CN114387531A - Ground surface temperature downscaling method based on improved geographic weighted regression model - Google Patents

Ground surface temperature downscaling method based on improved geographic weighted regression model Download PDF

Info

Publication number
CN114387531A
CN114387531A CN202210062674.9A CN202210062674A CN114387531A CN 114387531 A CN114387531 A CN 114387531A CN 202210062674 A CN202210062674 A CN 202210062674A CN 114387531 A CN114387531 A CN 114387531A
Authority
CN
China
Prior art keywords
surface temperature
spatial resolution
data
regression model
regression
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210062674.9A
Other languages
Chinese (zh)
Inventor
罗小波
王植
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing University of Post and Telecommunications
Original Assignee
Chongqing University of Post and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Chongqing University of Post and Telecommunications filed Critical Chongqing University of Post and Telecommunications
Priority to CN202210062674.9A priority Critical patent/CN114387531A/en
Publication of CN114387531A publication Critical patent/CN114387531A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
    • G06T17/05—Geographic models

Abstract

The invention belongs to the technical field of remote sensing image enhancement, and relates to a ground surface temperature downscaling method based on an improved geographical weighted regression model, which comprises the steps of obtaining remote sensing data for preprocessing, and calculating a scale factor and a temperature difference coefficient according to the preprocessed remote sensing data; an improved geographical weighting regression model is built by utilizing the temperature difference coefficient, and a regression relation between the subsurface temperature and the scale factor under low spatial resolution is built to obtain a regression coefficient and a residual error; improving the low spatial resolution of the regression coefficient and the residual error to the target high spatial resolution by common kriging interpolation; the scale is reduced, an improved geographical weighting regression model is input, the earth surface temperature value under high spatial resolution is obtained according to the regression relation between the earth surface temperature and the scale factor, the temperature conditions under different earth surface coverage types are added to represent the difference between different earth surface types, the weight matrix in the model is optimized through the temperature difference coefficients of the different earth surface coverage types, and the precision of the scale reduction result is improved.

