A method for generating global surface temperature based on a pixel-by-pixel daily temperature cycle model

Through the image-by-cell intraday temperature cycle model and the Tyson polygon method, combined with global meteorological stations and remote sensing data, the problems of low time resolution and cloud coverage of traditional remote sensing data acquisition methods are solved, and the high accuracy of global surface temperature data is achieved. Efficient calculation.

CN118761236BActive Publication Date: 2024-11-15AEROSPACE INFORMATION RES INST CAS
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411237248.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-05
Publication Date
2024-11-15
Estimated Expiration
2044-09-05

AI Technical Summary

Technical Problem

The traditional remote sensing surface temperature data acquisition method has low time resolution and cannot continuously capture dynamic changes in surface temperature. The cloud cover leads to missing data. The existing intraday temperature cycle model uses the same model parameters in large areas, resulting in errors and cannot accurately reflect them. Data characteristics in different regions around the world.

Method used

Using a cell-by-cell intraday temperature cycle model, combining global meteorological station temperature data, surface type data and remote sensing surface temperature inversion data, data classification and storage are carried out through the Tyson polygon method, and a regression model is constructed to calculate the number of each cell. Unique parameters to improve data accuracy and efficiency.

Benefits of technology

It has achieved high temporal resolution and accuracy of global cell-by-cell surface temperature data, improved data search efficiency and accuracy, and expanded the application space and time coverage of intraday temperature cycle models.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118761236B_ABST
    Figure CN118761236B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for generating global surface temperature based on a pixel-by-pixel intraday temperature cycle model, and belongs to the field of remote sensing application technology. The method comprises using Thiessen polygons to classify and spatialize weather station data, and using remote sensing images, weather stations and surface type data to construct a regression model of the parameters of the pixel-by-pixel intraday temperature cycle model, and finally using the pixel-by-pixel intraday temperature cycle model to generate a global hourly surface temperature data set. The present invention is conducive to improving the accuracy of a large-scale surface temperature survey, and has unique advantages in obtaining a large-scale surface temperature data set with high temporal and spatial resolution.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of remote sensing application, and in particular relates to a method for generating global surface temperature based on a pixel-by-pixel intraday temperature cycle model. Background Art

[0002] In recent years, surface temperature data has been widely used in climate change research, agricultural management, environmental monitoring, and disaster prevention and mitigation. At the same time, autonomous satellite operation is becoming increasingly important in satellite design. The satellite can autonomously and accurately set the imaging parameters of visible light and infrared spectrum, which can improve the quality of remote sensing data and greatly reduce the complexity of satellite operation control. Since the entrance pupil signal of the infrared channel of the space camera is inseparable from the temperature, emissivity and other characteristics of the surface objects, and has the characteristics of a wide dynamic range and high spatiotemporal energy variability, the construction of surface temperature data can provide support for the autonomous estimation of the entrance pupil radiation signal of the infrared channel of the onboard space camera and the selection of the optimal imaging parameters.

[0003] Traditional methods for acquiring remote sensing surface temperature data mainly rely on satellite observations. However, these methods usually have the disadvantage of low temporal resolution and cannot continuously capture the dynamic changes of surface temperature. In addition, cloud cover is a major challenge in remote sensing data acquisition. Clouds block satellite observations, resulting in the loss of surface temperature data. These problems limit the application of traditional remote sensing data in high temporal resolution and continuous monitoring.

[0004] In order to make up for these deficiencies, researchers have proposed using interpolation methods to fill in the missing data. This method uses spline functions, universal kriging interpolation and other methods to calculate and fill in the missing data caused by short-term cloud cover, thereby improving the temporal resolution of surface temperature data. However, interpolation methods have certain limitations when dealing with complex surface temperature changes and are prone to errors, especially in periods and areas with drastic temperature changes. In addition, this method cannot accurately reflect the nonlinear characteristics of surface temperature in diurnal changes.

[0005] In contrast, the intraday temperature cycle model provides a more accurate and effective method for generating high temporal resolution surface temperature data. Using these models, model parameters closely related to the thermal properties of the surface can be obtained by fitting the measured surface temperature data. These models usually use sine, cosine or hyperbolic functions to describe the changes in surface temperature during the day and night, which can better reflect the diurnal variation of surface temperature. However, the method of establishing intraday temperature cycle models according to surface types in the prior art makes a large area use the same model parameters. For example, the same surface type may cover an area of ​​more than tens of thousands of square kilometers, and such a large area only uses one model parameter, which may cause errors; at the same time, a lookup table is used to store data of a meteorological station corresponding to different objects. If this method is used to generate surface temperature data on a global scale, no distinction is made between data from different regions of the world, which is not in line with reality and may also cause errors. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a method for generating global surface temperature based on a pixel-by-pixel intraday temperature cycle model. The global weather station temperature data, surface type data and remote sensing surface temperature inversion data are used to construct a pixel-by-pixel intraday temperature cycle model, which realizes the parameter differentiation between different surface types and different pixels, and improves the accuracy of constructing monthly global hourly surface temperature data; the method of using surface type and weather station location to establish Thiessen polygons improves the efficiency and accuracy of data selection when establishing the model, and at the same time expands the spatial and temporal coverage of the application of the intraday temperature cycle model.

[0007] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0008] A method for generating global surface temperature based on a pixel-by-pixel daily temperature cycle model, the method comprising the following steps:

[0009] Step 1: Data preprocessing: obtain remote sensing temperature data, weather station temperature data, and surface type data and preprocess them respectively;

[0010] Step 2: Spatial classification of weather station temperature data based on Thiessen polygons: According to the preprocessed surface type data, the preprocessed weather station temperature data is divided into eight surface types and grouped. A TIN triangulated network is generated based on each group of weather station temperature data, and the corresponding Thiessen polygons are constructed.

[0011] Step 3: Temperature model construction and parameter calculation based on regression analysis: A regression model is constructed based on the intraday temperature cycle model, and the pre-processed remote sensing temperature data is obtained pixel by pixel by combining the Thiessen polygons of the temperature data of the meteorological station. The regression parameters are iteratively calculated and stored according to the month and pixel position.

[0012] Step 4: Calculation of surface temperature data based on the intraday temperature cycle model: perform regression parameter search pixel by pixel, input the found regression parameters into the regression model, and output global surface temperature data for all months and all times.

[0013] The beneficial effects of the present invention are:

[0014] By using global weather station temperature data, surface type data and remote sensing surface temperature inversion data to establish a daily temperature cycle model, it is possible to use different parameters for temperature calculation for each pixel around the world; the Thiessen polygons based on surface type and spatial location of weather stations are used for data storage, classification and search, which improves the efficiency and accuracy of data search and enables rapid selection of weather station data of the same surface type in the nearest area when searching for spatial data. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic flow chart of a method for generating global surface temperature based on a pixel-by-pixel intraday temperature cycle model of the present invention;

[0016] Figure 2 Thiessen polygon grouping structure diagram constructed for the present invention;

[0017] Figure 3 This is a schematic diagram of a temperature sample at 14:00 on January based on the results of the present invention. DETAILED DESCRIPTION

[0018] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0019] like Figure 1 As shown, the present invention provides a method for generating global surface temperature based on a pixel-by-pixel daily temperature cycle model. Based on the daily temperature cycle model, the surface temperature data inverted from satellite remote sensing data, the temperature data of ground meteorological stations, and the surface type data are coupled to achieve pixel-by-pixel establishment of a daily temperature cycle model to improve the calculation accuracy of the parameters of the daily temperature cycle model; the parameter accuracy of the daily temperature cycle model is improved by refining the ground station data and surface type classification by establishing a Thiessen polygon method, and the spatial and temporal scope of the application of the daily temperature cycle model is expanded. The specific implementation process of the present invention is introduced step by step below:

[0020] Step 1: Data preprocessing

[0021] Step 1.1: Remote sensing data preprocessing

[0022] Step 1.1.1. Obtain the MODIS L1B data at four times (11:00, 13:00, 23:00, 1:00) of each month in the past two years to obtain the surface temperature data of the remote sensing source;

[0023] Step 1.1.2, obtain the corrected remote sensing data that can be used for surface temperature data inversion through radiometric calibration and FLASH rapid atmospheric correction, geometric correction, and geocoding;

[0024] Step 1.1.3, the surface temperature inversion data of land and ocean are obtained by split window algorithm and nonlinear multi-channel algorithm, and fused by weight-based data fusion method to obtain the remote sensing data of surface temperature inversion of two years at four moments of each month;

[0025] Step 1.1.4: Based on the above steps, the remote sensing data inverted from the four-moment monthly surface temperature data for two years are averaged to obtain the remote sensing data inverted from the four-moment remote sensing surface temperature data after averaging.

[0026] Step 1.2: Preprocessing of temperature data from weather stations

[0027] Step 1.2.1, obtain the daily global weather station NCDC temperature data for the past five years;

[0028] Step 1.2.2, organize the acquired weather station temperature data into a data table classified by date, time, and weather station ID code;

[0029] Step 1.2.3, unify the time interval of the weather station temperature data into three hours, a total of eight times (2:00, 5:00, 8:00, 11:00, 14:00, 17:00, 20:00, 23:00), and exclude abnormal values;

[0030] Step 1.2.4: First, average the daily data of the weather station temperature data for five years by year to obtain the annual average weather station temperature data for 8 moments every day. Then, average these weather station data by month to obtain the monthly average weather station temperature data for 8 moments every month.

[0031] Step 1.3: Preprocessing of surface type data

[0032] Step 1.3.1, obtain GL30 (GlobeLand30) surface type data online;

[0033] Step 1.3.2: Merge the vegetation types in the surface type data into 8 types (water bodies, forests, grasslands, wetlands, farmlands, cities, glaciers, deserts, and bare land).

[0034] Step 2: Spatial classification method of weather station temperature data based on Thiessen polygons

[0035] Step 2.1: Meteorological station temperature data grouping method based on spatial attributes

[0036] In order to use the temperature data obtained from the weather station for spatial analysis, it must be processed and classified into data sets with different dimensional attributes. In this way, the grouped data can be quickly located and searched according to the attributes to clarify the values ​​of its specific attributes.

[0037] First, the method performs spatial overlay analysis on the meteorological station temperature data that have been processed as monthly averages. These data include the latitude and longitude of the meteorological station and the surface classification data. Through overlay analysis, the surface type to which each meteorological station belongs can be determined.

[0038] According to the surface classification data, the temperature data of the meteorological station is divided into eight surface types. Let the temperature data of the meteorological station be T, and let c represent the surface type of the meteorological station L. The temperature data of the meteorological station is divided into eight groups according to the surface type c of the meteorological station:

[0039]

[0040] The data of each land surface type are divided into 12 groups according to the month. Let the data of the jth month be T Lcj ,but:

[0041] ,

[0042] Among them, TLcj represents the grouping level, and its elements represent the data range. The data of each month is divided into eight groups according to the specific time. When the specific time h is 2, 5, 8, 11, 14, 17, 20, 23, the data of each month is grouped as follows:

[0043] ,

[0044] Among them, the subscript h=[1,8] represents the hth specific moment.

[0045] In this way, it is possible to ensure that each temperature data is searched and traced according to the primary key sequence of surface type-month-time-station ID. In this way, each data point can be uniquely located and searched by its surface type, month, time and station ID. The temperature data of the weather station classified by surface type, month, time and station ID can be obtained.

[0046] Step 2.2: TIN generation method based on Delaunay triangulation

[0047] In order to use the temperature data of the weather station for spatial analysis, it is necessary to process the temperature data of the weather station into samples with spatial significance, and then extrapolate the samples to the entire space. The spatial extrapolation algorithm selected by this method is the Thiessen polygon method. Before creating the Thiessen polygon, the spatial position of the temperature data of the weather station needs to be processed into a TIN triangulated network, and then the Thiessen polygon of the temperature data of the weather station is created on this basis.

[0048] The most commonly used algorithm for TIN triangulation is Delaunay triangulation. The characteristics of this triangulation are that the triangulation has relatively perfect geometric properties, and its network construction rules are unique, and the generation algorithm is relatively stable. When used to generate Thiessen polygons, it is easier to find the smallest mesh unit that constitutes Thiessen polygons, making the values ​​of Thiessen polygons more representative. The most common algorithm in its generation algorithm is the point-by-point insertion algorithm, and its algorithm characteristic is that it can quickly construct the entire triangulation in sequence.

[0049] In this method, a TIN triangulation network is generated for each set of weather station temperature data to obtain Thiessen polygons. That is, the spatial location of the input weather station temperature data , output the corresponding TIN triangulation network TIN(S), Represents the point-by-point plane coordinates of the weather station. This process can be directly expressed by the following formula:

[0050] ,

[0051] Delaunay represents a point-by-point insertion algorithm of the Delaunay triangulation network, which is used to convert the input spatial position S into a corresponding TIN triangulation network.

[0052] Step 2.3: Thiessen polygon generation method based on TIN

[0053] like Figure 2 As shown in the figure, after obtaining the TIN triangulation of the weather station temperature data, the TIN triangulation can be used to construct Thiessen polygons. Thiessen polygons are also called Voronoi diagrams. Voronoi diagrams include a set of continuous polygons composed of perpendicular bisectors connecting two adjacent points. It has the following characteristics: (1) The distance from each point in the polygon to the generator where the point is located is shorter than the distance to other generators; (2) The distances from the points on the polygon boundary to the two generators that generate this boundary are equal; (3) The generator in the polygon is the center of the circumscribed circle of the triangle formed by the three points that form the three sides, and all these circumscribed circles do not contain any vertices other than these three points (hollow circle characteristics). These properties give it an irreplaceable advantage in replacing the numerical values ​​of the adjacent area for spatial analysis.

[0054] In this method, the TIN triangulation network of each set of weather station temperature data is used to generate Thiessen polygons, the primary key of which is the station ID, and the attribute is the temperature data at 8 specific times of the day. It is then converted into raster data consistent with the size of remote sensing pixels, and finally the attributes at 8 specific times are converted into raster values ​​of 8 bands. That is:

[0055] ,

[0056] Thiessen represents the Thiessen polygon generation algorithm, TIN(S) is the TIN triangulation network of the weather station location, Polygon( ) represents the Thiessen polygon corresponding to the site ID and its temperature data at 8 specific moments.

[0057] ,

[0058] Among them, Raster represents the raster conversion function, PixelSize is the size of the remote sensing pixel, that is, 0.01°, and RasterBand(B1,B2,...,B8) represents the corresponding raster data and the temperature values ​​of its 8 bands.

[0059] Thus, the weather station temperature data that can be used for spatial analysis can be obtained by searching and tracing according to the attribute order of surface type-month-time.

[0060] Step 3: Construction of temperature model and parameter calculation method based on regression analysis

[0061] Step 3.1: Construction of regression model based on intraday temperature cycle model

[0062] After obtaining the basic temperature data of the meteorological station that can be used for spatial analysis, the next step is to fit the temperature change through regression to achieve time normalization. To fit the intraday temperature and regress it, it is necessary to select a physical or semi-empirical model to simulate the change of the intraday temperature curve. This method selects the intraday temperature cycle model to simulate the intraday temperature change. The intraday temperature cycle model (DTC) uses a piecewise function to describe the change of surface temperature over time during the day. The intraday temperature cycle model uses different function expressions to characterize the change of surface temperature in different time periods of the day. This method selects the intraday temperature cycle model GOT01 proposed by Göttsche. Based on the heat conduction equation, the change of daytime surface temperature over time is expressed by a cosine function. Other intraday temperature cycle models such as the GOT09 model and the INA08 model can also be used as alternatives. Assuming that the change of surface temperature obeys Newton's Law of Cooling, the change of surface temperature at night is described by an exponential function. The model formula is as follows:

[0063] ,

[0064] Where TEM1 / TEM2 are the two parts of the model representing the daytime / nighttime temperatures respectively. Where t is the specific time in the 24-hour system, TEM0 is the residual temperature before and after sunrise, and TEM a is the temperature amplitude, tm The time when the temperature reaches the maximum value, t s is the time when the temperature starts to decay freely, ΔTEM is the difference between TEM0 and the temperature value TEM at the end time, the end time represents the last value of the specific time t in the 24-hour system, and k represents the decay constant. There are 5 free variables in the model: TEM0, TEM a , t m , t s and ΔTEM. ω in the model is the half-cycle width of the cosine function, defined as the duration of solar energy input, which is determined by the latitude φ and the solar inclination angle δ, and is expressed as:

[0065] ,

[0066] ,

[0067] Where doy is the annual accumulated day. Since the daily temperature cycle model curve is continuous, the temperature simulation function of day and night is in t s The derivatives at should be the same, from which the decay constant k can be obtained:

[0068] ,

[0069] Among them, the parameters that need to be fitted in the model are TEM0, TEM a , t m , t s and ΔTEM, while other parameters need to be derived by calculation. For example, if φ is the latitude, the mapping relationship between the remote sensing raster data encoding and the latitude needs to be obtained:

[0070] ,

[0071] Among them, row represents the number of rows of raster data, φ represents latitude, and Δφ represents the latitude resolution of the raster image. Among them, the pixel resolution of the raster data is 0.01 degrees, so the latitude of each pixel can be derived. Therefore, Δφ=0.01. Indicates rounding down.

[0072] doy is the annual cumulative date. It is necessary to obtain the mapping relationship between the basic data and the annual cumulative date. The mapping relationship is as follows:

[0073] ,

[0074] Among them, DOY j represents the median of the annualized day of the jth month. j Represents the day of the jth month.

[0075] Given the complexity of the expression of the daily temperature cycle model, it is not a linear equation, so the regression algorithm used is the Levenberg-Marquardt (LM) method, which is an optimization algorithm for nonlinear least squares problems. It combines the advantages of the gradient descent method and the Gauss-Newton method. This method is very effective in fitting nonlinear model parameters, especially in parameter estimation problems. Its mathematical description is as follows:

[0076] Given a nonlinear model f(x,β), where x is the independent variable and β is the parameter vector to be estimated. The goal is to find the parameter vector β to be estimated so that the nonlinear model f best fits the observed data y, that is, to minimize the following objective function:

[0077] ,

[0078] The Levenberg-Marquardt method combines the fast convergence of the Gauss-Newton method and the robustness of the gradient descent method, achieving a balance between the two.

[0079] Step 3.2: Regression data search based on attribute matching

[0080] The regression calculation is based on each independent pixel, that is, each instantaneous field of view (0.01°) of the remote sensing temperature data. To this end, it is necessary to search for the corresponding attributes pixel by pixel.

[0081] The basic data required for pixel-by-pixel regression are the temperature values ​​at four times corresponding to the remote sensing temperature data corresponding to the pixel in that month (11:00, 13:00, 23:00, 1:00) and the temperature values ​​at eight times corresponding to the Thiessen polygons of the location class of the pixel in that month (2:00, 5:00, 8:00, 11:00, 14:00, 17:00, 20:00, 23:00). For each pixel, first determine the specific month, then select the corresponding remote sensing temperature data, then select the corresponding Thiessen polygon group by the surface type, and finally determine the specific Thiessen polygon by the pixel location to obtain the temperature value at the specific time. In this way, 12 specific data for regression can be obtained.

[0082] Step 3.3: Calculation of regression parameters based on LM algorithm

[0083] Steps for calculating regression parameters pixel by pixel: (1) Calculate the latitude and annual accumulation day of regression parameters based on the pixel position and the corresponding month; (2) Find the 12 basic data corresponding to the pixel through regression data search; (3) Input the basic parameters and basic data into the daily temperature cycle model as the initial parameters of the LM algorithm for regression calculation and iteration; (4) Output the five parameters TEM0, TEM a , t m , t s and ΔTEM.

[0084] The parameters of the five regression fits are output and stored according to month and pixel location.

[0085] Step 4: Calculation method of surface temperature data based on daily temperature cycle model

[0086] Step 4.1: Parameter search based on pixel-by-pixel search method

[0087] When outputting the surface temperature data, it is still necessary to calculate it through the intraday temperature cycle model, and calculate the basic parameters latitude and annual accumulation day again according to the mapping relationship, and obtain the five fitted parameters by reading each pixel month by month and regressing before, and obtain all the complete parameters of the intraday temperature cycle model used to calculate the specific temperature.

[0088] Step 4.2: Calculation of surface temperature data based on the daily temperature cycle model

[0089] All parameters are input into the daily temperature cycle model, and the surface temperature (Kelvin) of all months at all times (0-23 o'clock) is calculated pixel by pixel, and a float raster image projected according to WGS84 is output according to the month-time classification, such as Figure 3 The temperature sample diagram shown is at 14:00 on January. This time is based on the local time of all pixels, that is, the solar altitude angle of all pixels in the same raster image is consistent. In this way, the surface temperature data of all months and all times are obtained.

[0090] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. 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 generating global surface temperature based on a pixel-by-pixel daily temperature cycle model, characterized in that: The method comprises the following steps: Step 1: Data preprocessing: obtain remote sensing temperature data, weather station temperature data, and surface type data and preprocess them respectively; Step 2: Spatial classification of weather station temperature data based on Thiessen polygons: According to the preprocessed surface type data, the preprocessed weather station temperature data is divided into eight surface types and grouped. A TIN triangulated network is generated based on each group of weather station temperature data, and the corresponding Thiessen polygons are constructed. Step 3: Temperature model construction and parameter calculation based on regression analysis: A regression model is constructed based on the intraday temperature cycle model, and the pre-processed remote sensing temperature data is obtained pixel by pixel by combining the Thiessen polygons of the temperature data of the meteorological station. The regression parameters are iteratively calculated and stored according to the month and pixel position. Step 4: Calculation of surface temperature data based on the intraday temperature cycle model: perform regression parameter search pixel by pixel, input the found regression parameters into the regression model, and output global surface temperature data for all months and all times; The step 2 specifically includes the following steps: Step 2.1: Group the temperature data of the weather stations based on spatial attributes at specific times for each month: , in, represents the temperature data of the weather station, the subscript L represents the weather station index, the subscript c represents the surface type of the weather station, the subscript j represents the jth month, and the subscript h=[1,8] represents the hth specific moment; Step 2.2: Generate TIN triangulation based on Delaunay method: Generate a TIN triangulation for each set of weather station temperature data: , Among them, TIN(S) represents the output TIN triangulation network, and the spatial location of the temperature data of the weather station , represents the point-by-point plane coordinates of the weather station, and Delaunay represents the point-by-point insertion algorithm of the Delaunay triangulation; Step 2.3: Generate Thiessen polygons based on TIN triangulation: Generate Thiessen polygons using the TIN triangulated network generated from each set of weather station temperature data: , Among them, Thiessen represents the Thiessen polygon generation algorithm, Polygon( ) represents the Thiessen polygon corresponding to the weather station site ID and its temperature data at 8 specific times; The temperature data at 8 specific moments are converted into raster data with the same size as the remote sensing pixel, and the attributes of the temperature data at 8 specific moments are converted into raster values ​​of 8 bands: , Among them, Raster represents the raster conversion function, PixelSize represents the size of the remote sensing pixel, and RasterBand (B1, B2, ..., B8) represents the corresponding raster data and the raster values ​​of its 8 bands.

2. The method for generating global surface temperature based on a pixel-by-pixel daily temperature cycle model according to claim 1, characterized in that: The step 3 comprises: Step 3.1: Construction of regression model based on intraday temperature cycle model: The GOT01 model is used to build a regression model to describe the daily temperature changes. The GOT01 model formula is as follows: , Where TEM1 / TEM2 represent the daytime / nighttime temperatures in the model, t represents the specific time in a 24-hour system, TEM0 represents the residual temperature before and after sunrise, and TEM a represents the temperature amplitude, t m Indicates the time when the temperature reaches the maximum value, t s represents the time when the temperature starts to decay freely, e represents the exponential function, ΔTEM represents the difference between the residual temperature TEM0 before and after sunrise and the temperature value TEM at the end time, and the end time represents the last value of the specific time t in the 24-hour system; k represents the decay constant, , ω represents the half-cycle width of the cosine function, which is defined as the duration of solar energy input and is determined by the latitude φ and the solar inclination angle δ, and is expressed as: , , Wherein, doy represents the annual accumulation day; there is a mapping relationship between the basic data and the annual accumulation day, and the basic data includes the temperature value of the pre-processed remote sensing temperature data and the temperature values ​​of 8 specific moments corresponding to the Thiessen polygon: , Among them, DOY j represents the annualized daily median of the jth month; Indicates rounding down, d j represents the day of the jth month, i represents the index of the day; The mapping relationship between the number of rows and latitude of remote sensing raster data is: , Among them, row represents the number of rows of raster data, and Δφ represents the latitude resolution of the raster image; Step 3.2: Regression data search based on attribute matching: For each pixel, determine the specific month, select the temperature value of the corresponding pre-processed remote sensing temperature data, then select the corresponding Thiessen polygon group according to the surface type, and finally determine the specific Thiessen polygon according to the pixel position, so as to obtain the specific temperature values ​​at 8 specific moments; Step 3.3: Calculation of regression parameters based on LM algorithm: The regression parameters latitude and annual cumulative day are calculated by pixel location and corresponding month; By searching the regression data, the basic data corresponding to the pixel is obtained; The calculated latitude and annual cumulative daily basic data are input into the daily temperature cycle model as the initial parameters of the LM algorithm for regression calculation and iteration: Output iteration completes five parameters of each pixel: residual temperature TEM0 before and after sunrise, temperature amplitude TEM a , the time t when the temperature reaches the maximum value m , the time t when the temperature starts to decay freely s The difference ΔTEM between the residual temperature TEM0 before and after sunrise and the temperature value TEM at the end time.

3. The method for generating global surface temperature based on a pixel-by-pixel daily temperature cycle model according to claim 1, characterized in that: The step 4 comprises: Step 4.1: Parameter search based on pixel-by-pixel search method: The latitude and annual accumulated days are calculated according to the mapping relationship, and the five fitted parameters are read month by month and pixel by pixel to obtain all the complete parameters of the daily temperature cycle model used to calculate the specific temperature; Step 4.2: Calculation of surface temperature data based on the daily temperature cycle model: All complete parameters are input into the daily temperature cycle model, and the surface temperature at all times of all months is calculated pixel by pixel. Float raster images based on the WGS84 projection are output according to the month-time classification to obtain global surface temperature data for all months and all times.