A GWDM-based land surface temperature downscaling method

Abstract

The present invention claims a method for spatial downscaling of land surface temperature based on the Geographically Weighted Durbin Model (GWDM). Land surface temperature with high temporal and spatial resolution is crucial in the study of environmental ecology and climate systems. However, due to technical limitations, satellite thermal sensors cannot provide thermal infrared images with high temporal resolution and high spatial resolution at the same time. The spatial downscaling method combines low spatial resolution LST and auxiliary data with high spatial resolution from other satellite sensors to improve the temporal and spatial resolution of temperature images under the premise of assuming scale invariance. However, spatial non-stationarity and spatial autocorrelation coexist in most spatial variables, and the spatial characteristics of land surface temperature should be fully considered as spatial variables in the downscaling process. However, previous studies on downscaling of land surface temperature only considered spatial non-stationarity and ignored the spatial autocorrelation between independent variables and dependent variables.

Description

Technical Field

[0001] The present invention belongs to the technical field of remote sensing image processing, is subdivided into a surface temperature downscaling direction, and relates to a surface temperature downscaling algorithm based on a GWDM (Geographically Weighted Durbin Model). Background Art

[0002] Land surface temperature is one of the key physical parameters that affect the environmental ecology and climate system. It is widely used in various scientific research, including environmental modeling from local to global scales, soil moisture research, surface urban heat island and environmental monitoring, environmental biogeochemical process simulation and climate change assessment. In recent years, the importance of LST has received increasing attention.

[0003] At present, the main way to obtain surface temperature over a large range is still through thermal sensors carried by satellites, which can detect and describe the spatial and temporal differences in surface temperature from both macroscopic and dynamic aspects), while conventional ground station observation methods make it difficult to obtain the spatiotemporal distribution of surface temperature at regional and global scales. However, due to the constraints of satellite remote sensing imaging technology, the thermal infrared data provided by the commonly used thermal infrared sensors usually have a contradiction between spatial resolution and temporal resolution, and it is impossible to simultaneously obtain high spatial resolution surface temperature data on a single satellite or sensor. At the same time, the influence of thermal mixing effects and cloud cover further reduces the available thermal infrared images. In order to obtain high-resolution spatiotemporal data, it is necessary to use selected methods to process thermal infrared data, and surface temperature downscaling is one of the effective and commonly used methods.

[0004] Although researchers have tried to improve the accuracy and reliability of LST downscaling algorithms through various statistical regression methods, from linear to nonlinear, from global to local, there are still some issues that need further study. The most critical issue is that these statistical downscaling algorithms mainly solve the spatial non-stationarity between surface temperature and scale factor, without considering the spatial autocorrelation between surface temperature and surface temperature, and between surface temperature and scale factor. Summary of the invention

[0005] The present invention aims to solve the above problems of the prior art. A surface temperature downscaling method based on GWDM is proposed. The technical solution of the present invention is as follows:

[0006] A surface temperature downscaling method based on GWDM, comprising the following steps:

[0007] The spatial Durbin model SDM is introduced into the geographically weighted regression, and the spatial non-stationarity between regions and the spatial autocorrelation between surface temperature data and between surface temperature and scale factors are considered at the same time, which is used to describe the relationship between surface temperature and scale factors; the geographically weighted Durbin model GWDM is an extension of the geographically weighted regression model GWR. The spatial Durbin model SDM is introduced into the geographically weighted regression model GWR to describe the geographical non-stationarity of geographical variables and the spatial autocorrelation between surface temperature, as well as the spatial autocorrelation between surface temperature and scale factors; spatial autocorrelation means that geographical variables may show spatial dependence effects, there is heat transfer between airflow and the surface, which causes thermal movement, LST will affect each other in space, and there may be autocorrelation, and the weight matrix is ​​calculated based on the geographical space distance between two observation points.

[0008] Furthermore, the GWDM model is expressed as follows:

[0009]

[0010] In the formula, (u i ,v i ) represents the location information of point i in geographic space coordinates; y i and x ik is the dependent variable at point i and the kth independent variable of the GWDM method; β0(u i ,v i ) and β k (u i ,v i ) are the independent variables x k The intercept and local regression coefficient at point i; ρ(u i ,v i ) is the autoregression coefficient of the dependent variable; is the autoregression coefficient of the kth independent variable; ε i is the regression residual of point i; is the spatial adjacency matrix of size (u×v)×(u×v); Y is the dependent variable matrix; n is the total number of pixels.

