A method for predicting regional distribution of surface air temperature based on space-time coupling model

Abstract

The present invention provides a method for predicting the regional distribution of ground temperature based on a space-time coupling model, and the steps are as follows: collect hourly temperature data of several meteorological stations from a historical database to obtain collected data; construct a three-dimensional space-time coupling model of temperature according to the time series distribution law of temperature data and the spatial correlation of temperature data; train the three-dimensional space-time coupling model using the collected data to determine the parameters of the three-dimensional space-time coupling model; collect spatial position parameters and time parameters of the area to be predicted, and input them into the trained three-dimensional space-time coupling model to predict the temperature. The present invention fully considers the time and space laws of temperature distribution, and can accurately predict the temperature at any location and any time in the region, with high prediction accuracy.

Description

Technical Field

[0001] The invention relates to the technical field of ground temperature prediction, in particular to a method for predicting historical temperature distribution in a blank area without an observation station. Background Art

[0002] Many materials, products, and equipment in life will suffer from corrosion, aging, and damage to varying degrees after being subjected to various environmental factors, thus affecting their performance and service life. Among these environmental factors, the influence of temperature is particularly obvious. For example, high temperature accelerates the thermal motion of molecules and the diffusion of elements in organic coatings, providing sufficient energy for the breakage of molecular chains, thereby accelerating the degradation of organic coating materials. In addition, alternating changes in temperature will cause the volume of the coating to shrink and expand, causing internal stress in the organic coating, affecting its adhesion, and ultimately accelerating the aging process of the coating, such as cracking, blistering, and shedding. Therefore, in order to grasp the severity of temperatures in various places, a large number of environmental data manuals and standards (DEF STAN 00-35, GJB1172) have been published at home and abroad to guide the environmental adaptability design of products and equipment.

[0003] With the requirement of using material products in all regions, the temperature effects they will experience throughout their life cycle will be more diverse and severe. At present, the use of automatic weather station observation technology can achieve continuous and long-term temperature observation. However, due to the high cost of equipment purchase and maintenance, it is impossible to observe at high density, which also leads to the inability to obtain accurate temperature data. In addition, traditional studies on the spatiotemporal distribution of meteorology are mostly based on ordinary time series analysis and spatial difference analysis. Their individual analyses cannot fully explore the value of spatiotemporal data, resulting in poor accuracy of predicted data. Therefore, in order to support the need to determine the temperature value in the process of material testing and evaluation, it is necessary to construct an accurate space-time coupling model to realize temperature prediction at any spatial location and any time in the region. Summary of the invention

[0004] The purpose of the present invention is to provide a method for predicting the regional distribution of ground air temperature based on a space-time coupling model, which can be used to predict the temperature at any spatial position and any time in a region.

[0005] The objective of the present invention is achieved through such a technical solution, and the specific steps are as follows:

[0006] 1) Data collection: collect hourly temperature data of several weather stations from the historical database to obtain collected data;

[0007] 2) Model construction: Construct a three-dimensional space-time coupling model of temperature based on the time series distribution law of temperature data and the spatial correlation of temperature data;

[0008] 3) Model training: using the collected data in step 1) to train the three-dimensional space-time coupling model to determine the parameters of the three-dimensional space-time coupling model;

[0009] 4) Temperature prediction: Collect the spatial position parameters and time parameters of the area to be predicted, and input them into the three-dimensional space-time coupling model trained in step 3) to predict the temperature.

[0010] Further, the specific steps of constructing the three-dimensional space-time coupling model of temperature in step 2) are:

[0011] 2-1) According to the hourly time variation trend and fluctuation characteristics of the temperature data, a trend term model of the temperature time series and a residual term model of the temperature time series are constructed, wherein the trend term model of the temperature time series is fitted by an M-spline function, and the residual term model of the temperature time series is fitted by a Gaussian process;

[0012] 2-2) According to the spatial characteristics of temperature, the Gaussian process model is used to model the unknown parameters in the time series in the three-dimensional coordinate plane of longitude, latitude and altitude, and the longitude dimension and altitude parameters are added to the coordinates of the residual model of the temperature time series.

[0013] Further, the specific method of fitting the trend term model of the temperature time series by the M-spline function in step 2-1) is:

[0014] Taking the hourly temperature values ​​at the observation point within a certain period of time as input, the trend term model of the temperature time series is constructed using the M-spline function:

[0015]

[0016] In formula (1), β0 is the temperature trend term without unknown parameters, t is the time, is the largest integer not exceeding t, is an unknown parameter, β k is the kth parameter of the time series, and the least square method is used to calculate the unknown parameter α. is the daily temperature variation function, is the daily temperature mean function, is the function of the daily temperature fluctuation amplitude in a year, lam is the number of basis functions of the spline family in the annual periodic function of the mean temperature, ld is the number of basis functions of the spline family in the daily periodic function of the temperature value, laf is the number of basis functions of the spline family in the annual periodic function of the daily temperature fluctuation, is the kth basis function in the M-spline family of the s-order mean annual temperature period, is the kth basis function in the M-spline family of daily periodic temperature values ​​of order s, It is the kth basis function in the M-spline family of the s-order daily temperature fluctuation with an annual period.

[0017] Further, the specific method of fitting the residual term model of the temperature time series through the Gaussian process in step 2-1) is:

[0018] Taking the hourly temperature values ​​within a certain period of time at the observation point and the estimated temperature time series value calculated by the time series trend term model as input, the residual term model of the temperature time series is constructed using the Gaussian process as follows:

[0019] y R (t; θ)=μ+K(X,x(t))′[K(X,X)+σ 2 I] -1 (Y R -μ1) (2)

[0020] In formula (2), θ={μ,σ,k,γ1,...,γ d} is the residual sequence model parameter, and the maximum likelihood estimation method is used to calculate the unknown parameter θ, μ is the temperature residual mean, μ1 is the temperature residual mean vector, σ is the random error, I is the n×n unit matrix with 1 on the diagonal and 0 on the other elements, σ 2 I represents the random fluctuation at each observation point, K(X,x(t))′ is the 1×n square matrix of each element in the time collection X and x(t), is the five-dimensional time vector corresponding to any time t; X = (x1′, x2′, ..., x n ′) is given n time instants t1, t2, ..., t n The corresponding five-dimensional time vector, n is the number of given moments, K(X,x(t))′ describes the temperature correlation between any moment and the adjacent n moments. For the observation value y and time series estimate given n moments The difference collection of , K(X,X) is the n×n square matrix of the (i,j)th element:

[0021]

[0022] In formula (3), k is the standard deviation parameter, x i is the collection of all time dimensions at any time i, x i,s is the independent variable x i The sth element of , d represents the time dimension at any time i (d = 5), γ s is the scale parameter of the sth dimension, and K(X,X) describes the correlation between the temperatures at any adjacent n moments.

[0023] Furthermore, in step 2-2), the spatial correlation of the unknown parameter α in the time trend term model is considered, and the specific method of modeling the unknown parameters in the time series using the Gaussian process model in the three-dimensional coordinate plane of latitude, longitude and altitude is as follows:

[0024] Assume that the trend term model of the temperature time series is y T (t; α η (x long ,x lat ,x h )) The Gaussian process model is used to model the unknown parameters in the time series in the three-dimensional coordinate plane of latitude, longitude and altitude:

[0025] α η (x long ,x lat ,x h )=μ′+cov(X p ,x p )[K(X p ,X p )+σ 2 I′] -1 (β η -μ′1) (4)

[0026] In formula (4), x long ,x lat ,x h are the longitude, latitude, and altitude parameters of the temperature data, β η =(β 1,k ,β 2,k ,...,β n,k )′,k=0,...,q,q=l a +l b +l f is the number of skeleton points of each single observation point time series in the residual term model of the temperature time series, η=n×q is the total number of skeleton points of n observation stations, μ′ is β η The mean of η The mean vector of , I′ is the n-order unit matrix with 1 on the diagonal and 0 on the other elements, σ 2 I′ is the random fluctuation at each observation point, X p is the three-dimensional space-time position of the known observation point, x p is the three-dimensional space-time position of any location in the region.

[0027] Further, in step 2-2), the specific method of adding the longitudinal dimension and the height parameter to the coordinates of the residual term model of the temperature time series is:

[0028] Add longitude, latitude, and altitude parameters to the five-dimensional time vector corresponding to any time t at any location i:

[0029]

[0030] Furthermore, the three-dimensional space-time coupling model of temperature in step 2) is the sum of the trend term model of the temperature time series including longitude, latitude, and altitude parameters and the residual term model of the temperature time series including longitude, latitude, and altitude parameters:

[0031] y(t;x long ,x lat ,x h )=y T (t; α(x long ,x lat ,x h ))+y R (t;θ(x long ,x lat ,x h )) (6).

[0032] Due to the adoption of the above technical solution, the present invention has the following advantages:

[0033] 1. The present invention fully considers the temporal and spatial laws of temperature distribution, and can accurately predict the temperature at any spatial position and any time in the region with high prediction accuracy.

[0034] 2. The temperature data of the area without observation points predicted by the present invention can provide more basic temperature data for temperature data products, solve the problem of describing the temperature distribution law from a point to a whole, and thus improve the precision of the study of the temperature spatiotemporal distribution law.

[0035] 3. The temperature data of the area without observation points predicted by the present invention can be used to guide the environmental adaptability design of the product and has a wide range of application scenarios.

[0036] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description and claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings of the present invention are as follows.

[0038] Figure 1 The figure is a flow chart of the method of the present invention.

[0039] Figure 2This is a distribution map of national meteorological observation stations in the humid and hot ocean area in an embodiment of the present invention.

[0040] Figure 3 This is the distribution of temperature in the hot and humid ocean area predicted at 12:00 on January 1, 2019 in an embodiment of the present invention.

[0041] Figure 4 This is the comparison result between the temperature prediction value and the actual recorded data for the whole month of May 2019 at the Chengmai Observation Station in an embodiment of the present invention.

[0042] Figure 5 This is the distribution of the absolute deviation between the predicted temperature value and the actual temperature value for the whole month of May 2019 at the Chengmai Observation Station in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0044] like Figure 1 The specific steps of the surface temperature regional distribution prediction method based on the space-time coupling model are as follows:

[0045] 1) Data collection: Collect hourly temperature data of several weather stations from the historical database to obtain collected data; the specific method is:

[0046] The hourly temperature data of 24 national meteorological observation stations in the humid and hot ocean area (Hainan Island + Leizhou Peninsula) from 2010 to 2020 were obtained from the National Meteorological Information Center. The distribution information of the 24 national meteorological observation stations in the humid and hot ocean area is as follows: Figure 2 shown.

[0047] 2) Model construction: Construct a three-dimensional space-time coupling model of temperature based on the time series distribution law of temperature data and the spatial correlation of temperature data; the specific steps are:

[0048] 2-1) According to the hourly time variation trend and fluctuation characteristics of the temperature data, a trend term model of the temperature time series and a residual term model of the temperature time series are constructed, wherein the trend term model of the temperature time series is fitted by an M-spline function, and the residual term model of the temperature time series is fitted by a Gaussian process; the specific steps are:

[0049] 2-1-1) Taking the hourly temperature values ​​at the observation point within a certain period of time as input, the trend term model of the temperature time series is constructed using the M-spline function:

[0050]

[0051] In formula (7), β0 is the temperature trend term without unknown parameters, t is the time, is the largest integer not exceeding t, is an unknown parameter, β k is the kth parameter of the time series, and the least square method is used to calculate the unknown parameter α. is the daily temperature variation function, is the daily temperature mean function, is the function of the daily temperature fluctuation amplitude in a year, lam is the number of basis functions of the spline family in the annual periodic function of the mean temperature, ld is the number of basis functions of the spline family in the daily periodic function of the temperature value, laf is the number of basis functions of the spline family in the annual periodic function of the daily temperature fluctuation, is the kth basis function in the M-spline family of the s-order mean annual temperature period, is the kth basis function in the M-spline family of daily periodic temperature values ​​of order s, It is the kth basis function in the M-spline family of the s-order daily temperature fluctuation with an annual period.

[0052] In the example of the present invention, a third-order M-spline function is selected to characterize the daily variation of temperature and the annual variation of the daily mean and daily fluctuation amplitude. The basis function of the third-order M-spline family is a second-order polynomial with a continuous first-order derivative, which meets the requirement for smoothness of fitting of the time trend term.

[0053] 2-1-2) Taking the hourly temperature values ​​at the observation point within a certain period of time and the estimated temperature time series value calculated by the time series trend term model as input, the residual term model of the temperature time series is constructed using the Gaussian process as follows:

[0054] y R (t; θ)=μ+K(X,x(t))′[K(X,X)+σ 2 I] -1 (Y R -μ1) (8)

[0055] In formula (8), θ={μ,σ,k,γ1,...,γ d} is the residual sequence model parameter, and the maximum likelihood estimation method is used to calculate the unknown parameter θ, μ is the temperature residual mean, μ1 is the temperature residual mean vector, σ is the random error, I is the n×n unit matrix with 1 on the diagonal and 0 on the other elements, σ 2 I represents the random fluctuation at each observation point, K(X,x(t))′ is the 1×n square matrix of each element in the time collection X and x(t), is the five-dimensional time vector corresponding to any time t; X = (x1′, x2′, ..., x n ′) is given n time instants t1, t2, ..., t nThe corresponding five-dimensional time vector, n is the number of given moments, K(X,x(t))′ describes the temperature correlation between any moment and the adjacent n moments. For the observation value y and time series estimate given n moments The difference collection of , K(X,X) is the n×n square matrix of the (i,j)th element:

[0056]

[0057] In formula (9), k is the standard deviation parameter, x i is the collection of all time dimensions at any time i, x i,s is the independent variable x i The sth element of , d represents the time dimension at any time i (d = 5), γ s is the scale parameter of the sth dimension, and K(X,X) describes the correlation between the temperatures at any adjacent n moments.

[0058] 2-2) According to the spatial characteristics of temperature, the unknown parameters in the time series are modeled using the Gaussian process model in the three-dimensional coordinate plane of latitude, longitude and altitude, and the longitude dimension and altitude parameters are added to the coordinates of the residual term model of the temperature time series. The specific steps are as follows:

[0059] 2-2-1) Assume that the trend term model of the temperature time series is y T (t; α η (x long ,x lat ,x h )) The Gaussian process model is used to model the unknown parameters in the time series in the three-dimensional coordinate plane of latitude, longitude and altitude:

[0060] α η (x long ,x lat ,x h )=μ′+cov(X p ,x p )[K(X p ,X p )+σ 2 I′] -1 (β η -μ′1) (10)

[0061] In formula (10), x long ,x lat ,x h are the longitude, latitude, and altitude parameters of the temperature data, β η =(β 1,k ,β 2,k ,...,β n,k)′,k=0,...,q,q=l a +l b +l f is the number of skeleton points of each single observation point time series in the residual term model of the temperature time series, η=n×q is the total number of skeleton points of n observation stations, μ′ is β η The mean of η The mean vector of , I′ is the n-order unit matrix with 1 on the diagonal and 0 on the other elements, σ 2 I′ is the random fluctuation at each observation point, X p is the three-dimensional space-time position of the known observation point, x p is the three-dimensional space-time position of any location in the region.

[0062] 2-2-2) Add longitude and altitude parameters to the coordinates of the residual term model of the temperature time series, that is, add longitude, latitude, and altitude parameters to the five-dimensional time vector corresponding to any time t at any location i:

[0063]

[0064] 2-3) The three-dimensional space-time coupling model of temperature is the sum of the trend term model of the temperature time series including longitude, latitude, and altitude parameters and the residual term model of the temperature time series including longitude, latitude, and altitude parameters:

[0065] y(t;x long ,x lat ,x h )=y T (t; α(x long ,x lat ,x h ))+y R (t;θ(x long ,x lat ,x h )) (12).

[0066] 3) Model training: using the collected data in step 1) to train the three-dimensional space-time coupling model to determine the parameters of the three-dimensional space-time coupling model;

[0067] In the example of the present invention, 8 nodes are selected for the annual spline and 4 nodes are selected for the daily period spline. The model contains a total of 20 unknown parameters. The hourly temperature data of 24 national gas phase observation stations in the humid and hot ocean area are substituted into the prediction model of the present invention. The Gaussian process is used for regression based on the spatial longitude and latitude altitude coordinates. The obtained Gaussian process model parameters are shown in Table 1; the prediction effect of the temperature data of the 24 stations is shown in Table 2.

[0068] Table 1 Parameter fitting results of the trend term model of temperature time series

[0069]

[0070]

[0071] Table 2 Prediction results of temperature data from 24 stations

[0072]

[0073] 4) Temperature prediction: Collect the spatial position parameters and time parameters of the area to be predicted, and input them into the three-dimensional space-time coupling model trained in step 3) to predict the temperature.

[0074] In the present example, the hourly temperature data of 24 national gas phase observation stations in the humid and hot ocean area are substituted into the prediction model of the present invention to obtain the temperature distribution of the humid and hot ocean area at 12:00 on January 1, 2019 as follows: Figure 3 The comparison between the predicted values ​​and the actual values ​​of each observation station is shown in Table 3.

[0075] Table 3 Comparison of actual and recorded temperatures at each observation station at 12:00 on January 1, 2019

[0076]

[0077] 5) Model verification: The model of the present invention was verified using the leave-one-out method.

[0078] In this example, the data from Chengmai station was selected as test data, and the data from the other 23 stations were used as training data for verification. Figure 4 , the distribution of absolute deviations from the actual value is shown in Figure 5 .

[0079] The results of the embodiment show that the predicted value is highly consistent with the real recorded data, not only matching the actual temperature change trend, but also the deviation between the predicted value and the real value is very small. Therefore, the regional distribution prediction method of ground temperature proposed in the present invention can realize accurate temperature modeling and prediction at any spatial position and any time in the region based on the input temperature observation station temperature hourly data. Thereby, the calculation precision of basic products such as temperature distribution law and extreme value in the region can be improved, and the environmental adaptability design of the product can be guided.

[0080] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0081] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0082] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0083] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0084] 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 above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting regional distribution of surface air temperature based on a space-time coupling model, characterized in that: The specific steps are as follows: 1) Data collection: Collect hourly temperature data of several weather stations from the historical database to obtain collected data; 2) Model construction: Construct a three-dimensional space-time coupling model of temperature based on the time series distribution law of temperature data and the spatial correlation of temperature data; 3) Model training: using the collected data in step 1) to train the three-dimensional space-time coupling model to determine the parameters of the three-dimensional space-time coupling model; 4) Temperature prediction: collect the spatial position parameters and time parameters of the area to be predicted, and input them into the three-dimensional space-time coupling model trained in step 3) to predict the temperature; The specific steps of constructing the three-dimensional space-time coupled model of temperature in step 2) are: 2-1) According to the hourly time variation trend and fluctuation characteristics of the temperature data, a trend term model of the temperature time series and a residual term model of the temperature time series are constructed, wherein the trend term model of the temperature time series is fitted by an M-spline function, and the residual term model of the temperature time series is fitted by a Gaussian process; 2-2) According to the spatial characteristics of temperature, the Gaussian process model is used to model the unknown parameters in the time series in the three-dimensional coordinate plane of longitude, latitude and altitude, and the longitude dimension and altitude parameters are added to the coordinates of the residual model of the temperature time series.

2. A method for predicting regional distribution of surface air temperature based on a space-time coupling model as claimed in claim 1, characterized in that: The specific method of fitting the trend term model of the temperature time series by the M-spline function in step 2-1) is: Taking the hourly temperature values ​​at the observation point within a certain period of time as input, the trend term model of the temperature time series is constructed using the M-spline function: In formula (1), β0 is the temperature trend term without unknown parameters, t is the time, is the largest integer not exceeding t, is an unknown parameter, β k is the kth parameter of the time series, and the least square method is used to calculate the unknown parameter α. is the daily temperature variation function, is the daily temperature mean function, is the function of the daily temperature fluctuation amplitude in a year, lam is the number of basis functions of the spline family in the annual periodic function of the mean temperature, ld is the number of basis functions of the spline family in the daily periodic function of the temperature value, laf is the number of basis functions of the spline family in the annual periodic function of the daily temperature fluctuation, is the kth basis function in the M-spline family of the s-order mean annual temperature period, is the kth basis function in the M-spline family of daily periodic temperature values ​​of order s, It is the kth basis function in the M-spline family of the s-order daily temperature fluctuation with an annual period.

3. A method for predicting regional distribution of surface air temperature based on a space-time coupling model as claimed in claim 1, characterized in that: The specific method of fitting the residual term model of the temperature time series through the Gaussian process in step 2-1) is: Taking the hourly temperature values ​​within a certain period of time at the observation point and the estimated temperature time series value calculated by the time series trend term model as input, the residual term model of the temperature time series is constructed using the Gaussian process as follows: y R (t;θ)=μ+K(X,x(t))′[K(X,X)+σ 2 I] -1 (Y R -μ1) (2) In formula (2), θ={μ,σ,k,γ1,...,γ d } is the residual sequence model parameter, and the maximum likelihood estimation method is used to calculate the unknown parameter θ, μ is the temperature residual mean, μ1 is the temperature residual mean vector, σ is the random error, I is the n×n unit matrix with 1 on the diagonal and 0 on the other elements, σ 2 I represents the random fluctuation at each observation point, K(X,x(t))′ is the 1×n square matrix of each element in the time collection X and x(t), is the five-dimensional time vector corresponding to any time t; X = (x1′, x2′, ..., x n ′) is given n time instants t1, t2, ..., t n The corresponding five-dimensional time vector, n is the number of given moments, K(X,x(t))′ describes the temperature correlation between any moment and the adjacent n moments. For the observation value y and time series estimate given n moments The difference collection of , K(X,X) is the n×n square matrix of the (i,j)th element: In formula (3), k is the standard deviation parameter, x i is the collection of all time dimensions at any time i, x i,s is the independent variable x i The sth element of , d represents the time dimension at any time i (d = 5), γ s is the scale parameter of the sth dimension, and K(X,X) describes the correlation between the temperatures at any adjacent n moments.

4. A method for predicting regional distribution of surface air temperature based on a space-time coupling model as claimed in claim 1, characterized in that: In step 2-2), the spatial correlation of the unknown parameter α in the time trend term model is considered. The specific method of modeling the unknown parameters in the time series using the Gaussian process model in the three-dimensional coordinate plane of latitude, longitude and altitude is as follows: Assume that the trend term model of the temperature time series is y T (t; α η (x long, x lat, x h )) The Gaussian process model is used to model the unknown parameters in the time series in the three-dimensional coordinate plane of latitude, longitude and altitude: a η (x long, x lat, x h )=μ′+cov(X p ,x p )[K(X p ,X p )+s 2 I'] -1 (b η -μ′1) (4) In formula (4), x long, x lat, x h are the longitude, latitude, and altitude parameters of the temperature data, β η =(β 1,k ,β 2,k ,...,β n,k )′,k=0,...,q,q=l a +l b +l f is the number of skeleton points of each single observation point time series in the residual term model of the temperature time series, η=n×q is the total number of skeleton points of n observation stations, μ′ is β η The mean of η The mean vector of , I′ is the n-order unit matrix with 1 on the diagonal and 0 on the other elements, σ 2 I′ is the random fluctuation at each observation point, X p is the three-dimensional space-time position of the known observation point, x p is the three-dimensional space-time position of any location in the region.

5. The method for predicting regional distribution of surface air temperature based on a space-time coupling model according to claim 1, characterized in that: The specific method of adding the longitude dimension and height parameters to the coordinates of the residual term model of the temperature time series in step 2-2) is: Add longitude, latitude, and altitude parameters to the five-dimensional time vector corresponding to any time t at any location i:

6. A method for predicting regional distribution of surface air temperature based on a space-time coupling model as claimed in claim 1, characterized in that: The three-dimensional space-time coupling model of temperature in step 2) is the sum of the trend term model of the temperature time series including longitude, latitude and altitude parameters and the residual term model of the temperature time series including longitude, latitude and altitude parameters: y(t;x long ,x lat ,x h )=y T (t;α(x long ,x lat ,x h ))+y R (t;θ(x long ,x lat ,x h )) (6)。

Images