An integrated method for spatial reconstruction and downscaling of all-sky surface temperature

Abstract

A method for spatial reconstruction and downscaling of all-sky surface temperature belongs to the field of remote sensing science and technology. It contains the following main steps: acquisition and preprocessing of multi-source remote sensing data and atmospheric driven data, spatial reconstruction of all-sky surface temperature data, and spatial downscaling of all-sky surface temperature data. The operability and universality of the present invention in practical applications will promote the popularization and application of surface temperature data in the fields of climate change, dynamic monitoring and assessment of hydrological and ecological environment, land surface radiation budget and energy balance, and further promote the development of thermal infrared remote sensing science.

Description

Technical Field

[0001] The invention belongs to the field of remote sensing science and technology, and relates to an integrated method for all-sky surface temperature spatial reconstruction and downscaling. Background Art

[0002] Land surface temperature (LST) is a key parameter in the interaction between the surface and the atmosphere. It is of great significance in the study of climate change, dynamic monitoring and assessment of hydrological and ecological environment, land surface radiation budget and energy balance. For a long time, satellite thermal infrared remote sensing has been the main way to obtain surface temperature in a large area. However, the current satellite thermal infrared remote sensing data is contradictory in spatial resolution and temporal resolution, and cannot be taken into account. For example, Landsat can provide high spatial resolution (~100 m) thermal infrared channel data, and the revisit cycle requires 16 days; MODIS can provide transit time data twice a day, with a low spatial resolution (1000m). In addition, since the thermal infrared channel is easily interfered by adverse weather conditions such as clouds and rain, it is impossible to obtain surface radiation information under cloud cover, resulting in the spatial loss of surface temperature products. Due to the widespread existence of cloud cover, the temporal resolution of satellite thermal infrared remote sensing surface temperature is reduced. With the continuous development of remote sensing science and the needs of application fields, people hope to obtain accurate, high temporal resolution (daily), and high spatial resolution (100m) surface temperature data products that are not affected by cloud cover.

[0003] In fact, low spatial resolution thermal infrared remote sensing data cannot provide an effective expression of the heterogeneity of land surface moisture conditions and energy exchange at a fine scale. In order to improve the spatial resolution of surface temperature data, statistical regression methods are often used to establish the correlation between surface temperature and subsidiary variables related to surface temperature, such as vegetation index and elevation, for spatial downscaling. In recent years, the use of machine learning algorithms (such as gradient boosting, artificial neural networks, and random forests) to establish quantitative relationships between multiple scaling factors and surface temperature has shown great advantages and has received widespread attention. However, this type of method lacks a clear physical mechanism, which in turn limits the promotion and application of surface temperature downscaling methods to complex land surfaces.

[0004] In addition, in order to overcome the difficulty of filling the spatial gaps in thermal infrared remote sensing data, there are currently two main approaches to reconstructing the surface temperature of the entire sky: (1) One is reconstruction based on effective satellite thermal infrared observation information. This approach usually uses methods such as spatiotemporal interpolation and energy balance equations to reconstruct the surface temperature by combining satellite thermal infrared remote sensing information under clear sky conditions, surface radiation energy information, and atmospheric driving data of the adjacent surface. It has the advantages of clear thermal infrared information and simple operation, and is mostly suitable for filling short-term and small-area spatial gaps in thermal infrared surface temperature. However, there are many limitations in actual use, such as the obvious systematic difference between the surface temperature of the entire sky reconstructed by spatiotemporal interpolation and the actual surface temperature of the missing pixels, the invalid reconstruction of the surface temperature when the missing pixels are surrounded by large areas of missing pixels, and the accuracy of net shortwave radiation. (2) Another approach is based on multi-source data integration. This approach specifically includes the integration of thermal infrared and passive microwave remote sensing, and the integration of thermal infrared and reanalysis data. Integration methods are often limited by the transit time of multi-source satellite remote sensing data, the coarse spatial resolution of passive microwave remote sensing and its thermal sampling depth, the large uncertainty and low spatial resolution of reanalysis data, etc.