[0011] The calculation method is as follows:

[0012]

[0013] Where W ij represents the adjacency relationship between the dependent variables of point i and point j, 1 represents adjacent, 0 represents non-adjacent, assuming that point i is not adjacent to itself, b is the bandwidth, d ij is the spatial distance between point i and point j.

[0014] make β, ρ, and Represent the regression coefficient, the autoregression coefficient of the dependent variable and the autoregression coefficient of the independent variable respectively; where δ(u i ,v i ) is estimated as δ(u i ,v i ) represents the coefficient matrix, which is estimated as follows:

[0015]

[0016] In the formula, represents the variable matrix, is the estimate of Z, Y is the dependent variable matrix, and X is the independent variable matrix, which is expressed as follows:

[0017] Where n = (u × v);

[0018] W(u i ,v i ) is a weight matrix, whose diagonal elements represent the weights provided by other points to a certain point, and its size is (u×v)×(u×v), which is expressed as follows:

[0019]

[0020] The improved GWAR model optimizes the calculation method of W. The original calculation of W is:

[0021]

[0022] The optimal bandwidth is obtained by the cross-validation method of local regression analysis. The optimal bandwidth can be obtained by changing the value of b to minimize CV, which is expressed as follows:

[0023]

[0024] In the downscaling algorithm, it is necessary to select appropriate explanatory variables.

[0025] Furthermore, NDVI, NDBI and NDWI were selected as scale factors.

[0026]

[0027] b NIR Indicates the near infrared band band5 of Landsat8TIRS, Indicates the second shortwave infrared band band6 of Landsat8TIRS.

[0028]

[0029] Indicates the first short-wave infrared band band6 of Landsat8TIRS, b NIR Represents the near-infrared band band5 of Landsat8TIRS.

[0030]

[0031] Where b NIR Indicates the near-infrared band band5 of Landsat8TIRS, b Red Represents the red band band4 of Landsat8TIRS.

[0032] Furthermore, the introduction of the spatial Durbin model (SDM) into the geographically weighted regression also includes the following surface temperature downscaling steps:

[0033] (1) Scale factor calculation and upscaling steps;

[0034] (2) Steps for establishing regression relationships at low spatial resolution;

[0035] (3) Regression coefficient and residual interpolation steps;

[0036] (4) Steps for calculating surface temperature at high spatial resolution.

[0037] Furthermore, the step (2) of establishing a regression relationship at low spatial resolution specifically includes:

[0038] The average surface temperature of each land cover type in the study area was calculated using the Moderate Resolution Imaging Spectroradiometer (MODIS) land cover data, and a downscaling algorithm based on GWDM was constructed to establish the regression relationship between LST and scale factor at 1000 m, and the estimated values ​​of regression coefficients and residuals were obtained, which are expressed as follows:

[0039]

[0040] Among them, LST i 1000 Indicates the surface temperature with a spatial resolution of 1000 meters, that is, MODIS LST; NDVI i 1000 NDWI i 1000 NDBI i 1000 is the NDVI, NDWI, and NDBI value of point i at a spatial resolution of 1000 meters; β0 1000 , β1 1000 , β2 1000 and ρ 1000 is the regression coefficient of point i to be estimated; The autoregression coefficient of point i to be estimated; εi 1000 is the residual at point i.

[0041] Furthermore, the (3) regression coefficient and residual interpolation step specifically includes: interpolating the estimated regression coefficient and residual from 1000 meters to 100 meters by ordinary Kriging interpolation method.

[0042] Furthermore, the (4) calculation of surface temperature at high spatial resolution specifically includes:

[0043] Based on scale invariance, the regression relationship between LST and scale factor at 1000 meters is applied to 100 meters, and the 100-meter LST is calculated, which is expressed as follows:

[0044]

[0045] Among them, LST i 100 Indicates the surface temperature at 100 meters, i.e. the downscaling result, NDVI i 100 NDWI i 100 NDBI i 100 is the NDVI, NDWI, and NDBI value at point i at 100 meters, β0 100 , β1 000 , β2 100 and ρ 100 is the regression coefficient at point i, is the autoregressive coefficient at point i, ε i 100 is the residual at point i.

