042] Figure 3 The distribution map of meteorological stations in a certain study area in the embodiment;

[0043] Figure 4 It is the elevation and slope result map of the study area in the embodiment;

[0044] Figure 5 This is an example diagram of the inversion results of meteorological data in the study area in the embodiment;

[0045] Figure 6 This is an example diagram of the land use results of the study area in the embodiment;

[0046] Figure 7 This is an example diagram of the TVDI inversion results of the study area in the embodiment;

[0047] Figure 8 This is an example of the inversion results of the mining drought index SDI in the study area in 2019 in the embodiment;

[0048] Fig. 9 It is a relationship diagram of the direct driving relationship data and the indirect driving relationship data of the study area in the embodiment. DETAILED DESCRIPTION

[0049] The present invention is further described in detail below in conjunction with embodiments:

[0050] Example

[0051] like Figure 1 As shown, a mining area drought index inversion method based on multi-source data includes:

[0052] S1. Construct a multi-source data sample set containing multi-source data and drought index labels; the sources of multi-source data include at least remote sensing image data, DEM data, meteorological data, surface albedo data, surface solar radiation data, land use data, and of course other data related to drought. Remote sensing image data are preprocessed using the Google Earth Engine (GEE) cloud platform, meteorological data (including temperature, precipitation, surface wind speed and relative humidity RH data) are from the Meteorological Data Center (such as the China Meteorological Data Network (http: / / data.cma.cn / wa)), DEM data are from NASA Shuttle Radar Topography Mission (STRM) data, and slope and aspect data can be obtained after ARCGIS processing. Surface albedo data are from MODIS surface albedo product MCD43A3. Surface solar radiation data are from the spatiotemporal three-level environmental big data platform (http: / / poles.tpdc.ac.cn / zh-hans / ). After the relevant data of multi-source data are collected, the spatial resolution matching the Landsat remote sensing image data can be obtained through resampling. After the resolution is matched, the data can be interpolated and cropped to obtain multi-source data of the mining area.

[0053] This embodiment takes Xiwan open-pit coal mine in Shenmu City, Yulin City, Shaanxi Province as an example of the research area. The distribution of meteorological stations in the research mining area in Shaanxi Province is as follows: Figure 3 As shown, Figure 4 The elevation and slope results of the study area are shown (the left picture is elevation, the right picture is slope). Figure 5 The following figure shows the example of the inversion results of meteorological data in the study area (the left figure is temperature, and the right figure is precipitation). Figure 6 An example map of land use results for the study area is shown.

[0054] S2. Construct a drought index inversion model. The drought index inversion model uses the multi-source data of the multi-source data sample set to invert the temperature vegetation drought index, vegetation condition index, temperature condition index, and precipitation condition index as prediction factors. The drought index inversion model uses the multi-source data sample set to perform model training according to the following expression:

[0055] SDI i =α×VCI i +β×TCI i +γ×PCI i +δ×TVDI i +b, where SDI i Represents the drought index of pixel i, VCI i Represents the vegetation condition index of pixel i, TCI i Represents the temperature condition index of pixel i, PCI i Represents the precipitation condition index of pixel i, TVDIi represents the temperature vegetation drought index of pixel i; α, β, γ, δ are coefficients respectively, and b is a constant term. The Synthesized Drought Index (SDI) in this embodiment is an agricultural drought index for monitoring the spatiotemporal characteristics of agricultural drought. It is composed of VCI i 、TCI i 、PCI i 、TVDI i It is a weighted combination of multiple single factor indexes; VCI i 、TCI i 、PCI i 、TVDI i Respectively represent vegetation condition index, temperature condition index, precipitation condition index and temperature vegetation drought index, α+β+γ+δ=1; VCI is converted into i 、TCI i 、PCI i and TVDI i Perform a weighted combination.