[0005] In summary, the existing methods still have the following two shortcomings: First, the surface temperature obtained by the current full-sky surface temperature reconstruction method based on multi-source remote sensing data lacks a unified physical meaning, and is obviously insufficient in filling the gaps in long time intervals and large areas. Second, the current full-sky surface temperature reconstruction method still cannot directly obtain surface temperature data with full sky, high temporal resolution, and high spatial resolution, which makes it difficult to meet the application requirements on a fine scale, and it is difficult to promote and use it in a large area of ​​complex land surfaces, and the consistency of data product quality is difficult to guarantee. Summary of the invention

[0006] In view of the problems in the prior art, the present invention provides an integrated method for spatial reconstruction and downscaling of all-sky surface temperature.

[0007] A method for integrating spatial reconstruction and downscaling of all-sky surface temperature includes the following steps:

[0008] Step 1: Acquisition and preprocessing of multi-source remote sensing data and atmospheric driving data.

[0009] Step 2: Spatial reconstruction of all-sky surface temperature data.

[0010] Step 3: Spatial downscaling of all-sky surface temperature data.

[0011] An integrated method for spatial reconstruction and downscaling of all-sky surface temperature is proposed. Based on multi-source remote sensing data and atmospheric driving data, a comprehensive scale factor ( ), the spatial reconstruction of the surface temperature of the entire sky (clear sky and cloudy and rainy conditions) is completed by calculating the total differential of the surface radiation balance equation, using the geographically weighted regression method and combining Factors are used to perform spatial downscaling of the full sky surface temperature, and finally obtain the full sky surface temperature with high spatial resolution. ) combines all-sky surface temperature spatial reconstruction and spatial downscaling into an integrated algorithm.

[0012] The advantages of the present invention are: operability and universality in practical applications, which will promote the application of surface temperature data in the fields of climate change, dynamic monitoring and assessment of hydrological ecological environment, land surface radiation budget and energy balance, and further promote the application and development of thermal infrared remote sensing.

[0013] The full sky surface temperature obtained by this invention is based on the simulation experiment of a large number of station observation data and the physical constraints of the surface radiation balance equation, completing the reconstruction of the spatial full coverage surface temperature. In addition, combined with the geographically weighted regression method, a comprehensive scale factor is established. The spatially weighted regression relationship between the high and low resolution surface temperatures is transformed into the difference between the two resolutions. The differences between the factors have realized the spatial downscaling of a set of data and a physical model method, ensuring the stability of the method and the accuracy of the results. It has achieved the accurate acquisition of the surface temperature of the entire sky and the high spatial resolution surface temperature product.

[0014] The spatial distribution of the surface temperature image reconstructed using this method is basically consistent with the spatial distribution of the original clear sky surface temperature, and the distribution of the reconstructed surface temperature under clouds is reasonable; the verification accuracy RMSE of reconstructing the full sky surface temperature using station data is 2.98K. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. As shown in the figure:

[0016] Figure 1 Flowchart of the spatial reconstruction and spatial downscaling of all-sky surface temperature of the present invention.

[0017] Figure 2 The present invention is based on the comprehensive scale factor Validation results of reconstructing all-sky surface temperature using station data.

[0018] Figure 3 A graph of the surface temperature before reconstruction of the present invention.

[0019] Figure 4 Two graphs of the surface temperature before reconstruction of the present invention.

[0020] Figure 5 A graph of the reconstructed full-sky surface temperature of the present invention.

[0021] Figure 6 Two graphs of the reconstructed full-sky surface temperature of the present invention.

[0022] Figure 7 A graph of the reconstructed full-sky and spatially downscaled surface temperature of the present invention.

[0023] Figure 8 Two images of the reconstructed full-sky and spatially downscaled surface temperature of the present invention.