Description

Ground surface temperature downscaling method based on improved geographic weighted regression model
Technical Field
The invention belongs to the technical field of remote sensing image enhancement, and relates to a ground surface temperature downscaling method based on an improved geographical weighted regression model.
Background
Surface Temperature (LST) is one of the key parameters affecting environmental ecology and climate systems, and is of great importance in applications such as Surface energy balance analysis, wildfire detection, and research on climate change of different spatial scales. Since the surface temperature is closely related to the urban heat island, the method is also widely used for researching the urban heat island effect. The surface temperature and its spatial distribution differences can be detected dynamically by thermal infrared remote sensing. Many satellite sensors can provide thermal infrared spectroscopy data, including Landsat TM/ETM + and Terra ASTER, among others, from which many publicly available surface temperature products are derived. The polar orbit satellite earth surface temperature product has good radiation precision and medium spatial resolution, but the acquisition time interval is longer. Geostationary satellites provide higher frequency detection, but with lower spatial resolution, typically 3 or 4 kilometers. For a single satellite sensor, there are inherent conflicts in the temporal and spatial resolution of the acquired surface temperature data due to satellite loading and sensor technology limitations.
The current inability of a single satellite sensor to provide both high spatial and temporal resolution thermal infrared data limits the potential applications of the earth's surface temperature obtained thereby in various fields. Various solutions have been proposed to this problem, where spatial downscaling is currently widely used. From the perspective of scale factors, the research is focused on researching physical factors influencing LST change, and the consistency of heat radiation information before and after the scale reduction is ensured. The method is to improve the spatial resolution of the temperature image by combining the low spatial resolution earth surface temperature and auxiliary data with high spatial resolution on the premise of assuming the scale invariance. These auxiliary data may be physical or ecological parameters that affect changes in the surface temperature. The MODIS LST was first scaled using the downscaling Distrad algorithm with normalized vegetation index (NDVI) as proposed in Kustas W P, Norman J M, Anderson M C, et al. The literature [ hanging W, Yang Y, Yue X, et al. A spatial down scaling for satellite-based prediction over the Tibet plan based on NDVI, DEM, and land surface temperature [ J ]. removal Sensing,2016, 8(8):655 ] uses Digital Elevation Model (DEM) data for down-scaling studies and achieves better results on a variety of down-scaling algorithms. The downscaling model includes a global regression model and a local regression model. The global regression model comprises a DisTrad algorithm, a TsHARP algorithm and the like. For the study of local regression models, the first time the geo-weighted regression model was applied to the surface temperature down-scaling in the literature [ Duan S B, Li Z L.spatial down scaling of MODIS land surface temporal using the geographic weighted regression [ J ]. IEEE Transactions on Geoscience and removal Sensing,2016,54(11): 6458-; the document [ Yang C, Zhan Q, Lv Y, et al, descending and surface temporal using multiscale geological information weighted regression over heterologous landsizes in Wuhan, China [ J ]. IEEE Journal of Selected topocs in Applied Earth space and Remote Sensing,2019,12(12): 5213) uses a multi-scale geographical weighted regression model to improve the spatial resolution of the surface temperature, and the model is better in a plurality of spatial resolution experiments; the document [ Peng Y, Li W, Luo X, et al. A. geographic and temporal weighted regression model for spatial down scaling of MODIS land surface temporal errors relating to [ J ]. IEEE transactions on geographic and motion sensing,2019,57(7): 5012-; the ground Temperature down-scaling algorithm is constructed by using a geographical weighting spatial Autoregressive Model in consideration of the spatial autocorrelation of ground Temperature data in documents [ Wang S, Luo X, Peng Y.spatial regression of MODIS Land Surface Temperature Based on geographic Weighted Autoregressive Model [ J ]. IEEE Journal of Selected topocs in Applied Earth temperatures and subsurface estimation, 2020, PP (99):1-1 ].
In the heterogeneous region, the local regression model considers the characteristics of the surface temperature such as spatial non-stationarity and spatial autoregressive property at different geographic positions, and improves the accuracy of the scale reduction result to a certain extent. However, past studies have found that different surface feature types exhibit different radiation temperatures, and that the target area may contain multiple complex surface coverage types. However, the current local weighting models applied to the surface temperature down-scaling only consider temporal or spatial distances to calculate weights, and do not consider the temperature differences of data points under different surface coverage types.
