Station temperature interpolation method, device and equipment and storage medium

Abstract

The application discloses a station temperature interpolation method and device, equipment and a storage medium. The linear temperature data sequence of the acquired weather station is decomposed by a wavelet packet, so as to be divided into a background sequence and a fluctuation sequence. After the interpolation operation of the coupled terrain is performed on the background sequence, a low-frequency background temperature value is obtained. After the time sequence alignment is performed on the fluctuation sequence, a high-frequency fluctuation temperature value is obtained by performing a natural neighbor interpolation operation. Then, the target temperature is obtained according to the low-frequency background temperature value and the high-frequency fluctuation temperature value. Therefore, without requiring too many computing resources, the terrain spatial characteristics of the weather station and the temperature fluctuation characteristics of the target time to be interpolated are considered at the same time, and the interpolation accuracy of the temperature data sequence of the ground weather station in various complex terrain environments and weather mutation scenes is greatly improved.

Description

Technical Field

[0001] This application relates to the field of meteorological data processing technology, and in particular to a method, apparatus, equipment and storage medium for station temperature interpolation. Background Technology

[0002] When using ground-based meteorological stations to observe continuous, high-precision hourly temperature data for a target area, data gaps often occur in certain time periods or regions. Therefore, interpolation processing is needed to obtain the continuous temperature distribution within the target area. However, existing interpolation methods generally suffer from low interpolation accuracy: interpolation techniques based on spatial distance relationships struggle to quantify the impact of steady-state factors such as topography on temperature, making it difficult to obtain accurate interpolation results in complex and variable terrain areas; multivariate interpolation techniques that introduce covariates fail to consider the high-frequency, short-term nonlinear fluctuations in temperature caused by cold air passages and strong convection, making it difficult to guarantee consistent interpolation accuracy across different terrain regions and during sudden weather changes; and deep learning-based interpolation prediction models are susceptible to noise interference, consuming excessive computational resources without effectively improving interpolation accuracy, thus lacking practicality. Summary of the Invention

[0003] This application aims to propose a method, apparatus, device, and storage medium for interpolating station temperature data, which can significantly improve the interpolation accuracy of temperature data in complex terrain areas.

[0004] The station temperature interpolation method according to a first aspect of this application is applied to a target adjustment region, wherein the target adjustment region has multiple meteorological stations, and the station temperature interpolation method includes: In response to the interpolation instruction, the target station to be interpolated at the target time is determined, wherein the target station is one of the plurality of meteorological stations; Wavelet packet decomposition is performed on the linear temperature data sequence of each of the pre-acquired meteorological stations to obtain a background sequence and a fluctuation sequence, wherein the frequency of the background sequence is less than the frequency of the fluctuation sequence; the target time is the time in the time interval corresponding to the linear temperature data sequence. An initial low-frequency candidate station set is determined based on the location data and terrain dataset of the target station. The geographical distance weight and terrain influence weight of each candidate station in the initial low-frequency candidate station set are calculated to obtain the initial comprehensive weight of each candidate station in the initial low-frequency candidate station set. The corresponding cumulative weight sum and total weight sum are obtained. When the proportion of the cumulative weight sum to the total weight sum reaches a preset weight threshold, an accumulated low-frequency candidate station set is determined. The low-frequency background temperature value of the target station at the target time is obtained based on the temperature value of each candidate station in the accumulated low-frequency candidate station set at the target time and its corresponding initial comprehensive weight. A time window centered on the target time and a set of natural neighbors of the target station are obtained. High-frequency fluctuation sequence segments of each neighboring station in the set of natural neighbors and the target station within the time window are extracted. The high-frequency fluctuation sequence segments of each neighboring station in the set of natural neighbors are time-aligned with the high-frequency fluctuation sequence segments of the target station to obtain the phase delay time corresponding to each neighboring station in the set of natural neighbors. The high-frequency fluctuation temperature value of each neighboring station in the set of natural neighbors is time-phase compensated according to the phase delay time corresponding to each neighboring station in the set of natural neighbors. The high-frequency fluctuation temperature value of the target station at the target time is obtained by using natural neighbor interpolation. The target temperature of the target station at the target time is obtained based on the low-frequency background temperature value and the high-frequency fluctuating temperature value.

[0005] A station temperature interpolation device according to a second aspect embodiment of this application includes: The target station determination module is used to determine the target station to be interpolated at the target time in response to the interpolation instruction, wherein the target station is one of multiple meteorological stations; The wavelet packet decomposition module is used to perform wavelet packet decomposition on the linear temperature data sequence of each of the pre-acquired meteorological stations to obtain a background sequence and a fluctuation sequence, wherein the frequency of the background sequence is less than the frequency of the fluctuation sequence; the target time is the time in the time interval corresponding to the linear temperature data sequence. The low-frequency background temperature acquisition module is used to determine an initial low-frequency candidate station set based on the location data and terrain dataset of the target station, calculate the geographical distance weight and terrain influence weight of each candidate station in the initial low-frequency candidate station set to obtain the initial comprehensive weight of each candidate station in the initial low-frequency candidate station set, and obtain the corresponding cumulative weight sum and total weight sum. When the proportion of the cumulative weight sum to the total weight sum reaches a preset weight threshold, the cumulative low-frequency candidate station set is determined, and the low-frequency background temperature value of the target station at the target time is obtained based on the temperature value of each candidate station in the cumulative low-frequency candidate station set at the target time and its corresponding initial comprehensive weight. The high-frequency fluctuating temperature acquisition module is used to acquire a time window centered on the target time and a set of natural neighbors of the target station, extract high-frequency fluctuating sequence segments of each neighboring station in the set of natural neighbors and the target station within the time window, align the high-frequency fluctuating sequence segments of each neighboring station in the set of natural neighbors with the high-frequency fluctuating sequence segments of the target station in time sequence, obtain the phase delay time corresponding to each neighboring station in the set of natural neighbors, and perform time-phase compensation on the high-frequency fluctuating temperature value of each neighboring station in the set of natural neighbors at the target time based on the phase delay time corresponding to each neighboring station in the set of natural neighbors, and then use natural neighbor interpolation to obtain the high-frequency fluctuating temperature value of the target station at the target time. The target temperature acquisition module is used to obtain the target temperature of the target station at the target time based on the low-frequency background temperature value and the high-frequency fluctuating temperature value.

[0006] An electronic device according to a third aspect of this application includes a processor and a memory storing a computer program; The computer program executed by the processor is a site temperature interpolation method as described in the first aspect embodiment above.

[0007] According to a fourth aspect embodiment of this application, a computer storage medium stores computer-executable instructions for causing a computer to perform the station temperature interpolation method as described in the first aspect embodiment above.

[0008] This application provides a method, apparatus, device, and storage medium for interpolating station temperature data. By performing wavelet packet decomposition on the acquired linear temperature data sequence of a meteorological station, it divides the data into a background sequence and a fluctuation sequence. After performing terrain-coupled interpolation on the background sequence, a low-frequency background temperature value is obtained. After temporal alignment of the fluctuation sequence, a natural neighbor interpolation is performed to obtain a high-frequency fluctuation temperature value. Finally, the target temperature is obtained based on the low-frequency background temperature value and the high-frequency fluctuation temperature value. Thus, without requiring excessive computing resources, it simultaneously considers the topographical spatial characteristics of the meteorological station and the temperature fluctuation characteristics of the target time to be interpolated, significantly improving the interpolation accuracy of temperature data sequences observed by ground meteorological stations under various complex terrain environments and sudden weather change scenarios.