[0046] The advantages and beneficial effects of the present invention are as follows:

[0047] This paper proposes a new downscaling algorithm for land surface temperature based on the GWDM model, selects NDVI, NDWI, and NDBI as scale factors, and improves the spatial resolution of MODIS LST from 1000 meters to 100 meters. The algorithm introduces the spatial Durbin model to express the autocorrelation between land surface temperature and land surface temperature, and land surface temperature and scale factor, which can better describe the relationship between LST and scale factor.

[0048] The main innovation of the present invention is to consider the spatial heterogeneity of the study area and the spatial autocorrelation of surface temperature and scale factors, and propose a new surface temperature downscaling model - Geographically Weighted Durbin Model (GWDM). The normalized water index NDWI, the normalized building index NDBI and the normalized vegetation index NDVI are selected as explanatory variables, and the spatial resolution of the MODIS temperature product is downscaled to a finer resolution, so as to obtain high-precision and high-resolution surface temperature in cloudless conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 The present invention provides a flow chart of a preferred embodiment of a surface temperature downscaling method based on GWDM. DETAILED DESCRIPTION

[0050] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0051] The technical solution of the present invention to solve the above technical problems is:

[0052] like Figure 1 As shown, the present invention proposes a new surface temperature downscaling algorithm based on the geographically weighted Durbin model (GWDM), introduces the spatial Durbin model (SDM) into the geographically weighted regression, and considers the spatial non-stationarity between regions and the spatial autocorrelation between surface temperature data and between surface temperature and scale factor, which can better describe the relationship between surface temperature and scale factor.

[0053] The GWDM model is an extension of the GWR model. It is proposed to introduce SDM into the GWR model, which can well describe the geographic non-stationarity of geographic variables and the spatial autocorrelation between surface temperature and the spatial autocorrelation between surface temperature and scale factors. Spatial autocorrelation means that geographic variables may show spatial dependence effects. There is heat transfer between airflow and the surface, which causes thermal movement. LST will affect each other in space and may have autocorrelation. The weight matrix is ​​calculated based on the geographic spatial distance between two observation points. The general form of the GWDM model is expressed as follows:

[0054]

[0055] In the formula, (u i ,v i ) represents the location information of point i in geographic space coordinates; y i and x ik is the dependent variable at point i and the kth independent variable of the GWDM method; β0(ui ,v i ) and β k (u i ,v i ) are the independent variables x k The intercept and local regression coefficient at point i; ρ(u i ,v i ) is the autoregression coefficient of the dependent variable; is the autoregression coefficient of the kth independent variable; ε i is the regression residual of point i; is the spatial adjacency matrix of size (u×v)×(u×v); Y is the dependent variable matrix; n is the total number of pixels.

[0056] The calculation method is as follows:

[0057]

[0058] Where W ij Indicates the adjacency relationship between the dependent variables of point i and point j. 1 represents adjacent and 0 represents non-adjacent. It is assumed that point i is not adjacent to itself. b is the bandwidth. d ij is the spatial distance between point i and point j.

[0059] GWDM uses weighted least squares method to estimate parameters. β, ρ, and Represent the regression coefficient, the autoregression coefficient of the dependent variable and the autoregression coefficient of the independent variable respectively; where δ(u i ,v i ) is estimated as δ(u i ,v i ) represents the coefficient matrix, which is estimated as follows:

[0060]

[0061] In the formula, represents the variable matrix, is the estimate of Z, Y is the dependent variable matrix, and X is the independent variable matrix, which is expressed as follows:

[0062] Where n = (u × v);

[0063] W(u i ,v i ) is a weight matrix, whose diagonal elements represent the weights provided by other points to a certain point, and its size is (u×v)×(u×v), which is expressed as follows:

[0064]

[0065] The improved GWAR model proposed in this paper mainly optimizes the calculation method of W. The original calculation of W is:

[0066]

[0067] The weight function calculates the weight based on the geographic spatial distance between two points. The first law of geography shows that in the spatial relationship of geographic variables, the observation data closer to point i has a closer relationship with point i than the data farther away from point i. When applied to parameter estimation, the observation data closer to point i has a greater impact on the estimation of the parameter of point i than the data farther away from point i. Therefore, the observation data closer to point i is weighted more than the data farther away from point i. The weight given to observation points at different locations varies with their distance from point i. Specifically for LST, due to the influence of the surface cover type, the temperature data located in different land feature categories will have different degrees of difference. The observation data in the land feature category with a smaller temperature difference with point i may have a closer relationship with point i than the land feature category with a larger temperature difference. Accordingly, the former observation data has a greater impact on the estimation of the regression coefficient of point i than the latter. Therefore, for LST data, not only the distance difference, but also the temperature difference between the observation point and point i due to the difference in surface cover type should be used to consider the weight change.