Disclosure of Invention
Aiming at a target area with multiple earth surface coverage types, the earth surface temperature downscaling method based on the improved geographic weighted regression model provided by the invention comprises the following steps:
s1, obtaining remote sensing data, preprocessing the remote sensing data, and calculating a scale factor and a temperature difference coefficient according to the preprocessed remote sensing data;
s2, constructing an improved geographical weighted regression model by using the temperature difference coefficient, establishing a regression relation between the surface temperature and the scale factor under low spatial resolution through the improved geographical weighted regression model, and obtaining a regression coefficient and a residual error under the low spatial resolution according to the regression relation;
s3, improving the low spatial resolution of the regression coefficient and the residual error to the target high spatial resolution by common Krigin interpolation;
and S4, reducing the scale of the preprocessed remote sensing data, inputting an improved geographical weighted regression model, and obtaining the earth surface temperature value under high spatial resolution according to the regression relation between the earth surface temperature under low spatial resolution and the scale factor.
Furthermore, the obtained remote sensing data comprise MODIS earth surface temperature data, MODIS land coverage data and Landsat8 earth surface reflectivity data, and the remote sensing data are preprocessed by adopting geometric cutting and reprojection.
Further, calculating scale factors by using Landsat8 earth surface reflectivity data, wherein the scale factors comprise normalized vegetation index NDVI, normalized water body index NDWI and normalized building index NDBI, and the calculation mode is as follows:
Figure BDA0003478889690000041
Figure BDA0003478889690000042
Figure BDA0003478889690000043
wherein R isRED，RNIR，RSWIR1，RGREENThe earth surface reflectivities of the Landsat8 red wave band, the near infrared wave band, the first short wave infrared wave band and the green wave band are respectively.
Furthermore, data points under each ground cover type are counted according to MODIS land cover data, then the average ground surface temperature, the highest average ground surface temperature and the lowest average ground surface temperature of each ground cover type are calculated according to MODIS ground surface temperature data, and therefore temperature difference coefficients of different ground cover types are calculated, and the calculation mode is as follows:
Figure BDA0003478889690000044
wherein the content of the first and second substances,
Figure BDA0003478889690000045
and
Figure BDA0003478889690000046
respectively representing the average surface temperatures of the types of surface coverage in which data point i and data point j are located,
Figure BDA0003478889690000047
and
Figure BDA0003478889690000048
highest average surface respectively representing all surface coverage types in MODIS surface temperature dataTemperature and lowest average surface temperature.
Further, the improved geographical weighted regression model is constructed as follows:
Figure BDA0003478889690000049
wherein (u)i,vi) Representing the spatial coordinates of data points i, n being the number of scale factors, yiIs the observed value of the surface temperature at data point i, xikIs the observed value, β, of the k-th scale factor at data point i0(ui,vi) And betak(ui,vi) Is the regression parameter, beta, to be estimatedk(ui,vi) Is the kth parameter value, ρ (u) for data point ii,vi) Is the autoregressive parameter of the data point i, Y is the surface temperature vector, εiIs the residual error of the data point i,
Figure BDA00034788896900000410
is a spatial adjacency matrix having a size of (u × v) × (u × v),
Figure BDA00034788896900000411
representing the adjacency of the dependent variable of data point i and data point j.
Further, the calculation formula of all regression parameters in the improved geo-weighted regression model is as follows:
Figure BDA00034788896900000412
Figure BDA0003478889690000051
where Y is the surface temperature vector, X is the scale factor matrix, W (u)i,vi) Is a weight matrix of size (u x v) x (u x v),
Figure BDA0003478889690000052
is a spatial adjacency matrix having a size of (u × v) × (u × v),
Figure BDA0003478889690000053
are estimates of all regression parameters.
Further, the calculation method of the weight function is optimized and expressed as follows:
Figure BDA0003478889690000054
μijis based on the temperature difference coefficient of different earth surface coverage types, b is the bandwidth, (u)i,vi) Representing the spatial coordinates of data point i.
Further, the improved geographical weighted regression model establishes a regression relationship between the earth surface temperature and the scale factor at a low spatial resolution as follows:
Figure BDA0003478889690000055
wherein, LSTi LThe surface temperature representing data point i at low spatial resolution, i.e. the MODIS surface temperature,
Figure BDA0003478889690000056
and
Figure BDA0003478889690000057
respectively are a normalized vegetation index value, a normalized water index value and a normalized building index value of a data point i under low spatial resolution,
Figure BDA0003478889690000058
and ρLIs the regression coefficient, LST, of the data point i to be estimatedLFor the surface temperature vector at low spatial resolution,
Figure BDA0003478889690000059
is a number ofFrom the residual error at the point i, it is,
Figure BDA00034788896900000510
is a spatially contiguous matrix.
Further, according to the regression coefficient and the estimated value of the residual error calculated in the regression relationship, calculating the earth surface temperature value under the high spatial resolution, wherein the calculation formula is as follows:
Figure BDA00034788896900000511
wherein, LSTi HRepresenting the high spatial resolution surface temperature, i.e. the downscaling result,
Figure BDA00034788896900000512
and
Figure BDA00034788896900000513
normalized vegetation index value, normalized water index value and normalized building index value of data point i under high spatial resolution, LSTHSurface temperature vector, BETA, at low spatial resolution0 H、β1 H、β2 H、β3 HAnd ρHIs the regression coefficient, ε, of the data points ii HThe residual error of data point i.
The invention has the beneficial effects that:
according to the remote sensing earth surface temperature data downscaling method based on the improved geographic weighted regression model, the temperature conditions under different earth surface coverage types are added to quantitatively represent the difference between different earth surface types, and the calculation of the weight matrix in the temperature difference coefficient optimization model of the different earth surface coverage types enables the calculation of the weight value to simultaneously consider two factors of the spatial distance and the earth surface coverage type difference, so that the accuracy of the downscaling result is improved.
The method disclosed by the invention has the advantages that the ground surface temperature is subjected to scale reduction based on the scale factor, the consistency of the heat radiation information in the process is effectively ensured, meanwhile, the optimization is carried out based on the temperature difference of different ground surface coverage types, the effect of a target area with a mixture of multiple ground surface coverage types is better than that of a classical method, and the high-spatial-resolution ground surface temperature can be provided for urban heat islands, ground surface energy balance analysis and other researches.
Drawings
FIG. 1 is a general flow diagram of a method for reducing the surface temperature scale based on an improved geoweighted regression model according to the present invention;
FIG. 2 is a schematic diagram illustrating an implementation of an improved geo-weighted regression model based surface temperature downscaling method in accordance with an embodiment of the present invention;
FIG. 3 is a graph of the 100 meter spatial resolution down-scaling results of an embodiment of the improved geoweighted regression model based surface temperature down-scaling method and three classical methods in accordance with the present invention;
fig. 4 is a graph of the 100-meter spatial resolution down-scaling result and the Landsat8 earth surface temperature absolute error based on an embodiment of the improved geo-weighted regression model-based earth surface temperature down-scaling method and three classical methods according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
A method for reducing the scale of the earth surface temperature based on an improved geo-weighted regression model, as shown in fig. 1, comprises the following steps:
s1, obtaining remote sensing data, preprocessing the remote sensing data, and calculating a scale factor and a temperature difference coefficient according to the preprocessed remote sensing data;
s2, constructing an improved geographical weighted regression model by using the temperature difference coefficient, establishing a regression relation between the surface temperature and the scale factor under low spatial resolution through the improved geographical weighted regression model, and obtaining a regression coefficient and a residual error under the low spatial resolution according to the regression relation;
s3, improving the low spatial resolution of the regression coefficient and the residual error to the target high spatial resolution by common Krigin interpolation;
and S4, reducing the scale of the preprocessed remote sensing data, inputting an improved geographical weighted regression model, and obtaining the earth surface temperature value under high spatial resolution according to the regression relation between the earth surface temperature under low spatial resolution and the scale factor.
