A radar precipitation bias correction algorithm based on the combination of probability density matching and climatological ratio method

Abstract

This invention discloses a radar precipitation bias correction algorithm based on a combination of probability density matching and climatological ratio method. The algorithm includes the collection and preprocessing of ground station observation data, PDF bias correction of radar precipitation, and climatological bias correction of radar precipitation. This invention achieves effective utilization of precipitation observation data from ground meteorological stations. By using a ground station grid analysis field and a ground-radar climatological ratio analysis field to correct biases in radar network-estimated precipitation products, it effectively corrects the discrepancy between radar precipitation estimates and actual precipitation, reduces the discontinuity and systematic deviation in the spatiotemporal distribution of fused products, enhances its adaptability to complex terrain and rapidly evolving precipitation scenarios, and achieves collaborative optimization correction of multi-scale and multi-type biases, significantly improving the reliability and operational applicability of radar precipitation products.

Description

Technical Field

[0001] This invention relates to the field of precipitation bias correction technology. Specifically, it relates to a radar precipitation bias correction algorithm that combines probability density matching and climatological ratio methods. Background Technology

[0002] In multi-source precipitation fusion and real-time analysis, the quality of radar precipitation data significantly impacts the accuracy of the fusion results. However, due to limitations such as radar detection principles, environmental interference, and terrain obstruction, systematic deviations often exist between radar precipitation estimates and actual precipitation. Particularly in complex terrain areas, radar beams are obstructed by mountains, buildings, etc., leading to incomplete or distorted precipitation echo signals. In large-scale precipitation systems, differences in radar hardware performance, attenuation effects, and errors in precipitation particle phase identification can cause non-independent systematic biases. Existing research indicates that without effective bias correction of radar precipitation data, the fused real-time product is prone to significant discontinuities and systematic deviations in its spatiotemporal distribution, affecting the effectiveness of short-term forecasts, heavy rain warnings, and forecast model validation.

[0003] The formation mechanism of radar precipitation bias is complex, and its influencing factors can be summarized into two categories: First, dynamic non-independent system bias, which is closely related to radar hardware performance, precipitation system evolution characteristics (such as occurrence and dissipation time, movement speed), and regional station density differences. For example, the detection range of a single radar is limited, and different regional station densities (such as dense stations in eastern China and sparse stations in western China) lead to significant differences in the spatial matching accuracy between ground observations and radar data. Second, static terrain obstruction bias, caused by obstruction of the radar beam propagation path, manifests as a systematic underestimation or overestimation of terrain-related areas in long-term climate statistics. In addition, the spatiotemporal distribution characteristics of radar precipitation bias are also affected by the superposition of factors such as seasonal variations, precipitation intensity distribution, and sensor upgrades and iterations.

[0004] Current methods for correcting radar precipitation biases still have limitations. Traditional methods mainly fall into two categories: the first is statistical correction techniques based on real-time ground observations, such as dynamic calibration or regression analysis. While these methods can partially correct the instantaneous biases between radar and ground observations, they rely excessively on the spatial representativeness of real-time station data, are prone to interpolation errors in sparsely networked areas, and fail to consider the spatiotemporal evolution characteristics of precipitation systems, making it difficult to reliably capture the long-term variation patterns of non-independent biases. The second is correction schemes based on climatological statistics, such as constructing static correction factors using historical precipitation ratios. While this method can alleviate systematic biases caused by topographic obstruction, it cannot effectively handle dynamic biases in short-term precipitation processes, and the spatial resolution and timeliness of climatological factors are insufficient, making it difficult to meet the rapid response requirements of extreme weather events. Existing methods often apply these two types of techniques independently, resulting in a disconnect in the coordinated processing of dynamic systematic biases and static topographic biases in the correction results. This may lead to problems such as imbalances in correction magnitudes or decreased spatial consistency. Radar precipitation data plays an irreplaceable role in meteorological operations. On the one hand, it provides high spatiotemporal resolution precipitation structure information for multi-source fusion real-time products, supporting refined weather monitoring and disaster early warning. On the other hand, as an important input for numerical model assimilation, its quality directly affects the accuracy of precipitation forecasts. However, existing correction techniques struggle to balance the combined effects of dynamic and static biases, especially in complex terrain and rapidly evolving precipitation scenarios, where corrected radar data still exhibits significant uncertainties. For example, 202210066596.X discloses a multi-source precipitation data fusion method that utilizes CMORPH satellite data, radar data, and automatic weather station data for fusion processing. However, it does not consider dynamic terrain adjustments. For special weather conditions (such as short-duration severe convection) or extreme terrain (such as high-altitude mountains and plateaus), the representativeness of a single data source combination is insufficient, leading to a decrease in fusion accuracy. Moreover, its temporal resolution is only 1 hour, which is insufficient to meet the requirements for short-duration heavy precipitation. Therefore, there is an urgent need to develop new fusion correction algorithms that combine the advantages of dynamic probability density matching and climate-state ratio adjustment, break through the limitations of single methods, and achieve collaborative optimization correction of multi-scale and multi-type biases, so as to provide technical support for improving the reliability and operational applicability of multi-source precipitation fusion products. Summary of the Invention

[0005] Therefore, the technical problem to be solved by the present invention is to provide a radar precipitation deviation correction algorithm based on a combination of probability density matching and climatological ratio method, which can effectively solve the problem that the existing technology is difficult to coordinate the processing of dynamic and static deviations, and the corrected radar data still has significant uncertainty.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] A radar precipitation bias correction algorithm based on a combination of probability density matching and climatological ratio method includes:

[0008] S1. The collection and preprocessing of ground station observation data yields the ground precipitation grid analysis field;

[0009] S2, PDF bias correction for radar precipitation

[0010] S21. Perform a consistency matching analysis between ground-observed precipitation and radar-estimated precipitation;

[0011] S22. Construct a correction model and adjust the spatiotemporal matching window between ground and radar precipitation PDF samples;

[0012] S23. Use the ground precipitation grid analysis field to perform PDF bias correction on radar network-estimated precipitation data;

[0013] S3. Climatic bias correction of radar precipitation

[0014] S31. For areas where radar precipitation is severely obstructed by terrain and has poor quality, the INCA algorithm is used to correct the precipitation by calculating the climate proportion correction factor (RFC). The calculation formula is as follows:

[0015]

[0016] Among them, P i,j For ground-observed precipitation values ​​at specified grid locations, P Radar,i,j Precipitation values ​​are estimated by radar at specified grid locations;

[0017] S32. Construct a correction model, use the ground-radar climatological ratio analysis field to correct the bias in the radar network-estimated precipitation data, apply the RFC to the original radar-estimated precipitation in the grid, and obtain the climatological bias-corrected radar-estimated precipitation. The calculation formula is as follows:

[0018] P * Radar,i,j =RFC i,j *P Radar,i,j (2).

[0019] Preferably, in step S1 above, precipitation data from ground meteorological observation stations are collected, and quality control processing is performed to obtain quality-controlled ground observation precipitation data. Using the optimal interpolation method, the quality-controlled ground observation precipitation data is processed to generate a 1km ground grid analysis precipitation product, i.e., a ground precipitation grid analysis field.

[0020] Preferably, the method for processing ground-observed precipitation data after quality control in step S1 above is to use a single data source control module, a multi-source data collaborative quality control module, and a dynamic blacklist module for data quality control.

[0021] Preferably, the aforementioned single data source control module includes metadata checking, feature value checking, boundary value checking, and dead value checking, which can eliminate gross errors and dead values ​​that have not changed for a long time in real-time precipitation data; the multi-source data collaborative quality control module is a QC algorithm developed based on the consistency of meteorological observation data related to various precipitation phenomena such as radar and weather phenomena, which can accurately identify false clear-sky precipitation that is difficult to identify by rapid quality control and false 0 value data in typical precipitation areas; the data after quality control by the single data source control module and the multi-source data collaborative quality control module is further processed by the dynamic blacklist module to remove observation stations with high error ratios and long error durations. At the same time, a flexible dynamic evaluation mechanism is adopted to remove the data from the blacklist after the data quality is restored, and select data with quality control codes of 0, 1, 3, and 4 to participate in subsequent fusion analysis.

[0022] Preferably, step S1 above, which uses optimal interpolation to generate precipitation products for 1km ground grid analysis, includes:

[0023] 1) Establish a grid background field for precipitation climate values;

[0024] 2) Calculate the precipitation ratio data of each station and interpolate to generate the corresponding grid field. The ratio data is a new element defined with the help of the climate background field: Precipitation ratio = Station observed precipitation / Corresponding grid precipitation climate value;

[0025] 3) The precipitation grid field is generated by multiplying the precipitation ratio grid field with the corresponding climate background field. The interpolation method used in step 2) to generate the grid field is the optimal interpolation method, and the calculation formula (3) is as follows:

[0026]

[0027] That is, the analysis value A of the grid. k The initial estimate F at that point k Including the deviation between the observed value and the initial estimate at that point, the deviation is calculated from n known initial estimates F within the specified analysis range. i With the observed value O i The bias-weighted estimate is obtained.

[0028] Preferably, in step S21 above, the consistency matching analysis includes: (1) data preparation and time alignment, (2) consistency index calculation, (3) optimal lag time determination, analyzing the difference between the cumulative precipitation observed by the ground automatic station in the first 10 minutes and the radar QPE at the lag time of 0 minutes during the precipitation process, and using the correlation coefficient, root mean square error and relative deviation index to determine the optimal lag time.

[0029] Preferably, in step S22 above, the spatiotemporal matching window parameters are set to 1 hour and 35 km, and the minimum number of valid sample pairs participating in PDF matching is 120.

[0030] Preferably, in step S1 above, the ground precipitation grid analysis field is an isotropic latitude and longitude grid interpolated precipitation field generated using precipitation data from ground stations.

[0031] Preferably, in step S32 above, the radar network estimated precipitation data is a radar network estimated precipitation product developed by the National Meteorological Information Center using basic data from more than 200 operational weather radars across the country.

[0032] Preferably, in step S32 above, the ground-radar climate ratio analysis field is an analysis grid climate ratio correction factor calculated using the ratio of historical ground observations and radar-estimated cumulative precipitation over the past three months.

[0033] The technical solution of the present invention achieves the following beneficial technical effects:

[0034] 1. This invention enables the effective utilization of precipitation observation data from ground meteorological stations. By using the ground station grid analysis field and the ground-radar climate ratio analysis field to correct the bias of radar network-estimated precipitation products, it effectively corrects the deviation between radar precipitation estimates and actual precipitation, reduces the discontinuity and systematic deviation of the spatiotemporal distribution of fused products, enhances its adaptability to complex terrain and rapidly evolving precipitation scenarios, and achieves collaborative optimization and correction of multi-scale and multi-type biases, significantly improving the reliability and operational applicability of radar precipitation products.

[0035] 2. This invention can help improve the level of meteorological operational forecasting and services. The corrected and accurate radar precipitation data, as an important input for numerical model assimilation, can improve the accuracy of precipitation forecasts, with a temporal resolution of up to 10 minutes and a spatial resolution of up to 1 km. At the same time, it provides more accurate high spatiotemporal resolution precipitation structure information for multi-source fusion real-time products, optimizes refined weather monitoring, helps to detect precipitation anomalies in a timely manner and issue disaster warnings in advance, and provides strong data support for meteorological operations such as intelligent grid short-term forecasting, rainstorm warnings and decision-making services, thereby promoting the overall improvement of meteorological operations. Attached Figure Description

[0036] Figure 1 This invention presents a radar precipitation bias correction algorithm based on a combination of probability density matching and climatological ratio method, showing the relationship between radar QPE and cumulative precipitation observed by ground automatic stations for the first 10 minutes under different lag times (where 1a is the correlation coefficient relationship; 1b is the root mean square error relationship; and 1c is the relative bias relationship). Detailed Implementation

[0037] This embodiment utilizes a ground precipitation grid analysis field, a ground-radar climatological ratio analysis field, and radar network-estimated precipitation data to perform a radar precipitation bias correction algorithm. Specifically:

[0038] S1. Collection and preprocessing of ground station observation data

[0039] S11. Collect precipitation data from ground meteorological observation stations, perform quality control processing, and obtain quality-controlled ground observation precipitation data;

[0040] Data quality control is achieved using a single data source control module, a multi-source data collaborative quality control module, and a dynamic blacklist module.

[0041] The single data source control module includes metadata checking, feature value checking, limit value checking, and dead value checking, which can quickly and efficiently remove gross errors and dead values ​​that have not changed for a long time from real-time precipitation data.

[0042] The multi-source data collaborative quality control module is a QC algorithm developed based on the consistency of meteorological observation data related to various precipitation phenomena, including radar and weather phenomena. It can accurately identify false clear-sky precipitation and false 0-value data in typical precipitation areas, which are difficult to identify by rapid quality control.

[0043] Data that has undergone quality control by the single data source control module and the multi-source data collaborative quality control module is further processed by the dynamic blacklist module to remove observation stations with high error rates and long error durations. At the same time, a flexible dynamic evaluation mechanism is adopted to remove the data from the blacklist only after the data quality is restored. Data with quality control codes of 0, 1, 3, and 4 are selected for subsequent fusion analysis (i.e., correct, questionable, corrected, and modified observation data, which is the same as the selection strategy for hourly surface precipitation observation data in the ART_1km business system).