[0024] Fig. 9 Flowchart of the present invention. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] Example 1: Figure 1 , Figure 2 , Figure 3 , Figure 4 , Figure 5 , Figure 6 , Figure 7 , Figure 8 and Fig. 9 As shown, an integrated method of all-sky surface temperature spatial reconstruction and downscaling can efficiently obtain accurate, all-sky, high temporal resolution and high spatial resolution surface temperature data.

[0027] The technical problems mainly solved by the present invention include the following two points:

[0028] First, in order to solve the problem of large-scale missing of satellite thermal infrared surface temperature products, a comprehensive scale is proposed based on the radiation balance equation for calculating the net radiation of the surface and a large number of station data experiments. Factor (ratio of net radiation to net shortwave radiation), derived from the surface radiation balance equation The analytical relationship between the LST and the LST is calculated by calculating the total differential of the surface temperature, which converts the difference between the effective observation (clear sky) and the surface temperature in the missing area (under the cloud) of the thermal infrared remote sensing surface temperature into The influence of factors, cloud back radiation and net shortwave radiation is taken into account, and then the surface temperature of the entire sky is calculated to achieve spatial reconstruction of the surface temperature of the entire sky based on physical processes, thereby improving the consistency and accuracy of data quality.

[0029] Second, low spatial resolution surface temperature remote sensing data cannot provide an effective expression of the heterogeneity of land surface moisture conditions and energy exchange at a fine scale. Using the low spatial resolution all-sky surface temperature obtained in the previous step, we recalculate the comprehensive scale factor. As a function of vegetation, temperature and humidity, the geographically weighted regression (GWR) method was used to establish The spatial weighted relationship between the factors converts the difference between the surface temperature at high and low spatial resolutions into the difference between the surface temperature at two resolutions. Finally, the high spatial resolution By bringing back the surface radiation balance equation, the surface temperature with high spatial resolution can be obtained.

[0030] After the above two steps, the spatial reconstruction and spatial downscaling of the surface temperature of the whole sky are completed. These two steps are defined by a comprehensive scale factor ( ) ties them all together and calls it an integrative approach.

[0031] A method for integrating all-sky surface temperature spatial reconstruction and downscaling includes the following steps:

[0032] Step 1: Acquisition and preprocessing of multi-source remote sensing data, atmospheric driving data and auxiliary data.

[0033] Step 2: Spatial reconstruction of all-sky surface temperature data.

[0034] Step 3: Spatial downscaling of all-sky surface temperature data.

[0035] In step 1, data needs to be obtained according to the input requirements of the surface radiation balance equation, including:

[0036] Step 1.1) Meteorological data: low spatial resolution geostationary meteorological satellite observation hourly radiation data, atmospheric driven data set including air temperature, humidity, atmospheric pressure and downlink shortwave radiation;

[0037] Step 1.2) Low spatial resolution surface temperature products such as MODIS or FY-3 and Normalized Difference Vegetation Index (NDVI); High spatial resolution surface temperature products and reflectance products such as Landsat Level 2;

[0038] Step 1.3) Auxiliary data, mainly including latitude and longitude, digital elevation (DEM), etc.;

[0039] Step 1.4) Data processing: Extract the atmospheric driving data of the corresponding study area based on the satellite transit time data in low-resolution surface temperature products such as MODIS or FY-3, and use DEM and longitude and latitude auxiliary data to perform terrain and altitude correction on atmospheric driving data such as air temperature; uniformly crop the MODIS, FY-3 and Landsat remote sensing data according to the longitude and latitude range of the study area; for MODIS and FY-3 surface temperature data, the albedo and normalized difference vegetation index (NDVI) can be directly obtained from their corresponding products; for Landsat, its reflectance product can be used to calculate the required surface remote sensing parameters, such as albedo and normalized difference vegetation index (NDVI).

