Angle effect correction method for remote sensing surface temperature based on emissivity kernel driven model
By constructing a directional emissivity model and GSW algorithm based on a parameter kernel driven model, the inversion accuracy problem caused by the emissivity angle effect in satellite remote sensing technology is solved, and higher-precision surface temperature data acquisition is achieved.
Patent Information
- Application Number
- CN202411012125.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-26
- Publication Date
- 2025-09-16
- Estimated Expiration
- 2044-07-26
AI Technical Summary
Existing satellite remote sensing technology ignores the angular effect of emissivity when acquiring surface temperature data, resulting in unsatisfactory inversion accuracy, especially in large-scale or regional monitoring.
A directional emissivity model is constructed based on the parameter kernel driven model. By calculating the directional emissivity and combining it with the atmospheric profile data and the GSW algorithm model, multiple linear regression is performed to construct a coefficient lookup table and invert the pixel-by-pixel land surface temperature (LST).
The inversion accuracy of land surface temperature (LST) is improved, the uncertainty caused by angle effect is reduced, and higher-precision data acquisition is achieved.
Smart Images
Figure CN118916585B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of surface temperature inversion technology, in particular to a remote sensing surface temperature angle effect correction method based on an emissivity kernel driven model, and particularly to the field of quantitative remote sensing technology. Background Art
[0002] Currently, land surface temperature (LST) is a key driver of the energy budget and water cycle in surface-atmosphere interactions. There are two common methods for obtaining LST data. The first is to establish ground-based observation stations (which are expensive to deploy and not widely feasible). This method is only applicable to obtaining LST data in the area where the station is located, making it very limited in scope and unsuitable for monitoring LST over large areas. The second method is to simulate LST using satellite remote sensing technology (such as satellite-based sensor products). LST is an essential input for climate change, environmental monitoring, and evapotranspiration estimation. However, thermal infrared sensors observe different radiation from different viewing angles, a phenomenon known as thermal radiation directionality (TRD). This TRD can lead to significant discrepancies between the accuracy of the retrieved LST and the true value. Studies have found that LST observed from different viewing angles can vary by up to 6-12 K. Sensor products from associated satellites use the observed zenith angle (VZA) and emissivity based on classification (for example, the feature classification generally includes eight representative feature types: evergreen coniferous forest (ENF), deciduous broadleaf forest (DBF), open shrubland (OS), woody savanna (WS), grassland (GL), cultivated land (CL), urban and built-up land (UBL), and wasteland (BR). The observed zenith angle VZA and the classification also interact (for example, the angular effect of emissivity is ignored). For another example, emissivity is a directional quantity that is related to the observed zenith angle (VZA). Natural surfaces are highly heterogeneous, with composition, roughness, structure, and spatial heterogeneity leading to anisotropy in emissivity. For non-isothermal pixels, the directionality of thermal radiation is due to two factors: 1) the variation in the observation angle between components with different temperatures and emissivities; and 2) multiscattering within different components. For isothermal pixels, the anisotropy of thermal radiation stems from the anisotropy of the ground surface emissivity (LSE). Ignoring the anisotropy of emissivity will introduce significant uncertainty into the subsequent surface temperature inversion, so the angular effect of emissivity cannot be ignored. Therefore, the second method of obtaining surface temperature data has many technical challenges (such as the angular effect) and still suffers from unsatisfactory accuracy. Acquiring higher-precision surface temperature data is a current technical challenge and a key area of research and development in this field. Summary of the Invention
[0003] The purpose of the present invention is to provide a remote sensing surface temperature angle effect correction method based on an emissivity kernel driven model. A directional emissivity model is constructed based on a parameter kernel driven model. A hotspot kernel or / and a basic shape kernel are used to establish a directional emissivity model for the study area data to calculate the directional emissivity. At the same time, the channel atmospheric parameters of the two channels of band 31 and band 32 are obtained according to the atmospheric profile data. The observed zenith angle, the atmospheric bottom boundary temperature, the average emissivity, and the emissivity difference are grouped and the channel on-board brightness temperature is calculated. Then, a GSW algorithm model is constructed to obtain a coefficient lookup table under multiple grouping conditions. Then, the directional emissivity, the observed zenith angle, the atmospheric water vapor content, and the brightness temperature are combined and the split window coefficients under the corresponding group are traversed and extracted from the coefficient lookup table and input into the GSW algorithm model to invert the surface temperature LST of each pixel in the study area, thereby improving the inversion accuracy.