In an embodiment, as shown in fig. 2, a surface temperature of 100 meters on the day is estimated by obtaining a relevant data set of 2017, 9,12 and showing a remote sensing surface temperature data scale reduction method based on an improved geography weighted regression model (IGWAR model), where the embodiment takes beijing area as an example, 1000 meters of MODIS surface temperature data are reduced to 100 meters.
Preferably, collecting remote sensing data of a Beijing research area, wherein the obtained remote sensing data comprises MODIS surface temperature data, MODIS land cover data and Landsat8 surface reflectivity data, the MODIS surface temperature data is from a temperature product MOD11A1, and the spatial resolution is 1000 meters; MODIS land cover data is from MCD12Q1 with a spatial resolution of 500 meters; landsat8 surface reflectivity data with a spatial resolution of 30 meters. The MODIS surface temperature data needs to convert an LST _ Day _1km data set in a product into surface temperature in units of centigrade, and the calculation method is as follows:
TLST＝data·scale_factor-273.15；
wherein, scale _ factor is a conversion coefficient, and data is an LST _ Day _1km data set; and preprocessing the acquired remote sensing data, including geometric cutting, reprojection and other methods.
And calculating related parameters by adopting the preprocessed remote sensing data.
Preferably, calculating a scale factor by utilizing Landsat8 surface reflectivity data; the scale factors comprise normalized vegetation index NDVI, normalized water body index NDWI and normalized building index NDBI, and the calculation mode is as follows:
Figure BDA0003478889690000081
Figure BDA0003478889690000082
Figure BDA0003478889690000083
in RRED，RNIR，RSWIR1，RGREENThe earth surface reflectivities of a Landsat8 red band (band 4), a near infrared band (band 5), a first short wave infrared band (band 6) and a green band (band 3) respectively. Since the spatial resolution of the surface reflectivity of Landsat8 used in this embodiment is 30 meters, the calculated index spatial resolution is also 30 meters, and therefore, in this embodiment, the scale factors need to be scaled up to 1000 meters and 100 meters, respectively.
Preferably, since the spatial resolution of the MODIS land cover data is 500 meters, the MODIS land cover data MCD12Q1 is firstly scaled up to the spatial resolution of 1000 meters, then data points under each land cover type are counted, then the average earth surface temperature, the highest average earth surface temperature and the lowest average earth surface temperature of each land cover type are calculated according to the MODIS LST data classification, and therefore the temperature difference coefficient mu of different land cover types is calculatedijThe calculation method is as follows:
Figure BDA0003478889690000084
wherein the content of the first and second substances,
Figure BDA0003478889690000085
and
Figure BDA0003478889690000086
respectively representing the average surface temperatures of the types of surface coverage in which data point i and data point j are located,
Figure BDA0003478889690000087
and
Figure BDA0003478889690000088
respectively, the highest average surface temperature and the lowest average surface temperature for all surface coverage types in the MODIS surface temperature data.
An IGWAR model is constructed by utilizing the temperature difference coefficient, a regression relation between the surface temperature and the scale factor under low spatial resolution is established, and estimation values of the regression coefficient and the residual error under the low spatial resolution are obtained, wherein the low spatial resolution is the corresponding MODIS surface temperature spatial resolution which is 1000 meters;
specifically, the improved geo-weighted regression model constructed is represented as:
Figure BDA0003478889690000091
wherein (u)i,vi) Representing the spatial coordinates of data points i, n being the number of scale factors, yiIs the observed value of the surface temperature at data point i, xikIs the observed value, β, of the k-th scale factor at data point i0(ui,vi) And betak(ui,vi) Is the regression parameter, beta, to be estimatedk(ui,vi) Is the kth parameter value, ρ (u) for data point ii,vi) Is the autoregressive parameter of the data point i, Y is the surface temperature vector, εiIs the residual error of the data point i,
Figure BDA0003478889690000092
is a spatial adjacency matrix having a size of (u × v) × (u × v),
Figure BDA0003478889690000093
representing the adjacency of the dependent variable of data point i and data point j.
Figure BDA0003478889690000094
Expressed as:
Figure BDA0003478889690000095
where 1 represents adjacent, 0 represents non-adjacent, and the set data point i is not adjacent to itself.
Let δ (u)i,vi)＝(β0(ui,vi),βk(ui,vi),ρ(ui,vi))T，δ(ui,vi) Is expressed as
Figure BDA0003478889690000096
The improved geographical weighted regression model has the calculation formula of the regression coefficient as follows:
Figure BDA0003478889690000097
Figure BDA0003478889690000098