[0040] In step 2, after a large number of experimental comparative analyses, it was found that the ratio of net radiation to net shortwave radiation on sunny days (where surface temperature is valid) and rainy days (where surface temperature is invalid) is very small in close time. Based on this, a comprehensive scale factor ( ). Combined Factor, derived from the surface radiation balance equation, is the LST Next, using only clear days (effective values ​​of surface temperature), we calculate Factor, combined with linear interpolation method to reconstruct the time series factor (usually one year is chosen), and the surface temperature derived above is The analytical expression of the factor can be used to obtain the predicted full-sky surface temperature. The full-sky surface temperature reconstruction based on the surface radiation balance equation is as follows:

[0041] The radiation balance equation for net surface radiation is given by:

[0042] (1)

[0043] Where R n is the net radiation from the surface, SWD is the downward shortwave radiation; α is the surface albedo; ε m is the emissivity; σ is the Stefan-Boltzmann constant, and LWD is the downlink longwave radiation.

[0044] First, define a comprehensive scale factor ( ), that is, net radiation (Rn) and net shortwave radiation (S net = (1-α) SWD). Calculated using clear day (effective surface temperature) data Factor, combined with linear interpolation method to reconstruct the time series factor (usually one year is chosen) and is substituted into the following derived surface temperature The predicted full-sky surface temperature can be obtained by using the analytical expression of the factor.

[0045] Combined with the surface radiation balance equation, LST and , SWD, LWD, α and ε m The expression between is as follows,

[0046] (2)

[0047] The difference between the clear sky effective pixel and the missing pixel land surface temperature (LST) mainly comes from the cloud blocking of shortwave radiation, the downward longwave radiation of cloud radiation, the change of the comprehensive scale factor and the change of emissivity. Here, it is obtained by calculating the total differential of equation (2), as shown below:

[0048] (3)

[0049] The differential change is obtained by integrating the scale factor ( ), net shortwave radiation (S net = (1-α)SWD)), downlink longwave radiation (LWD) and emissivity (ε m ) and obtain the partial derivative of LST. Next, consider using the method recommended by FAO to calculate the downward shortwave radiation of a sunny day, and use the actual downward shortwave radiation of the surface obtained by reanalysis data (this quantity is easier to obtain). The difference between the two can be used to obtain the change in radiation, so that the impact of radiation can be quantified; use the difference in downward longwave radiation between sunny and cloudy days to obtain dL, and combine it with the previous derivation to quantify the downward inverse radiation impact of clouds; for the comprehensive scale factor , full coverage of space First, use the clear sky effective pixels to calculate and then reconstruct (generally simple linear interpolation is sufficient. After testing with station data, it is found that the ∇ results calculated on sunny days and days with missing LST are very close on nearby days, which is also the key to proposing this innovative method). Calculate d The process is similar to During the calculation process, different shortwave downlink radiation inputs obtain different , the difference between the two is d ; For the surface emissivity (ε m ), the difference between the same pixel under clear sky conditions and non-clear sky conditions on that day is extremely small and can be ignored in the calculation.

[0050] In summary, shortwave radiation, longwave radiation and The reconstruction of the land surface temperature of the whole sky (LSTAK) is completed by taking into account the influence of the factors as follows:

[0051] (4)

[0052] Where LSTAK represents the total sky surface temperature.

[0053] In step 3, the spatial downscaling of the all-sky surface temperature data is performed as follows:

[0054] Using the low spatial resolution all-sky surface temperature data obtained in the previous step, recalculate the all-sky low spatial resolution Factor, while calculating the high spatial resolution of the satellite transit day Factor. Daily low spatial resolution Factors as temperature, humidity and high spatial resolution satellite transit time The function of the daily low spatial resolution is established by combining the geographically weighted regression (GWR) method. Factor and high spatial resolution Factor resampling to lower resolution The spatially weighted regression relationship between the high and low resolutions is thus converted into the difference between the surface temperatures at the two resolutions. The specific process is as follows: First, establish The geographically weighted statistical relationship of the factors at low resolution is shown below.