[0068] The optimal bandwidth is obtained by the cross-validation method of local regression analysis. The optimal bandwidth can be obtained by changing the value of b to minimize CV, which is expressed as follows:

[0069]

[0070] In the downscaling algorithm, it is necessary to select appropriate explanatory variables. There are many scaling factors that can be used for LST downscaling, among which NDVI and NDBI are the most widely used and easiest to obtain. Using these factors to establish a downscaling model can improve the adaptability of the model to different thermal data. In addition, the relationship between NDVI and LST in water areas is different from that in other areas, and NDWI is more sensitive to water bodies. NDBI is an urban impervious index that can quantitatively describe the intensity of urban areas. The NDBI value has a high consistency with LST in spatial distribution. In addition, the relationship between NDBI and LST is less affected by seasons, and can be used to compensate for the error NDVI and LST caused by the change in the relationship between NDBI and LST. Considering the land cover types and the above conditions in the two study areas, this paper selects NDVI, NDBI and NDWI as scaling factors.

[0071] The Geographically Weighted Durbin Model (GWDM) takes into account the spatial autocorrelation of the surface temperature data as well as the spatial heterogeneity of the surface temperature data. This paper uses the GWDM model to downscale the MODIS LST product MOD11A1 with a spatial resolution of 1000 meters using three scale factors: NDVI, NDWI, and NDBI as explanatory variables, and downscales the MOD11A1 data in Beijing and Zhangye to a spatial resolution of 100m. The specific steps are as follows:

[0072] (1) Calculation of scale factors and upscaling

[0073] Since the spatial resolution of the surface reflectance of the Landsat-8 satellite used is 30 meters, the obtained index is also 30 meters. In the experiment, each index is first upscaled to 1000 meters and 100 meters respectively.

[0074] (2) Establishing regression relationship at low spatial resolution

[0075] The average surface temperature of each land cover type in the study area was calculated using MODIS land cover data, and a downscaling algorithm based on GWDM was constructed to establish the regression relationship between LST and scale factor at 1000 m, and the estimated values ​​of regression coefficients and residuals were obtained, which are expressed as follows:

[0076]

[0077] Among them, LST i 1000 Indicates the surface temperature with a spatial resolution of 1000 meters, that is, MODIS LST; NDVI i 1000 NDWI i 1000 NDBI i 1000 is the NDVI, NDWI, and NDBI value of point i at a spatial resolution of 1000 meters; β0 1000 , β1 1000 , β2 1000 and ρ 1000 is the regression coefficient of point i to be estimated; The autoregression coefficient of point i to be estimated; ε i 1000 is the residual at point i.

[0078] (3) Regression coefficients and residual interpolation

[0079] The estimated regression coefficients and residuals were interpolated from 1000 m to 100 m using ordinary kriging interpolation.

[0080] (4) Calculation of surface temperature at high spatial resolution

[0081] Based on scale invariance, the regression relationship between LST and scale factor at 1000 meters is applied to 100 meters, and the 100-meter LST is calculated, which is expressed as follows:

[0082]

[0083] Among them, LST i 100 Indicates the surface temperature at 100 meters, i.e. the downscaling result. i 100 NDWI i 100 NDBI i 100 NDVI, NDWI, and NDBI values ​​at point i at 100 meters. β0 100 , β1 000 , β2 000 and ρ 100 is the regression coefficient of point i. i 100 is the residual at point i.

