A method for correcting angle effects of thermal infrared land surface temperature data from geostationary meteorological satellites

Abstract

The present invention discloses an angle effect correction method for geostationary meteorological satellite thermal infrared surface temperature data, which belongs to the field of thermal infrared remote sensing technology. The present invention first obtains geostationary satellite thermal infrared remote sensing surface temperature data of the target area and converts it to the target coordinate system; then calculates the angle parameters of each pixel based on the satellite sub-satellite data; performs system error elimination processing on the nighttime surface temperature data of two satellites screened based on the satellite zenith angle; then uses the nighttime surface temperature data to simplify the kernel-driven model and solve the emissivity kernel parameters; solves the solar nuclear coefficient and hotspot height based on the daytime surface temperature data at multiple consecutive moments, and combines the solved parameters to obtain vertical observations of each pixel. The present invention can quickly and accurately solve the parameters of the kernel-driven model and then accurately solve the vertically observed surface temperature, overcoming the defects of the traditional method such as large data demand, low calculation efficiency, and low accuracy, and can serve a large number of thermal infrared remote sensing surface temperature application scenarios.

Description

Technical Field

[0001] The present invention relates to the field of thermal infrared remote sensing, and in particular to an angle effect correction method for thermal infrared surface temperature data of a geostationary meteorological satellite. Background Art

[0002] As an indicator of the energy interaction between the Earth's surface and the atmosphere, land surface temperature (LST) can provide feedback on surface-atmospheric processes at various spatial and temporal scales, thereby influencing and objectively reflecting climate change. LST is also widely used in the study and application of land surface processes and models, such as climate and meteorological change, hydrological and disaster monitoring, and forest fire prevention. With the advancement of detection technology, the use of satellite data for LST inversion offers advantages such as convenient data acquisition, high efficiency, high temporal resolution, low cost, and wide coverage. Because thermal infrared remote sensing data is directly related to temperature, it is often used to infer LST data. Currently, international satellites with thermal infrared detection capabilities include Terra / Aqua, MSG, NOAA, Landsat, ASTR, NPP, and Himawari. Domestic satellites carrying thermal infrared sensors include the Environmental Satellite (HJ) series and the Fengyun series (FY-2, FY-3, and FY-4).

[0003] However, surface temperatures derived from thermal infrared remote sensing data typically represent surface temperatures within a range of tens of meters to several kilometers. When different satellites observe the same area from different angles, they may observe different surface components, resulting in discrepancies in remote sensing surface temperature measurements. This discrepancy is known as the angle effect. This angle effect increases inconsistencies between surface temperature products from different satellites, posing challenges to the application of multi-source surface temperature data, such as data assimilation and cross-validation.

[0004] Commonly used methods for normalizing the angular effect of surface temperature include component temperature synthesis based on vegetation cover, machine learning-based methods, and kernel-driven models. However, the component temperature synthesis method based on vegetation cover faces the difficulty of accurately calculating the temperatures of each surface component, while the machine learning-based method relies on accurate and extensive surface temperature data. Therefore, semi-empirical kernel-driven models have been widely used in research on normalizing the angular effect of surface temperature. However, kernel-driven models also require multiple sets of simultaneous observational data for the same area, which greatly limits their application. Therefore, it is necessary to rationally simplify the kernel-driven model based on its physical meaning to achieve a balance between accurate application and ease of solution. Summary of the Invention

[0005] The purpose of the present invention is to address the problem that existing kernel-driven models require multiple sets of observation data for the same area at the same time. Based on the physical meaning of the kernel-driven model parameters and combined with the diurnal temperature cycle (DTC) model of the surface temperature, an angle effect correction method is provided to simplify the solution process of the kernel-driven model, thereby efficiently and accurately correcting the surface temperature angle effect to achieve surface temperature angle normalization.