[0009] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing the application. Attached Figure Description

[0010] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart illustrating the site temperature interpolation method according to an embodiment of this application. Figure 2 This is a schematic diagram of a site grid according to an embodiment of this application; Figure 3 This is an ASTER GDEM topographic map of meteorological stations in Hunan Province used in the embodiments of this application; Figure 4 In the embodiments of this application, the R value of the temperature interpolation result using only the background sequence is... 2 Comparison chart; Figure 5 This is a comparison chart of the observed values ​​of meteorological station 57678 in this embodiment of the application and the interpolation results using this application and RFSP respectively; Figure 6 This is a comparison chart of the observed values ​​of meteorological station P1434 in the embodiments of this application and the interpolation results using this application and RFSP respectively; Figure 7 This is a comparison chart of the observed values ​​of meteorological station P1534 in the embodiments of this application and the interpolation results using this application and RFSP respectively; Figure 8 A schematic diagram of the structure of an embodiment of the electronic device provided in this application. Detailed Implementation

[0011] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0012] In the description of this application, the use of terms such as "first," "second," etc., is for the purpose of distinguishing technical features only and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or the order of the technical features indicated.

[0013] In the description of this application, it should be understood that the orientation descriptions, such as up, down, etc., are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0014] In the description of this application, it should be noted that, unless otherwise explicitly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.

[0015] The technical solution of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are some embodiments of this application, not all embodiments.

[0016] like Figure 1 As shown, an embodiment of this application provides a station temperature interpolation method, which is applied to a target adjustment area. The target adjustment area has multiple meteorological stations. The station temperature interpolation method includes steps S100 to S500: Step S100: In response to the interpolation instruction, determine the target station to be interpolated at the target time. The target station is one of multiple meteorological stations. Step S200: Perform wavelet packet decomposition on the pre-acquired linear temperature data sequence of each meteorological station to obtain the background sequence and the fluctuation sequence. The frequency of the background sequence is less than the frequency of the fluctuation sequence. The target time is the time in the time interval corresponding to the linear temperature data sequence. Step S300: Determine the initial low-frequency candidate station set based on the location data and terrain dataset of the target station. Calculate the geographical distance weight and terrain influence weight of each candidate station in the initial low-frequency candidate station set to obtain the initial comprehensive weight of each candidate station in the initial low-frequency candidate station set. Obtain the corresponding cumulative weight sum and total weight sum. If the proportion of the cumulative weight sum to the total weight sum reaches a preset weight threshold, determine the cumulative low-frequency candidate station set. Obtain the low-frequency background temperature value of the target station at the target time based on the temperature value of each candidate station in the cumulative low-frequency candidate station set at the target time and its corresponding initial comprehensive weight. Step S400: Obtain the time window centered on the target time and the set of natural neighbors of the target station. Extract the high-frequency fluctuation sequence segments of each neighboring station in the set of natural neighbors and the target station within the time window. After aligning the high-frequency fluctuation sequence segments of each neighboring station in the set of natural neighbors with the high-frequency fluctuation sequence segments of the target station in time, obtain the phase delay time corresponding to each neighboring station in the set of natural neighbors. Then, perform time-phase compensation on the high-frequency fluctuation temperature value of each neighboring station in the set of natural neighbors at the target time based on the phase delay time corresponding to each neighboring station in the set of natural neighbors. Finally, use the natural neighbor interpolation method to obtain the high-frequency fluctuation temperature value of the target station at the target time. Step S500: Based on the low-frequency background temperature value and the high-frequency fluctuating temperature value, obtain the target temperature of the target station at the target time.

[0017] In this embodiment, the linear temperature data sequence of the acquired meteorological station is divided into a background sequence and a fluctuation sequence by wavelet packet decomposition. The background sequence is then subjected to interpolation with the terrain to obtain the low-frequency background temperature value. The fluctuation sequence is then time-aligned and subjected to natural neighbor interpolation to obtain the high-frequency fluctuation temperature value. Finally, the target temperature is obtained based on the low-frequency background temperature value and the high-frequency fluctuation temperature value. This significantly improves the interpolation accuracy of the temperature data sequence observed by the ground meteorological station under various complex terrain environments and sudden weather changes, without requiring excessive computing resources, while taking into account both the topographic spatial characteristics of the meteorological station and the temperature fluctuation characteristics of the target time to be interpolated.

[0018] The target adjustment area can be any region where temperature data from meteorological stations need to be interpolated, and where multiple meteorological stations are located. For example, the target adjustment area can be an administrative region such as a province, city, or county, or it can be a natural geographical region such as a river basin, mountainous area, or plain.

[0019] The aforementioned meteorological stations can be any ground-based meteorological observation station capable of continuously observing and recording hourly temperature data. For example, such stations may include national meteorological observation stations and regional automatic weather stations. The observation data from multiple meteorological stations can be uniformly collected and managed by the data center of the meteorological operational department. While continuously collecting temperature data, the aforementioned meteorological stations can also output basic geographic information such as station latitude, longitude, and altitude. The station deployment area roughly covers various landform units such as plains, hills, and mountains within the target adjustment area.

[0020] The aforementioned interpolation instruction can be any instruction used to instruct the interpolation of temperature data sequences from meteorological stations. For example, the interpolation instruction could be automatically generated by an electronic device when it detects missing data points in the temperature observation data of any meteorological station. For instance, after each hourly interval, the electronic device performs an integrity check on the previous hour's observation data from multiple meteorological stations, and automatically generates an interpolation instruction for that meteorological station if it detects missing data points. Alternatively, the interpolation instruction could be an instruction issued by a meteorological service platform connected to the electronic device. Or, the interpolation instruction could be an instruction triggered by a staff member through a human-computer interaction interface for a specific meteorological station and time. The interpolation instruction can carry indication information for locating the target station. For example, this indication information could include the target station's identifier and the target time. After receiving the interpolation instruction, the relevant equipment can parse the station identifier and target time from it and identify the meteorological station corresponding to the identifier as the target station. It should be noted that when there are multiple consecutive missing measurement times at the target site, the interpolation instruction can carry the start and end times of the missing time period, so that the electronic device can determine multiple consecutive target times that need to be interpolated.

[0021] The target time mentioned above can be any moment in the time interval corresponding to the linear temperature data series where data is missing. For example, if the temperature observation data of the meteorological station is observed hourly, the target time can be the hour on the hour of a certain hour of a certain day.

[0022] The number of target stations can be one or more. When the interpolation instruction only marks a single target station with a missing temperature reading, the interpolation process is only performed on that single target station. If the relevant equipment identifies multiple meteorological stations in the target area with simultaneous observation gaps based on the interpolation instruction, all target stations can be traversed in parallel, and the temperature interpolation work can be completed independently for each target station in turn.

[0023] The aforementioned linear temperature data sequence can be a continuous temperature data sequence with equal time intervals, meaning that the time interval between adjacent temperature values ​​is equal and there are no missing measurements. For example, this linear temperature data sequence can be a continuous temperature data sequence with a time interval of 1 hour.

[0024] The aforementioned time interval is the continuous time range fully covered by meteorological observation data in the pre-collected and stored linear temperature data series. The linear temperature data series of all meteorological stations in the target area correspond to the same time interval. Moreover, the time interval can be flexibly set according to business needs. It can be a long-term time series interval over many years, or a short-term interval such as a single month, a single week, or a single climate change process.

[0025] The wavelet packet decomposition described above can be any signal decomposition method capable of simultaneously performing multi-scale partitioning of the signal in both the time and frequency domains. When decomposing the signal layer by layer, wavelet packet decomposition not only continues to decompose the low-frequency components but also the high-frequency components, thus simultaneously capturing the time-domain and frequency-domain features of the temperature data sequence, making it suitable for processing the scale fluctuation characteristics corresponding to the temperature data sequence. The Daubechies-4 wavelet basis, which has good balance and strong continuity, can be used to split the linear temperature data sequence into low-frequency and high-frequency sequences for separate processing. Let the time index be t and the sequence length be... The linear temperature data sequence of any station s is denoted as sampling frequency times / hour, according to the Nyquist sampling theorem, the effective frequency band is: times / hour. After j-level wavelet packet decomposition, the following can be obtained: The decomposition coefficients of each subband. Let... Let represent the coefficient sequence of the k-th subband after the j-th level decomposition, and let the initial conditions of the decomposition be... And it is recursively calculated using the following formula: ; in, and Let be the low-pass filter coefficients and high-pass filter coefficients corresponding to the Daubechies-4 wavelet basis, respectively. The relationship between them satisfies the orthogonal mirror filter relationship and is expressed by the following equation: .