[0004] The purpose of the present invention is achieved through the following technical solutions:
[0005] A method for correcting the angle effect of remote sensing surface temperature based on an emissivity kernel driven model, the method comprising:
[0006] S1. Construct a directional emissivity model based on a parameter kernel driven model. The parameter kernel driven model includes a hotspot kernel and a basic shape kernel. Collect data for the study area, including MYD21, MYD03, MYD10_L2, MYD021 km, MYD11_L2, and MCD12Q1 products (using sufficient satellite thermal infrared data at various observation zenith angles (VZA) within 16 days as multi-angle observation data) and input them into the parameter kernel driven model. The directional emissivity model is calculated according to the following formula:
[0007] ,in is the isotropy coefficient;
[0008] S2. Atmospheric profile data were collected and input into MODTRAN 5.2 software to simulate atmospheric parameters. The data were then convolved with the channel's spectral response function to obtain channel atmospheric parameters for Bands 31 and 32. The channel atmospheric parameters included the effective transmittance of the atmosphere, the atmospheric downward radiation, and the atmospheric upward radiation at the observation zenith angle θ. The observation zenith angle, the bottom boundary temperature of the atmosphere, the average emissivity, and the emissivity difference were grouped, and then an inverse Planck operation was performed to obtain the channel brightness temperature for each group.
[0009] S3. Construct the GSW algorithm model, input the atmospheric bottom boundary temperature, channel satellite brightness temperature, average emissivity, and emissivity difference into the GSW algorithm model, and obtain multiple sets of equations according to the following algorithm formula. Perform multiple linear regression on them to obtain the split window coefficients of the general split window algorithm. to , and then construct a coefficient lookup table for multiple grouping cases;
[0010]
[0011] in, represents the boundary temperature of the bottom atmosphere, and Respectively represent the channel on-board brightness temperature of bands 31 and 32, represents the difference in emissivity between bands 31 and 32, represents the average emissivity of bands 31 and 32, to is the split window coefficient in the coefficient lookup table;
[0012] S4. Extract the observation zenith angle, atmospheric water vapor content, and brightness temperature from the study area data and combine them with the directional emissivity obtained in step S1. Extract the split window coefficients under the corresponding group from the coefficient lookup table. Then input the directional emissivity, split window coefficients, and brightness temperature into the GSW algorithm model. The surface temperature LST of each pixel in the study area is obtained according to the following inversion formula:
[0013] in, represents the average emissivity of bands 31 and 32, represents the difference in emissivity between bands 31 and 32, and Represent the brightness temperature of bands 31 and 32 respectively, to is the split window coefficient obtained from the coefficient lookup table.
[0014] In order to better implement the present invention, the present invention also includes the following method:
[0015] S5. Collect data in the study area, including SURFRAD station data and / or ERA5_LAND data, and obtain the true LST value. Construct a loss constraint function in the GSW algorithm model to constrain the loss between the surface temperature LST and the true LST value.
[0016] Preferably, in step S5, the method of obtaining the LST true value from the SURFRAD station data of the present invention is as follows: the SURFRAD station data are uniformly converted to Coordinated Universal Time, and then the LST true value is obtained according to the following formula:
[0017] ,in represents the true value of LST, and represent the upward radiation and downward radiation from the surface, represents the Stefan Boltzmann constant, represents the broadband emissivity; The calculation method is as follows: ,in and They represent the narrow-band emissivity of band 29 and band 31 obtained from the MYD21 product, respectively.
[0018] Preferably, the parameter kernel driven model constructs several parameter kernel driven units according to the hot spot kernel and / or the basic shape kernel, selects the parameter kernel driven unit with the smallest root mean square error (RMSE) to establish a directional emissivity model, inputs the study area data, and calculates the directional emissivity.
[0019] Preferably, the hotspot nuclear data of the present invention is obtained according to the following method:
[0020] in, represents hotspot nuclear data, represents the solar zenith angle, represents the observation zenith angle, represents the relative azimuth angle between the sun and the sensor, Indicates the angular distance between the sun and the sensor.
[0021] Preferably, the basic shape kernel data of the present invention is obtained according to the following method:
[0022]
[0023] in Represents the basic shape kernel data, Indicates the phase angle between the sun, the earth's surface, and the sensor.