[0006] The technical solution adopted in the present invention is:

[0007] A method for correcting angle effects of thermal infrared land surface temperature data from a geostationary meteorological satellite, the method comprising the following steps:

[0008] Satellite surface temperature data preprocessing steps:

[0009] Obtain surface temperature data from two geostationary satellites in the target area and convert them to the target coordinate system;

[0010] Calculate the angle data of each pixel in the target coordinate system, including: satellite zenith angle, satellite azimuth angle, solar zenith angle, and solar azimuth angle;

[0011] Preferably, in this step, the coordinate system of one of the geostationary satellites is selected as the target coordinate system, and then the longitude and latitude of each pixel can be calculated based on the satellite sub-satellite point, orbital altitude and other data, and the projection information of the two surface temperature data can be reprojected to unify them; that is, the surface temperature data of the non-target coordinate system is reprojected to the target coordinate system to achieve the unification of the geospatial information of the surface temperature data of the two geostationary satellites;

[0012] Preferably, reprojecting the surface temperature data of another satellite to the target coordinate system specifically includes:

[0013] 1) Find the row and column numbers of the pixel points in the target area in the satellite surface temperature data;

[0014] 2) Based on the satellite's orbital position and orbital radius, calculate the observation angle θ of the two geostationary satellites to the target area;

[0015] 3) Calculate the longitude and latitude coordinates of each pixel based on the pixel row and column number and observation angle θ of the geostationary satellite;

[0016] 4) For each pixel of the satellite in the non-target coordinate system, find the pixel with the closest geographical distance in the projection grid, and project the surface temperature data of the nearest pixel to the current pixel.

[0017] The systematic error elimination step is performed based on the linear regression relationship between the surface temperature data of the two satellites:

[0018] According to the satellite zenith angle, a screening area is selected where the zenith angles of two satellites are less than or equal to the target angle;

[0019] In the screening area, a linear regression relationship is established between the surface temperature data of the two geostationary satellites, using the surface temperature data of one of the satellites as a reference value; based on the linear regression relationship, the surface temperature data of the other satellite is corrected, that is, the surface temperature data that is not the reference value is corrected; that is, the established linear regression relationship is used to eliminate the systematic error between the surface temperature data of the two satellites;

[0020] Because the difference in the surface temperature data of the two satellites in the screening area is mainly due to the sensor, inversion algorithm, etc., a linear regression is performed in the screening area to eliminate the systematic difference. Preferably, the expression of the linear regression relationship is: LST a =k*LST b +d, where LST represents the land surface temperature data, the subscripts a and b are used to distinguish the two satellites, and k and d represent the slope and intercept, respectively; that is, in the present invention, the land surface temperature data of one of the satellites is selected as the reference value, and then the data is used to establish a linear regression relationship to eliminate the systematic error of the two satellite data. Preferably, when the coordinate system of one of the geostationary satellites is selected as the target coordinate system, the land surface temperature data of the satellite is correspondingly selected as the reference value, corresponding to LST in the expression a ; Then, the obtained linear regression expression is applied to the surface temperature data of the screened area to obtain a multi-angle observation surface temperature dataset;

[0021] Preferably, in this step, when screening out a screening area where two satellite zenith angles are less than or equal to the target angle based on the satellite zenith angles, the target angle is set to 5±Δ1 degrees, where Δ1 is the allowable angle deviation.

[0022] The kernel-driven model is simplified based on nighttime surface temperature data and the calculation steps for the emissivity kernel coefficient A are performed:

[0023] In the target area, the nighttime surface temperature data are screened based on the solar zenith angle. Since there is no light-shadow effect at night, the kernel-driven model is simplified based on the nighttime surface temperature data as follows:

[0024]

[0025] Where, T N represents the temperature observed at the zenith, θ v ,θ s and are the observation zenith angle, solar zenith angle and sun-satellite relative azimuth angle, respectively. Indicates the angle parameter The emissivity kernel under ;

[0026] At night, for two T in different directions at the same time, a,t1 and T b,t1 It can be expressed by formulas (3) and (4):

[0027]

[0028] The emissivity kernel coefficient A is affected by the composition of the surface components, so it can be considered a constant within a day. The coefficient A is calculated pixel by pixel based on formulas (3) and (4); that is, the emissivity kernel coefficient A of each pixel is fitted based on the nighttime surface temperature of the same pixel at the same time from two satellites;

[0029] If the A value calculated for the day is abnormal due to missing data or low data quality, it is necessary to use other time data to fill in the gaps based on the characteristic that A does not change much within a few days. That is, based on the baseline range of coefficient A, the coefficient A of each pixel calculated is checked to see if it is abnormal. If the coefficient A of the current pixel is an abnormal value, the coefficient A of the next few days near the current pixel is selected to fill in the gaps. Preferably, the coefficient A of the 4 days before and after the current pixel is selected, that is, the selected filling time threshold is set to 8 days. After the filling process, a more complete data distribution map of each pixel A can be obtained.