where Y is the surface temperature vector, X is the scale factor matrix, W (u)i,vi) Is a weight matrix of size (u × v) × (u × v), whose diagonal elements represent the weights provided by other points to a point, expressed as:
Figure BDA0003478889690000099
the calculation method of the weight function is optimized and expressed as:
Figure BDA00034788896900000910
preferably, the improved geo-weighted regression model establishes a regression relationship of the earth surface temperature and the scale factor at a low spatial resolution represented by:
Figure BDA0003478889690000101
wherein, LSTi 1000Represents the 1000 meter surface temperature of data point i, i.e., the MODIS surface temperature. NDVIi 1000、NDWIi 1000And NDBIi 1000NDVI, NDWI and NDBI values, BETA, of data point i at 1000 m0 1000、β1 1000、β2 1000、β3 1000And ρ1000Are the different regression coefficients of the data point i, the four regression coefficients need to be estimated by the regression relationship, LST1000Is a surface temperature vector of 1000 m, epsiloni 1000The residual error of data point i.
Preferably, the spatial resolution of the regression coefficients and residuals are increased to the target high spatial resolution by ordinary kriging interpolation. In this embodiment, a tif image of a regression coefficient and a residual is introduced into ArcGis software, and is converted into shp data by transformation. Selecting shp data from a kriging interpolation plug-in unit of ArcGis, and selecting common kriging interpolation to obtain a high-spatial-resolution regression coefficient and a residual tif image.
Based on the scale invariance, applying the regression relationship between the surface temperature and the scale factor under the spatial resolution of 1000 meters to the spatial resolution of 100 meters to obtain the surface temperature value under the spatial resolution of 100 meters, which is expressed as:
Figure BDA0003478889690000102
wherein, LSTi 100Representing the surface temperature of 100 meters, i.e. the downscaling result, NDVIi 100、NDWIi 100And NDBIi 100NDVI, NDWI and NDBI values, BETA, of data point i at 100 m0 100、β1 000And beta2 000And ρ100Is the regression coefficient, ε, of the data points ii 100The residual error of data point i.
Figure 3 shows a comparison of the downscaling results obtained according to the invention for a spatial resolution of 100 meters and obtained using an algorithm based on the TsHARP, GWR and GWAR models. Wherein, (a) is MODIS surface temperature, (b) is Landsat8 surface temperature, (c) is downscaling result obtained by using TsHARP algorithm, (d) is downscaling result obtained by using GWR model-based algorithm, (e) is downscaling result obtained by using GWIR model-based algorithm, and (f) is downscaling result obtained by using IGWAR model-based algorithm. Specifically, the downscaling result of the 100-meter spatial resolution acquired by the method has similar spatial texture and distribution details compared with the actual Landsat8 ground surface temperature of the 100-meter spatial resolution, the displayed mountain area has clearer veins, and the low-temperature area and the high-temperature area are more accurate in place; the downscaling results obtained using algorithms based on TsHARP, GWR and GGAR models are better, but still inferior to the present invention. The downscaling result obtained by using the TsHARP algorithm has a rectangular block effect, and the result obtained by using the GWR algorithm has a smoothing effect.
Figure 4 further shows a plot of the downscaling results obtained according to the present invention for a spatial resolution of 100 meters and the absolute error of the downscaling results obtained using an algorithm based on the TsHARP, GWR and GWAR models. Wherein, (a) is an absolute error map of a downscaling result obtained by using a TsHARP algorithm, (b) is an absolute error map of a downscaling result obtained by using an algorithm based on a GWR model, (c) is an absolute error map of a downscaling result obtained by using an algorithm based on a GWIR model, and (d) is an absolute error map of a downscaling result obtained by using an algorithm based on an IGWAR model. As can be seen from the figure, the overall spatial distribution of the absolute errors of the downscaling result obtained by the IGWAR algorithm provided by the invention is relatively uniform, the high error part is less, and the overall error is the lowest. The overall error of the downscaling result obtained by the GWIR and GWR algorithms is higher. The high error part of the downscaling result obtained by the TsHARP algorithm is more, and a block effect exists.
Table 1 further compares the downscaling results obtained according to the present invention for a spatial resolution of 100 meters with downscaling results obtained using algorithms based on TsHARP, GWR and GWAR models, using Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). The downscaling results with a resolution of 100 meters obtained with the present invention have the best accuracy compared to the other three methods, with an RMSE of 1.26 ℃ and an MAE of 0.91 ℃.
TABLE 1 MAE and RMSE results for different algorithms
Figure BDA0003478889690000111
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (9)

