A window self-adjusting satellite precipitation data correction method

Abstract

The application discloses a satellite precipitation data correction method with a self-adjusting window, comprising the following steps: S1, using satellite grid data and ground observation station measured precipitation data, calculating the ground observation station density in each grid of the satellite grid data; S2, combining the ground observation station density in each grid, determining the optimal correction window width corresponding to the distribution state of each ground observation station; S3, setting an initial window based on the optimal correction window width, moving the initial window to successively correct the satellite grid data in each window, obtaining the function relationship between the satellite grid data and the ground observation station measured precipitation data, and inputting the satellite grid data into the function relationship to obtain the corrected satellite grid data. The method can consider the space-time distribution characteristics of the satellite precipitation data, take the ground observation station measured precipitation data as a reference, locally correct the grid data, reduce the correction error, and make the correction result have relatively high precision.

Description

Technical Field

[0001] This invention relates to the field of meteorological and hydrological satellite precipitation data correction technology, and in particular to a method for correcting satellite precipitation data with self-adjusting windows. Background Technology

[0002] The existing technology is a window-sliding GPM data correction method that considers spatial distribution. This method can take into account the spatiotemporal distribution characteristics of satellite precipitation data, using measured precipitation data from ground observation stations as a reference to perform local corrections on the satellite raster data, reducing correction errors and resulting in relatively high accuracy. However, it also has the following limitations:

[0003] (1) Determining the correction window. The existing window setting method is relatively simple and fails to fully consider the distribution characteristics of ground observation stations in the watershed. It is difficult to apply to different research areas. In particular, for watersheds with complex shapes, such as long and narrow watersheds with uneven distribution and inconsistent distribution characteristics of upstream and downstream ground observation stations, the data correction effect produced by this technology will not be obvious.

[0004] (2) Selection of Correction Algorithm. Existing techniques calculate the ratio between the arithmetic mean of measured precipitation data from ground observation stations within the initial window and the arithmetic mean of the included satellite raster data, using this ratio directly as a correction coefficient for data correction. However, this approach presents the following problems when applied to different study areas: Before applying the correction algorithm, it is necessary to calculate the arithmetic mean of both the satellite raster data and the measured ground data within the window, which introduces some error. Therefore, the selection of the correction algorithm is crucial for clarifying the correlation between the two types of data. However, existing techniques only calculate the ratio between satellite precipitation data and measured ground values, which is insufficient to meet the requirements of spatial scale. The correlation between satellite precipitation data and measured ground values ​​should be complex and diverse; existing techniques simplify this relationship to a ratio, which introduces some error into data correction.

[0005] In summary, how to intelligently determine the correction window to be applicable to different research areas, and how to efficiently analyze the correlation between satellite precipitation data and ground-based measurements to obtain a more applicable correction algorithm are crucial issues. Solving these problems and optimizing existing correction technologies are therefore of paramount importance. Thus, this invention addresses these challenges by proposing a more applicable technology to better correct satellite precipitation products, thereby obtaining more accurate precipitation data. Summary of the Invention

[0006] The purpose of this invention is to provide a self-adjusting satellite precipitation data correction method to solve the aforementioned problems in the prior art.

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

[0008] A method for correcting satellite precipitation data with window self-adjustment includes the following steps:

[0009] S1. Calculate the density of ground observation stations within each grid of the satellite grid data using satellite grid data and measured precipitation data from ground observation stations.

[0010] S2. Based on the density of ground observation stations in each grid, determine the optimal correction window width corresponding to the distribution of ground observation stations in each grid.

[0011] S3. Set the initial window based on the optimal correction window width, move the initial window to correct the satellite raster data in each window in turn, obtain the functional relationship between the satellite raster data and the measured precipitation data of the ground observation station, and input the satellite raster data into the functional relationship to obtain the corrected satellite raster data.

[0012] Preferably, step S1 specifically includes the following:

[0013] S11. Determine the relative positional relationship between the satellite grid data and the latitude and longitude coordinates of the measured data from the ground observation stations, i.e., the ground observation stations are scattered within the satellite grid.