[0044] S22. Using the optimal interpolation method, the quality-controlled ground observation precipitation data is processed to generate a 1km ground grid analysis precipitation product.

[0045] The optimal interpolation (OI) gridded analysis technique based on the climate background field ratio (Xie et al. 2007; Chen et al. 2002; Shen et al. 2010, Shen et al. 2012) was used to process and generate precipitation products with a 1km surface grid, including:

[0046] 1) Establish a grid background field for precipitation climate values;

[0047] 2) Calculate the precipitation ratio data of each station and interpolate to generate the corresponding grid field. The ratio data is a new element defined with the help of the climate background field: Precipitation ratio = Station observed precipitation / Corresponding grid precipitation climate value;

[0048] 3) A precipitation grid field is generated by multiplying the precipitation ratio grid field with the corresponding climate background field. The interpolation method used in step 2) to generate the grid field is the optimal interpolation (OI) method, and the calculation formula (3) is as follows:

[0049]

[0050] That is, the analysis value A of the grid. k The initial estimate F at that point k Including the deviation between the observed value and the initial estimate at that point, the deviation is calculated from n known initial estimates F within the specified analysis range. i With the observed value O i The bias-weighted estimate is obtained;

[0051] S2, PDF bias correction for radar precipitation

[0052] S21. Perform consistency matching analysis between ground-observed precipitation and radar-estimated precipitation, including:

[0053] (1) Data preparation and time alignment; (2) Consistency index calculation; (3) Determination of optimal lag time. The relationship between the cumulative precipitation observed by the ground automatic station in the first 10 minutes and the radar QPE at 0 minutes lag was analyzed during the precipitation process. The optimal lag time was determined by using the correlation coefficient, root mean square error, and relative deviation index. This analysis verified the reliability of time matching on the 10-minute scale by quantifying the consistency between radar and ground observations under different time lags, and provided the prerequisite for the application of PDF matching method. Its core value lies in eliminating the interference of time misalignment on deviation correction and ensuring that the correction result truly reflects the error characteristics of radar data, rather than the pseudo-bias caused by time sampling differences.

[0054] S22. Construct a correction model and adjust the spatiotemporal matching window between ground and radar precipitation PDF samples;

[0055] If the matching window is not set properly, the following problems may occur:

[0056] (1) Characteristics of precipitation system formation and dissipation: Short-duration heavy precipitation systems (such as thunderstorms) have short lifecycles (usually <1 hour). If the matching window is too large (such as 1 hour), it will mix observation data from different precipitation stages and mask the true bias; if the window is too small (such as a single minute), the sample size will be insufficient and the statistical reliability will be reduced. Therefore, the spatiotemporal matching window parameter is set to 1 hour.

[0057] (2) Radar detection range limitations: The effective detection range of a single radar is limited (typically with a diameter of approximately 200-300 km). Precipitation in peripheral areas may be biased due to beam obstruction or distance attenuation, requiring adjustment of spatial representativeness through a matching window. Furthermore, there are regional differences in station density: the density of rain gauges is high in eastern China (approximately 10-20 km spacing) and sparse in the west (>50 km spacing). A matching window is needed to balance spatial coverage and sample representativeness, avoiding overfitting due to an excessively small window in the east and smoothing distortion due to an excessively large window in the west. Therefore, the spatial window parameter was set to 35 km, and the minimum number of effective sample pairs participating in PDF matching was 120 to ensure statistical reliability and avoid extreme bias caused by small samples. By comprehensively considering the generation and dissipation time characteristics of precipitation systems every 10 minutes, the detection range of a single radar, the stability of radar bias changes, and the variations in station density in different regions of eastern and western China, the spatiotemporal matching window parameters for ground and radar precipitation PDF samples were adjusted to improve the universality and accuracy of bias correction.