[0024] Preferably, in step S2, the brightness temperature of the channel star of the present invention is Obtained as follows:
[0025] in, represents the Planck equation, Indicates channel i, at the observation zenith angle Channel brightness temperature at ; Indicates channel i, at the observation zenith angle The channel emissivity at Indicates channel i, temperature is The emission radiance of a black body under Planck's law is, represents the effective atmospheric transmittance of channel i, represents the atmospheric downward radiation of channel i, represents the atmospheric upward radiation of channel i when the observation zenith angle is θ.
[0026] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0027] The present invention constructs a directional emissivity model based on a parameter kernel-driven model, inputs study area data, and calculates directional emissivity. Simultaneously, atmospheric profile data is input into MODTRAN 5.2 software to simulate atmospheric parameters and convolute with the channel's spectral response function to obtain channel atmospheric parameters for Bands 31 and 32. The observed zenith angle, atmospheric bottom boundary temperature, average emissivity, and emissivity difference are grouped, and the channel on-board brightness temperature is calculated. A GSW algorithm model is then constructed to perform multiple sets of equations and multivariate linear regression to obtain the split-window coefficients of the general split-window algorithm and construct a coefficient lookup table for multiple groupings. The split-window coefficients for the corresponding groups are then extracted from the coefficient lookup table and input into the GSW algorithm model to invert the land surface temperature (LST) of each pixel in the study area, thereby improving the inversion accuracy of the LST. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 Schematic diagram of the process of the present invention;
[0029] Figure 2 Schematic maps of 8 IGBP land types in the study area are selected as examples in the embodiments;
[0030] Figure 3 Schematic diagram of the principle of grouping according to the observation zenith angle, bottom atmospheric boundary temperature, average emissivity, and emissivity difference in the embodiment;
[0031] Figure 4 A schematic diagram illustrating the principle of hierarchical extension grouping in an embodiment;
[0032] Figure 5 As an example, the polar plot of directional emissivity of woody savanna in band 31 during summer daytime in the study area is shown;
[0033] Figure 6 For example, the polar plot of directional emissivity of urban and built-up areas in band 32 during the summer daytime in the study area is shown;
[0034] Figure 7 For the embodiment, the RMSE of the directional emissivity of the kernel-driven unit of each parameter in the 31 band of the study area is displayed;
[0035] Figure 8 For the embodiment, the RMSE of the directional emissivity of the kernel-driven unit of each parameter in the 32 bands of the study area is displayed;
[0036] Figure 9For example, a comparison chart of LST of three types of inversions is used to verify the SURFRAD station data in the study area as an example;
[0037] Figure 10 This is a comparison chart of LST of the three types of inversions verified using ERA5-Land data, taking the study area as an example. DETAILED DESCRIPTION
[0038] Below in conjunction with embodiment, the present invention is described in further detail:
[0039] Example
[0040] like Figures 1 to 8 As shown, a method for correcting the angle effect of remote sensing surface temperature based on an emissivity kernel driven model includes:
[0041] S1. Construct a directional emissivity model based on a parameter kernel driven model. The parameter kernel driven model includes a hotspot kernel and a basic shape kernel. Collect data of the study area including MYD21, MYD03, MYD10_L2, MYD021 km, MYD11_L2, and MCD12Q1 products (using sufficient satellite thermal infrared data at each observation zenith angle VZA within 16 days as multi-angle observation data) and input them into the parameter kernel driven model. In this embodiment, Figure 2 For example, in a certain study area (covering the satellite data from 2019 to 2021 for this study), some data (including observation azimuth angle VAA, observation zenith angle VZA, solar azimuth angle SAA, solar zenith angle SZA; quality assurance QA, and atmospheric precipitable water PWV) were obtained using products such as MYD21, MYD03, MYD10_L2, MYD021 km, MYD11_L2, and MCD12Q1. The following table shows the data:
[0042] Product Type Spatial resolution Temporal resolution Variable MYD21 Swath 1 km 5-minute emis31, emis32, LST, QA, PWV MYD03 Swath 1 km 5-minute SZA, SAA, VZA, VAA MYD021 km Swath 1 km 5-minute EV_1KM_Emissive MYD11_L2 Swath 1 km 5-minute emis31, emis32 MCD12Q1 Tile 500 m Yearly IGBP classification MYD10_L2 Swath 500 m 5-minute NDSI_Snow_Cover
[0043] See also Figure 2Using a certain area as an example study area, the different colored land cover corresponds to eight IGBP land classes, with the triangulation points representing seven SURFRAD stations (which can be used for data verification). The MYD21 product was reprojected based on the geographic location information of the MYD03 product. To reduce uncertainty in LSE (Land Surface Emissivity in remote sensing and Earth sciences) retrieval, only pixels with the mandatory QA flag and good quality were used. The Precipitable Water Volume (PWV) threshold was set below 1.5 cm to filter out pixels in relatively dry areas. In addition, MYD10A1 was used to remove snow-covered pixels. The IGBP data extracted from the MCD12Q1 product has a spatial resolution of 500 meters. To match the 1 km spatial resolution of the MYD21 data, the IGBPs were clustered from 500 meters to 1 km. The rule is that only if four 500-meter cells have the same land cover type are they aggregated into a new 1-kilometer cell; otherwise, they are set to a null value.