[0030] The steps for calculating the angle-normalized surface temperature at night based on the emissivity kernel coefficient A are as follows:

[0031]

[0032] in, represents the angle-normalized surface temperature at night;

[0033] Preferably, in this step, the surface temperature data with a solar zenith angle greater than or equal to a nighttime angle reference value is filtered as the nighttime surface temperature data, where the nighttime angle reference value is set to 90±Δ2 degrees, where Δ2 is the allowable angle deviation. The remaining surface temperature data is then filtered based on the nighttime surface temperature data to obtain the daytime surface temperature data.

[0034] The following steps are used to calculate the angle-normalized surface temperature during the day based on the DTC model and the emissivity kernel parameter A:

[0035] First, the daytime surface temperature data is filtered out based on the solar zenith angle, and the solar nuclear coefficient k and hotspot height ΔT of each pixel are calculated based on the daytime surface temperature difference of the same pixel at adjacent times between the two satellites. HS ;

[0036] Characterization of daytime surface temperature data based on a kernel-driven model including hotspot effects:

[0037]

[0038] Among them, e is the natural base, k is the solar core coefficient representing the width of the hotspot, d represents the relative position between the sun and the satellite, and the hotspot height ΔT HS Characterizes the difference between the hotspot temperature and the temperature observed at the zenith;

[0039] Among them, for adjacent time t i and t i+1 , calculate the daytime surface temperature difference between adjacent moments based on the DTC model:

[0040]

[0041] Among them, T a It represents the difference between the maximum temperature and the initial temperature T0 of the DTC model, that is, the temperature variation in a day, ω represents the width of the cosine half cycle; t m Indicates the time when the temperature reaches its peak in a day (this parameter is the data to be fitted, and the daytime LST data can be used to fit each pixel);

[0042] Based on the fact that the solar nuclear coefficient K does not change with time, the hot spot height ΔT HS The criterion that does not change with the observation angle is based on The solar nuclear coefficient kK and hotspot height ΔT are solved by the daytime surface temperature difference of at least three consecutive moments. HS ; That is, in the present invention, the solar nuclear coefficient k and hotspot height ΔT are fitted based on the DTC model HS .

[0043] For example, based on Jointly solve the solar nuclear coefficient k and hotspot height ΔT HS , where t i-1 is time t i The next moment, They represent satellites s1 and s2 at time t i+1 The daytime surface temperature is characterized based on formula (6). That is, in the calculation, the solar nuclear coefficient k in formula (6) is a constant, and the hot spot height ΔT HS The hotspot heights ΔT of the two satellites at the same moment change only with time. HS The value is the same; finally, the hot spot height ΔT at multiple times is solved HS .

[0044] Then, based on the emissivity coefficient A, the solar coefficient K and the hotspot height ΔT HS , the angle-normalized surface temperature during the day is calculated based on the kernel-driven model including the hotspot effect. Then, based on the angle-normalized surface temperature T at night and during the day, N night 、TN day Get the vertical observation of the surface temperature for each pixel.

[0045] The technical solution provided by the present invention brings at least the following beneficial effects:

[0046] This paper proposes a comprehensive method for correcting the angular effect of thermal infrared land surface temperature based on a DTC-kernel-driven model. Compared to the commonly used kernel-driven model method, this method can accurately correct the angular effect of land surface temperature using only land surface temperature data from two geostationary meteorological satellites. This method reduces the difficulty of solving the kernel-driven model and overcomes the high data requirements and difficulty in solving the existing kernel-driven model. This method lays the foundation for more accurate and efficient use of thermal infrared land surface temperature data. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 paying any creative work.