1. An improved geography weighted regression model-based surface temperature downscaling method comprises the following steps:
s1, obtaining remote sensing data, preprocessing the remote sensing data, and calculating a scale factor and a temperature difference coefficient according to the preprocessed remote sensing data;
s2, constructing an improved geographical weighted regression model by using the temperature difference coefficient, establishing a regression relation between the surface temperature and the scale factor under low spatial resolution through the improved geographical weighted regression model, and obtaining a regression coefficient and a residual error under the low spatial resolution according to the regression relation;
s3, improving the low spatial resolution of the regression coefficient and the residual error to the target high spatial resolution by common Krigin interpolation;
and S4, reducing the scale of the preprocessed remote sensing data, inputting an improved geographical weighted regression model, and obtaining the earth surface temperature value under high spatial resolution according to the regression relation between the earth surface temperature under low spatial resolution and the scale factor.
2. The earth surface temperature downscaling method based on the improved geographical weighted regression model according to claim 1, wherein the obtained remote sensing data comprise MODIS earth surface temperature data, MODIS land cover data and Landsat8 earth surface reflectivity data, and the remote sensing data are preprocessed by geometric cutting and reprojection.
3. The improved geography weighted regression model-based ground surface temperature downscaling method of claim 2, wherein Landsat8 ground surface reflectivity data is used for calculating scale factors, the scale factors include normalized vegetation index NDVI, normalized water body index NDWI and normalized building index NDBI, and the calculation method includes:
Figure FDA0003478889680000011
Figure FDA0003478889680000012
Figure FDA0003478889680000013
wherein R isRED，RNIR，RSWIR1，RGREENThe earth surface reflectivities of the Landsat8 red wave band, the near infrared wave band, the first short wave infrared wave band and the green wave band are respectively.
4. The earth surface temperature de-scaling method based on the improved geographical weighted regression model as claimed in claim 2, wherein data points under each earth surface coverage type are counted according to MODIS land coverage data, and then the average earth surface temperature, the highest average earth surface temperature and the lowest average earth surface temperature of each earth surface coverage type are calculated according to MODIS earth surface temperature data, so as to calculate the temperature difference coefficients of different earth surface coverage types, wherein the calculation method is as follows:
Figure FDA0003478889680000021
wherein the content of the first and second substances,
Figure FDA0003478889680000022
and
Figure FDA0003478889680000023
respectively representing the average surface temperatures of the types of surface coverage in which data point i and data point j are located,
Figure FDA0003478889680000024
and
Figure FDA0003478889680000025
respectively, the highest average surface temperature and the lowest average surface temperature for all surface coverage types in the MODIS surface temperature data.
5. The earth's surface temperature de-scaling method based on the improved geographical weighted regression model as claimed in claim 1, wherein the improved geographical weighted regression model is constructed by:
Figure FDA0003478889680000026
wherein (u)i,vi) Representing the spatial coordinates of data points i, n being the number of scale factors, yiIs the observed value of the surface temperature at data point i, xikIs the observed value, β, of the k-th scale factor at data point i0(ui,vi) And betak(ui,vi) Is a regression parameter, betak(ui,vi) Is the kth parameter value, ρ (u) for data point ii,vi) Is the autoregressive parameter of the data point i, Y is the surface temperature vector, εiIs the residual error of the data point i,
Figure FDA0003478889680000027
is a spatial adjacency matrix having a size of (u × v) × (u × v),
Figure FDA0003478889680000028
representing the adjacency of the dependent variable of data point i and data point j.
6. The earth's surface temperature de-scaling method based on the improved geographical weighted regression model as claimed in claim 5, wherein the calculation formula of all regression parameters in the improved geographical weighted regression model is as follows:
Figure FDA0003478889680000029
Figure FDA00034788896800000210
where Y is the surface temperature vector, X is the scale factor matrix, W (u)i,vi) Is a weight matrix of size (u x v) x (u x v),
Figure FDA0003478889680000031
is a spatial adjacency matrix having a size of (u × v) × (u × v),
Figure FDA0003478889680000032
are estimates of all regression parameters.
7. The earth's surface temperature de-scaling method based on the improved geography weighted regression model as claimed in claim 6, wherein the weight matrix is optimized, and the optimization formula is represented as:
Figure FDA0003478889680000033
μijis based on the temperature difference coefficient of different earth surface coverage types, b is the bandwidth, (u)i,vi) Representing the spatial coordinates of data point i.
8. The earth's surface temperature de-scaling method based on the improved geographical weighted regression model as claimed in claim 1, wherein the improved geographical weighted regression model establishes the regression relationship between the earth's surface temperature and the scale factor at low spatial resolution as follows:
Figure FDA0003478889680000034
wherein, LSTi LThe surface temperature representing data point i at low spatial resolution, i.e. the MODIS surface temperature,
Figure FDA0003478889680000035
and
Figure FDA0003478889680000036
respectively are a normalized vegetation index value, a normalized water index value and a normalized building index value of a data point i under low spatial resolution,
Figure FDA0003478889680000037
and ρLIs the regression coefficient, LST, of the data point iLFor the surface temperature vector at low spatial resolution,
Figure FDA0003478889680000038
is the residual error of the data point i,
Figure FDA0003478889680000039
is a spatially contiguous matrix.
9. The earth surface temperature de-scaling method based on the improved geographical weighted regression model as claimed in claim 1, wherein the earth surface temperature value at high spatial resolution is calculated according to the estimated values of the regression coefficient and the residual error calculated in the regression relationship, and the calculation formula is as follows:
Figure FDA00034788896800000310
wherein the content of the first and second substances,
Figure FDA0003478889680000041
representing the high spatial resolution surface temperature, i.e. the downscaling result,
Figure FDA0003478889680000042
and
Figure FDA0003478889680000043
normalized vegetation index value, normalized water index value and normalized building index value of data point i under high spatial resolution, LSTHSurface temperature vector, BETA, at low spatial resolution0 H、β1 H、β2 H、β3 HAnd ρHIs the regression coefficient of the data point i,
Figure FDA0003478889680000044
the residual error of data point i.
CN202210062674.9A 2022-01-19 2022-01-19 Ground surface temperature downscaling method based on improved geographic weighted regression model Pending CN114387531A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210062674.9A CN114387531A (en) 2022-01-19 2022-01-19 Ground surface temperature downscaling method based on improved geographic weighted regression model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210062674.9A CN114387531A (en) 2022-01-19 2022-01-19 Ground surface temperature downscaling method based on improved geographic weighted regression model