[0055] (5)

[0056] The superscript C indicates low resolution. , , , and is the regression coefficient, is the residual at low resolution, High spatial resolution Factor resampling to lower spatial resolution.

[0057] Use linear interpolation or Kriging interpolation to interpolate the regression coefficients and residuals of low spatial resolution to high spatial resolution to obtain the downscaling comprehensive scale factor of high spatial resolution. , as shown below,

[0058] (6)

[0059] The superscript F indicates high spatial resolution. High spatial resolution of satellite transit days factor.

[0060] Finally, the daily high spatial resolution comprehensive scale factor calculated by formula (6) is Bringing back equation (2), we can obtain the surface temperature with high spatial resolution. So far, after the above two steps, the spatial reconstruction and spatial downscaling of the surface temperature of the whole sky are completed. These two steps are achieved through the comprehensive scale factor ( ) brings them together, which is an integrated approach.

[0061] Example 2: Figure 1 , Figure 2 , Figure 3 , Figure 4 , Figure 5 , Figure 6 , Figure 7 , Figure 8 and Fig. 9 As shown, an integrated method for all-sky surface temperature spatial reconstruction and downscaling is divided into the following steps.

[0062] Step 1: Acquisition and preprocessing of multi-source remote sensing data and atmospheric driving data;

[0063] (1) Obtain and process reanalysis data. For example, obtain temperature, humidity, atmospheric pressure and downwelling shortwave radiation through the reanalysis dataset (ERA5-Land) produced by the European Centre for Medium-Range Weather Forecasts (ECMWF);

[0064] (2) Obtain low-resolution surface temperature products such as MODIS or FY-3 and high-resolution surface temperature products such as Landsat Level 2. For example, download surface temperature products, reflectance products, and Normalized Difference Vegetation Index (NDVI) products from the China Fengyun Satellite Remote Sensing Data Service Network (http: / / satellite.nsmc.org.cn / PortalSite / Default.aspx) or the NASA EARTHDATA website (https: / / ladsweb.modaps.eosdis.nasa.gov); download Landsat Level 2 surface temperature products and surface reflectance products from the USGS website (https: / / earthexplorer.usgs.gov / ).

[0065] The data is preprocessed, including band extraction, projection conversion, and image cropping.

[0066] (3) Auxiliary data, mainly including latitude and longitude, and digital elevation (DEM). For example, DEM data uses the ASTE Global Digital Elevation Model version 3 data with a spatial resolution of 1 arc second (30m), which can be obtained from https: / / lpdaac.usgs.gov / products / astgtmv003 / .

[0067] The reanalysis data of the corresponding study area are extracted based on the satellite transit time data in low-resolution surface temperature products such as MODIS or FY-3; the MODIS, FY-3 and Landsat remote sensing surface temperature product data are uniformly clipped according to the longitude and latitude range of the study area; for the surface temperature data of MODIS and FY-3, the albedo, normalized difference vegetation index (NDVI) and leaf area index products (LAI) can be directly obtained from their corresponding products; for Landsat, its reflectivity product can be used to calculate the required surface remote sensing parameters, such as albedo and normalized difference vegetation index (NDVI).

[0068] Step 2: Spatial reconstruction of all-sky surface temperature data;

[0069] The surface radiation balance equation is as follows:

[0070] (1)

[0071] Where R n is the net surface radiation, SWD is the downlink shortwave radiation, obtained from the reanalysis dataset (ERA5-Land); α is the surface albedo, obtained from data products such as MODIS or FY-3;

[0072] Data products such as MODIS or FY-3 are surface temperature products released by the following networks.

[0073] MODIS: https: / / ladsweb.modaps.eosdis.nasa.gov.

[0074] FY-3: http: / / satellite.nsmc.org.cn / PortalSite / Default.aspx.

[0075] ε m is the emissivity, which can be calculated using the Normalized Difference Vegetation Index (NDVI); σ is the Stefan-Boltzmann constant.