[0048] Figure 1 A schematic flow chart of a method for correcting angle effects of thermal infrared land surface temperature data from geostationary meteorological satellites provided by an embodiment of the present invention;

[0049] Figure 2 A DEM image of the target area according to an embodiment of the present invention;

[0050] Figure 3 is the linear regression relationship between two satellites in the observation angle proximity region in an embodiment of the present invention;

[0051] Figure 4 An image of parameter A (emissivity kernel coefficient A) obtained by solving the target area in an embodiment of the present invention;

[0052] Figure 5 is the night-time angle-normalized thermal infrared remote sensing surface temperature data in the embodiment of the present invention, wherein (5a) is the temperature image before correction, and (5b) is the image after correction;

[0053] Figure 6 is the daytime angle-normalized thermal infrared remote sensing surface temperature data in the embodiment of the present invention, wherein (6a) is the temperature image before correction, and (6b) is the image after correction;

[0054] Figure 7 The figure shows the test results of the angle-normalized LST product using Landsat LST data in the examples of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be described in detail and completely in conjunction with the drawings in the implementation of the present invention. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings can be arranged and designed using different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present invention.

[0056] The embodiment of the present invention provides a method for correcting the angle effect of thermal infrared surface temperature data from a geostationary meteorological satellite. Figure 1 The method proposed in the embodiment of the present invention first obtains the geostationary satellite thermal infrared remote sensing surface temperature data of the target area, and pre-processes it to convert it into the target coordinate system, such as the commonly used WGS-84 projection; then, the satellite zenith angle, satellite azimuth angle, solar azimuth angle and solar zenith angle of each pixel are calculated based on the satellite sub-satellite point data; then, the area where the difference in observation angles of the two satellites is small is screened out based on the satellite zenith angle, and the nighttime surface temperature data of the area is obtained based on the solar zenith angle, and then a linear regression relationship is established using the data to eliminate the systematic error of the two satellite data; next, the nighttime surface temperature data is screened out in the target area, and the kernel-driven model is simplified by using the characteristic that the nighttime surface temperature data has no hotspot effect, and the emissivity kernel parameter A (also called the emissivity kernel coefficient) of the kernel-driven model is solved based on the simplified kernel-driven model; then, during the day, reasonable assumptions are made based on the physical meaning of each parameter of the kernel-driven model, assuming that the solar kernel coefficient K does not change with time and the hotspot height ΔT HS The kernel-driven model is reasonably simplified and does not change with the observation angle. Then, the surface temperature data of two satellites at multiple consecutive moments (preferably 3 moments) are used, and the emissivity kernel parameter A obtained previously is combined to solve K and ΔT. HS Finally, the kernel driven model coefficients are used to solve the vertical observation T of each pixel. N The embodiment of the present invention is based on the thermal infrared kernel-driven model, fully considering the physical meaning of each parameter of the kernel-driven model. It can quickly and accurately solve each parameter of the kernel-driven model and then accurately solve the vertically observed surface temperature. It overcomes the defects of traditional methods such as large data requirements, low computational efficiency, and low accuracy, and can serve a large number of thermal infrared remote sensing surface temperature application scenarios.

[0057] As a possible implementation, an embodiment of the present invention provides a method for correcting angle effects of thermal infrared land surface temperature data from a geostationary meteorological satellite, including the following steps:

[0058] Step 1: Obtain thermal infrared remote sensing surface temperature data of the target area

[0059] In the embodiment of the present invention, the experimental area is selected as Sichuan Province and its surrounding areas. In the WGS-84 projection coordinate system, the specific coordinate interval is: (97.35°E-108.57°E, 26.05°N-34.31°N). The experimental data used are the thermal infrared surface temperature data of the geostationary meteorological satellite GK-2A provided by the Korea Meteorological Administration and the Himawari-8 thermal infrared surface temperature data produced by Chiba University in Japan, both of which have a spatial resolution of 2km. The surface temperature dataset used for verification is Landsat LST data, which has a spatial resolution of 30m. Since its observation zenith angle is close to 0° and its spatial resolution is high, it is suitable for result verification.

[0060] (1) Download GK-2A LST data and Himawari-8 LST data with a spatial resolution of 2 km in the target area, with a time range of July to August 2023. During this period, the target area has relatively less cloud cover, making it more suitable for subsequent angle normalization experimental research. The DEM of the experimental area is as follows: Figure 2 shown.

[0061] (2) Since the geostationary satellite LST data uses a disk projection, which is different from the commonly used WGS-84 projection, each pixel does not have specific longitude and latitude information. Therefore, before processing, it is necessary to calculate the longitude and latitude of each pixel based on the row and column numbers and the longitude and latitude lookup table. The specific steps are as follows:

[0062] 1) Obtain geostationary satellite data and find the row and column numbers of the required pixel points in the data.

[0063] 2) Calculate the sun-synchronous orbit of a geostationary satellite and determine the satellite's orbital position and orbital radius.

[0064] 3) Using the Earth’s longitude and latitude and the satellite’s orbital position, calculate the observation angle θ of the target area.

[0065] 4) Calculate the longitude and latitude coordinates of the pixel point based on the pixel row and column number and θ value of the geostationary satellite.

[0066] (3) Since the two geostationary satellites have different subsatellite points, their projection grids are also different, so the projection grids of the two satellites need to be unified. In this example, since the subsatellite point of the GK-2A satellite is closer to the target area than that of the Himawari-8 satellite, the Himawari-8 LST data is projected onto the GK-2A LST projection grid. The specific method is: for each pixel in the Himawari-8 LST data, find the pixel with the closest geographical distance in the GK-2A projection grid and project the Himawari-8 LST data onto that pixel. After this process, the two LST data will have the same projection information.

[0067] Step 2: Eliminate the systematic error between the two satellites

[0068] Through step 1, LST data with the same projection information can be obtained. However, in addition to the different projection information, the two different LST data may also have differences caused by factors such as the sensor and the inversion algorithm. Therefore, it is necessary to eliminate these differences to better analyze the errors caused by the observation angle. The specific method for eliminating errors in the embodiment of the present invention is as follows:

[0069] (1) The satellite zenith angle, satellite azimuth angle, solar zenith angle, and solar azimuth angle are calculated pixel by pixel based on the satellite subsatellite point, orbit information, and solar orbit information. The areas where the two satellites observe a smaller zenith angle (θv ≤ 5°) are selected. In this example, the specific areas are Japan and its surrounding areas and northern Australia.

[0070] (2) Using GK-2A LST as the benchmark value in the region, a linear regression relationship between GK-2A LST and Himawari-8 LST was established. GK2A =a*LST Himawari +b, linear regression relationship is as follows Figure 3 The established linear regression relationship was applied to the Himawari-8LST data, and the Himawari-8LST data with systematic differences eliminated were obtained.

[0071] Step 3: Calculate the emissivity kernel parameter A:

[0072] The core model used in this example is the VIN-RL model, whose formula is:

[0073]

[0074] Where, T N represents the temperature observed at the zenith, θ v ,θ s and are the observation zenith angle, solar zenith angle and sun-satellite relative azimuth angle respectively; ΔT HS Indicates the hot spot temperature THS With T N The difference between the satellite and the sun; d is used to represent the relative position between the satellite and the sun. The d value of different satellites and the sun will change at different times. The numerical subscripts will be used to distinguish the relative positions at different times. k is the solar core K RL Hotspot width parameters; A, ΔT HS 、T N and k are the values ​​to be solved. To ensure the accuracy of the solved parameters, this example adopts a step-by-step solution method.

[0075] At night, since there is no hot spot, ΔT in formula (8) HS The term can be omitted and the kernel-driven model can be rewritten as:

[0076]

[0077] According to the solar zenith angle (θ s ≥90°) to filter out the nighttime LST data. According to formula (12) and the observation values ​​of two satellites on the same area at the same time t1, equations (13) and (14) are obtained:

[0078]

[0079] According to (13) and (14), the A of a specific pixel can be calculated. Since the A value is affected by the ground components of the observation area, it can be assumed that it does not change much in a short period of time. Therefore, the A value obtained by solving the nighttime LST data can be used to correct the angle effect throughout the day. If there is a pixel that has insufficient data or poor data quality due to cloud cover on a certain day, resulting in missing or abnormal A values, the A values ​​of the pixel in the nearby days are used to fill the gap. In this example, the threshold of the number of days is selected as 4 days before and after. The reason for selecting this threshold is that the normalized vegetation index, which is often used to reflect the ground component situation, usually has a time resolution of 8 days. Figure 4 The A calculation results for the target area are shown.