[0056] In some embodiments, the method for inverting the drought index inversion model to obtain the temperature vegetation drought index is as follows:

[0057] A10. Extract the reflectance NIR of the near infrared band and the reflectance R of the red light band from the multi-source data. The reflectance NIR of the near infrared band and the reflectance R of the red light band mainly come from the remote sensing image data in the multi-source data. In this embodiment, the remote sensing images of Landsat5 / 7 / 8 are obtained year by year through the Google Earth Engine cloud platform. For example, the remote sensing images of the study area in August are selected to perform the inversion of the Normalized Difference Vegetation Index (NDVI). The Normalized Difference Vegetation Index NDVI is calculated according to the following method:

[0058]

[0059] A11. Extracting the surface emissivity T from multi-source data s , the transmittance of the atmosphere in the thermal infrared band T, the downward radiation brightness of the atmosphere L1, the upward radiation brightness of the atmosphere L2, and the surface emissivity T s The transmittance T of the atmosphere in the thermal infrared band, the downward radiation brightness L1 of the atmosphere, and the upward radiation brightness L2 of the atmosphere are mainly derived from the surface albedo data and surface solar radiation data of the remote sensing image data in the multi-source data. The atmospheric correction method is used to invert the blackbody radiation brightness. The inversion expression is as follows:

[0060] B(T s ) represents the black body radiation brightness, ε represents;

[0061] The real surface temperature T is calculated based on the inverse function of Planck's formula s :

[0062] Where K1 and K2 are both constants;

[0063] A12. The temperature vegetation drought index is calculated by inversion as follows:

[0064]

[0065] T max =a 1 +b 1 ×NDVI

[0066] T min =a 2 +b 2 ×NDVI, where T max Indicates the highest true surface temperature; T min Indicates the lowest true surface temperature; a 1 、a 2 , b 1 , b 2 is the regression coefficient.

[0067] In the example of Shaanxi Province as the study area, the TVDI inversion results of the study area are as follows: Figure 7 shown.

[0068] In some embodiments, the drought index inversion model inverts and calculates the temperature vegetation drought index according to the following formula:

[0069] VCI i Represents the temperature vegetation drought index of pixel i, NDVI i Represents the NDVI value of pixel i; NDVI max and NDVI min They represent the maximum and minimum NDVI values ​​during the study period, respectively;

[0070] The drought index inversion model is used to invert and calculate the temperature condition index according to the following formula:

[0071] TCI i Represents the temperature condition index of pixel i, LST i Indicates the true surface temperature of pixel i; LST max and LST minThey represent the maximum and minimum values ​​of the true surface temperature during the study period, respectively.

[0072] In some embodiments, the drought index inversion model extracts precipitation data from multi-source data and inverts and calculates the precipitation condition index according to the following formula:

[0073] PCI i represents the precipitation condition index of pixel i, PRE i represents the precipitation in pixel i, PRE max 、PRE min They represent the maximum and minimum precipitation during the study period, respectively.

[0074] S3. Collect multi-source data sets in the study area and input them into the drought index inversion model to obtain the predicted drought index corresponding to each location in the study area. In the example of Shaanxi Province as the study area, the SDI inversion results of the study area in 2019 are as follows: Figure 8 shown.

[0075] In some embodiments, the present invention further comprises the following method:

[0076] S4. Extract data source factors of multi-source data for drought index inversion model (extract data source factors from multi-source data including remote sensing image data, DEM data, meteorological data, surface albedo data, surface solar radiation data, and land use data); use structural equation modeling to analyze drought driving of all data source factors, all predictors and drought index, and obtain direct driving relationship data of data source factors directly driving drought index, and indirect driving relationship data of data source factors driving drought index through predictors. In the example of Shaanxi Province as the study area, Fig. 9 The relationship diagram of the direct driving relationship data and the indirect driving relationship data of the study area is represented in the form of icons.

[0077] In some embodiments, the structural equation modeling in step S4 uses the Mantel test method, the geographic detector method, and the Pearson correlation analysis method to screen out data source factors that are closely related to the drought index from all data source factors before the drought drive analysis. The Mantel test method is as follows:

[0078] The present invention uses the Mantel normalized statistic "r" to determine the key data source factors affecting drought:

[0079]

[0080] Where i≠j, n is the distance number of one of the matrices (except the diagonal), and are the means of x and y respectively. xand S y are the standard deviations of x and y respectively; the closer the value of r is to 1, the stronger the correlation is, and the closer the value of p is to 1, the more significant the correlation between the two is. The Mantel test is used to determine the key driving factors of the drought index SDI (i.e., test all data source factors), especially the variables that are significantly correlated under different drought levels (i.e., data source factors), and further screen out the data source factors that are closely related to the drought index SDI in space.