[0026] Setting up considering time scale and computational complexity After a 3-level decomposition, the original sequence was decomposed into... The coefficient sequence of each sub-band Effective frequency band range The frequency per hour is divided into 8 equal intervals by 8 sub-bands, and the coefficients of the 4 sub-bands with the lowest frequency are used to reconstruct the background sequence. The four most frequent subbands were reconstructed into wave sequences. That is, it can be expressed by the following formula: ; in The wavelet packet reconstruction operator can reconstruct the subband coefficients into time-domain signals of the same length as the original sequence, and then add them point by point to obtain the background sequence. and fluctuation sequence .

[0027] Among them, respectively targeting the fluctuation sequence With background sequence In the subsequent interpolation process, different methods are used for interpolation: low-frequency sequences have strong spatial continuity, so a spatial weight-based method is used for interpolation; high-frequency sequences have phase delays, so interpolation is performed after the high-frequency sequences are aligned.

[0028] It should be noted that the number of wavelet packet decomposition layers, wavelet basis type, and the division method of low-frequency and high-frequency sub-bands can all be adaptively set according to the fluctuation characteristics of the temperature data sequence, and are not limited here. It should also be noted that wavelet packet decomposition requires the input sequence to be continuous; therefore, the above linear temperature data sequence is a continuous sequence that has undergone pre-filling processing. The method for obtaining this sequence can be found in the implementation method of data preprocessing below, and will not be elaborated upon here.

[0029] The temperature change frequency of the background sequence is lower than that of the fluctuation sequence. The background sequence reflects the gradual trend of temperature change, such as low-frequency components like diurnal and seasonal variations. It has strong spatial continuity and is suitable for interpolation based on spatial weights. The fluctuation sequence reflects short-term fluctuations in temperature, such as short-term cooling or warming caused by cold air passage or cloud and rain processes. There may be phase delays between different meteorological stations, so time alignment technology needs to be introduced before interpolation.

[0030] The aforementioned topographic dataset can be any collection of data capable of quantifying the impact of topography on temperature. For example, the dataset may include data on elevation, aspect, and topographic relief. Elevation directly determines the regional baseline temperature; the higher the elevation, the lower the overall baseline temperature. Aspect affects the total amount of solar radiation received by the surface; sunny slopes are consistently warmer than shady slopes. Topographic relief characterizes the complexity of the regional terrain; the greater the relief, the more significant the spatial differences in local temperature. When performing interpolation on the background sequence, this application incorporates all types of topographic data into the weight calculation process, addressing the problems of traditional spatial interpolation's inability to quantify topographic effects and poor accuracy in mountain interpolation. It is understood that the types of topographic data selected are not limited to the three mentioned above; topographic factors such as slope, valley shading, and local water system elevation can also be added based on the geomorphological characteristics of the target area. The types of factors can be flexibly adapted, and this application does not impose any limitations on this.

[0031] The aforementioned initial low-frequency candidate station set can be a collection of candidate stations selected from multiple meteorological stations in the target adjustment area, used to participate in the low-frequency background temperature interpolation calculation of the target station. For example, the electronic device can select a preset initial number of meteorological stations from multiple meteorological stations in order of geographical distance from the target station, from closest to furthest, to form the initial low-frequency candidate station set. This preset initial number can be 20. Alternatively, the electronic device can also include meteorological stations whose location data is within the same preset spatial range as the target station and whose terrain dataset differs from the target station by less than a preset difference threshold in the initial low-frequency candidate station set.

[0032] The aforementioned preset weight threshold can be a pre-set threshold used to measure the extent to which the accumulated weights cover all weights. For example, the preset weight threshold can be 0.95, meaning that when the sum of the accumulated weights after sorting accounts for 95% of the total weights, it is considered that the accumulated meteorological stations are sufficient to represent the overall contribution of the initial low-frequency candidate station set, and further accumulation is stopped.

[0033] The above method of determining the low-frequency background temperature value based on the temperature value of each candidate station in the accumulated low-frequency candidate station set at the target time and its corresponding initial comprehensive weight can be achieved by normalizing the initial comprehensive weight of each candidate station in the accumulated low-frequency candidate station set, and then multiplying the background sequence value of each candidate station at the target time with the normalized initial comprehensive weight and summing the results, that is, using the weighted average value as the low-frequency background temperature value of the target station at the target time.

[0034] The aforementioned low-frequency background temperature values ​​are base temperature values ​​obtained by interpolating the background sequence based on the target station's topographic dataset and combining spatial and topographical weights. These values ​​represent the baseline temperature level determined by steady-state topographic and climatic factors such as altitude, terrain, and season, and do not include short-term, nonlinear temperature abrupt changes caused by cold waves, convection, or other short-term weather events. In this embodiment, the target station... Low-frequency background temperature value at target time t It can be expressed by the following formula: ; in, To accumulate the number of low-frequency candidate sites, To accumulate any candidate site in the low-frequency candidate site set, To compare with any candidate site in the accumulated low-frequency candidate site set The corresponding initial comprehensive weights are normalized. This involves accumulating the temperature value at the target time for each candidate station in the low-frequency candidate station set. The methods for obtaining these variables will be found later and will not be elaborated upon here.

[0035] The aforementioned time window can be a fixed-length time range centered on the target time, the length of which can be adapted to the scale of short-term temperature fluctuations. Specifically, the time window can be a 7-hour time window consisting of 3 hours before and after the target time, in order to adapt to the scale of short-term temperature fluctuations.

[0036] The aforementioned natural neighbor set can be any set of neighboring stations determined based on the spatial relationship between the target station and its surrounding meteorological stations. For example, electronic devices can perform Delonix triangulation on the location data of multiple meteorological stations, identify the meteorological stations directly connected to the target station in the triangulation as the target station's natural neighbors, and form a natural neighbor set from these natural neighbors.

[0037] The above-mentioned method involves aligning the high-frequency wave sequence segments of each neighboring station in the natural neighbor set with the high-frequency wave sequence segment of the target station in time to obtain the phase delay time. Soft dynamic time warping can be used to calculate the optimal alignment path between the high-frequency wave sequence segments of each pair of neighboring stations and the target station. This optimal alignment path can be represented as a set of matching point pairs, where each pair includes the time index of the wave sequence segment of the neighboring station within the time window and the time index of the wave sequence segment of the target station within the time window. Then, the index difference between the two time indices in each matching point pair is calculated, and the median of all index differences is taken as the phase delay time of that neighboring station relative to the target station. For example, when the total duration of the time window is 7 hours, the time index can be an integer from 0 to 6, and the path length of the optimal alignment path can be from 7 to 13.

[0038] The sign of the aforementioned phase delay time can characterize the sequential relationship between fluctuations at neighboring stations and the target station. For example, if the phase delay time is greater than 0, it means that the fluctuations at neighboring stations are ahead of the target station, and the fluctuation sequence of the neighboring station needs to be shifted backward by a corresponding number of time intervals before being included in the interpolation; if the phase delay time is less than 0, it means that the fluctuations at neighboring stations are behind the target station, and the fluctuation sequence of the neighboring station needs to be shifted forward by a corresponding number of time intervals before being included in the interpolation.

[0039] The aforementioned timing phase compensation can be a process of arbitrarily shifting and correcting the high-frequency fluctuating temperature values ​​of neighboring stations at the target time according to the phase delay time. For example, the high-frequency fluctuating temperature values ​​of neighboring stations at the target time can be replaced with the shifted high-frequency fluctuating temperature values ​​of the corresponding time period of the target time.

[0040] The method described above for obtaining the high-frequency fluctuating temperature value of the target station at the target time using natural neighbor interpolation can be achieved by determining the interpolation weights of each neighboring station in the natural neighbor set after completing temporal phase compensation, and then performing a weighted summation of the high-frequency fluctuating temperature values ​​of each neighboring station after phase compensation to obtain the high-frequency fluctuating temperature of the target station at the target time. Natural neighbor interpolation is characterized by strong locality and good continuity, making it suitable for spatial interpolation of high-frequency fluctuation sequences after phase compensation.

[0041] Preferably, in this embodiment, a fixed-length time window is formed by extracting 3 hours before and after the target time t, with a total time window length of 7 hours, to accommodate short-term temperature fluctuations. For the target station... Neighboring sites in each natural neighbor set Within the aforementioned 7-hour time window, the wavelet packet decomposition-derived wave sequence is extracted. The fluctuation sequence of the target station to be interpolated For each group The optimal alignment path is calculated using soft-DTW (soft dynamic programming), and this path can be represented as a set of matching point pairs by the following formula: ; in and Fluctuation sequences of each neighboring site Fluctuation sequence of the target site The time index within the time window takes the value of an integer from 0 to 6. The total path length satisfies 7 ≤ ≤13. Then calculate each group. The difference in time index is used to obtain the median of these differences to get the neighboring stations. Phase delay time between fluctuation sequences relative to the target site And expressed by the following formula: ; Obtain the corresponding delay time of neighboring sites Then, timing phase compensation begins: during the delay time When this occurs, it indicates that temperature fluctuations at neighboring stations are leading those at the target station, and the fluctuation sequences at neighboring stations need to be shifted backward. Interpolation is performed after each time step; This indicates that the fluctuations at neighboring stations lag behind the target station, and the fluctuation sequences at neighboring stations need to be shifted forward. Interpolation is performed after each time step. After time series alignment, the natural neighbor interpolation method, which has strong locality and good continuity, is used to interpolate the fluctuation sequence of the target station. The interpolation result of the fluctuation sequence of the target station can be expressed as: ; Where Ω represents the target site. The set of natural neighbors, The weights of each neighboring site are obtained through natural neighbor interpolation. For each nearby station Phase delay relative to the target site.

[0042] The aforementioned high-frequency fluctuating temperature values ​​are short-term temperature fluctuation components obtained by interpolation after the fluctuation sequences of each station are aligned in time. They are used to characterize short-term, nonlinear temperature abrupt changes caused by weather processes such as cold waves, strong convection, and rapid day-night cooling. They do not include the base temperature values ​​caused by topography and season.

[0043] The target temperature mentioned above is the final interpolated temperature result obtained by linearly superimposing low-frequency background temperature values ​​and high-frequency fluctuating temperature values ​​at the target station at the target time. It fully integrates the steady-state temperature base value of the terrain with short-term, nonlinear temperature abrupt fluctuation values. The target temperature of the target station at the target time can be expressed by the following formula: ; in, For target site The target temperature at the target time t For target site The low-frequency background temperature value at the target time t, For target site The high-frequency fluctuating temperature value at the target time t.

[0044] In some implementations, the station temperature interpolation method further includes: Acquire initial temperature data sequences and location data from multiple meteorological stations within a time period; Linear interpolation is performed on the initial temperature data sequence of each meteorological station to obtain a linear temperature data sequence; The terrain dataset for each weather station is determined based on location data.

[0045] In this embodiment, the missing time points are initially filled by linear interpolation of the initial temperature data sequence, which can ensure the sequence continuity required for wavelet packet decomposition. The terrain dataset of each meteorological station is determined by location data and elevation data, which can provide terrain factor input for subsequent interpolation operations coupled with terrain, thereby providing a data foundation for improving the temperature interpolation accuracy in complex terrain areas.

[0046] The aforementioned initial temperature data sequence can be the raw temperature data sequence obtained from actual observations at meteorological stations. This raw sequence may be affected by various factors such as equipment component failures, wireless communication signal interruptions, abnormal power supply to the station, periodic instrument calibration and maintenance, and interference from extreme weather. Random, scattered missing data points or continuous missing periods ranging from several hours to several days may appear in the raw sequence, making it impossible to form complete and continuous time-series data. Since wavelet packet decomposition requires the input time sequence to be a complete sequence with equal time intervals and no missing time points, the initial temperature data sequence with missing data cannot be directly input into the decomposition algorithm. Therefore, it is necessary to preprocess and fill the initial temperature data sequence to generate a complete linear temperature data sequence, providing compliant input data for subsequent frequency division decomposition. It should be noted that this linear filling is only an intermediate preprocessing step, used only to ensure the normal execution of wavelet packet decomposition, and is not the final missing data interpolation result. For continuous missing periods, the final high-precision interpolation value will be output through the frequency division interpolation scheme of this application, replacing the coarse filling value from the preprocessing stage.

[0047] The aforementioned location data can be any data used to characterize the spatial location of a meteorological station. For example, the location data may include the longitude and latitude of the meteorological station. Location data is the core geographical basis for the spatial interpolation operation of this application. In the low-frequency background temperature interpolation stage, the spherical geographic distance is calculated based on the latitude and longitude coordinates of the target station and candidate stations. Geographic distance weights are then generated based on this distance and participate in the initial comprehensive weight fusion calculation. In the high-frequency fluctuating temperature interpolation stage, the set of natural neighbors corresponding to the target station is identified based on the location data of all meteorological stations in the entire region, providing spatial topological support for the natural neighbor interpolation operation. It is understood that the specific form of the location data is not limited to latitude and longitude coordinates; other spatial representations such as planar projected coordinates and grid points can also be used according to computational needs. This application does not impose any limitations on this.

[0048] The above-mentioned linear interpolation of the initial temperature data sequence for each meteorological station to obtain a linear temperature data sequence can be achieved by linearly interpolating the temperature values ​​of the adjacent valid observation times for each missing time period in the initial temperature data sequence, thereby obtaining a pre-filled temperature value for that missing time period, thus completing the initial temperature data sequence into a continuous sequence with equal intervals. For example, for hourly temperature data sequences, after linear interpolation, the linear temperature data sequences for each meteorological station are continuous sequences with equal intervals of 1 hour. It should be noted that the pre-filled temperature values ​​obtained by linear interpolation are only intermediate results to meet the continuity requirements of wavelet packet decomposition, and they will ultimately be overridden by the interpolation results of this application embodiment.

[0049] The aforementioned determination of the terrain dataset for each meteorological station based on location data can refer to using the latitude and longitude coordinates of each meteorological station as a spatial positioning reference, matching the target area's digital elevation model (DEM) raster data, and generating a multi-dimensional terrain attribute set corresponding to each station through spatial matching extraction and neighborhood statistical operations. This provides quantitative parameter support for subsequent terrain-coupled interpolation operations of low-frequency background temperature values. In this embodiment, the terrain dataset may include terrain factors such as elevation, aspect, and topographic relief. Specifically, the electronic device can first map the latitude and longitude coordinates of all meteorological stations to the raster coordinate system of the DEM data, such as... Figure 2 As shown, the grid cell corresponding to each station is located, and the grid cell corresponding to each meteorological station is set as e. A 3x3 window of 9 grid cells centered on grid cell e is obtained, and the elevation value of the grid cell is extracted as the altitude of the meteorological station. Then, the elevation values ​​of nine grids in a 3x3 window centered on the grid where the weather station is located are calculated. Based on this, the Zevenbergen-Thorne formula is used to calculate the slope change rate in the east-west direction and the slope change rate in the north-south direction, respectively. The east-west slope change rate... The following formula can be used for calculation: ; North-South Slope Change Rate The following formula can be used for calculation: ; Raster resolution Then the slope aspect of the weather station The calculation formula is as follows: ; The formula for calculating the terrain relief (TRI) is as follows: ; in This represents the average elevation of all grid points within the window. After the above processing, each terrain factor in the terrain data can be stored as a static attribute associated with the temperature data sequence of meteorological stations for subsequent interpolation calculations coupled with the terrain. It is understood that the terrain dataset can also select other terrain factors according to actual conditions and needs; when different terrain factors are selected, the formulas in this application will also change accordingly.

[0050] In some implementations, the initial comprehensive weight of each candidate site in the initial low-frequency candidate site set is obtained after calculating the geographical distance weight and terrain influence weight of each candidate site in the initial low-frequency candidate site set, including: Based on the location data of each candidate site in the initial low-frequency candidate site set and the target site, determine the geographical distance weight of each candidate site in the initial low-frequency candidate site set. The terrain influence weight of each candidate site in the initial low-frequency candidate site set is determined based on the difference between the terrain dataset of each candidate site in the initial low-frequency candidate site set and the target site. The initial comprehensive weight of each candidate site in the initial low-frequency candidate site set is determined based on the weights of geographical distance and terrain influence.

[0051] In this embodiment, the interpolation weights are split into two parts: geographical distance weights and terrain influence weights. The geographical distance weights characterize the spatial proximity between meteorological stations, while the terrain influence weights characterize the terrain similarity between them. Combining these two weights simultaneously considers the impact of both spatial distance and terrain conditions on temperature, overcoming the limitation of interpolation techniques based solely on spatial distance that struggle to quantify terrain influences, thereby improving interpolation accuracy in complex terrain areas. In this embodiment, for the target station... Assume that the initial set of K candidate meteorological stations in the low-frequency candidate station set is Each can be calculated using the following formula. Initial composite weights To measure its impact on the target site: ; The value of K can be a preset initial quantity M at the beginning of the interpolation operation, and the value of K can be redefined in subsequent calculations. Geographic distance weighting, The terrain influence weights are constructed based on the terrain dataset.

[0052] The aforementioned geographical distance weight can be any weight negatively correlated with the geographical distance between the candidate and target stations; the closer the geographical distance, the greater the geographical distance weight. For example, an electronic device can calculate the geographical distance between two meteorological stations based on their longitude and latitude, and use the reciprocal of that geographical distance or the reciprocal of the square of that geographical distance as the geographical distance weight. In this embodiment, the geographical distance weight... It can be determined by the following formula: ; in Candidate sites in the initial low-frequency candidate site set To the target site Geographical distance.

[0053] The difference values ​​in the aforementioned terrain dataset can be any numerical value used to quantify the degree of difference in terrain conditions between two meteorological stations. For example, the difference value could be the terrain difference degree. Topographical differences The relationship with the terrain dataset is shown in the following formula: ; in, , These represent the differences in altitude, slope aspect, and topographic relief between each candidate site in the initial low-frequency candidate site set and the target site. , , This is the maximum difference value of the corresponding terrain factor within the range, and the above maximum difference value is used for normalization calculation; , , The weighting coefficients determined using the entropy weighting method satisfy... The relationship, and in this embodiment , , The value can vary depending on the geographical environment of each meteorological station in the initial low-frequency candidate station set and the target station. Preferably, in plain areas... =0.60、 =0.35、 =0.05, in hilly areas =0.50、 =0.20、 =0.30, in mountainous areas =0.42、 =0.25、 =0.33.

[0054] The aforementioned terrain influence weights can be any weights that decrease as the difference in the terrain dataset increases. For example, relevant equipment can be based on terrain dissimilarity. The terrain influence weight is calculated using a Gaussian kernel function, and can be expressed by the following formula: ; in, This is a scale parameter representing the impact of terrain differences on the weights. This scale parameter controls the decay rate of terrain differences on the weights; the greater the terrain difference, the smaller the weight of the terrain's influence. In this embodiment, this scale parameter is set to a value through cross-validation. .

[0055] In some implementations, obtaining the corresponding cumulative weight sum and the total weight sum includes: The initial comprehensive weights of each candidate site in the initial low-frequency candidate site set are sorted from high to low and then accumulated sequentially to obtain the total accumulated weights. The initial comprehensive weights of each candidate site in the initial low-frequency candidate site set are summed to obtain the total weight.

[0056] In this embodiment, by combining descending sorting with successive accumulation processing logic, all calculations can be completed with only a single sorting and a single traversal, resulting in high computational efficiency. At the same time, this mechanism naturally ensures that high-weight stations are included in the interpolation calculation first, which is in line with the weight allocation rules of low-frequency background temperature spatial interpolation and avoids low-contribution redundant stations interfering with the interpolation accuracy.

[0057] In some implementations, determining the cumulative low-frequency candidate site set when the proportion of the cumulative weighted sum to the total weight reaches a preset weight threshold includes: If the quotient of the sum of accumulated weights and the sum of all weights is not less than a preset weight threshold, obtain the accumulated quantity and the accumulated low-frequency candidate site set corresponding to the accumulated quantity. The accumulated quantity is the number of candidate sites accumulated in the initial low-frequency candidate site set. If the accumulated quantity is lower than the first quantity threshold, the accumulated quantity is the first quantity threshold. If the accumulated quantity is higher than the second quantity threshold, the accumulated quantity is the second quantity threshold.

[0058] In this embodiment, since the terrain conditions and station densities of different target stations vary, it is difficult to adapt to all target stations using a fixed number of reference stations. The number of stations participating in the interpolation is dynamically determined by accumulating the contribution of weights, which allows for the adaptive selection of an appropriate number of reference stations for each target station. At the same time, by limiting the range of the number of stations through a first number threshold and a second number threshold, it is possible to avoid unstable interpolation results due to too few reference stations or reduced interpolation accuracy due to too many reference stations introducing weakly correlated stations, thereby further improving the reliability of the interpolation results.

[0059] The first threshold can be the minimum number of meteorological stations participating in the interpolation, for example, the first threshold can be 3; the second threshold can be the maximum number of meteorological stations participating in the interpolation, for example, the second threshold can be 12. That is to say, the cumulative number of meteorological stations participating in the final interpolation can range from 3 to 12.

[0060] The cumulative low-frequency candidate station set corresponding to the cumulative number mentioned above can be a set composed of the first cumulative number of meteorological stations taken from the sorted initial low-frequency candidate station set.

[0061] Specifically, the number of low-frequency candidate sites accumulated is limited. The value range is between the first and second quantity thresholds, that is, [3, 12], which is the initial low-frequency candidate site set. If the initial weights of the candidate sites are arranged in descending order, then the cumulative number of low-frequency candidate sites is calculated. It can be expressed using the following formula: ; Where M=20 is the initial number of candidates. To set a preset weight threshold; to obtain the cumulative number of low-frequency candidate sites. Afterwards, As the K value, the initial comprehensive weight of each candidate site in the accumulated low-frequency candidate site set is normalized using the following formula to obtain the normalized initial comprehensive weight of each candidate site in the accumulated low-frequency candidate site set. : ; Then target site Low-frequency background temperature value at target time t It can be represented as: ; in, This refers to any candidate site in the cumulative low-frequency candidate site set.

[0062] In some implementations, obtaining the target temperature of the target station at the target time based on the low-frequency background temperature value and the high-frequency fluctuating temperature value includes: The sum of the low-frequency background temperature value and the high-frequency fluctuating temperature value is taken as the target temperature of the target station at the target time.

[0063] In this embodiment, the summation operation corresponds to the target temperature reconstruction stage under the wavelet packet decomposition architecture. After wavelet packet decomposition, the initial temperature data sequence is split into a background sequence and a fluctuation sequence with mutually independent physical meanings. The low-frequency sequence corresponds to the temperature base determined by slow-changing factors such as large-scale meteorological background and diurnal temperature rhythm, while the high-frequency sequence corresponds to the temperature deviation caused by fast-changing factors such as local topographic disturbances and short-term weather fluctuations. After the two types of sequence components are spatially extrapolated and missing data are filled in using adaptive interpolation algorithms, they are numerically superimposed in a one-to-one correspondence manner at each time step to reconstruct a complete and continuous target temperature sequence.

[0064] In some implementations, the temperature interpolation method at this station further includes: When there are multiple consecutive target times that need to be interpolated at the target site, interpolation is performed hourly from both ends of the missing time period toward the middle.

[0065] In this embodiment, when the target site experiences multiple consecutive data gaps due to equipment failure, transmission interruption, or other reasons, the time interval between the time gaps in the middle of the missing time period and the valid observation time periods is relatively large, resulting in low reliability of direct interpolation. By interpolating time by time from both ends of the missing time period toward the middle, each interpolation can be based on known or interpolated temperature values ​​that are adjacent to the target time and have high reliability. This avoids the accumulation of errors caused by directly interpolating the missing middle time periods over long distances, thereby improving the interpolation accuracy and reliability in scenarios with continuous data gaps.

[0066] The above interpolation, which proceeds hourly from both ends of the missing time period towards the middle, can specifically be achieved by setting a left pointer. Right pointer The initial position of the left pointer is the first time interval of the missing time interval, and the initial position of the right pointer is the last time interval of the missing time interval. The above interpolation process is performed sequentially on the time intervals corresponding to the left and right pointers, so that the interpolation result covers the pre-filled value of the corresponding time interval. After completing one round of interpolation, the left pointer is moved one time interval to a later time, and the right pointer is moved one time interval to an earlier time. The process ends when the position of the left pointer exceeds the position of the right pointer. At this time, all the pre-filled values ​​in the missing time interval are covered by the interpolation result.

[0067] In this embodiment, wavelet packet decomposition is performed on the linear temperature data sequence of each acquired meteorological station to divide the temperature data sequence into a low-frequency background sequence and a high-frequency fluctuation sequence. Low-frequency background temperature values ​​are obtained by performing terrain-coupled interpolation on the background sequence, and high-frequency fluctuation temperature values ​​are obtained by performing temporal alignment and interpolation on the fluctuation sequence. The target temperature is then obtained based on the low-frequency background temperature values ​​and the high-frequency fluctuation temperature values. Thus, on the one hand, the coupling of the terrain dataset quantifies the influence of terrain on temperature, overcoming the shortcomings of related technologies where interpolation techniques based on spatial distance relationships are difficult to quantify terrain influence and obtain accurate interpolation results in complex terrain areas. On the other hand, temporal alignment allows the interpolation process to adapt to the nonlinear and short-term fluctuation characteristics of temperature changes, overcoming the shortcomings of related technologies where multivariate interpolation techniques introducing covariates cannot guarantee consistent interpolation accuracy in different terrain areas and during sudden weather changes. Simultaneously, the entire interpolation process is based on signal decomposition and weighted interpolation, eliminating the need to train a deep learning model and avoiding the problem of model susceptibility to noise interference without requiring excessive computational resources, thus enhancing practicality. Therefore, the embodiments of this application can simultaneously take into account the topographical spatial characteristics of meteorological stations and the temperature fluctuation characteristics of the target time to be interpolated without requiring excessive computing resources. This overcomes the shortcomings of existing technologies that cannot simultaneously take into account the steady-state base of the terrain and short-term nonlinear fluctuations when using a unified spatial strategy. It achieves a balance between accuracy and computational efficiency in complex terrain and sudden weather scenarios, and significantly improves the interpolation accuracy of temperature data sequences observed by ground meteorological stations in various complex terrain environments and sudden weather scenarios.

[0068] To facilitate a better understanding of the station temperature interpolation method in the embodiments of this application, examples of the station temperature interpolation method in practical application scenarios are provided below.

[0069] An application example using temperature interpolation in Hunan Province is provided. In this example, the target region is adjusted to Hunan Province. This region encompasses diverse and complex terrains, including the Wuling Mountains in western Hunan, the hilly areas in central Hunan, and the Dongting Lake Plain. The region experiences significant elevation differences, ranging from approximately 20 meters to over 1600 meters, resulting in a substantial influence of topography on the spatial distribution of temperature. Furthermore, Hunan has a subtropical monsoon climate, exhibiting distinct diurnal and seasonal temperature variations, making it an ideal region for verifying the interpolation method's adaptability to scenarios of abrupt weather changes.

[0070] Data Acquisition and Preprocessing. Hourly temperature observation data from 2375 national and regional stations with temperature elements in Hunan Province from 2021 to 2025 were acquired as the initial temperature data sequence; and for any meteorological station within this target area, based on the station's location data, such as... Figure 3As shown, ASTER GDEM topographic data maps of meteorological stations in Hunan Province were obtained, and the elevation of the grid cell containing the meteorological station was directly extracted from the ASTER GDEM. Based on the elevation values ​​of 9 grid cells in a 3x3 window centered on the station, the Zevenbergen-Thorne formula was used to calculate the slope change rate in the east-west and north-south directions, respectively. Then, the slope aspect of the station was calculated by combining the grid resolution. The difference between the elevation value of each grid cell in the window and the average elevation value of all grid cells in the window was calculated to obtain the topographic relief of the station. Linear interpolation was performed on the missing time periods in the initial temperature data series to obtain a continuous series with equal 1-hour intervals. Elevation, slope aspect, and topographic relief were stored as static attributes associated with the temperature data series for subsequent scale decomposition and coupled topographic interpolation.

[0071] Wavelet packet decomposition. A three-layer Daubechies-4 wavelet basis was used to perform wavelet packet decomposition on the linear temperature data series of each station. The original sequence was decomposed into eight equal-width sub-bands. The first four sub-bands of the third layer were reconstructed into a low-frequency background sequence, and the last four sub-bands of the third layer were reconstructed into a high-frequency wave sequence.

[0072] Background sequence interpolation coupled with terrain. For the target station to be interpolated, 20 nearby reference stations are selected according to geographical distance from near to far to form an initial low-frequency candidate station set. The geographical distance weight and terrain influence weight of each reference station are calculated separately, and the two are multiplied to obtain the initial comprehensive weight of each reference station. Among them, the terrain difference is obtained by weighted summation of normalized differences of three factors: altitude, aspect, and terrain relief. The weight coefficients corresponding to the three factors are determined by the entropy weight method. The terrain influence weight is calculated by Gaussian kernel function based on the terrain difference. The scale parameter is determined by cross-validation. The initial comprehensive weights of each reference station are sorted from high to low and accumulated sequentially. Accumulation stops when the quotient of the sum of accumulated weights and the sum of all weights is not lower than a preset weight threshold. The number of stations participating in the interpolation is limited to between 3 and 12 to obtain the accumulated low-frequency candidate station set. After normalizing the initial comprehensive weights of each station in the accumulated low-frequency candidate station set, the background sequence values ​​of each station at the target time are weighted and averaged to obtain the low-frequency background temperature value of the target station at the target time.

[0073] Temporal alignment and interpolation of the fluctuation sequence. A fixed-length alignment window of 7 hours is formed by extracting 3 hours before and after the target time. For each neighboring station in the natural neighbor set of the target station, a high-frequency fluctuation sequence segment within this alignment window is extracted. Soft dynamic time warping is used to calculate the optimal alignment path. The median of the index difference between matching point pairs is taken as the phase delay time of the neighboring station relative to the target station. Based on this, temporal phase compensation is performed on the high-frequency fluctuation temperature value of the neighboring station at the target time. Then, natural neighbor interpolation is used to interpolate the compensated high-frequency fluctuation temperature values ​​of each neighboring station to obtain the high-frequency fluctuation temperature value of the target station at the target time.

[0074] Sequence merging and result output. The low-frequency background temperature value is added to the high-frequency fluctuating temperature value to obtain the target temperature of the target station at the target time, and this target temperature is used to overwrite the pre-filled value at the target time. For consecutive missing time periods, interpolation is performed hourly from both ends of the missing time period towards the middle until all pre-filled values ​​within the missing time period are covered by the interpolation result.

[0075] Interpolation effectiveness verification. Four representative methods were selected as benchmarks: Inverse Distance Weighted Interpolation (AIDW) with 12 reference points, Co-kriging (CK) with altitude as a covariate and a step size of 0.01, Thin Disk Smoothing Spline (TPSS) with a smoothing parameter of 1, and Random Forest Spatial Interpolation (RF) with 500 decision trees. Two types of missing data were constructed for interpolation verification: one was random single-point missing data, where 10% of the time intervals were randomly removed from the complete temperature series of each station to simulate sporadic missing data; the other was blocky continuous missing data, where three time intervals were randomly selected from the complete temperature series of each station, each segment being continuously missing for 12 hours to simulate long-term missing data due to equipment failure and transmission interruption. All methods were repeated 10 times using different random seeds, and the average mean absolute error (MAE) and root mean square error (RMSE) were calculated as the verification results to reduce the impact of randomness. The interpolation error comparison results of each method are shown in Table 1. Table 1

[0076] As shown in Table 1, the hourly average MAE and hourly average RMSE of the embodiments of this application are the best among all the methods compared, and are reduced by 0.19℃ and 0.30℃ respectively compared with the second best method. This indicates that the embodiments of this application have significantly improved the control of interpolation error compared with the comparison methods, and the significant reduction in RMSE reflects the ability of the embodiments of this application to suppress extreme interpolation errors. At the same time, it can be seen from the error ratio that the interpolation error of the embodiments of this application is basically controlled within 2.0℃.

[0077] Furthermore, referring to the basic landform classification standards of China, and using the topographic relief extracted by ASTER GDEM as the indicator, the stations were divided into three topographic zoning categories: mountains (topographic relief greater than 200 meters), hills (topographic relief between 30 and 200 meters), and plains (topographic relief less than 30 meters). The statistical errors of each method were compared under different terrains and seasons. The results of the interpolation error comparison under different terrains are shown in Table 2. Table 2

[0078] As shown in Table 2, the mean MAE of all methods increases with the increase of terrain relief. This is because the complex terrain and sparse stations lead to an increase in interpolation error. Among them, AIDW, as a pure geographic distance method, has the largest error. Although CK introduces altitude as a covariate, the error caused by terrain is still significant. TPSS and RF perform better in mountainous areas than CK and AIDW, but are still inferior to the embodiment of this application. The embodiment of this application shows the best MAE results for plains, hills, and mountains. Moreover, the error increases from plains to hills and then to mountains are 0.16℃ and 0.18℃, respectively, which are significantly smaller than other comparative methods, demonstrating the advantages of scale decomposition and coupled terrain weights.

[0079] Table 3 shows the comparison of the mean MAE under different seasons: Table 3

[0080] As can be seen from Table 3, when the MAE mean of each method is statistically analyzed by season, the error of each method shows that the error is greater in winter than in spring, greater in spring than in autumn, and greater in autumn than in summer. However, the temperature interpolation method of this application embodiment is the best in all four seasons, and the error fluctuates the least with seasonal changes, showing strong seasonal adaptability.

[0081] The distribution of errors in the station temperature interpolation method in this application under different terrain conditions is further analyzed, as shown in Table 4: Table 4

[0082] As shown in Table 4, the average MAE of the temperature interpolation method in this application embodiment has been controlled within 2.0℃ in plain areas. Although the proportion of small errors (MAE < 1.0℃) decreases with the increase of terrain complexity, it is generally within a controllable range, indicating good extreme error control capability.

[0083] Temperature interpolation during a cold wave is used as an application example. In this example, the temperature interpolation results obtained by the complete interpolation method of this application and the interpolation method using only the background sequence are compared with the observed values ​​using the correlation coefficient R². The correlation coefficient R is calculated monthly using data from 2021 to 2025. 2 The average value, the result is as follows Figure 4 As shown, the correlation coefficients of the embodiments of this application are superior to the interpolation results using only the background sequence in all months, with both exhibiting a characteristic of being higher in summer and autumn and lower in winter and spring. Since low-frequency background components dominate and temperature fluctuations are smaller in summer, interpolation using only the background sequence can also achieve results close to those of the embodiments of this application. However, the correlation coefficients of interpolation results using only the background sequence decrease significantly in the winter and spring months, while the correlation coefficients of the embodiments of this application change relatively gradually with the months. This indicates that during periods of frequent cold air activity and increased temperature fluctuations, relying solely on the background sequence is insufficient to effectively capture temperature changes. Introducing high-frequency fluctuation sequence interpolation allows the embodiments of this application to better adapt to the influence of temperature fluctuations and provides a better fit to the observed data.

[0084] Furthermore, the temperature series from 12:00 on the 27th to 11:00 on the 29th of March 2025, during the period of most drastic temperature drop in Hunan during the cold wave, were selected for comparison. Three stations were selected: Changsha Station (57678, altitude 68 meters), Xiangxi Huayuan Station (P1434, altitude 292 meters), and Zhangjiajie Sangzhi Station (P1534, altitude 1328 meters). The interpolation results of the method in this application and the overall suboptimal Random Forest Spatial Interpolation (RFSP) were compared with the observed values. Figures 5 to 7 As can be seen, during the three periods of cold wave cooling at the plain stations, the interpolation results of RFSP all showed significant lag. The embodiment of this application captured the cooling process more accurately. During the other periods, both methods captured the overall temperature change trend well, indicating that the random forest method also exhibits phase lag even in environments with low spatial variability. At hilly and mountain stations, temperature observations showed two phases: rapid cooling and stabilization. At hilly stations, RFSP produced a more significant lag error during the rapid cooling period than in the plains. At mountain stations, the temperature dropped faster, and the lag error of RFSP was also more pronounced during the rapid cooling period. In summary, as terrain complexity increases, the error increase of the embodiment of this application remains gradual, while the lag bias of RFSP increases positively correlated with the increase in terrain undulation. This indicates that the embodiment of this application, through frequency division processing and phase alignment, has better adaptability in environments with sudden weather changes and complex terrain.

[0085] This application embodiment also provides a station temperature interpolation device, which includes: The target station determination module is used to determine the target station to be interpolated at the target time in response to the interpolation instruction. The target station is one of multiple meteorological stations. The wavelet packet decomposition module is used to perform wavelet packet decomposition on the linear temperature data sequence of each pre-acquired meteorological station to obtain the background sequence and the fluctuation sequence. The frequency of the background sequence is less than the frequency of the fluctuation sequence, and the target time is the time in the time interval corresponding to the linear temperature data sequence. The low-frequency background temperature acquisition module is used to determine an initial low-frequency candidate station set based on the location data and terrain dataset of the target station. After calculating the geographical distance weight and terrain influence weight of each candidate station in the initial low-frequency candidate station set, the module obtains the initial comprehensive weight of each candidate station in the initial low-frequency candidate station set and obtains the corresponding cumulative weight sum and total weight sum. When the proportion of the cumulative weight sum to the total weight sum reaches a preset weight threshold, the module determines the cumulative low-frequency candidate station set. Based on the temperature value of each candidate station in the cumulative low-frequency candidate station set at the target time and its corresponding initial comprehensive weight, the module obtains the low-frequency background temperature value of the target station at the target time. The high-frequency fluctuating temperature acquisition module is used to acquire a time window centered on the target time and a set of natural neighbors of the target station. It extracts the high-frequency fluctuating sequence segments of each neighboring station in the natural neighbor set and the target station within the time window. After aligning the high-frequency fluctuating sequence segments of each neighboring station in the natural neighbor set with the high-frequency fluctuating sequence segments of the target station in time, it obtains the phase delay time corresponding to each neighboring station in the natural neighbor set. Based on the phase delay time corresponding to each neighboring station in the natural neighbor set, it performs time-phase compensation on the high-frequency fluctuating temperature value of each neighboring station in the natural neighbor set at the target time. Finally, it uses the natural neighbor interpolation method to obtain the high-frequency fluctuating temperature value of the target station at the target time. The target temperature acquisition module is used to obtain the target temperature of the target station at the target time based on the low-frequency background temperature value and the high-frequency fluctuating temperature value.

[0086] The aforementioned target station determination module, wavelet packet decomposition module, low-frequency background temperature acquisition module, high-frequency fluctuating temperature acquisition module, and target temperature acquisition module can be software functional modules within the station temperature interpolation device, or they can be functional modules implemented by hardware circuits; no limitation is imposed here. For example, each of the above modules can be a program module running in a processor.

[0087] The station temperature interpolation device provided in this application embodiment can perform the processing procedure in the above-described station temperature interpolation method embodiment and achieve the same beneficial effect. To avoid repetition, it will not be described again here.

[0088] like Figure 8The diagram shows a structural schematic of an electronic device provided in this application embodiment. The electronic device 600 includes: one or more processors 810; and a memory 820 storing a computer program. When the computer program is executed by one or more processors 810, the one or more processors 810 implement the station temperature interpolation method described in any of the above embodiments.

[0089] Memory 820, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory 820 may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 820 may optionally include memory remotely located relative to processor 810, and these remote memories can be connected to processor 810 via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0090] The memory 820 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 820 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 820 and is called and executed by the processor 810.

[0091] The processor 810 can be implemented using a general-purpose central processing unit (CPU), microprocessor, application specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0092] In some embodiments, the electronic device further includes: an input / output interface for inputting and outputting information; a communication interface for communication and interaction between the device and other devices, which can be implemented via wired means (e.g., USB, Ethernet cable, etc.) or wireless means (e.g., mobile network, WIFI, Bluetooth, etc.); and a bus for transmitting information between various components of the device (e.g., processor 810, memory 820, input / output interface, and communication interface); wherein the processor 810, memory 820, input / output interface, and communication interface can be interconnected within the device via the bus.

[0093] This application also provides a computer-readable storage medium storing computer-executable instructions for causing a computer to perform the station temperature interpolation method described in any of the above embodiments.

[0094] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0095] The functional blocks shown in the above structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. The programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM, floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0096] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0097] The aspects of this application have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer program products according to embodiments of this application. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, 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, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0098] The above are merely specific embodiments of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for interpolating station temperature, characterized in that, Applied to a target adjustment region, wherein the target adjustment region has multiple meteorological stations, the temperature interpolation method for these stations includes: In response to the interpolation instruction, the target station to be interpolated at the target time is determined, wherein the target station is one of the plurality of meteorological stations; Wavelet packet decomposition is performed on the linear temperature data sequence of each of the pre-acquired meteorological stations to obtain a background sequence and a fluctuation sequence, wherein the frequency of the background sequence is less than the frequency of the fluctuation sequence; the target time is the time in the time interval corresponding to the linear temperature data sequence. An initial low-frequency candidate station set is determined based on the location data and terrain dataset of the target station. The geographical distance weight and terrain influence weight of each candidate station in the initial low-frequency candidate station set are calculated to obtain the initial comprehensive weight of each candidate station in the initial low-frequency candidate station set. The corresponding cumulative weight sum and total weight sum are obtained. When the proportion of the cumulative weight sum to the total weight sum reaches a preset weight threshold, an accumulated low-frequency candidate station set is determined. The low-frequency background temperature value of the target station at the target time is obtained based on the temperature value of each candidate station in the accumulated low-frequency candidate station set at the target time and its corresponding initial comprehensive weight. A time window centered on the target time and a set of natural neighbors of the target station are obtained. High-frequency fluctuation sequence segments of each neighboring station in the set of natural neighbors and the target station within the time window are extracted. The high-frequency fluctuation sequence segments of each neighboring station in the set of natural neighbors are time-aligned with the high-frequency fluctuation sequence segments of the target station to obtain the phase delay time corresponding to each neighboring station in the set of natural neighbors. The high-frequency fluctuation temperature value of each neighboring station in the set of natural neighbors is time-phase compensated according to the phase delay time corresponding to each neighboring station in the set of natural neighbors. The high-frequency fluctuation temperature value of the target station at the target time is obtained by using natural neighbor interpolation. The target temperature of the target station at the target time is obtained based on the low-frequency background temperature value and the high-frequency fluctuating temperature value.

2. The station temperature interpolation method according to claim 1, characterized in that, Also includes: Acquire the initial temperature data sequence and the location data of multiple meteorological stations within the time period; Linear interpolation is performed on the initial temperature data sequence of each meteorological station to obtain the linear temperature data sequence; The terrain dataset for each of the meteorological stations is determined based on the location data.

3. The station temperature interpolation method according to claim 1, characterized in that, The initial comprehensive weight of each candidate site in the initial low-frequency candidate site set is obtained by calculating the geographical distance weight and terrain influence weight of each candidate site in the initial low-frequency candidate site set, including: Based on the location data of each candidate site in the initial low-frequency candidate site set and the target site, determine the geographical distance weight of each candidate site in the initial low-frequency candidate site set; Based on the difference value between the terrain dataset of each candidate site in the initial low-frequency candidate site set and the target site, the terrain influence weight of each candidate site in the initial low-frequency candidate site set is determined; The initial comprehensive weight of each candidate site in the initial low-frequency candidate site set is determined based on the geographical distance weight and the terrain influence weight.

4. The station temperature interpolation method according to claim 1, characterized in that, The process of obtaining the corresponding cumulative weight sum and the total weight sum includes: The initial comprehensive weights of each candidate site in the initial low-frequency candidate site set are sorted from high to low and then accumulated sequentially to obtain the total accumulated weights. The initial comprehensive weights of each candidate site in the initial low-frequency candidate site set are summed to obtain the total weight.

5. The station temperature interpolation method according to claim 1, characterized in that, The step of determining the accumulated low-frequency candidate site set when the proportion of the sum of accumulated weights to the total sum of all weights reaches a preset weight threshold includes: If the quotient of the sum of accumulated weights and the sum of all weights is not less than a preset weight threshold, the accumulated quantity and the accumulated low-frequency candidate site set corresponding to the accumulated quantity are obtained. The accumulated quantity is the number of candidate sites accumulated in the initial low-frequency candidate site set. If the accumulated quantity is lower than a first quantity threshold, the accumulated quantity is the first quantity threshold. If the accumulated quantity is higher than a second quantity threshold, the accumulated quantity is the second quantity threshold.

6. The station temperature interpolation method according to claim 1, characterized in that, The step of obtaining the target temperature of the target station at the target time based on the low-frequency background temperature value and the high-frequency fluctuating temperature value includes: The sum of the low-frequency background temperature value and the high-frequency fluctuating temperature value is taken as the target temperature of the target station at the target time.

7. The station temperature interpolation method according to claim 1, characterized in that, Also includes: When there are multiple consecutive target times that need to be interpolated at the target site, interpolation is performed hourly from both ends of the missing time period toward the middle.

8. A station temperature interpolation device, characterized in that, include: The target station determination module is used to determine the target station to be interpolated at the target time in response to the interpolation instruction, wherein the target station is one of multiple meteorological stations; The wavelet packet decomposition module is used to perform wavelet packet decomposition on the linear temperature data sequence of each of the pre-acquired meteorological stations to obtain a background sequence and a fluctuation sequence, wherein the frequency of the background sequence is less than the frequency of the fluctuation sequence; the target time is the time in the time interval corresponding to the linear temperature data sequence. The low-frequency background temperature acquisition module is used to determine an initial low-frequency candidate station set based on the location data and terrain dataset of the target station, calculate the geographical distance weight and terrain influence weight of each candidate station in the initial low-frequency candidate station set to obtain the initial comprehensive weight of each candidate station in the initial low-frequency candidate station set, and obtain the corresponding cumulative weight sum and total weight sum. When the proportion of the cumulative weight sum to the total weight sum reaches a preset weight threshold, the cumulative low-frequency candidate station set is determined, and the low-frequency background temperature value of the target station at the target time is obtained based on the temperature value of each candidate station in the cumulative low-frequency candidate station set at the target time and its corresponding initial comprehensive weight. The high-frequency fluctuating temperature acquisition module is used to acquire a time window centered on the target time and a set of natural neighbors of the target station, extract high-frequency fluctuating sequence segments of each neighboring station in the set of natural neighbors and the target station within the time window, align the high-frequency fluctuating sequence segments of each neighboring station in the set of natural neighbors with the high-frequency fluctuating sequence segments of the target station in time sequence, obtain the phase delay time corresponding to each neighboring station in the set of natural neighbors, and perform time-phase compensation on the high-frequency fluctuating temperature value of each neighboring station in the set of natural neighbors at the target time based on the phase delay time corresponding to each neighboring station in the set of natural neighbors, and then use natural neighbor interpolation to obtain the high-frequency fluctuating temperature value of the target station at the target time. The target temperature acquisition module is used to obtain the target temperature of the target station at the target time based on the low-frequency background temperature value and the high-frequency fluctuating temperature value.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory storing computer programs; The computer program executed by the processor implements the station temperature interpolation method as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions for causing a computer to perform the station temperature interpolation method as described in any one of claims 1 to 7.

Images