[0084] The following are preferred embodiments of the present invention, specifically:

[0085] A geographically weighted Durbin model (GWDM) takes into account the spatial autocorrelation of surface temperature data as well as the spatial heterogeneity of surface temperature data. This paper uses the GWDM model to downscale the MODIS LST product MOD11A1 with a spatial resolution of 1000 meters using three scale factors: NDVI, NDWI, and NDBI as explanatory variables, and downscales the MOD11A1 data in Beijing and Zhangye to a spatial resolution of 100m. The specific steps include:

[0086] (1) Calculation of scale factors and upscaling

[0087] Since the spatial resolution of the surface reflectance of the Landsat-8 satellite used is 30 meters, the obtained index is also 30 meters. In the experiment, each index is first upscaled to 1000 meters and 100 meters respectively.

[0088] (2) Establishing regression relationship at low spatial resolution

[0089] The average surface temperature of each land cover type in the study area was calculated using MODIS land cover data, and a downscaling algorithm based on GWDM was constructed to establish the regression relationship between LST and scale factor at 1000 m, and the estimated values ​​of regression coefficients and residuals were obtained, which are expressed as follows:

[0090]

[0091] Among them, LST i 1000 Indicates the surface temperature at 1000 meters, that is, MODIS LST. NDVI i 1000 NDWI i 1000 NDBI i 1000 is the NDVI, NDWI, and NDBI value of point i at 1000 meters. 1000 , β1 1000 , β2 1000 and ρ 1000 is the regression coefficient of point i to be estimated. i 1000 is the residual at point i.

[0092] (3) Regression coefficients and residual interpolation

[0093] The estimated regression coefficients and residuals were interpolated from 1000 m to 100 m using ordinary kriging interpolation.

[0094] (4) Calculation of surface temperature at high spatial resolution

[0095] Based on scale invariance, the regression relationship between LST and scale factor at 1000 meters is applied to 100 meters, and the 100-meter LST is calculated, which is expressed as follows:

[0096]

[0097] Among them, LST i 100 Indicates the surface temperature at 100 meters, i.e. the downscaling result, NDVI i 100 NDWI i 100 NDBI i 100 is the NDVI, NDWI, and NDBI value at point i at 100 meters, β0 100 , β1 100 , β2 100 and ρ 100 is the regression coefficient at point i, is the autoregressive coefficient at point i, ε i 100 is the residual at point i.

[0098] The present invention has the following beneficial effects:

[0099] A new downscaling algorithm for land surface temperature based on the GWDM model is proposed. NDVI, NDWI, and NDBI are selected as scale factors to increase the spatial resolution of MODIS LST from 1000 meters to 100 meters. The algorithm introduces the spatial Durbin model to express the autocorrelation between land surface temperature and land surface temperature, and land surface temperature and scale factor, which can better describe the relationship between LST and scale factor.

[0100] The main innovation of the present invention is to consider the spatial heterogeneity of the study area and the spatial autocorrelation of surface temperature and scale factors, and propose a new surface temperature downscaling model - Geographically Weighted Durbin Model (GWDM). The normalized water index NDWI, the normalized building index NDBI and the normalized vegetation index NDVI are selected as explanatory variables, and the spatial resolution of the MODIS temperature product is downscaled to a finer resolution, so as to obtain high-precision and high-resolution surface temperature in cloudless conditions.

