A method for spatiotemporal matching of radar observations with ground station precipitation data

Abstract

This invention discloses a spatiotemporal matching method for radar observations and ground station precipitation data, relating to the field of meteorological data processing technology. The method includes: acquiring a target ground precipitation station; determining a first data sampling area within a narrow radar observation ring containing the station, and performing gridding and normalization processing; calculating the spatial correlation coefficient between the gridded radar data at the initial time and the gridded precipitation data at the initial time and multiple subsequent comparison times, using the maximum coefficient to determine the optimal time for temporal matching; determining a secondary sampling area within the corresponding first data sampling area, calculating the temporal correlation coefficient between each spatial location within this area and the target station data, using the maximum coefficient to determine the optimal spatial matching location, and finally combining the results to obtain the optimal spatiotemporal matching data. This invention achieves point-to-point fine-grained matching of the two types of data, avoiding extreme value weakening caused by spatiotemporal averaging operations, and preserving the characteristics of heavy precipitation radar and automatic ground station observation data with high precision.

Description

Technical Field

[0001] This invention relates to the field of quantitative precipitation estimation technology using meteorological radar, and more specifically, to a method for spatiotemporal matching of radar observations and precipitation data from ground stations. Background Technology

[0002] Matching radar and ground station precipitation data integrates meteorological and computer science techniques. Weather radar, with its large observation coverage and high spatiotemporal resolution, has irreplaceable advantages in quantitative precipitation estimation. Traditional quantitative precipitation estimation models often establish a correlation between radar observations and ground rainfall intensity, and in recent years, many artificial intelligence models have been incorporated to discover this mapping relationship. However, correctly matching radar observation data with corresponding ground station precipitation observation data at both the temporal and spatial levels is the most crucial prerequisite for establishing this mapping relationship.

[0003] In real-world scenarios, precipitation takes time to fall from the air to the ground, and due to factors such as wind fields and the movement of the precipitation system itself, radar observations and ground precipitation data exhibit varying degrees of temporal and spatial misalignment. Current methods to mitigate these data misalignments mainly include: first, averaging radar or precipitation data over time while ensuring the closest possible spatial location; second, averaging data across multiple spatial points while ensuring the closest possible time; and third, determining the temporal correlation by calculating the temporal lag correlation between the two types of data. However, existing methods reveal several insurmountable shortcomings in practical applications:

[0004] First, the averaging method (time averaging or spatial averaging) can easily cause the two types of data to deviate from the true situation in terms of numerical value. Especially at some extreme heavy precipitation points, the spatiotemporal averaging operation will artificially weaken the maximum value of strong radar echoes and ground precipitation, significantly reducing the estimation accuracy under heavy precipitation conditions.

[0005] Second, radar observations often have an elevation angle, which, combined with the curvature of the Earth, leads to huge differences in the height of precipitation particles observed at different distances within the radar coverage area, and the time it takes for them to fall to the ground will inevitably be different. In addition, the uneven wind field makes it difficult to predict the degree of lateral drift of precipitation particles. The existing fixed time or spatial matching algorithms (such as the nearest distance principle) are simply unable to meet the requirement of accurate matching of all locations within the full coverage area of ​​the radar.

[0006] Third, existing time-series lag correlation methods require constant trial and error to determine different lag times, which is cumbersome and still fails to completely solve the spatial misalignment problem caused by wind field drift. Summary of the Invention

[0007] The present invention provides a spatiotemporal matching method for radar observations and ground station precipitation data, which can solve the technical problems in existing radar and ground precipitation data matching algorithms, such as the weakening of radar data maxima by using spatiotemporal averaging, the cumbersome trial time lag, and the inability of fixed matching algorithms to guarantee the global optimal matching due to radar observation elevation angle and uneven wind field drift.

[0008] To solve the above problems, the technical solution adopted by the present invention is as follows:

[0009] A spatiotemporal matching method for radar observations and ground station precipitation data includes:

[0010] Acquire several ground precipitation observation stations within the radar coverage area and identify them as target stations in sequence;

[0011] The radar observation range is divided into multiple narrow ring regions centered on the radar station location. A narrow ring region containing the target station is selected, and a first data sampling area centered on the target station is determined within the selected narrow ring region.