Publications (1)

Publication Number Publication Date
CN114387531A true CN114387531A (en) 2022-04-22

Family

ID=81202895

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210062674.9A Pending CN114387531A (en) 2022-01-19 2022-01-19 Ground surface temperature downscaling method based on improved geographic weighted regression model

Country Status (1)

Country Link
CN (1) CN114387531A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114596498A (en) * 2022-05-10 2022-06-07 湖北省地质调查院 Assignment method and system for geochemical sampling blind area and storage medium
CN117540530A (en) * 2023-10-10 2024-02-09 二十一世纪空间技术应用股份有限公司 Urban earth surface temperature downscaling method and device based on high-resolution satellite images

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114596498A (en) * 2022-05-10 2022-06-07 湖北省地质调查院 Assignment method and system for geochemical sampling blind area and storage medium
CN117540530A (en) * 2023-10-10 2024-02-09 二十一世纪空间技术应用股份有限公司 Urban earth surface temperature downscaling method and device based on high-resolution satellite images

Similar Documents

Publication Publication Date Title
Li et al. Evaluation of machine learning algorithms in spatial downscaling of MODIS land surface temperature
Son et al. Assessment of urbanization and urban heat islands in Ho Chi Minh City, Vietnam using Landsat data
Wang et al. Urban impervious surface detection from remote sensing images: A review of the methods and challenges
CN102521815B (en) Fast fusion system and fast fusion method for images
CN108268735A (en) Soil moisture NO emissions reduction method based on multi-source remote sensing satellite fused data
CN114387531A (en) Ground surface temperature downscaling method based on improved geographic weighted regression model
Song et al. An improved surface soil moisture downscaling approach over cloudy areas based on geographically weighted regression
CN103914678B (en) Abandoned land remote sensing recognition method based on texture and vegetation indexes
CN102681033A (en) Sea surface wind measurement method based on X-band marine radar
CN104698462B (en) Synthetic aperture radar Ocean Wind-field fusion method based on variation
Yoo et al. Spatial downscaling of MODIS land surface temperature: Recent research trends, challenges, and future directions
CN107688776A (en) A kind of urban water-body extracting method
CN109033543A (en) A kind of heterogeneous area's vegetation fraction estimation method, device and equipment of earth's surface
Harty et al. Intra-hour cloud index forecasting with data assimilation
Li et al. A new method for surface water extraction using multi-temporal Landsat 8 images based on maximum entropy model
Zhao et al. A robust and adaptive spatial-spectral fusion model for PlanetScope and Sentinel-2 imagery
CN111523451B (en) Method for constructing high space-time resolution NDVI data
CN112285808B (en) Method for reducing scale of APHRODITE precipitation data
CN109840539A (en) A kind of remote sensing space-time data fusion method based on plot figure spot
Chen et al. A stepwise framework for interpolating land surface temperature under cloudy conditions based on the solar-cloud-satellite geometry
CN114821095A (en) Landslide deformation analysis method based on Offset-Tracking technology
CN114202699A (en) Dynamic monitoring method for lunar change of surface water body
Jing et al. A Rigorously-Incremental Spatiotemporal Data Fusion Method for Fusing Remote Sensing Images
Wang et al. Spatiotemporal Fusion Algorithm for Single-Time Phase High Resolution Remote Sensing Image Based on Sparse Representationy.
Yang et al. Finer scale mapping with super resolved GF-4 satellite images

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination