Air temperature and rainfall data construction method without site observation data
By fusing observational data with reanalysis data and combining multivariate bias correction methods, high-resolution temperature and precipitation data sequences were generated, solving the problems of low interpolation accuracy and large errors caused by complex mountainous terrain and sparse stations, and achieving high-precision data construction.
Patent Information
- Application Number
- CN202511252739.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-03
- Publication Date
- 2025-11-18
AI Technical Summary
Existing spatial interpolation techniques are difficult to accurately reflect the influence of complex terrain in mountainous, high-altitude areas with no data, resulting in low accuracy of interpolation results. Furthermore, mountainous areas have few and unevenly distributed stations, making existing interpolation methods prone to large errors when processing sparse data.
By acquiring and preprocessing observational data and reanalysis datasets, the observational data are integrated into the reanalysis data using a weighted average method based on influence radius and distance. Statistical downscaling and multivariate bias correction are then performed to generate high-resolution temperature and precipitation data sequences.
It improves the accuracy of meteorological data interpolation in high-altitude mountainous areas with no data, reduces errors, and generates a reliable high-resolution gridded dataset, which is suitable for runoff studies in the context of climate change.
Smart Images
Figure 61CB8436-A64D-4090-9F83-B3DDCC78BF63 
Figure 7F4C288A-2C2E-4040-A29F-9ECB65489577 
Figure 97AF0015-A8D1-4AAD-ACA0-6EA272EDADE7
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of sequence construction, and particularly to a temperature and precipitation data construction method without station observation data. BACKGROUND
[0002] Due to the factors such as the lack of stations in mountainous areas and the difficulty in maintaining stations, the temperature and precipitation observation data in mountainous areas are seriously missing. At the same time, the problem of flood resource utilization in a typical watershed needs to use long-time scale and high-resolution data sets.
[0003] The meteorological data interpolation technology based on spatial interpolation is one of the common technical means in the field of meteorological data interpolation. This method estimates the meteorological element values of the data-free area according to the geographical spatial position relationship by using the data of the known observation stations. For example, Kriging is an interpolation method based on spatial autocorrelation, which can estimate the meteorological data of the unknown area according to the spatial distribution characteristics of the known station data.
[0004] However, the existing spatial interpolation technology has obvious shortcomings in the application of high-altitude data-free areas in mountainous areas. First, the terrain in mountainous areas is complex, and the elevation difference is large. The spatial distribution of meteorological elements is influenced by many factors such as terrain and altitude. The traditional spatial interpolation method is often difficult to accurately reflect the influence of such complex terrain, resulting in low accuracy of the interpolation results. Second, the stations in mountainous areas are sparse and unevenly distributed. The existing interpolation method is prone to large errors when dealing with sparse data, especially in high-altitude unpopulated areas, which lacks sufficient adjacent station data as support, making the reliability of the interpolation results greatly discounted. In addition, the existing spatial interpolation technology usually assumes that the spatial distribution of meteorological elements is stationary, but in mountainous areas, due to the dramatic changes in terrain, the spatial distribution of meteorological elements may have obvious non-stationarity, which further limits the application effect of the existing spatial interpolation method in data-free areas in mountainous areas.
[0005] In summary, a temperature and precipitation data construction method without station observation data is needed to solve the above problems. SUMMARY
[0006] The present application aims to provide a temperature and precipitation data construction method without station observation data to solve the problems raised in the background.
[0007] To achieve the above-mentioned purpose, the present application provides the following technical scheme: The present application provides a temperature and precipitation data construction method without station observation data, comprising the following steps: S1. Obtain and preprocess the observation data and reanalysis data set; S2. Use the weighted average method based on the influence radius and distance to fuse the observation data into the reanalysis data of the same period to generate the fused grid data; S3. Perform statistical downscaling and true value setting, taking the fused grid data as the true value reference for driving the statistical downscaling process; S4. Apply the N-dimensional probability density function transformation to match the bias correction method of multivariate distribution to the reanalysis data to generate the final data sequence.
[0008] Preferably: the step S1 acquires and preprocesses the observation data, and the implementation process is: collecting the measured meteorological data of the target area, including air temperature and precipitation, and data cleaning and formatting, which is used to process missing values and ensure time consistency.