[0081] The geographic detector method uses statistical data to reveal various influencing factors (i.e., data source factors) that lead to spatial heterogeneity by testing spatial correlation. The purpose of factor detection is to determine the spatial diversity of the dependent variable and to evaluate the extent to which the independent variable can explain this diversity; the q value ranges from 0 to 1 and is used to measure the explanatory power. The factor detector is used to determine how the driving factors affect the response variable. The q statistic of the factor detector measures the explanatory power. The higher the value, the more significant the spatial differentiation of the dependent variable and the higher the spatial correlation. Conversely, the lower the value, the weaker the spatial differentiation and the lower the spatial correlation. Factors with high q values ​​are important factors in explaining SDI changes in space and can be used as important observed variables in the SEM model.

[0082]

[0083] In the formula, L is the total number of stratified influencing factors, N and N are the number of samples in the i-th stratum and the total number of samples, respectively. and σ 2 are the local variance of SDI at layer i and the global variance of SDI at layer i, respectively.

[0084] In some embodiments, structural equation modeling is used to obtain the relationship between the observed variable and the latent variable, and the relationship between the two variables is determined according to the following expression:

[0085] X=λ x ω+μY=λ y ω+v

[0086] Where X represents the observed variables of exogenous variables (i.e., those observed variables that may have an impact on the drought index SDI, such as temperature, precipitation, DEM, etc.); Y represents the observed variables of endogenous variables (i.e., the comprehensive drought index SDI); η represents the endogenous latent variable; ω represents the exogenous latent variable (a variable that cannot be directly observed, but an implicit factor used in the model to explain the complex relationship between the influencing factors and the drought index SDI, which can be understood as a mediating variable). μ, υ and They represent the errors of X, Y, and η respectively (indicating the part of the exogenous observed variable X that cannot be explained by the latent factor b); x is the coefficient matrix between X and ω, λ yis the coefficient matrix between Y and ω, β is the coefficient matrix of η, and Ψ is the coefficient matrix between ω and η.

[0087] The path coefficient between two factors (i.e., data source factors) represents the direct impact, while the product of the path coefficients of the relevant impact paths represents the indirect impact. If there are multiple indirect effect paths, the sum of the indirect effects is the sum of these indirect effects. At the same time, this embodiment can establish a preliminary model to further understand the balance between vegetation drought in mining areas and the main driving factors; specifically, the lavaanSEM package in R is used to standardize and refine the model.

[0088]

[0089] The model was debugged using R software, and the goodness of fit of the SEM equation was measured using four parameters: comparative fit index (CFI), root mean square error of approximation (RSMEA), chi-square test (SHISQ) and P value. The statistic and degrees of freedom of the model were assumed to be Th and fh, respectively, and the statistic and degrees of freedom of the model were assumed to be T. b and f b , when CFI>0.9, RMSEA<0.08, CHISQ<0.1, p<0.05, it means that the model fits well. The model parameters are debugged in R software to meet the above requirements. Each model established by SEM can obtain a good fit for subsequent analysis. Formula 1: RMSE is an indicator used to measure the fitting error between the model and the data, and is used to test the goodness of model fit; Formula 2: CFI is an indicator to measure the goodness of model fit, which is usually used to compare the fitting effect of the fitted model with the unstructured model. The value range of CFI is from 0 to 1. The closer to 1, the better the fitting effect.

[0090] Direct effect: refers to the relationship between two variables directly represented by the path coefficient, usually the direct effect of one variable on another variable in the model. For example, the direct effect of variable X on Y is usually expressed by the path coefficient, which indicates how much X will change for every unit change in Y.

[0091] Indirect effect: refers to the effect of one variable on another variable through one or more mediating variables (i.e., mediating effect). It represents the multiple path effects through one or more mediating variables. For example, X affects M, and M further affects Y, then the indirect effect of X on Y is the effect of X through M. Indirect effects are usually calculated by multiplying the path coefficients.

[0092] In some embodiments, both the direct driving relationship data and the indirect driving relationship data include influence degree data, and the influence degree data is calculated using the Pearson correlation coefficient according to the following formula:

[0093] Where n represents the total number of time series of the research time, j represents the i-th node in the time series of the research time, and x j and j represents the two variables of the ith node in the time series of the research time, and Represent the mean respectively; By setting the Pearson correlation coefficient threshold between the drought index SDI and the data source factor, the data source factors with significant correlation with the drought index SDI are screened out.

[0094] Analysis of drought driving factors based on structural equation modeling

[0095] A long-term spatiotemporal analysis of the drought situation in the mining area was conducted based on the results of the calculated long-term comprehensive drought index (SDI), and a driving factor analysis was conducted based on the results of structural equation modeling (SEM).