[0012] The first data sampling area is divided into multiple grids at equal intervals, and the radar observation data and ground precipitation observation data within the first data sampling area are normalized.

[0013] The normalized radar observation data and ground precipitation observation data are allocated to corresponding grids to obtain gridded radar data and gridded precipitation data, respectively.

[0014] A radar observation data time is determined as the initial time, and the initial time and multiple precipitation observation times after the initial time are used as comparison times; the spatial correlation coefficient between the gridded radar data of the initial time and the gridded precipitation data of each comparison time is calculated, and the comparison time corresponding to the largest spatial correlation coefficient is recorded as the optimal time matching time;

[0015] Within the first data sampling area corresponding to the optimal time matching time, a secondary sampling area centered on the target station is determined;

[0016] A first time period is obtained by extending a preset time span before and after the initial time period; a second time period is obtained by extending the preset time span before and after the optimal time matching time period.

[0017] Calculate the temporal correlation coefficient between radar observation data of any spatial location within the secondary sampling area in the first time period and ground precipitation observation data of the target station in the second time period, and record the spatial location corresponding to the largest temporal correlation coefficient as the best spatial matching location;

[0018] The ground precipitation observation data of the target station at the optimal time-matching time and the radar observation data of the optimal spatial matching location at the initial time are determined as the optimal spatiotemporal matching data.

[0019] Furthermore, the first data sampling area is a rectangular area centered on the target station with a first preset length as its side length; the secondary sampling area is a circular area centered on the target station with a second preset length as its radius; the first data sampling area is divided into the multiple grids according to a resolution of one-quarter of the side length (which can be set as needed).

[0020] Furthermore, the method for allocating the normalized radar observation data to the corresponding grids includes: determining the maximum value of the radar echo intensity among all the normalized radar observation data falling within each grid as the radar echo value of the corresponding grid.

[0021] Furthermore, the method for allocating the normalized surface precipitation observation data to the corresponding grid includes: when the corresponding grid contains at least one surface precipitation observation station, taking the maximum precipitation value of all stations in the grid as the precipitation value of the corresponding grid; when the corresponding grid does not contain any surface precipitation observation station, taking the average precipitation value of all stations in the first data sampling area as the precipitation value of the corresponding grid for filling.

[0022] Further, the method for normalizing radar observation data and surface precipitation observation data within the first data sampling area specifically includes: subtracting a preset minimum value of radar echo intensity from the radar echo intensity of the radar observation data to obtain a first difference; subtracting the preset minimum value of radar echo intensity from the preset maximum value of radar echo intensity to obtain a second difference; calculating the ratio of the first difference to the second difference to obtain the normalized radar observation data; subtracting a preset minimum value of rainfall from the minute rainfall of the surface precipitation observation data to obtain a third difference; subtracting the preset minimum value of rainfall from the preset maximum value of rainfall to obtain a fourth difference; and calculating the ratio of the third difference to the fourth difference to obtain the normalized surface precipitation observation data.

[0023] Further, the spatial correlation coefficient for any of the comparison time periods is determined as follows: The spatial difference of radar data between the gridded radar data of each grid at the initial time period and the average value of its gridded radar data within the first data sampling area is calculated; the spatial difference of precipitation data between the gridded precipitation data of each grid at the comparison time period and the average value of its gridded precipitation data within the first data sampling area is calculated; the product of the spatial difference of radar data belonging to the same grid and the spatial difference of precipitation data is calculated, and the sum of the products of all grids is used as the spatial correlation numerator; the sum of squares of the spatial differences of radar data of all grids and the sum of squares of the spatial differences of precipitation data of all grids are calculated; the sum of squares of the spatial differences of radar data and the sum of squares of the spatial differences of precipitation data are multiplied together, and the square root of the product is taken to obtain the spatial correlation denominator; the spatial correlation numerator is divided by the spatial correlation denominator to obtain the spatial correlation coefficient corresponding to the comparison time period.

[0024] Furthermore, when calculating the time correlation coefficient, the radar observation data within the secondary sampling area is the normalized result of radar radial data without gridding processing.