[0044] The directional emissivity model is calculated according to the following formula:
[0045] ,in is the isotropy coefficient.
[0046] Preferably, the hotspot core data of this embodiment is obtained according to the following method:
[0047] in, represents hotspot nuclear data, represents the solar zenith angle, represents the observation zenith angle, represents the relative azimuth angle between the sun and the sensor, Indicates the angular distance between the sun and the sensor.
[0048] Preferably, the basic shape kernel data of this embodiment is obtained according to the following method:
[0049]
[0050] in Represents the basic shape kernel data, Indicates the phase angle between the sun, the earth's surface, and the sensor.
[0051] In some embodiments, the directional emissivity model constructs several parameter kernel drive units according to the hotspot kernel and / or the basic shape kernel, selects the parameter kernel drive unit with the smallest root mean square error (RMSE) to establish the directional emissivity model, inputs the study area data, and calculates the directional emissivity. As a more detailed technical content of this embodiment, the parameter kernel drive model includes 11 parameter kernel drive units. The directional emissivity formula of the first parameter kernel drive unit (ROU) is ; The directional emissivity formula of the second parameter core driver unit (VIN) is ; The formula for the directional emissivity of the third parameter nuclear driver unit (RTH) is ;
[0052] The fourth parameter, the directional emissivity formula of the kernel driver unit (RTK), is ; The formula for the directional emission rate of the fifth parameter nuclear driver unit (RVI) is ; The formula for the directional emissivity of the sixth parameter nuclear drive unit (RVUS) is ; The formula for the directional firing rate of the seventh parameter core driver unit (VTH) is ; The directional emission rate formula of the eighth parameter kernel driver unit (VTK) is ;
[0053] The directional firing rate formula of the ninth parameter core driver unit (VVI) is ; The directional firing rate formula of the eleventh parameter core drive unit (VUS) is ;
[0054] The directional emissivity formula of the twelfth parameter kernel driver unit (GUTA_sparse) is .in represents the isotropy coefficient, Both represent hotspot kernel coefficients, Represents the basic shape kernel coefficient.
[0055] Hot spot nuclear data Obtained as follows:
[0056] in represents the solar zenith angle, represents the observation zenith angle, represents the relative azimuth angle between the sun and the sensor, Indicates the angular distance between the sun and the sensor.
[0057] Hot spot nuclear data Obtained as follows:
[0058] .
[0059] The basic shape kernel data Obtained as follows:
[0060]
[0061] in Indicates the phase angle between the sun, the earth's surface, and the sensor.
[0062] Basic shape kernel data Obtained as follows:
[0063]
[0064] Basic shape kernel data Obtained as follows:
[0065]
[0066] Basic shape kernel data Obtained as follows:
[0067] .
[0068] exist In the formula, each parameter is calculated as follows:
[0069]
[0070] , , , and is the nuclear coefficient, which is determined by the surface structure and component temperature; , ,and There are three kernels that weight the contributions of surface background, directional effects, and shadows respectively.
[0071] In an embodiment, the parameter kernel driven model for constructing the directional emissivity model may select the directional emissivity with the minimum root mean square error RMSE (minimum on a time scale) among the 11 parameter kernel driven units as the final directional emissivity.