[0096] Variable selection and path hypothesis: Based on the selected key variables, the path structure of SEM is designed. It is assumed that the factors that directly affect SDI are variables with high q values ​​and variables with strong spatial correlation, such as temperature, precipitation, relative humidity, etc., while other factors such as land use type, DEM, etc. have indirect effects on SDI. In addition, interaction paths can be set based on the results of geographic detectors.

[0097] Model fitting and adjustment: Use SEM software to run the model and evaluate the model fit by comparing fit indicators; if the fit is not ideal, adjust the SEM path, increase or decrease the path, to ensure that the model has a better ability to explain the data.

[0098] Results analysis and interpretation: Interpret the coefficients of each path in SEM to determine the direct, indirect and total impact of each factor on SDI. Analyze the impact of each influencing factor on drought in the study area.

[0099] A mining area drought index inversion system based on multi-source data includes a data acquisition module, a multi-source data sample set and a drought index inversion model. The data acquisition module is used to collect multi-source data and mark drought index labels to form a multi-source data sample set. The multi-source data sample set is stored in association with the multi-source data and the drought index label; the drought index inversion model uses the multi-source data of the multi-source data sample set to invert the temperature vegetation drought index, the vegetation condition index, the temperature condition index, and the precipitation condition index as prediction factors. The drought index inversion model uses the multi-source data sample set to perform model training according to the following expression: SDI i =α×VCI i +β×TCI i +γ×PCI i +δ×TVDI i +b, where SDIi Represents the drought index of pixel i, VCI i Represents the vegetation condition index of pixel i, TCI i Represents the temperature condition index of pixel i, PCI i Represents the precipitation condition index of pixel i, TVDI i Represents the temperature vegetation drought index of pixel i; α, β, γ, δ are coefficients respectively, and b is a constant term; the data acquisition module also collects multi-source data sets of the study area and inputs them into the drought index inversion model to obtain the predicted drought index corresponding to each location in the study area.

[0100] The mining area drought index inversion system of the present invention also includes structural equation modeling, which extracts data source factors of multi-source data; uses structural equation modeling to analyze the drought driving of all data source factors, all prediction factors and drought index, and obtains direct driving relationship data of the data source factors directly driving the drought index and indirect driving relationship data of the data source factors driving the drought index through the prediction factors.

[0101] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A mining area drought index inversion method based on multi-source data, characterized by: The methods include: S1. Construct a multi-source data sample set containing multi-source data and drought index labels; S2. Construct a drought index inversion model. The drought index inversion model uses the multi-source data of the multi-source data sample set to invert the temperature vegetation drought index, vegetation condition index, temperature condition index, and precipitation condition index as prediction factors. The drought index inversion model uses the multi-source data sample set to perform model training according to the following expression: SDI i =α×VCI i +β×TCI i +γ×PCI i +δ×TVDI i +b, where SDI i Represents the drought index of pixel i, VCI i Represents the vegetation condition index of pixel i, TCI i Represents the temperature condition index of pixel i, PCI i Represents the precipitation condition index of pixel i, TVDI i Represents the temperature vegetation drought index of pixel i; α, β, γ, δ are coefficients respectively, and b is a constant term; S3. Collect multi-source data sets in the study area and input them into the drought index inversion model to obtain the predicted drought index corresponding to each location in the study area.

2. The mining area drought index inversion method based on multi-source data according to claim 1, characterized in that: Also includes the following methods: S4. Extract data source factors of multi-source data from the drought index inversion model; use structural equation modeling to analyze the drought driving relationship between all data source factors, all prediction factors and drought index, and obtain the direct driving relationship data of the data source factors directly driving the drought index and the indirect driving relationship data of the data source factors driving the drought index through the prediction factors.

3. The method for inverting the mining area drought index based on multi-source data according to claim 2, characterized in that: In step S4, structural equation modeling uses the Mantel test method, geographic detector method, and Pearson correlation analysis method to screen out data source factors that are closely related to the drought index from all data source factors before drought driving analysis.