[0076] Land surface temperature (LST) is a key physical quantity in the surface radiation balance equation. Based on the calculation of the surface radiation balance equation and a large number of station data experiments, a comprehensive scale is proposed. Factor (ratio of net radiation to net shortwave radiation), derived from the surface radiation balance equation The analytical relationship between the surface temperature data. First, define a comprehensive scale factor ( ), that is, the net radiation (R n ) and net shortwave radiation (S net = (1-α)SWD). Next, using only clear days (effective surface temperature) we calculate Factor, combined with linear interpolation method to reconstruct the time series factor (usually one year is chosen), and the surface temperature derived above is The predicted full-sky surface temperature can be obtained by using the analytical expression of the factor.

[0077] Combined with the surface radiation balance equation, LST and , SWD, LWD, α and ε m The expression between is as follows,

[0078] (2)

[0079] The difference between the effective pixel surface temperature (LST) in the clear sky and the missing pixel surface temperature (LST) is mainly due to the shielding of shortwave radiation by clouds, the downward longwave radiation of cloud radiation, the change of the comprehensive scale factor and the change of the emissivity. It is obtained by calculating the total differential of equation (2). By calculating the total differential of the surface temperature data, the difference between the effective observation (clear sky) of the thermal infrared remote sensing surface temperature data and the surface temperature data of the missing area (under the cloud) is converted into Factors, cloud back radiation and net shortwave radiation are used to calculate the surface temperature data for the entire sky, as shown below:

[0080] (3)

[0081] The differential change is obtained by integrating the scale factor ( ), net shortwave radiation (S net = (1-α)SWD)), downlink longwave radiation (LWD) and emissivity (ε m ) and obtain the partial derivative of LST. Next, consider using the method recommended by FAO to calculate the downward shortwave radiation of a sunny day, and use the actual downward shortwave radiation of the surface obtained by reanalysis data (this quantity is easier to obtain). The difference between the two can be used to obtain the change in radiation, so that the impact of radiation can be quantified; use the difference in downward longwave radiation between sunny and cloudy days to obtain dL, and combine it with the previous derivation to quantify the downward inverse radiation impact of clouds; for the comprehensive scale factor The spatial full coverage ∇ is first calculated using clear sky effective pixels and then reconstructed (generally simple linear interpolation is sufficient. After testing with station data, it was found that the ∇ results calculated on sunny days and days with missing LST are very close on nearby days, which is also the key to proposing this innovative method). Calculate d The process is similar to During the calculation process, different shortwave downlink radiation inputs obtain different , the difference between the two is d ; For the surface emissivity (ε m ), the difference between the same pixel under clear sky conditions and non-clear sky conditions on that day is extremely small and can be ignored in the calculation.

[0082] In summary, shortwave radiation, longwave radiation and The reconstruction of the land surface temperature of the whole sky (LSTAK) is completed by taking into account the influence of the factors as follows:

[0083] (4)

[0084] Where LSTAK represents the total sky surface temperature.

[0085] Step 3: spatial downscaling of all-sky surface temperature data;

[0086] First, a geographically weighted statistical relationship with the ∇ factor at low spatial resolution is established as follows:

[0087] (5)

[0088] The superscript C indicates low spatial resolution. , and is the regression coefficient, is the residual at low spatial resolution, and Factor resampling to lower spatial resolution.

[0089] Use linear interpolation or Kriging interpolation to interpolate the regression coefficients and residuals of low spatial resolution to high spatial resolution to obtain a high spatial resolution downscaling comprehensive scale Factors are as follows,

[0090] (6)

[0091] The superscript F indicates high spatial resolution. High spatial resolution of satellite transit days factor.

[0092] Finally, the high spatial resolution ∇ factor is brought back to equation (2) to obtain the high spatial resolution surface temperature. So far, after the above two steps, the spatial reconstruction and spatial downscaling of the full sky surface temperature have been completed.