[0025] Further, the temporal correlation coefficient of any sampling location is determined as follows, where the sampling location is any spatial location within the secondary sampling area: The temporal dimension radar data difference between the normalized radar radial data of the sampling location at each time point within the first time period and the average value of the normalized radar radial data of the sampling location within the first time period is calculated; the temporal dimension precipitation data difference between the normalized surface precipitation observation data of the target station at each time point within the second time period and the average value of the normalized surface precipitation observation data within the second time period is calculated; and the same... The time-dimension radar data difference at the time-series location is multiplied by the time-dimension precipitation data difference, and the sum of the products for all corresponding time periods is used as the numerator of the time correlation. The sum of squares of the time-dimension radar data differences for all time periods and the sum of squares of the time-dimension precipitation data differences for all time periods are calculated. The sum of squares of the time-dimension radar data differences and the sum of squares of the time-dimension precipitation data differences are multiplied by the sum of squares of the time-dimension precipitation data differences, and the square root of the product is calculated to obtain the denominator of the time correlation. The time correlation numerator is divided by the time correlation denominator to obtain the time correlation coefficient corresponding to the sampling location.

[0026] Furthermore, in determining the optimal time-matching time and the optimal spatial-matching location, the following filtering steps are also included: a preset spatial correlation threshold representing the lower limit of time matching is established; if the calculated maximum spatial correlation coefficient is less than the spatial correlation threshold, the radar observation data and ground precipitation observation data corresponding to the current calculation sequence are discarded; a preset time correlation threshold representing the lower limit of spatial matching is established; if the calculated maximum time correlation coefficient is less than the time correlation threshold, the corresponding radar observation data and ground precipitation observation data are discarded.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] (1) This invention determines temporal matching by calculating the spatial feature correlation at different times, and then determines spatial matching by calculating the temporal feature correlation at different spatial locations, completely eliminating the time or space averaging operation in traditional technology. This method finds a unique corresponding radar observation data for each ground precipitation station in space and time, avoiding the problem of radar maxima being weakened due to smoothing, significantly improving the sampling upper limit and model estimation accuracy including features such as extreme heavy precipitation, while eliminating the tedious process of manually trying the time lag.

[0029] (2) The present invention introduces a narrow ring area centered on the radar station for initial sampling. By selecting the first data sampling area within the circumferential narrow band, the consistency of the radar observation data at different spatial locations within the sampling area is guaranteed to the maximum extent. This effectively overcomes the problem of excessive differences in raindrop falling time caused by the radar observation elevation angle, and lays a physical foundation for accurate time matching.

[0030] (3) When performing secondary matching of spatial location, the present invention directly extracts the radar radial data that has not been gridded within the secondary sampling area to participate in the time correlation calculation, thus preserving the original high-resolution expression of the radar data to the greatest extent. This non-fixed search mode can adaptively find precipitation particles that are laterally drifted by the non-uniform wind field, overcoming the spatial mismatch caused by the traditional nearest distance method.

[0031] (4) The present invention sets spatial correlation threshold and temporal correlation threshold in the algorithm to represent the lower limit of matching, which can control the calculation results in both time and space, and eliminate inferior data pairs that cannot meet the correlation standard due to strong external interference or severe wind shear, thereby ensuring that the final output matching sample group has extremely high reliability and quality.

[0032] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, embodiments of the present invention are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0033] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a flowchart of the spatiotemporal matching method between radar observations and ground station precipitation data described in this invention;

[0035] Figure 2 This is a schematic diagram of the optimal time matching principle in this invention;

[0036] Figure 3 T is the present invention target A schematic diagram of the principle of time-optimal spatial matching. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0038] Currently, techniques such as proximity matching, spatial averaging, and temporal lag correlation cannot maximize the optimal match between radar observation data and ground precipitation observation data within the radar coverage area. This is because the vertical distance between ground stations at different distances from the radar station within the radar coverage area and the radar observation data varies significantly, and the spatial non-uniformity of wind speed makes it impossible to accurately estimate the horizontal drift of precipitation particles.

[0039] This invention can specifically find the radar observation data that best matches the target precipitation station in both time and space, thus achieving a higher degree of spatiotemporal matching between the determined radar data and ground precipitation observation data. This invention provides a spatiotemporal matching method for radar observations and ground station precipitation data. Its core idea is to: design a narrow ring region to maximize the consistency of the vertical distance between radar data and ground stations in the sampling area; then determine the temporal matching between the two by analyzing the correlation of spatial characteristics between radar data and ground precipitation data within the sampling area; finally, determine the spatial matching by analyzing the temporal correlation between radar data and ground precipitation data in a secondary sampling area. Ultimately, this achieves the goal of determining a unique radar observation data that matches each ground precipitation station's data both spatially and temporally, without requiring any averaging or artificial lag adjustments.

[0040] Combination Figures 1 to 3The spatiotemporal matching method for radar observations and ground station precipitation data described in this invention specifically includes the following steps:

[0041] S1. Obtain the ground precipitation observation stations within the radar coverage area, and sequentially determine the obtained ground precipitation observation stations as target stations.

[0042] Specifically, collect data from effective areal precipitation observation stations within the radar coverage area (generally national stations with good observation conditions), and determine all or part of them sequentially as target stations for sample production as needed.

[0043] S2. Establish the first data sampling area and perform data normalization and gridding processing.

[0044] Centered on the target station, the radar observation range is divided into multiple narrow ring regions centered on the radar station's location. A narrow ring region containing the target station is selected. Within the selected narrow ring region, a first data sampling area centered on the target station is determined. This first data sampling area is preferably rectangular, with a side length of a first preset length (generally set to 8 km, which can be appropriately increased in sparsely populated areas).

[0045] The first data sampling area is divided into multiple grids at equal intervals. Specifically, the resolution is generally 1 / 4 of the side length of the first data sampling area.

[0046] To facilitate the determination of the spatial correlation between the two data sets, the radar observation data and surface precipitation observation data within the first data sampling area are normalized. The normalization formula is as follows:

[0047] (1)

[0048] in, Radar echo intensity (unit: dBZ) of radar observation data. The maximum preset value for radar echo intensity is 70 dBZ. A minimum value is preset for radar echo intensity, which is -20dBZ; The radar observation data is after normalization processing.

[0049] (2)

[0050] in, This refers to minute-by-minute rainfall (unit: mm) from surface precipitation observation data. The maximum preset rainfall value is 10mm. Set a minimum rainfall value of 0mm. The above surface precipitation observation data has been normalized.

[0051] Subsequently, the normalized radar observation data and the ground precipitation observation data are allocated to the corresponding grids to obtain gridded radar data and gridded precipitation data, respectively.

[0052] For radar data: Since the spatial resolution of radar data is necessarily higher than the resolution of the multiple grids, the maximum value of the radar echo intensity in all normalized radar observation data falling within each grid is determined as the radar echo value of the corresponding grid.

[0053] For precipitation data: Due to the huge differences in the distribution density of ground observation stations, when the corresponding grid contains at least one ground precipitation observation station, the maximum precipitation value of all stations in the grid is taken as the precipitation value of the corresponding grid; when the corresponding grid does not contain any ground precipitation observation station, the average precipitation value of all stations in the first data sampling area is taken as the precipitation value of the corresponding grid for filling.

[0054] S3. Calculate the spatial correlation coefficient to determine the optimal time matching time.

[0055] The time of a radar observation is determined as the initial time (i.e., time T), and the initial time and multiple precipitation observation times after the initial time are used as comparison times (i.e., time T+x; phased array radars generally select 4-10 times after this).

[0056] Combined with appendix Figure 2 As shown, the principle of optimal time matching is illustrated. Figure 2 The left side shows the gridded radar echo data at the initial time (time T), and the bottom row shows the gridded station precipitation data at times T, T+1, T+2, T+3 and T+4 (among which the precipitation data at time T and multiple times after time T correspond to the gridded precipitation data of the aforementioned comparison times). Figure 2 The character 'r' attached to the arrow connecting two types of data t0 r t1 r t2 r t3 r t4 These represent the spatial correlation coefficients under different time comparisons, specifically: r t0 This represents the spatial correlation coefficient between the initial time (time T) gridded radar data and the gridded station precipitation data at time T; r t1 This represents the spatial correlation coefficient between the initial time (time T) gridded radar data and the gridded station precipitation data at time T+1; and so on, r t2 r t3 rt4 These represent the spatial correlation coefficients between the initial time (time T) gridded radar data and the corresponding lagging time (time T+2, time T+3, time T+4) gridded station precipitation data.

[0057] Calculate the spatial correlation coefficient between the gridded radar data at the initial time interval and the gridded precipitation data at each of the comparison time intervals. The calculation formula is as follows:

[0058] (3)

[0059] in, At time T , For time T+x (x can be an integer between 0 and 10), ... , for and Spatial correlation coefficient; The number of grids divided within the first data sampling area; For grid number;

[0060] The spatial anomaly value of the radar data is calculated, that is, the gridded radar data of each grid at time T is calculated respectively. ) and the spatial average of the number of gridded radars within the first data sampling area ( The difference between ) The spatial anomaly of the precipitation data is calculated by retrieving the gridded precipitation data for each grid at the comparison time (T+x time). ) and its spatial average value of gridded precipitation data within the first data sampling area ( The difference between ).

[0061] because This refers to the spatial similarity between two types of data at different times. Specifically, this method determines the temporal matching information of radar and precipitation data by analyzing the spatial correlation of their spatial features at different times. The comparison time corresponding to the highest spatial correlation coefficient is recorded as the optimal temporal matching time. ,Right now:

[0062] (4)

[0063] in, The time when the gridded precipitation data that best matches the initial time (time T) gridded radar data, determined based on spatial correlation, occurs; represents the spatial correlation coefficient for each corresponding comparison time.

[0064] To ensure high-quality matching, a spatial correlation threshold (e.g., 0.4, which can be customized) is preset to limit the validity of the optimal time matching sequence. If the calculated maximum spatial correlation coefficient is less than the spatial correlation threshold, it means that even if the maximum value is selected, it is invalid. In this case, the radar observation data and ground precipitation observation data corresponding to the current calculation sequence are discarded and not included in the subsequent calculation of the optimal spatial matching position.

[0065] S4. Calculate the time correlation coefficient to determine the optimal spatial matching location.

[0066] based on After determining the optimal matching relationship between radar and precipitation in the time dimension, a new circular secondary sampling area is delineated in the first data sampling area corresponding to the optimal time matching time, with the target station as the center and a second preset length (generally set to 1.5-2km) as the radius.

[0067] Combined with appendix Figure 3 As shown, it demonstrates The principle of time-optimal spatial matching. Figure 3 In this process, any reflectivity sampling point is any spatial location (i.e., sampling location) within the secondary sampling area, used to extract radar radial data that has not undergone gridding.

[0068] When calculating the time correlation coefficient, the radar observation data within the secondary sampling area is the normalized result of radar radial data without gridding processing, to ensure the original representation of the radar data. The radar radial data normalization formula is:

[0069] (5)

[0070] in, The radar radial data at any spatial location within the secondary sampling area is the unmeshed data. and These are the preset maximum and minimum values ​​for the radar's radial data, respectively. This is the normalized result of radar radial data.

[0071] In this invention, a preset time span (e.g., 3 time intervals before and after) is extended before and after the initial time interval to obtain the first time period (i.e., The second time period (i.e., the time span) is obtained by extending the preset time span before and after the optimal time matching time. (Time period). Calculate the temporal correlation coefficient between radar observation data of any spatial location within the secondary sampling area during the first time period and surface precipitation observation data of the target station during the second time period. The formula is as follows:

[0072] (6)

[0073] in, for Time period With secondary sampling area Any time period The time correlation coefficient; This represents each time interval within the first time period. This is the time offset relative to the initial time, and its value is an integer from -3 to 3; For the optimal time matching time; This represents each time interval within the second time period; For any spatial location within the secondary sampling area, in each time interval of the first time period ( The normalized results of radar radial data; The average value of the normalized radar radial data for the spatial location within the first time period; For the target site in each time period of the second time period ( Normalized surface precipitation observation data; This represents the average value of the normalized surface precipitation observation data for the target station during the second time period.

[0074] This invention does not employ the traditional multi-point averaging method to enhance the spatial matching between precipitation and radar data. Instead, it finds the optimal spatial match between radar and precipitation data based on the similarity of changes in the temporal dimension. The spatial location corresponding to the largest temporal correlation coefficient is recorded as the optimal spatial matching location. ,Right now:

[0075] (7)

[0076] in, Indicates the second sampling region. The temporal correlation coefficient corresponding to each spatial location.

[0077] Similarly, a time correlation threshold (e.g., 0.4, which can be customized) can be preset to limit the effectiveness of the optimal spatial matching location. If the calculated maximum time correlation coefficient is less than the time correlation threshold, the current radar observation data and ground precipitation observation data are discarded to further filter out radar and precipitation data with higher matching degree.

[0078] S5. Determine and output the best matching sample.

[0079] based on and Theoretically, this identified a unique pair of spatiotemporally matched precipitation and radar data. Finally, the optimal time-matching sequence was determined. The surface precipitation observation data of the target station, and the optimal spatial matching location at the initial time (time T). The radar observation data is determined as the best spatiotemporal matching data, and S2-S5 are repeated until the best spatiotemporal matching radar data for all target sites is determined.

[0080] As shown in Table 1, this assessment used minute-level precipitation data from the Chengdu phased array radar and ground precipitation stations during the heavy precipitation process to compare the present invention with existing conventional matching algorithms.

[0081] Table 1

[0082]

[0083] As shown in Table 1, the method of the present invention can achieve point-to-point matching of radar data and precipitation data, which greatly improves the spatiotemporal correlation compared with the nearest distance method (the spatial correlation coefficient reaches 0.66 and the temporal correlation coefficient reaches 0.71). At the same time, since there is no radar maximum weakening problem caused by the use of time or space averaging methods, its maximum echo average value is retained at 43.37 dBZ, which is significantly higher than the averaging method, effectively improving the sampling upper limit of radar observation data intensity.

[0084] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for spatiotemporal matching of radar observations and ground station precipitation data, characterized in that, include: Acquire several ground precipitation observation stations within the radar coverage area and identify them as target stations in sequence; The radar observation range is divided into multiple narrow ring regions centered on the radar station location. A narrow ring region containing the target station is selected, and a first data sampling area centered on the target station is determined within the selected narrow ring region. The first data sampling area is divided into multiple grids at equal intervals, and the radar observation data and ground precipitation observation data within the first data sampling area are normalized. The normalized radar observation data and ground precipitation observation data are allocated to the corresponding grids to obtain gridded radar data and gridded precipitation data, respectively. A radar observation data time is determined as the initial time, and the initial time and multiple precipitation observation times after the initial time are used as comparison times; the spatial correlation coefficient between the gridded radar data of the initial time and the gridded precipitation data of each comparison time is calculated, and the comparison time corresponding to the largest spatial correlation coefficient is recorded as the optimal time matching time; Within the first data sampling area corresponding to the optimal time matching time, a secondary sampling area centered on the target station is determined; A first time period is obtained by extending a preset time span before and after the initial time period; a second time period is obtained by extending the preset time span before and after the optimal time matching time period. Calculate the temporal correlation coefficient between radar observation data of any spatial location within the secondary sampling area in the first time period and ground precipitation observation data of the target station in the second time period, and record the spatial location corresponding to the largest temporal correlation coefficient as the best spatial matching location; The ground precipitation observation data of the target station at the optimal time of the optimal time matching is determined to be the optimal spatiotemporal matching data with the radar observation data of the optimal spatial matching location at the initial time.

2. The method according to claim 1, characterized in that, The first data sampling area is a rectangular area centered on the target station with a first preset length as its side length; the first data sampling area is divided into multiple grids with a resolution of one-quarter of the side length (which can be set according to requirements); the secondary sampling area is a circular area centered on the target station with a second preset length (which can be set according to requirements) as its radius.

3. The method according to claim 1, characterized in that, The method for allocating normalized radar observation data to corresponding grids includes: determining the maximum value of the radar echo intensity among all normalized radar observation data falling within each grid as the radar echo value of the corresponding grid.