[0014] S12. Calculate the density of ground observation stations within each grid cell of the satellite precipitation product; the calculation formula is as follows:

[0015]

[0016] Where N is the total number of ground observation stations included in the study area; m i γ represents the number of ground observation stations distributed within the i-th grid cell; i Let be the density of ground observation stations within the i-th grid.

[0017] Preferably, step S2 specifically includes the following:

[0018] S21. Calculate the density of ground observation stations within each satellite grid;

[0019] S22. For a specific moment without considering the time dimension of the satellite three-dimensional matrix, calculate the density of ground observation stations in each grid, and determine the optimal correction window width corresponding to the distribution of ground observation stations in each region by referring to the empirical discrimination formula.

[0020] Preferably, the empirical discrimination formula in step S22 is:

[0021]

[0022] Where ρ is the density of ground observation stations in the grid; k is the optimal correction window width; and a0 and b0 are the specific values ​​corresponding to the optimal correction window width.

[0023] Preferably, step S3 specifically includes the following:

[0024] S31. Calculate the arithmetic mean of the satellite raster data and the arithmetic mean of the precipitation data of the ground observation stations included in the initial window, respectively. Calculate a univariate linear function of the arithmetic mean of the two types of data included in the initial window. Input the satellite raster data in the initial window into the univariate linear function to obtain the correction result of the satellite raster data in the initial window.

[0025] S32. Move the initial window in steps of 1 grid unit, and correct the satellite raster data in each window in sequence according to the method in S31, so as to obtain n sets of univariate linear function relationships.

[0026] S33. Numerical solutions for the slope and intercept of the univariate linear function are obtained through numerical methods. For a single grid with multiple values ​​of a and b, their arithmetic mean is taken to obtain the univariate linear function relationship between the satellite raster data and the measured precipitation data from the ground observation station. The corrected satellite raster dataset is then obtained by substituting the satellite raster data of each grid into the univariate linear function relationship.

[0027] Preferably, step S0 precedes step S1, and step S0 includes:

[0028] Preprocessing of satellite data: Read the raw data file of micro-precipitation products, process it into a three-dimensional matrix, and process the latitude and longitude coordinates corresponding to each grid of the satellite raster data into matrix form respectively;

[0029] Preprocessing of ground data: The measured precipitation data of various ground observation stations at a certain time scale and the corresponding latitude and longitude coordinates of the ground observation stations are processed into matrix form.

[0030] Preferably, after step S3, there is a step S4. Specifically, step S4 involves using the measured precipitation data from ground observation stations as a reference, and evaluating the corrected satellite raster data using three evaluation indicators: correlation coefficient, root mean square error, and mean absolute error. The closer the correlation coefficient is to 1, the smaller the root mean square error, and the smaller the mean absolute error, the better the correction effect of the satellite raster data.

[0031] Preferably, the evaluation of the correction results in step S4 includes assessments on an hourly, daily, monthly, quarterly, and annual scale.

[0032] Preferably, in order to compare the accuracy of the corrected satellite raster data with that of the uncorrected satellite raster data, the relevant index data of the measured precipitation data of the ground observation station and the uncorrected satellite raster data are first calculated using the above three evaluation indicators. Then, the relevant index data of the measured precipitation data of the ground observation station and the corrected satellite raster data are calculated using the above three evaluation indicators. The calculation results are compared to determine the accuracy of the corrected satellite raster data with that of the uncorrected satellite raster data.

[0033] The beneficial effects of this invention are as follows: 1. The method of this invention can consider the spatiotemporal distribution characteristics of satellite precipitation data, use measured precipitation data from ground observation stations as a reference, and perform local correction on raster data to reduce correction errors and achieve relatively high accuracy in the correction results. 2. The method of this invention can fully consider the geographical and topographical conditions and the distribution characteristics of measured stations in different regions, and perform relatively targeted corrections for different regions, thereby effectively solving the problems of uneven distribution of ground observation stations and lack of precipitation data in some areas to a certain extent. 3. The method of this invention can study the correlation between satellite precipitation products and measured precipitation data within a small area in depth, thereby improving the effectiveness of data correction to a certain extent and being applicable to correction at multiple spatiotemporal scales. 4. The method of this invention improves the verification approach, enabling more comprehensive verification of the corrected dataset and possessing high practical value. In summary, this invention obtains a set of precipitation datasets with high accuracy and strong applicability within the watershed through correction, thereby better serving local precipitation forecasting, hydrological simulation, water resource management, and other work. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating the correction method in an embodiment of the present invention;

[0035] Figure 2 This is a schematic diagram of the data preprocessing process in an embodiment of the present invention;

[0036] Figure 3 This is a schematic diagram of the site density calculation process in an embodiment of the present invention;

[0037] Figure 4 This is a flowchart illustrating the process of determining the optimal correction window width in an embodiment of the present invention;

[0038] Figure 5 This is a schematic diagram of the window scrolling correction of satellite raster data in an embodiment of the present invention;

[0039] Figure 6 This is a schematic diagram of the correction effect evaluation process in an embodiment of the present invention;

[0040] Figure 7This is a schematic diagram of the spatial location of ground measurement stations in the Lancang River Basin in an embodiment of the present invention;

[0041] Figure 8 This is a schematic diagram of the corrected site spatial location in an embodiment of the present invention;

[0042] Figure 9 This is a schematic diagram of the spatial distribution of verification sites in an embodiment of the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0044] Example 1

[0045] This embodiment provides a method for correcting satellite precipitation data with a self-adjusting window, including the following steps:

[0046] S1. Calculate the density of ground observation stations within each grid of the satellite grid data using satellite grid data and measured precipitation data from ground observation stations.

[0047] S2. Based on the density of ground observation stations in each grid, determine the optimal correction window width corresponding to the distribution of ground observation stations in each grid.

[0048] S3. Set the initial window based on the optimal correction window width, move the initial window to correct the satellite raster data in each window in turn, obtain the functional relationship between the satellite raster data and the measured precipitation data of the ground observation station, and input the satellite raster data into the functional relationship to obtain the corrected satellite raster data.

[0049] Combination Figure 1 The method flow in the document provides a detailed description of the method of this invention:

[0050] I. Data Preprocessing

[0051] Before step S1, there is step S0, which involves preprocessing the satellite and ground data, as detailed below.

[0052] Preprocessing of satellite data: Read the raw data file of micro-precipitation products, process it into a three-dimensional matrix, and process the latitude and longitude coordinates corresponding to each grid of the satellite raster data into matrix form respectively;

[0053] Preprocessing of ground data: The measured precipitation data of various ground observation stations at a certain time scale and the corresponding latitude and longitude coordinates of the ground observation stations are processed into matrix form.

[0054] II. Calculate the density of ground observation stations within the grid.

[0055] This section corresponds to step S1, and specifically includes the following:

[0056] 1. Determine the relative positional relationship between satellite grid data and ground observation station measured data based on their latitude and longitude coordinates, i.e., the ground observation stations are scattered within the satellite grid;

[0057] 2. Calculate the density of ground observation stations contained within each grid cell (the side length of the rectangle represents the spatial resolution, e.g., 0.1° × 0.1°) of the satellite precipitation product; the calculation formula is as follows:

[0058]

[0059] Where N is the total number of ground observation stations included in the study area; m i γ represents the number of ground observation stations distributed within the i-th grid cell; i Let be the density of ground observation stations within the i-th grid.

[0060] III. Determining the Optimal Correction Window Width

[0061] This part corresponds to step S2, and specifically includes the following:

[0062] S21. Calculate the density of ground observation stations within each satellite grid;

[0063] S22. For a specific time (i.e., without considering the time dimension of the satellite three-dimensional matrix), calculate the density of ground observation stations in each grid, and determine the optimal correction window width corresponding to the distribution of ground observation stations in each region by referring to the empirical discrimination formula.

[0064] The aforementioned empirical discriminant formula refers to the optimal correction window width corresponding to different distributions of ground observation stations, based on machine learning methods. This involves calculating the density range of observation stations within each grid within a region (considering only the density range greater than 50%), and recording the optimal window width under different scenarios. For example, if the station density is ρ and the optimal window width is k, then through multiple machine learning iterations, the empirical discriminant formula becomes:

[0065]

[0066] Where ρ is the density of ground observation stations in the grid; k is the optimal correction window width; and a0 and b0 are the specific values ​​corresponding to the optimal correction window width.

[0067] IV. Corrections will be carried out window by window.

[0068] This section corresponds to step S3, and specifically includes the following:

[0069] S31. Calculate the arithmetic mean of the satellite raster data and the arithmetic mean of the precipitation data from the included ground observation stations within the initial window. Calculate a univariate linear function to represent the arithmetic means of the two types of data included in the initial window. Input the satellite raster data within the initial window into this univariate linear function to obtain the correction result for the satellite raster data within the initial window. The calculation formula is as follows:

[0070] Y0 = a·X0 + b

[0071] Where X0 is the arithmetic mean of the satellite raster data within the initial window; Y0 is the arithmetic mean of the precipitation data from the ground observation stations included in the initial window; a and b are constant parameters of the univariate function;

[0072] S32. Move the initial window in steps of 1 grid unit, and correct the satellite raster data in each window sequentially according to the method in S31, thus obtaining n sets of univariate linear function relationships; that is...

[0073] Y j =a·X j +b

[0074] Among them, X j Y is the arithmetic mean of the satellite raster data within the j-th window; j Let j be the arithmetic mean of precipitation data from ground observation stations contained in the j-th window; j = 1, 2, ..., n, where n is the total number of windows.

[0075] S33. After correction, numerical solutions for the slope and intercept of the univariate linear function are obtained through numerical solution methods (such as finite element method, numerical approximation, and interpolation). For a single grid with multiple values ​​of a and b, their arithmetic mean is taken respectively to obtain the univariate linear function relationship between satellite raster data and measured precipitation data from ground observation stations. Then, the satellite raster data of each grid is substituted into this univariate linear function relationship to obtain the corrected satellite raster dataset.

[0076] Y = a·X + b

[0077] Where X represents the satellite raster data for each grid; Y represents the corrected satellite raster data.

[0078] V. Evaluation and Correction Results

[0079] After step S3, there is also step S4. This part corresponds to step S4. Specifically, taking the measured precipitation data of ground observation stations as a reference, the corrected satellite grid data is evaluated using three evaluation indicators: correlation coefficient, root mean square error, and mean absolute error. The closer the correlation coefficient is to 1, the smaller the root mean square error, and the smaller the mean absolute error, the better the correction effect of the satellite grid data.

[0080] 1. Pearson correlation coefficient (CC)

[0081] The Pearson correlation coefficient reflects the strength of the linear relationship between the satellite grid data and the precipitation data of ground observation stations. The absolute value of its range is from 0 to 1. The closer it is to 1, the more consistent the information between the satellite precipitation data and the ground observation precipitation data, and the higher the reference value.

[0082] Generally, 0.8 < CC ≤ 1.0 means very strong correlation; 0.6 < CC ≤ 0.8 means strong correlation; 0.4 < CC ≤ 0.6 means medium correlation; 0.2 < CC ≤ 0.4 means weak correlation; 0.0 ≤ CC ≤ 0.2 indicates extremely weak or no correlation; CC ≤ 0.0 indicates negative correlation.

[0083]

[0084] 2. Root mean square error (RMSE)

[0085] The root mean square error is used to evaluate the deviation degree between the satellite grid data and the precipitation data of ground observation stations. Its value is always non - negative. The smaller the value, the smaller the observation error, and vice versa, the larger the error.

[0086]

[0087] 3. Mean absolute error (MAE)

[0088] The mean absolute error (MAE) is often used to describe the difference between the satellite grid data and the precipitation data of ground observation stations, and measure the size of the average error. The mean absolute error can avoid the problem of error cancellation, and thus can accurately reflect the size of the actual error.

[0089]

[0090] In the calculation formulas of the above three evaluation indicators, n represents the number of data pairs used in the accuracy evaluation; X i represents the measured precipitation data of the i - th ground observation station, and Y i represents the pixel value of the satellite precipitation data grid where the location of this ground observation station is located; is the mean value of X i , is the mean value of Y iThe mean.

[0091] In this embodiment, the evaluation of correction results includes assessments at the hourly, daily, monthly, quarterly, and annual scales.

[0092] The corrected satellite raster data that meets the evaluation requirements will be output as a dataset. The evaluation requirements can be set according to specific circumstances to better meet actual needs.

[0093] In this embodiment, in order to compare the accuracy of the corrected satellite raster data with that of the uncorrected satellite raster data, the relevant index data of the measured precipitation data of the ground observation station and the uncorrected satellite raster data are first calculated using the above three evaluation indicators. Then, the relevant index data of the measured precipitation data of the ground observation station and the corrected satellite raster data are calculated using the above three evaluation indicators. The calculation results are compared to determine the accuracy of the corrected satellite raster data with that of the uncorrected satellite raster data.

[0094] Example 2

[0095] In this embodiment, the Lancang River Basin is used as an example to specifically illustrate the advantages of the method of the present invention. A total of 176 ground observation stations are used in the Lancang River Basin, and their spatial distribution is as follows: Figure 7 As shown. To better demonstrate the correction effect of the method of this invention, the traditional verification approach is improved by dividing the 176 ground observation stations in the Lancang River Basin into a "correction group" and a "verification group," each with 88 stations (spatial distribution as shown in the figure). Figure 8 , 9 (As shown). The measured precipitation data from ground observation stations in the "correction group" are used to correct satellite precipitation data, while the measured precipitation data from ground observation stations in the "validation group" are used to evaluate the correction effect of the sliding window method. Through empirical discriminant analysis, the optimal correction window width for this watershed is determined to be 10 grid units (each grid unit represents the spatial resolution of the satellite precipitation product; in this invention, it is 0.1°).

[0096] This embodiment calculates the correlation coefficients of 88 verification sites before and after correction using both the existing method (a window sliding GPM data correction method considering spatial distribution) and the method of this invention, as shown in Tables 1 and 2, respectively.

[0097] Table 1. Statistical results of CC index before and after correction using the existing method.

[0098]

[0099]

[0100]

[0101] Table 2. Statistical results of CC index before and after technical correction of this invention.

[0102]

[0103]

[0104]

[0105] As can be seen from Tables 1 and 2, the method of the present invention is significantly improved compared with the existing method (a window sliding GPM data correction method considering spatial distribution). It significantly improves the correlation coefficient between satellite precipitation products and ground-measured data, has a good correction effect in the Lancang River Basin, and can obtain a set of high-quality precipitation datasets, thereby better serving the local precipitation forecasting and water resource management fields.

[0106] By adopting the above-disclosed technical solution of this invention, the following beneficial effects are obtained:

[0107] This invention provides a self-adjusting window-based satellite precipitation data correction method. This method considers the spatiotemporal distribution characteristics of satellite precipitation data, uses measured precipitation data from ground observation stations as a reference, and performs local correction on the raster data, reducing correction errors and resulting in relatively high accuracy. This method fully considers the geographical and topographical conditions and the distribution characteristics of measured stations in different regions, providing targeted corrections for different areas, thus effectively addressing the problems of uneven distribution of ground observation stations and lack of precipitation data in some areas. This method can deeply study the correlation between satellite precipitation products and measured precipitation data within a small area, thereby improving the effectiveness of data correction and being applicable to corrections at multiple spatiotemporal scales. This method improves the verification approach, enabling more comprehensive verification of the corrected dataset, and has high practical value. In summary, this invention provides a high-accuracy precipitation dataset with strong applicability within a watershed, better serving local precipitation forecasting, hydrological simulation, and water resource management.

[0108] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A window self-adjusting satellite precipitation data correction method, characterized in that: The method comprises the following steps, S1, calculating the density of ground observation stations in each grid of satellite grid data by using satellite grid data and measured precipitation data of ground observation stations; S2, determining the optimal correction window width corresponding to the distribution of each ground observation station in combination with the density of ground observation stations in each grid; S3, setting an initial window based on the optimal correction window width, moving the initial window to sequentially correct the satellite grid data in each window, obtaining the functional relationship between the satellite grid data and the measured precipitation data of ground observation stations, and inputting the satellite grid data into the functional relationship to obtain the corrected satellite grid data; Step S3 specifically includes the following contents, S31, calculating the arithmetic mean of the satellite grid data in the initial window and the arithmetic mean of the precipitation data of the ground observation stations contained therein, respectively, calculating a linear function of one variable of the arithmetic means of the above two types of data contained in the initial window, inputting the satellite grid data in the initial window into the linear function of one variable, and obtaining the correction result of the satellite grid data in the initial window; S32, moving the initial window by 1 grid unit, and sequentially correcting the satellite grid data in each window in the manner of S31, so as to obtain n sets of linear functions of one variable; S33, obtaining the numerical solution of the slope and intercept in the linear function of one variable by a numerical solution method; for multiple a and b values in a single grid, the arithmetic mean of the a and b values is taken respectively, and finally the linear function of one variable of the satellite grid data and the measured precipitation data of the ground observation stations is obtained; then, the satellite grid data of each grid is substituted into the linear function of one variable, so as to obtain the corrected satellite grid data set.

2. The window self-adjusting satellite precipitation data correction method of claim 1, wherein: Step S1 specifically includes the following contents, S11, determining the relative position relationship between the satellite grid data and the measured data of the ground observation stations according to the longitude and latitude coordinates of the satellite grid data and the measured data of the ground observation stations, i.e. the ground observation stations are dispersed in the satellite grid; S12, calculating the density of the ground observation stations contained in each grid unit of the satellite precipitation product; The calculation formula is, ； wherein, is the total number of ground observation sites included in the study area; is the number of ground observation sites distributed in the grid cell; is the density of ground observation sites in the grid.

3. The window self-adjusting satellite precipitation data correction method of claim 1, wherein: Step S2 specifically includes the following contents, S21, calculating the density of the ground observation stations in all satellite grids; S22, for a specific time not considering the time dimension of the three-dimensional matrix of the satellite, the density of the ground observation stations in each grid is counted, the optimal correction window width corresponding to the distribution of each ground observation station is determined by referring to an empirical discrimination formula.

4. The window self-adjusting satellite precipitation data correction method according to any one of claims 1 to 3, characterized in that: Before step S1, there is also step S0, which includes, Preprocessing of satellite data: reading the original data file of the micro precipitation product, processing it into a three-dimensional matrix form, and processing the longitude and latitude coordinates corresponding to each grid of the satellite grid data into a matrix form respectively; Preprocessing of ground data: processing the measured precipitation data of each ground observation station under a certain time scale and the longitude and latitude coordinates corresponding to the ground observation stations into a matrix form respectively.

5. The window self-adjusting satellite precipitation data correction method according to any one of claims 1 to 3, characterized in that: After step S3, there is also step S4, which specifically is, taking the measured precipitation data of the ground observation stations as a reference, respectively adopting three evaluation indexes of correlation coefficient, root mean square error and mean absolute error to evaluate the corrected satellite grid data; The closer the correlation coefficient is to 1, the smaller the root mean square error is, and the smaller the mean absolute error is, the better the correction effect of the satellite grid data is.

6. The window self-adjusting satellite precipitation data correction method of claim 5, wherein: The correction result evaluation in step S4 includes evaluation of the hourly scale, the daily scale, the monthly scale, the quarterly scale and the annual scale.

7. The window self-adjusting satellite precipitation data correction method of claim 5, wherein: In order to compare the accuracy of the corrected satellite grid data and the uncorrected satellite grid data, the correlation index data of the measured precipitation data of the ground observation station and the uncorrected satellite grid data is calculated by using the three evaluation indexes, the correlation index data of the measured precipitation data of the ground observation station and the corrected satellite grid data is calculated by using the three evaluation indexes, the calculation results are compared, and then the accuracy of the corrected satellite grid data and the accuracy of the uncorrected satellite grid data are determined.

Images