[0080] Step 4: DTC key parameter fitting

[0081] The daytime LST data are filtered out according to the solar zenith angle. The value A in formula (8) has been determined in step 3. The observation values ​​of the two satellites at time t1 during the day can be expressed as:

[0082]

[0083] After simplification, we can get:

[0084]

[0085] At time t2, we can get:

[0086]

[0087] T N2 With T N1 The difference can be calculated using the Diurnal Temperature Cycle (DTC) model.

[0088]

[0089] Define T N3 represents the surface temperature at time t3, then T N3 With T N2 The difference can also be obtained:

[0090]

[0091] The formula for the DTC model is:

[0092]

[0093] Where, T s (t) represents the change of surface temperature during the day; T0 is the initial temperature near sunrise; T a is the difference between T0 and the maximum temperature, that is, the temperature variation during the day; ω is the width of the cosine half cycle; t m is the time when the temperature reaches its peak in a day, t m The data to be fitted can be fitted pixel by pixel using the daytime LST data; s is the moment when the temperature starts to decay; δT is the difference between T0 and T(t→∞); k is the decay coefficient, and the parameter ω can be calculated from the solar geometry:

[0094]

[0095] Where λ is the latitude and δ is the solar inclination, which is calculated by DOY:

[0096] δ=23.45sin(0.986(284+DOY)) (25)

[0097] Step 5: Parameters k and ∆T HS calculate:

[0098] Combining formulas (17) and (18) we can get:

[0099]

[0100] Similarly, the equations at time t2 and t3 can be obtained:

[0101]

[0102] At the same time, the t obtained in step 4 can be used m Calculate (T N3 -T N2 ) / (T N2 -T N1 ):

[0103]

[0104] For a single pixel, equations (26), (27), (28), (29) and (30) can be combined to obtain K and △T HS .

[0105] Step 6: Normalized surface temperature calculation:

[0106] The solar zenith angle of each pixel is used to determine whether the pixel is at night. If it is at night, the normalized surface temperature is calculated using a simplified kernel-driven model:

[0107]

[0108] If the pixel is during the day, a kernel-driven model that includes hotspot effects is used for the calculation:

[0109]

[0110] Figure 5 The normalized results of the nighttime surface temperature angle in the target area of ​​this example are shown. Figure 6 The normalized results of the daytime surface temperature angle in the target area of ​​this example are shown. Figure 7 The comparison between the angle-normalized surface temperature obtained in this example and the Landsat LST surface temperature is shown.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

[0112] The above are only some embodiments of the present invention. For those skilled in the art, several modifications and improvements can be made without departing from the inventive concept of the present invention, which all fall within the scope of protection of the present invention.