[0009] Preferably: the implementation process of the step S1 reanalysis data is: using high-resolution reanalysis data set as the basic framework, and using ERA5_Land data set for document, which is used to provide full coverage of gridded meteorological elements.
[0010] Preferably: the implementation process of the step S2 is: S21. For each reanalysis grid point, calculate its geographical distance and influence radius from the adjacent observation stations. Through the weighted average algorithm, the weight is based on the reciprocal of the distance or similar function, and the observation value is fused into the reanalysis data, see formula (1): (1); In the formula, wherein, is the fused grid value, is the observation value of the i th station, is the original ERA5_Land value, is the weight, which is inversely proportional to the distance, and n is the number of stations within the influence radius; S22. Generate “fused simultaneous grid data” as the “true value” reference for the subsequent steps.
[0011] Preferably: the step S3 downscaling operation utilizes the high-resolution characteristics of the fused data to spatially refine the reanalysis data, and through the regression model or probability distribution matching, the coarse resolution reanalysis data is down-scaled to the target grid.
[0012] Preferably: the role of the step S3 true value is: the fused data as a reference is used to calibrate the accuracy of the subsequent bias correction, integrates the local truth of the observation and the global coverage of the reanalysis, and is used to solve the error caused by the sparse stations.
[0013] Preferably: the implementation process of the step S4 is: S41. The system bias of the multivariate is corrected using the MBCn method, and the N-pdf transformation adjusts the value range and spatial structure of the reanalysis data by matching the joint probability distribution of the fused data true value and the original reanalysis data, as shown in formula (2): (2); Wherein, is the cumulative distribution function of the original reanalysis data, is the inverse cumulative distribution function of the fused data true value; S42. The corrected grid data set is generated and directly used to construct high-resolution temperature and precipitation sequences.
[0014] Compared with the prior art, the present application has the beneficial effects that: the present application fuses the original observation data into the reanalysis data of the same period, and on this basis, the reanalysis data set is subjected to multivariate bias correction, so as to give full play to the advantages of observation data and reanalysis data, and solve the problem that in complex mountainous terrain, large elevation difference, and the spatial distribution of meteorological elements is comprehensively affected by multiple factors such as terrain and altitude, resulting in low accuracy of the interpolation result. Secondly, it also solves the problem that the traditional mountainous area is short of stations and unevenly distributed, and the existing interpolation method is prone to large errors when processing sparse data, especially in high-altitude unpopulated areas, lacking sufficient adjacent station data as support, so the present scheme considers the original observation data, and at the same time, the reanalysis data set is subjected to multivariate bias correction, so as to give full play to the advantages of observation data and reanalysis data. BRIEF DESCRIPTION OF DRAWINGS
[0015] Figure 1 (a) original station, (b) original ERA5_land and (c) fused ERA5_obs spatial distribution map; Figure 2 (a) original ERA5_land and (b) corrected product spatial distribution map; Figure 3 (a) original observation, (b) interpolated original ERA5_land, (c) interpolated corrected product, (d) original ERA5_land and (e) corrected product spatial distribution map; Figure 4 (a) observation and original ERA5_land, (b) observation and corrected product RMSE spatial distribution map; Figure 5The present application is illustrated on the example of 2m temperature on January 1, 2022, (a) raw observations, (b) raw ERA5_land after interpolation, (c) spatial distribution of the corrected product after interpolation. DETAILED DESCRIPTION
[0016] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of the present application.
[0017] Embodiments, please refer to Figures 1 to 5 The present application provides a method for constructing temperature and precipitation data without station observation data, comprising the following steps: S1. obtaining and preprocessing observation data and reanalysis data set; S2. using a weighted average method based on influence radius and distance, fusing the observation data into the reanalysis data of the same period to generate the fused grid data; S3. performing statistical downscaling and true value setting, taking the fused grid data as the true value reference for driving the statistical downscaling process; S4. applying the deviation correction method of N-dimensional probability density function transformation to match the multivariate distribution to correct the reanalysis data and generate the final data sequence.
[0018] In this embodiment, it is also necessary to explain that the implementation process of step S1 of obtaining and preprocessing observation data is to collect the measured meteorological data of the target area, including temperature and precipitation, and data cleaning and formatting are performed to process missing values and ensure time consistency.
[0019] In this embodiment, it is also necessary to explain that the implementation process of step S1 of reanalysis data is to use a high-resolution reanalysis data set as a basic framework, and the document uses the ERA5_Land data set to provide full-coverage gridded meteorological elements.
[0020] In this embodiment, it is also necessary to explain that the implementation process of step S2 is: S21. For each reanalysis grid point, calculate its geographical distance and influence radius from the adjacent observation stations. Through a weighted average algorithm, the weight is based on the reciprocal of the distance or a similar function, and the observation value is fused into the reanalysis data, see equation (1): (1); In the formula, wherein, is the fused grid value, is the observation value of the i th station, is the original ERA5_Land value, is the weight, inversely proportional to the distance, and n is the number of stations within the influence radius; S22. Generate "fused simultaneous period grid data" as the "true value" reference for subsequent steps.
[0021] In this embodiment, it should also be noted that step S3 downscaling operation utilizes the high resolution characteristics of the fused data to spatially refine the reanalysis data, and through a regression model or probability distribution matching, the coarse resolution reanalysis data is down scaled to the target grid.
[0022] In this embodiment, it should also be noted that step S3 true value is used as a reference to calibrate the accuracy of subsequent bias correction, integrating the local truth of observations and the global coverage of reanalysis, to solve the error caused by sparse stations.
[0023] In this embodiment, it should also be noted that the implementation process of step S4 is: S41. Use the MBCn method to correct the system bias of multiple variables, and use the N-pdf transformation to match the joint probability distribution of the fused data true value and the original reanalysis data, adjust the value range and spatial structure of the reanalysis data, see equation (2): (2); wherein, is the cumulative distribution function of the original reanalysis data, is the inverse cumulative distribution function of the fused data true value; S42. Generate the corrected grid data set, which is directly used to construct high resolution temperature and precipitation sequences.
[0024] Embodiment 1, in practical application, a temperature and precipitation data construction method without station observation data, comprising the following steps: Embodiment, please refer to Figures 1 to 5 , the method is aimed at the problem of missing meteorological data in high-altitude data-free areas in mountainous areas (such as the upper reaches of a watershed), and is designed to construct high-resolution temperature and precipitation data sequences; The terrain in mountainous areas is complex, and stations are few and unevenly distributed, so traditional spatial interpolation methods (such as Kriging interpolation) are difficult to accurately process non-stationary meteorological element distribution; The method fuses observation data and reanalysis data, and combines multivariate bias correction to overcome the shortcomings of the prior art, generate reliable high-resolution grid data sets, and is suitable for runoff research under the background of climate change and other applications, the present application proposes a temperature and precipitation data construction method without station observation data, comprising the following steps: (1) High-altitude data-free area data fusion method The distribution of observation stations in high-altitude data-scarce areas is uneven, and the observation data are seriously missing. Therefore, based on the weighted average method of influence radius and distance, the existing observation data are fused into the reanalysis data of the same period, and the fused grid data of the same period are taken as the true value for further statistical downscaling. Due to the dispersion of stations, the multi-variable bias correction MBCn that matches the multi-variable distribution by N-pdft is used to correct the bias of reanalysis data. (2) Test data ERA5_Land reanalysis dataset, spatial resolution is 0.1*0.1, time period is 1991-2023 daily temperature and precipitation data; Observation data is daily temperature data from September 2021 to August 2024 (with more missing data); Figure 1 The spatial distribution of original stations, original ERA5_land and fused ERA5_obs; Figure 2 Take the average 2m temperature in January 2021 as an example, the spatial distribution of original ERA5_land and corrected fused product; It can be seen that after correction, the fused product considers the station data at the same time, effectively reducing the error; (3) Dataset evaluation: Figure 3 Take the average 2m temperature in January 2021 as an example, the spatial distribution of original observation, interpolated original ERA5_land and interpolated corrected product; Among them, in order to evaluate and compare, Figure 3 (b-c) data are uniformly interpolated to stations; From Figure 3 (a, b) can be seen, the temperature of original ERA5 data is lower than the observation, and the corrected fused product is basically consistent with the observation ( Figure 3 c); Among them, the RMSE of original ERA5 data and observation is 7.09, and that of corrected fused product is 3.28; At the same time, Figure 3 (d, e) gives the spatial distribution of original ERA5_land and corrected product; It can be seen that the temperature distribution of the corrected fused product is more uniform; Figure 4 The spatial distribution of RMSE between observation and original ERA5_land, observation and corrected product at each station is given; Compared with the original ERA5 land data, the RMSE of the revised fusion product is significantly reduced. It is shown that the fusion + revision scheme can effectively improve the simulation capability of the gridded dataset to the observation data. As Figure 5 It should be pointed out that the daily ERA5 land and the revised product after interpolation cannot well simulate the observed temperature. Therefore, the monthly average temperature is used for evaluation.