[0101] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0102] The above embodiments should be understood to be only used to illustrate the present invention and not to limit the protection scope of the present invention. After reading the contents of the present invention, technicians can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A surface temperature downscaling method based on GWDM, characterized in that: The following steps are involved: The spatial Durbin model SDM is introduced into the geographically weighted regression, and the spatial non-stationarity between regions and the spatial autocorrelation between surface temperature data and between surface temperature and scale factors are considered at the same time, which is used to describe the relationship between surface temperature and scale factors; the geographically weighted Durbin model GWDM is an extension of the geographically weighted regression model GWR. The spatial Durbin model SDM is introduced into the geographically weighted regression model GWR to describe the geographical non-stationarity of geographical variables and the spatial autocorrelation between surface temperature, as well as the spatial autocorrelation between surface temperature and scale factors; spatial autocorrelation means that geographical variables may show spatial dependence effects, there is heat transfer between airflow and the surface, which causes thermal movement, LST will affect each other in space, and there may be autocorrelation, and the weight matrix is ​​calculated based on the geographical space distance between two observation points; The GWDM model is expressed as follows: In the formula, (u i ,v i ) represents the location information of point i in geographic space coordinates; y i and x ik is the dependent variable at point i and the kth independent variable of the GWDM method; β0(u i ,v i ) and β k (u i ,v i ) are the independent variables x k The intercept and local regression coefficient at point i; ρ(u i ,v i ) is the autoregression coefficient of the dependent variable; is the autoregression coefficient of the kth independent variable; ε i is the regression residual of point i; is the spatial adjacency matrix of size (u×v)×(u×v); Y is the dependent variable matrix; n is the total number of pixels; The calculation method is as follows: Where W ij represents the adjacency relationship between the dependent variables of point i and point j, 1 represents adjacent, 0 represents non-adjacent, assuming that point i is not adjacent to itself, b is the bandwidth, d ij is the spatial distance between point i and point j; The GWDM uses the weighted least squares method to estimate the parameters, β, ρ, and Represent the regression coefficient, the autoregression coefficient of the dependent variable and the autoregression coefficient of the independent variable respectively; where δ(u i ,v i ) is estimated as δ(u i ,v i ) represents the coefficient matrix, which is estimated as follows: In the formula, represents the variable matrix, is the estimate of Z, Y is the dependent variable matrix, and X is the independent variable matrix, which is expressed as follows: Where n = (u × v); W(u i ,v i ) is a weight matrix, whose diagonal elements represent the weights provided by other points to a certain point, and its size is (u×v)×(u×v), which is expressed as follows: Where W ij The calculation is: The optimal bandwidth is obtained by the cross-validation method of local regression analysis. The optimal bandwidth can be obtained by changing the value of b to minimize CV, which is expressed as follows: In the downscaling algorithm, it is necessary to select appropriate explanatory variables; The introduction of the spatial Durbin model (SDM) into the geographically weighted regression also includes the following surface temperature downscaling steps: (1) Scale factor calculation and upscaling steps; (2) Steps for establishing regression relationships at low spatial resolution; (3) Regression coefficient and residual interpolation steps; (4) Calculation steps of surface temperature at high spatial resolution; The step (2) of establishing a regression relationship at low spatial resolution specifically includes: The average surface temperature of each land cover type in the study area was calculated using the land cover data of the Moderate Resolution Imaging Spectroradiometer (MODIS). A downscaling algorithm based on GWDM was constructed to establish the regression relationship between LST and the scale factor at 1000 m, and the estimated values ​​of the regression coefficient and residual were obtained, which are expressed as follows: Among them, LST i 1000 Indicates the surface temperature with a spatial resolution of 1000 meters, that is, MODIS LST; NDVI i 1000 NDWI i 1000 NDBI i 1000 is the NDVI, NDWI, and NDBI value of point i at a spatial resolution of 1000 meters; β0 1000 , β1 1000 , β2 1000 and ρ 1000 is the regression coefficient of point i to be estimated; The autoregression coefficient of point i to be estimated; ε i 1000 is the residual of point i; The (4) calculation of land surface temperature at high spatial resolution specifically includes: Based on scale invariance, the regression relationship between LST and scale factor at 1000 meters is applied to 100 meters, and the 100-meter LST is calculated, which is expressed as follows: Among them, LST i 100 Indicates the surface temperature at 100 meters, i.e. the downscaling result, NDVI i 100 NDWI i 100 NDBI i 100 is the NDVI, NDWI, and NDBI value at point i at 100 meters, β0 100 , β1 100 , β2 100 and ρ 100 is the regression coefficient at point i, is the autoregressive coefficient at point i, ε i 100 is the residual at point i.

2. A surface temperature downscaling method based on GWDM according to claim 1, characterized in that: NDVI, NDBI and NDWI were selected as scale factors; b NIR Indicates the near infrared band band5 of Landsat8TIRS, Indicates the second short-wave infrared band band6 of Landsat8TIRS; Indicates the first short-wave infrared band band6 of Landsat8TIRS, b NIR Indicates the near-infrared band band5 of Landsat8TIRS; Where b NIR Indicates the near-infrared band band5 of Landsat8TIRS, b Red Represents the red band band4 of Landsat8TIRS.

3. The surface temperature downscaling method based on GWDM according to claim 1, characterized in that: The (3) regression coefficient, autoregressive coefficient and residual interpolation step specifically includes: interpolating the estimated regression coefficient, autoregressive coefficient and residual from 1000 meters to 100 meters by ordinary Kriging interpolation method.

Images