Claims

1. A method for correcting angle effects of thermal infrared surface temperature data from geostationary meteorological satellites, characterized in that: The following steps are involved: Satellite surface temperature data preprocessing steps: Obtain surface temperature data from two geostationary satellites in the target area and convert them to the target coordinate system; Calculate the angle data of each pixel in the target coordinate system, including: satellite zenith angle, satellite azimuth angle, solar zenith angle, and solar azimuth angle; The systematic error elimination step is performed based on the linear regression relationship between the surface temperature data of the two satellites: According to the satellite zenith angle, a screening area is selected where the zenith angles of two satellites are less than or equal to the target angle; In the screening area, the surface temperature data of one of the satellites is used as a reference value to establish a linear regression relationship between the surface temperature data of the two geostationary satellites; based on the linear regression relationship, the surface temperature data of the other satellite is corrected; The kernel-driven model is simplified based on nighttime surface temperature data and the calculation steps for the emissivity kernel coefficient A are performed: The nighttime surface temperature data are filtered based on the solar zenith angle in the target area, and the kernel-driven model is simplified as follows: in, Indicates the angle parameter Nighttime surface temperature under N represents the temperature observed at the zenith, θ v ,θ s and are the observation zenith angle, solar zenith angle and sun-satellite relative azimuth angle, respectively. Indicates the angle parameter The emissivity kernel under ; The emissivity kernel coefficient A of each pixel is fitted based on the nighttime surface temperature of the same pixel at the same time from two satellites; The steps for calculating the angle-normalized surface temperature at night based on the emissivity kernel coefficient A are as follows: in, represents the angle-normalized surface temperature at night; The following steps are used to calculate the daytime angle-normalized surface temperature based on the surface temperature daily cycle model and the emissivity kernel coefficient A: The daytime surface temperature data is filtered out based on the solar zenith angle, and the solar nuclear coefficient k and hotspot height ΔT of each pixel are calculated based on the daytime surface temperature difference of the same pixel at adjacent moments between two satellites. HS : Characterization of daytime surface temperature data based on a kernel-driven model including hotspot effects: Among them, e is the natural base, k represents the solar nuclear coefficient used to characterize the hotspot width, d represents the relative position between the sun and the satellite, and the hotspot height ΔT HS Characterizes the difference between the hotspot temperature and the temperature observed at the zenith; For adjacent time t i and t i+1 , calculate the daytime surface temperature difference between adjacent moments based on the surface temperature daily cycle model: Among them, T a It represents the difference between the maximum temperature and the initial temperature T0 of the DTC model, that is, the temperature variation in a day, ω represents the width of the cosine half cycle; t m Indicates the time of day when the temperature reaches its peak; Based on the fact that the solar nuclear coefficient k does not change with time, the hot spot height ΔT HS The criterion that does not change with the observation angle is based on The solar nuclear coefficient k and hotspot height ΔT are solved by the daytime surface temperature difference of at least three consecutive moments. HS ; Based on the emissivity nuclear coefficient A, the solar nuclear coefficient k and the hotspot height ΔT HS , calculates the daytime angle-normalized surface temperature based on a kernel-driven model that includes hotspot effects.

2. The method according to claim 1, wherein In the satellite surface temperature data preprocessing step, the coordinate system of one of the geostationary satellites is selected as the target coordinate system, and then the surface temperature data of the other satellite is reprojected to the target coordinate system.

3. The method according to claim 2, wherein In the satellite surface temperature data preprocessing step, the surface temperature data of another satellite is reprojected to the target coordinate system, specifically including: 1) Find the row and column numbers of the pixel points in the target area in the satellite surface temperature data; 2) Based on the satellite's orbital position and orbital radius, calculate the observation angle θ of the two geostationary satellites to the target area; 3) Calculate the longitude and latitude coordinates of each pixel based on the pixel row and column number and observation angle θ of the geostationary satellite; 4) For each pixel of the satellite in the non-target coordinate system, find the pixel with the closest geographical distance in the projection grid, and project the surface temperature data of the nearest pixel to the current pixel.

4. The method according to claim 3, wherein When establishing a linear regression relationship between the surface temperature data of two geostationary satellites, the surface temperature data of the satellite corresponding to the target coordinate system is used as the reference value.

5. The method according to claim 1, wherein When filtering out a screening area where two satellite zenith angles are less than or equal to a target angle according to the satellite zenith angles, the target angle is set to 5±Δ1 degrees, where Δ1 is an allowable angle deviation.

6. The method according to claim 1, wherein The calculation steps for solving the emissivity kernel coefficient A also include: The coefficient A of each pixel calculated based on the reference range of parameter A is detected to see if it is abnormal. If the coefficient A of the current pixel is an abnormal value, the coefficient A of the next few days near the current pixel is selected to fill in and correct the coefficient A of the current pixel.

7. The method according to claim 6, wherein When the coefficient A of a pixel is padded and corrected, it is padded and corrected based on the coefficient A of the current pixel 4 days before and after.

8. The method according to claim 1, wherein The surface temperature data with a solar zenith angle greater than or equal to the nighttime angle reference value are filtered as nighttime surface temperature data, where the nighttime angle reference value is set to 90±Δ2 degrees, and Δ2 is the allowable angle deviation.

Images