[0072] Taking the data from 2019 to 2021 in the study area as an example, Figure 5 As shown, (a) is the firing rate of MYD21, (b)-(I) are the directional firing rates simulated by 11 parameter nuclear driven unit types, and the label in the upper right corner is the RMSE between the simulated firing rate and the original firing rate. Figure 5, the root mean square error (RMSE) of the directional emissivity obtained by using the ninth parameter kernel drive unit and the eleventh parameter kernel drive unit is relatively minimal. The directional emissivity with the smallest RMSE can be selected as the final directional emissivity (the directional emissivity of the woody savanna in band 31 during the day in summer). Taking the data from 2019 to 2021 in the study area as an example, Figure 6 As shown in the figure, the directional emissivity simulation effect of band 32 for urban and built-up areas during summer daytime is displayed. The root mean square error (RMSE) of the directional emissivity is minimized by using the tenth parameter kernel driving unit. It is selected as the directional emissivity model, and the data of the study area are input to obtain the directional emissivity of the city and built-up area in band 32 during summer daytime in the study area.
[0073] Taking the data from 2019 to 2021 in the study area as an example, we can Figure 7 、 Figure 8 The RMSE of the directional emissivity of each parameter kernel driver unit in eight land types, band 31 or band 32, day and night, and four seasons are displayed separately for quick selection.
[0074] S2. Collect atmospheric profile data and input them into MODTRAN 5.2 software to simulate atmospheric parameters and convolve them with the spectral response function of the channel to obtain the channel atmospheric parameters of band 31 and band 32 ( , and ), the channel atmospheric parameters include the effective transmittance of the atmosphere, the downward atmospheric radiation and the upward atmospheric radiation when the observation zenith angle is θ; the observation zenith angle, the bottom boundary temperature of the atmosphere, the average emissivity and the emissivity difference are grouped (see Figure 3 , observe the zenith angle VZA, the bottom boundary temperature of the atmosphere , average emissivity , emissivity difference Then perform Planck inverse operation to obtain the channel star brightness temperature under each group. , emissivity difference An example of grouping is as follows: (0.90, 0.96), (0.94, 1.0), 2 groups. An example of grouping for the zenith angle is as follows: [1, 2.8], with a step size of 0.2, 10 groups. An example of grouping for the bottom boundary temperature is as follows: ≤280 K, [275, 295), [290, 310), [305, 325), ≥320, 5 groups. If the atmospheric water vapor content (WVC) is introduced, an example of grouping for the atmospheric water vapor content (WVC) is as follows: [0, 1.5], [1.0, 2.5], [2.0, 3.5], [3.0, 4.5], [4.0, 5.5], [5.0, 6.5], 6 groups.
[0075] In some embodiments, as Figure 4 As shown, in addition to "grouping the observation zenith angle, bottom atmospheric boundary temperature, average emissivity, and emissivity difference", the grouping in this embodiment can also be based on Figure 4 The data are divided into eight land surface types, four seasons, and day and night (see Figure 1).
[0076] In some embodiments, the brightness temperature on the channel star Obtained as follows:
[0077] in, represents the Planck equation, Indicates channel i, at the observation zenith angle Channel brightness temperature at ; Indicates channel i, at the observation zenith angle The channel emissivity at Indicates channel i, temperature is The emission radiance of a black body under Planck's law is, represents the effective atmospheric transmittance of channel i, represents the atmospheric downward radiation of channel i, represents the atmospheric upward radiation of channel i when the observation zenith angle is θ.
[0078] S3. Construct a GSW algorithm model, input the atmospheric bottom boundary temperature, the brightness temperature on the channel satellite, the average emissivity, and the emissivity difference into the GSW algorithm model, obtain multiple sets of equations according to the following algorithm formula, and perform multiple linear regression on them to obtain the split window coefficients of the general split window algorithm (the split window algorithm in this embodiment preferably uses two split window channels): to , and then construct a coefficient lookup table for multiple grouping cases;
[0079]
[0080] in, represents the boundary temperature of the bottom atmosphere, and Respectively represent the on-board brightness temperature of bands 31 and 32, represents the difference in emissivity between bands 31 and 32, represents the average emissivity of bands 31 and 32, to is the split window coefficient from the coefficient lookup table.