[0093] A method for spatial reconstruction and spatial downscaling of all-sky surface temperature based on the surface radiation balance equation. The obtained all-sky surface temperature is based on a large number of simulation experiments of station observation data and the physical constraints of the surface radiation balance equation to complete the reconstruction of the spatial full coverage surface temperature. In addition, combined with the geographically weighted regression method, a comprehensive scale The spatially weighted regression relationship of the factors transforms the difference between the surface temperature at high and low spatial resolutions into the difference between the surface temperature at two resolutions. The differences between the factors enable spatial downscaling using a set of data and a physical model method, ensuring the stability of the method and the accuracy of the results.

[0094] like Figure 1 As shown, remote sensing data such as MODIS or FY-3, remote sensing data such as Landsat, atmospheric driving data and auxiliary data are obtained, low spatial resolution remote sensing parameters and atmospheric driving data preprocessing and high spatial resolution remote sensing parameters and atmospheric driving data preprocessing are performed, and the low spatial resolution comprehensive scale factor ( ) and complete the low spatial resolution daily all-sky surface temperature spatial reconstruction; at the same time, use the surface radiation balance equation and geographically weighted regression method to complete the spatial downscaling of the daily all-sky surface temperature, and finally obtain the high spatial resolution daily all-sky surface temperature product.

[0095] The surface temperature before and after reconstruction, such as Figure 3 and Figure 4 The surface temperature before reconstruction, Figure 5 and Figure 6 The reconstructed all-sky surface temperature, Figure 7 and Figure 8 The reconstructed full sky and spatially downscaled surface temperature is shown in the figure. The results show that the spatial distribution of the surface temperature image reconstructed using this method is basically consistent with the spatial distribution of the original clear sky surface temperature, and the distribution of the reconstructed surface temperature under the clouds is reasonable; the field station data (N=12911) was used for verification, and the results showed that the verification accuracy RMSE of the reconstructed high spatial resolution full sky surface temperature reached 2.98K, which is higher than the accuracy of the existing full sky surface temperature reconstruction method, and further promoted the application and development of thermal infrared surface temperature.

[0096] 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 method for integrating spatial reconstruction and downscaling of all-sky surface temperature, characterized in that: Contains the following steps: Step 1: Acquisition and preprocessing of multi-source remote sensing data and atmospheric driving data. Step 2: Spatial reconstruction of all-sky surface temperature data, Step 3: Spatial downscaling of all-sky surface temperature data; The whole sky surface temperature reconstruction based on the surface radiation balance equation is as follows: The radiation balance equation for net surface radiation is given by: （1） Where Rn is the net radiation of the surface, SWD is the downlink shortwave radiation; α is the surface albedo; ε m is the specific emissivity; σ is the Stefan-Boltzmann constant, LWD is the downlink long-wave radiation, Taking into account the physical relationship between the land surface temperature LST and the net surface radiation in the radiation balance equation, the comprehensive scale factor is defined as , that is, the net radiation R n and net shortwave radiation S net = (1-α) SWD ratio, calculated using the effective value of the surface temperature on a clear day factor, Combined with the surface radiation balance equation, LST and , SWD, LWD, α and ε m The expression between is as follows, （2） The difference between the surface temperature LST of the clear sky effective pixel and the missing pixel mainly comes from the shielding of shortwave radiation by clouds, the downward longwave radiation of cloud radiation, the change of the comprehensive scale factor and the change of the emissivity. It is obtained by calculating the total differential of formula (2), as shown below: （3） The differential change is calculated by the comprehensive scale factor , Net shortwave radiation S net = (1-α)SWD, downlink longwave radiation LWD and emissivity ε m The partial derivatives of LST are obtained separately. Consider using the method recommended by FAO to calculate the downlink shortwave radiation on a sunny day. The actual downlink shortwave radiation of the surface obtained using the reanalysis data can be used to obtain the radiation change dR through the difference between the two, and quantify the impact of radiation. The difference in downward longwave radiation between sunny and cloudy days is used to obtain dL, and combined with the previous derivation, the downward back radiation effect of clouds is quantified; For the comprehensive scale factor , full coverage of space It is first calculated using the clear sky effective pixels, and then the clear sky is used for cloud and rain conditions. The value is linearly interpolated to obtain the comprehensive scale factor of the entire sky ,calculate The process is similar to the dR calculation process. Different shortwave downlink radiation inputs obtain different , after subtracting the two, ; Superimpose shortwave radiation, longwave radiation and The influence of the factors is used to complete the reconstruction of the full sky surface temperature LSTAK, as follows: （4） Where LSTAK represents the total sky surface temperature.