4. The mining area drought index inversion method based on multi-source data according to claim 1, characterized in that: The method for inverting the temperature vegetation drought index using the drought index inversion model is as follows: A10. Extract the reflectance NIR of the near infrared band and the reflectance R of the red light band from the multi-source data, and calculate the normalized vegetation index NDVI according to the following method: A11. Extracting the surface emissivity T from multi-source data ε , the transmittance T of the atmosphere in the thermal infrared band, the downward radiation brightness L1 of the atmosphere, and the upward radiation brightness L2 of the atmosphere. The blackbody radiation brightness is inverted using the atmospheric correction method. The inversion expression is as follows: B(T S ) represents the black body radiation brightness, ε represents; The real surface temperature T is calculated based on the inverse function of Planck's formula S : Where K1 and K2 are both constants; A12. The temperature vegetation drought index is calculated by inversion as follows: T max =a1+b1×NDVI T min =a2+b2×NDVI, where T max Indicates the maximum true surface temperature; T min represents the lowest true surface temperature; a1, a2, b1, b2 are regression coefficients.

5. The mining area drought index inversion method based on multi-source data according to claim 4 is characterized in that: The drought index inversion model is calculated by inversion according to the following formula to obtain the temperature vegetation drought index: VCI i Represents the temperature vegetation drought index of pixel i, NDVI i Represents the NDVI value of pixel i; NDVI max and NDVI min They represent the maximum and minimum NDVI values ​​during the study period, respectively; The drought index inversion model is used to invert and calculate the temperature condition index according to the following formula: TCI i Represents the temperature condition index of pixel i, LST i Indicates the true surface temperature of pixel i; LST max and LST min They represent the maximum and minimum values ​​of the true surface temperature during the study period, respectively.

6. The mining area drought index inversion method based on multi-source data according to claim 1, characterized in that: The drought index inversion model extracts precipitation data from multi-source data and calculates the precipitation condition index according to the following formula: PCI i represents the precipitation condition index of pixel i, PRE i represents the precipitation in pixel i, PRE max 、PRE min They represent the maximum and minimum precipitation during the study period, respectively.

7. The method for inverting the mining area drought index based on multi-source data according to claim 2, characterized in that: Structural equation modeling is used to obtain the relationship between observed variables and latent variables. The relationship between two variables is determined according to the following expression: X=λ x ω+μY=λ y ω+v Where X represents the observed variable of the exogenous variable; Y represents the observed variable of the endogenous variable; η represents the endogenous latent variable; ω represents the exogenous latent variable. μ, υ and Represent the errors of X, Y and η respectively; λ x is the coefficient matrix between X and ω, λ y is the coefficient matrix between Y and ω, β is the coefficient matrix of η, and Ψ is the coefficient matrix between ω and η.

8. The mining area drought index inversion method based on multi-source data according to claim 2 is characterized in that: The direct driving relationship data and the indirect driving relationship data both include influence degree data, and the influence degree data is calculated using the Pearson correlation coefficient according to the following formula: Where n represents the total number of time series of the research time, j represents the i-th node in the time series of the research time, and x j and j represents the two variables of the ith node in the time series of the research time, and Represent the mean respectively; By setting the Pearson correlation coefficient threshold between the drought index SDI and the data source factor, the data source factors with significant correlation with the drought index SDI are screened out.

9. A mining area drought index inversion system based on multi-source data, characterized by: The invention comprises a data acquisition module, a multi-source data sample set and a drought index inversion model. The data acquisition module is used to collect multi-source data and mark drought index labels to form a multi-source data sample set. The multi-source data sample set is stored in association with the multi-source data and the drought index label. The drought index inversion model uses the multi-source data of the multi-source data sample set to invert the temperature vegetation drought index, the vegetation condition index, the temperature condition index and the precipitation condition index as prediction factors. The drought index inversion model uses the multi-source data sample set to perform model training according to the following expression: SDI i =α×VCI i +β×TCI i +γ×PCI i +δ×TVDI i +b, where SDI i Represents the drought index of pixel i, VCI i Represents the vegetation condition index of pixel i, TCI i Represents the temperature condition index of pixel i, PCI i Represents the precipitation condition index of pixel i, TVDI i Represents the temperature vegetation drought index of pixel i; α, β, γ, δ are coefficients respectively, and b is a constant term; the data acquisition module also collects multi-source data sets of the study area and inputs them into the drought index inversion model to obtain the predicted drought index corresponding to each location in the study area.

10. The mining area drought index inversion system based on multi-source data according to claim 9, characterized in that: It also includes structural equation modeling, which extracts data source factors of multi-source data; uses structural equation modeling to analyze the drought driving of all data source factors, all predictive factors and drought index, and obtains direct driving relationship data of data source factors directly driving drought index and indirect driving relationship data of data source factors driving drought index through predictive factors.

Images