[0081] S4. Extract the observation zenith angle, atmospheric water vapor content, and brightness temperature from the study area data and combine them with the directional emissivity obtained in step S1. Extract the split window coefficients under the corresponding group from the coefficient lookup table. Then input the directional emissivity, split window coefficients, and brightness temperature into the GSW algorithm model and obtain the surface temperature LST of each pixel in the study area according to the following inversion formula:
[0082] in, represents the average emissivity of pixel bands 31 and 32 in the study area, It represents the difference in emissivity between pixel bands 31 and 32 in the study area. and Respectively represent the brightness temperature of bands 31 and 32 (brightness temperature on the same channel), to The split window coefficients obtained in the coefficient lookup table. Multiple groups of the coefficient lookup table (grouping can further introduce atmospheric water vapor content WVC) and the split window coefficients under the groups have been obtained in step S3. This embodiment performs grouping, and a rough estimate can be made. Then based on , emissivity, etc., find the corresponding split window coefficient in the lookup table, and obtain the VZA corresponding coefficient by interpolation. to , and finally the accurate .
[0083] Taking the study area as an example, see Figure 9 、 Figure 10 , where LST_GSW_DE is the surface temperature LST considering the angle effect based on the method of the present invention. It is the LST data obtained by inputting into the GSW algorithm based on the directional reflectivity inverted by the method of the present invention combined with the coefficients in the coefficient lookup table under the grouping condition; LST_TES is the surface temperature LST considering the angle effect, which is derived from the LST product of MYD21; LST_GSW_CE does not consider the angle effect, and is the LST data obtained by inputting into the GSW algorithm based on the classified emissivity (LSE product of MYD11_L2) combined with the coefficients in the coefficient lookup table under the grouping condition.
[0084] S5. Collect data from the study area, including SURFRAD station data and / or ERA5_LAND data, and obtain the true LST. Construct a loss constraint function within the GSW algorithm model to constrain the loss between the LST and the true LST. For example, SURFRAD station data provides accurate, continuous, and long-term surface radiation budget data. SURFRAD is comprised of seven observation stations with different climates covering the study area. SURFRAD provides minute-by-minute surface radiation observations that can be used to evaluate satellite products. The geographic information of the SURFRAD stations and the IGBP surface coverage are shown in the following table:
[0085] Site name latitude longitude Elevation (m) IGBP land cover BND Bon***s 4*.**92°N 8*.**9°W 230 arable land TBL Tab***ado 4*.**98°N 10*.**0°W 1689 grassland DRA De***ada 3*.**73°N 11*.**7°W 1007 Barren land / open shrubland FPK For***ana 4*.**3°N 10*.**0°W 634 grassland GWN Goo***ppi 3*.**7°N 8*.**9°W 98 Woody savanna PSU Pen***ania 4*.**2°N 7*.**5°W 376 deciduous broad-leaved forest SXF Siou***ta 4*.**3°N 9*.**8°W 473 arable land
[0086] During ground truth verification, both ground observation and satellite imaging times are unified to UTC (Coordinated Universal Time), with site longitude used as an auxiliary condition for conversion between ground observation point times and UTC. Furthermore, this embodiment employs a three-standard-deviation method to eliminate outliers in both site and inverted LST. In some preferred embodiments, the true LST value is derived from SURFRAD site data as follows: SURFRAD site data is uniformly converted to Coordinated Universal Time, and then the true LST value is derived using the following formula: ,in represents the true value of LST, and represent the upward radiation and downward radiation from the surface, represents the Stefan Boltzmann constant, represents the broadband emissivity; The calculation method is as follows: ,in and The ERA5-LAND data are climate reanalysis data produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) that provide high-resolution information on land surface variables at an hourly grid spacing of approximately 9 kilometers. In the ERA5-Land data, the surface temperature represents the temperature of the uppermost surface layer, measured in degrees Kelvin, with an hourly temporal resolution and a spatial resolution of 11,132 meters.
[0087] This method utilizes data including SURFRAD station data and / or ERA5_LAND data to obtain the true LST value. First, the method can be used to obtain the LST corresponding to the true LST value area, thereby verifying data accuracy. Field verification has demonstrated high data accuracy. Second, a loss constraint function can be constructed within the GSW algorithm model to constrain the loss between the LST and the true LST value. This loss constraint further improves the accuracy of the LST inversion method.
[0088] 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 scope of protection of the present invention.