4. The method according to claim 1, characterized in that, Methods for allocating normalized surface precipitation observation data to corresponding grids include: When the corresponding grid contains at least one of the ground precipitation observation stations, the maximum precipitation value of all stations in the grid is taken as the precipitation value of the corresponding grid. When the corresponding grid does not contain the ground precipitation observation station, the average precipitation of all stations in the first data sampling area is taken as the precipitation value of the corresponding grid to fill it.

5. The method according to claim 1, characterized in that, The method for normalizing radar observation data and surface precipitation observation data within the first data sampling area specifically includes: The radar echo intensity of the radar observation data is subtracted from the preset minimum value of the radar echo intensity to obtain a first difference; the preset maximum value of the radar echo intensity is subtracted from the preset minimum value of the radar echo intensity to obtain a second difference; the ratio of the first difference to the second difference is calculated to obtain the normalized radar observation data. The third difference is obtained by subtracting the preset minimum rainfall value from the minute rainfall of the surface precipitation observation data; the fourth difference is obtained by subtracting the preset minimum rainfall value from the preset maximum rainfall value; the ratio of the third difference to the fourth difference is calculated to obtain the normalized surface precipitation observation data.

6. The method according to claim 1, characterized in that, The spatial correlation coefficient for any of the aforementioned comparison time points is determined as follows: The spatial difference of radar data between the gridded radar data of each grid at the initial time point and the average value of its gridded radar data within the first data sampling area is calculated; the spatial difference of precipitation data between the gridded precipitation data of each grid at the comparison time point and the average value of its gridded precipitation data within the first data sampling area is calculated; the product of the spatial difference of radar data belonging to the same grid and the spatial difference of precipitation data is calculated, and the sum of the products of all grids is used as the spatial correlation numerator; Calculate the sum of squares of the spatial differences of the radar data across all grids, and the sum of squares of the spatial differences of the precipitation data across all grids; multiply the sum of squares of the spatial differences of the radar data by the sum of squares of the spatial differences of the precipitation data, and take the square root of the product to obtain the spatial correlation denominator; divide the spatial correlation numerator by the spatial correlation denominator to obtain the spatial correlation coefficient corresponding to the comparison time.

7. The method according to claim 1, characterized in that, When calculating the time correlation coefficient, the radar observation data in the secondary sampling area is the normalized result of radar radial data without gridding processing.

8. The method according to claim 7, characterized in that, The temporal correlation coefficient for any sampling location is determined as follows, where the sampling location is any spatial location within the secondary sampling area: The temporal dimension radar data difference between the normalized radar radial data of the sampling location at each time point within the first time period and the average value of the normalized radar radial data of the sampling location within the first time period is calculated; the temporal dimension precipitation data difference between the normalized surface precipitation observation data of the target station at each time point within the second time period and the average value of the normalized surface precipitation observation data within the second time period is calculated; the temporal dimension radar data difference at the same time-series location is multiplied by the temporal dimension precipitation data difference, and the sum of the products at all corresponding time points is used as the temporal correlation numerator; the sum of squares of the temporal dimension radar data differences at all time points and the sum of squares of the temporal dimension precipitation data differences at all time points are calculated; the sum of squares of the temporal dimension radar data differences is multiplied by the sum of squares of the temporal dimension precipitation data differences, and the square root of the product is calculated to obtain the temporal correlation denominator; the temporal correlation numerator is divided by the temporal correlation denominator to obtain the temporal correlation coefficient corresponding to the sampling location.

9. The method according to claim 1, characterized in that, In the process of determining the optimal time matching time and the optimal spatial matching location, the following filtering steps are also included: a preset spatial correlation threshold representing the lower limit of time matching is set; if the calculated maximum spatial correlation coefficient is less than the spatial correlation threshold, the radar observation data and ground precipitation observation data corresponding to the current calculation sequence are discarded. A time correlation threshold is preset to represent the lower limit of spatial matching. If the calculated maximum time correlation coefficient is less than the time correlation threshold, the corresponding radar observation data and ground precipitation observation data are discarded.

Images