2. The method for integrating spatial reconstruction and downscaling of all-sky surface temperature according to claim 1, characterized in that: In step 1, data needs to be obtained according to the input requirements of the surface radiation balance equation, including the following steps: Step 1.1) Obtain meteorological data: low spatial resolution geostationary meteorological satellite observation hourly radiation data, atmospheric driven data set including temperature, humidity, atmospheric pressure and downlink shortwave radiation; Step 1.2) Obtain MODIS or FY-3 low-resolution surface temperature products and Landsat Level 2 high-resolution surface temperature products; Step 1.3) Obtain high- and low-resolution remote sensing auxiliary data, mainly including longitude and latitude, digital elevation DEM, MODIS / FY-3 and Landsat band reflectance products and Normalized Difference Vegetation Index NDVI; Step 1.4) Data processing: Extract the atmospheric driving data of the corresponding study area based on the satellite transit time data in the MODIS or FY-3 low-resolution surface temperature product; uniformly crop the MODIS or FY-3 and Landsat remote sensing data according to the latitude and longitude range of the study area; for the MODIS or FY-3 surface temperature data, directly obtain the albedo and normalized difference vegetation index NDVI from its corresponding products; for Landsat, use its reflectance product to calculate the required surface remote sensing parameters, albedo and normalized difference vegetation index NDVI.

3. The method for integrating spatial reconstruction and downscaling of all-sky surface temperature according to claim 1, characterized in that: In step 2, define a composite scale factor , combined with Factor, from first establishing Factors are used to derive LST from geographically weighted statistical relationships at low resolution. The analytical expression of the factor is calculated using the effective value of the surface temperature on a clear day. Factor, combined with linear interpolation method to reconstruct the time series Factor, into the surface temperature derived above about Analytical expressions for the factors to obtain the predicted all-sky surface temperature.

4. The method for integrating spatial reconstruction and downscaling of all-sky surface temperature according to claim 1, characterized in that: In step 3, the spatial downscaling of the all-sky surface temperature data is performed as follows: Obtained low spatial resolution all-sky surface temperature data, recalculated all-sky low resolution Factor, while calculating the high spatial resolution of the satellite transit day Factor, daily low spatial resolution Factors as temperature, humidity and high spatial resolution satellite transit time The function is combined with the geographically weighted regression (GWR) method to establish a daily low spatial resolution Factor and high spatial resolution Factor resampling to lower spatial resolution The spatially weighted regression relationship between the high and low spatial resolutions is transformed into the difference between the surface temperatures at the two resolutions. The difference between the factors, The specific process is as follows: First, establish The geographically weighted statistical relationship of the factors at low spatial resolution is shown below. （5） The superscript C indicates low spatial resolution. , , , and is the regression coefficient, is the residual term, Represents high spatial resolution Factor resample to lower resolution, Use linear interpolation or Kriging interpolation to interpolate the regression coefficients and residuals of low spatial resolution to high spatial resolution to obtain the downscaling comprehensive scale factor of high spatial resolution. , as shown below, （6） The superscript F indicates high spatial resolution. It is the highest spatial resolution of the satellite transit day. Factor, the comprehensive scale factor of high spatial resolution calculated by formula (6) Bringing back formula (2), we can obtain the surface temperature with high spatial resolution.

Images