[0025] Through the above steps, the original observation data is fused into the reanalysis data at the same period, and on this basis, the multi-variable bias correction is performed on the reanalysis dataset, so as to give full play to the advantages of observation data and reanalysis data, and solve the problem that in the complex terrain of mountainous area, the large elevation difference, and the spatial distribution of meteorological elements is comprehensively affected by many factors such as terrain and altitude, resulting in low precision of the interpolation result. Secondly, it also solves the problem that the traditional mountainous area is short of stations and the distribution is uneven, and the existing interpolation method is prone to large error when processing sparse data, especially in the high-altitude unmanned area, and there is a lack of sufficient adjacent station data as support, therefore, the original observation data is considered in the scheme, and the multi-variable bias correction is performed on the reanalysis dataset, so as to give full play to the advantages of observation data and reanalysis data.
[0026] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.
Claims
1. A method for constructing temperature and precipitation data without station observation data, characterized in that, Includes the following steps: S1. Acquire and preprocess observation data and reanalysis datasets; S2. Using a weighted average method based on influence radius and distance, the observed data are fused into the reanalysis data of the same period to generate fused grid data; S3. Perform statistical downscaling and set the truth value, using the fused grid data as the truth reference to drive the statistical downscaling process; S4. Apply the bias correction method of N-dimensional probability density function transformation to match multivariate distribution, correct the reanalysis data, and generate the final data sequence.
2. The method for constructing temperature and precipitation data without station observation data according to claim 1, characterized in that: The implementation process of step S1, acquiring and preprocessing observation data, is as follows: collect measured meteorological data of the target area, including temperature and precipitation, and clean and format the data to process missing values and ensure time consistency.
3. The method for constructing temperature and precipitation data without station observation data according to claim 2, characterized in that: The implementation process of the reanalysis data in step S1 is as follows: a high-resolution reanalysis dataset is used as the basic framework, and the ERA5_Land dataset is used to provide full coverage of gridded meteorological elements.
4. The method for constructing temperature and precipitation data without station observation data according to claim 3, characterized in that, The implementation process of step S2 is as follows: S21. For each reanalysis grid point, calculate its geographical distance and influence radius to neighboring observation stations. Using a weighted average algorithm, with weights based on the reciprocal of the distance or a similar function, the observations are integrated into the reanalysis data, as shown in Equation (1): （1）； In the formula, where, These are the merged grid point values. It is the observation value of the i-th station. It is the original ERA5_Land value. It is the weight, which is inversely proportional to the distance, and n is the number of stations within the influence radius; S22. Generate "fused grid data of the same period" as a "truth value" reference for subsequent steps.
5. The method for constructing temperature and precipitation data without station observation data according to claim 4, characterized in that: The downscaling operation in step S3 utilizes the high-resolution characteristics of the fused data to spatially refine the reanalysis data. Through regression models or probability distribution matching, the coarse-resolution reanalysis data is downscaled to the target grid.
6. The method for constructing temperature and precipitation data without station observation data according to claim 5, characterized in that: The purpose of the truth value in step S3 is to: use the fused data as a benchmark to calibrate the accuracy of subsequent bias corrections, integrate the local authenticity of observations with the global coverage of reanalysis, and address errors caused by sparse stations.
7. The method for constructing temperature and precipitation data without station observation data according to claim 6, characterized in that, The implementation process of step S4 is as follows: S41. Using the MBCn method to correct multivariate systematic bias, the N-pdf transformation adjusts the range and spatial structure of the reanalysis data by matching the joint probability distribution of the true values of the fused data with the original reanalysis data, as shown in equation (2): （2）； in, It is the cumulative distribution function of the original reanalysis data. It is the inverse cumulative distribution function of the true values of the fused data; S42. Generate the corrected grid dataset, which can be directly used to construct high-resolution temperature and precipitation sequences.