Claims
1. A method for correcting the angle effect of remotely sensed surface temperature based on an emissivity kernel driven model, characterized by: The methods include: S1. Construct a directional emissivity model based on the parameter kernel driven model. The parameter kernel driven model includes a hotspot kernel and a basic shape kernel. Collect data from the study area, including MYD21, MYD03, MYD10_L2, MYD021 km, MYD11_L2, and MCD12Q1 products, and input them into the parameter kernel driven model. The directional emissivity model is calculated according to the following formula: ,in is the isotropy coefficient; The parameter kernel driven model constructs several parameter kernel driven units according to the hot spot kernel and / or the basic shape kernel, and selects the parameter kernel driven unit with the smallest root mean square error (RMSE) to establish the directional emissivity model. The study area data is input and the directional emissivity is calculated. S2. Atmospheric profile data were collected and input into MODTRAN 5.2 software to simulate atmospheric parameters. The data were then convolved with the channel's spectral response function to obtain channel atmospheric parameters for Bands 31 and 32. The channel atmospheric parameters included the effective transmittance of the atmosphere, the atmospheric downward radiation, and the atmospheric upward radiation at the observation zenith angle θ. The observation zenith angle, the bottom boundary temperature of the atmosphere, the average emissivity, and the emissivity difference were grouped, and then an inverse Planck operation was performed to obtain the channel brightness temperature for each group. S3. Construct the GSW algorithm model, input the atmospheric bottom boundary temperature, channel satellite brightness temperature, average emissivity, and emissivity difference into the GSW algorithm model, and obtain multiple sets of equations according to the following algorithm formula. Perform multiple linear regression on them to obtain the split window coefficients of the general split window algorithm. to , and then construct a coefficient lookup table for multiple grouping cases; ； in, represents the boundary temperature of the bottom atmosphere, and Respectively represent the channel on-board brightness temperature of bands 31 and 32, represents the difference in emissivity between bands 31 and 32, represents the average emissivity of bands 31 and 32, to is the split window coefficient in the coefficient lookup table; S4. Extract the observation zenith angle, atmospheric water vapor content, and brightness temperature from the study area data and combine them with the directional emissivity obtained in step S1. Extract the split window coefficients under the corresponding group from the coefficient lookup table. Then input the directional emissivity, split window coefficient, and brightness temperature into the GSW algorithm model and obtain the surface temperature LST of each pixel in the study area according to the following inversion formula: in, represents the average emissivity of bands 31 and 32, represents the difference in emissivity between bands 31 and 32, and Respectively represent the on-board brightness temperature of bands 31 and 32, to is the split window coefficient obtained from the coefficient lookup table; S5. Collect data in the study area, including SURFRAD station data and ERA5_LAND data, and obtain the true value of LST. Construct a loss constraint function in the GSW algorithm model to constrain the loss between the surface temperature LST and the true value of LST.
2. The method for correcting the angle effect of remotely sensed surface temperature based on the emissivity kernel driven model according to claim 1, characterized in that: In step S5, the LST true value is obtained from the SURFRAD station data as follows: the SURFRAD station data are uniformly converted to Coordinated Universal Time, and then the LST true value is obtained according to the following formula: ,in represents the true value of LST, and represent the upward radiation and downward radiation from the surface, represents the Stefan Boltzmann constant, represents the broadband emissivity; The calculation method is as follows: ,in and They represent the narrow-band emissivity of band 29 and band 31 obtained from the MYD21 product, respectively.
3. The method for correcting the angle effect of remotely sensed surface temperature based on the emissivity kernel driven model according to claim 1, characterized in that: Hotspot kernel data is obtained as follows: in, represents hotspot nuclear data, represents the solar zenith angle, represents the observation zenith angle, represents the relative azimuth angle between the sun and the sensor, Indicates the angular distance between the sun and the sensor.
4. The method for correcting the angle effect of remotely sensed surface temperature based on the emissivity kernel driven model according to claim 1, characterized in that: The basic shape kernel data is obtained as follows: ； in Represents the basic shape kernel data, Indicates the phase angle between the sun, the earth's surface, and the sensor.
5. The method for correcting the angle effect of remotely sensed surface temperature based on the emissivity kernel driven model according to claim 1, characterized in that: In step S2, the brightness temperature of the channel star Obtained as follows: in, represents the Planck equation, Indicates channel i, at the observation zenith angle Channel brightness temperature at ; Indicates channel i, at the observation zenith angle The channel emissivity at Indicates channel i, temperature is The emission radiance of a black body under Planck's law is, represents the effective atmospheric transmittance of channel i, represents the atmospheric downward radiation of channel i, represents the atmospheric upward radiation of channel i when the observation zenith angle is θ.