[0058] S23. Use the ground precipitation grid analysis field to perform PDF bias correction on radar network-estimated precipitation data;

[0059] S3. Climatic bias correction of radar precipitation

[0060] S31. For areas where radar precipitation is severely obstructed by terrain and of poor quality, the INCA algorithm is used to correct the radar precipitation by calculating the Climate Radar Scaling Factor (RFC). The RFC is calculated using the ratio of historical ground observations and radar-estimated cumulative precipitation over the past three months. The formula is as follows:

[0061]

[0062] Among them, P i,j For ground-observed precipitation values ​​at specified grid locations, P Radar,i,j Precipitation values ​​are estimated by radar at specified grid locations;

[0063] S32. Construct a correction model, use the ground-radar climatological ratio analysis field to correct the bias in the radar network-estimated precipitation data, apply the RFC to the original radar-estimated precipitation in the grid, and obtain the climatological bias-corrected radar-estimated precipitation. The calculation formula is as follows:

[0064] P * Radar,i,j =RFC i,j *P Radar,i,j (2).

[0065] Among them, the ground precipitation grid analysis field is an interpolated precipitation field with equal latitude and longitude grids generated using precipitation data from ground stations.

[0066] The ground-radar climate ratio analysis field is a climate ratio correction factor for the analysis grid calculated using the ratio of historical ground observations and radar-estimated cumulative precipitation over the past three months.

[0067] The radar network-based precipitation estimation data is a radar network-based precipitation estimation product developed by the National Meteorological Information Center using basic data from more than 200 operational weather radars across the country.

[0068] The present invention is provided in the following information for specific implementation.

[0069] like Figure 1 This study examines the relationship between the cumulative precipitation observed by automatic ground stations over the first 10 minutes and the radar QPE at a lag of 0 minutes during a precipitation event, checking the consistency between radar QPE and ground-based precipitation observations on a 10-minute timescale. The correlation coefficient ( Figure 1 a) Root mean square error ( Figure 1 b) Relative deviation ( Figure 1 c) Indicators show that the cumulative precipitation observed on the ground in the first 10 minutes with a lag of 4 minutes is most consistent with the instantaneous precipitation estimated by the radar at a lag of 0 minutes. Within the lag of 0-9 minutes, the time difference in cumulative time sampling has a relatively small impact on the bias, only 0.2% overall. This indicates that the consistency between radar-detected precipitation and ground observation time matching is good at the 10-minute scale, making it suitable for using the PDF matching method for bias correction. Secondly, considering the characteristics of the formation and dissipation time of precipitation systems every 10 minutes, the detection range of a single radar, the stability of radar bias changes, and the variations in station density in different regions of eastern and western China, the spatiotemporal matching window for ground and radar precipitation PDF samples was adjusted.

[0070] By analyzing the 24-hour cumulative precipitation (unit: mm) observed on July 31, 2021, and comparing it with the 24-hour cumulative precipitation (unit: mm) before and after the bias correction of the radar QPE product PDF on July 31, 2021, it can be seen that, overall, the spatial morphology of precipitation estimated by the radar and the analysis of ground observations are quite similar. However, the radar estimates the strong precipitation centers in Henan, Shandong, and southern Hebei (Hebei-Shandong-Henan region), while the precipitation in Northeast China is underestimated. After bias correction, the intensity of the precipitation center in the Hebei-Shandong-Henan region by the radar is significantly weakened, while the precipitation in Northeast China is slightly strengthened, which is more consistent with ground observations. Statistics are shown in Table 1.

[0071] Table 1. Statistical values ​​of errors in radar and satellite precipitation products before and after correction, taken every 10 minutes as of July 31, 2021.

[0072]

[0073] Statistically, after the PDF deviation of radar QPE products, the CC (Correlation Coefficient) increases, the RMSE (Root Mean Square Error) decreases, and the Bias (deviation) decreases, indicating that deviation correction has a certain improvement on the accuracy of radar products.

[0074] Before the correction, the original radar precipitation data in the southwestern mountainous area of ​​the Sichuan Basin was significantly understated due to topographic obstruction, corresponding to a higher RFC value for this region. After the correction, the precipitation intensity in this region increased, approaching the precipitation intensity of the western part of the basin.

[0075] When selecting climatological bias corrections, comparing precipitation observations from ground stations and radar QPE products sometimes reveals an "overcorrection" problem in local areas. This is because when radar obstruction is severe, an excessively small radar accumulation as the denominator results in an abnormally large RFC value. Furthermore, radar has relatively good detection capabilities for heavy precipitation, and it's important to avoid weakening the detection of heavy precipitation by RFC corrections. Therefore, the concept of PDF correction is introduced to adjust and limit the RFC for heavy precipitation (greater than 1 mm / 10 min) and light precipitation (less than 0.2 mm / 10 min) reaching a certain intensity threshold. This avoids overcorrection of radar-detected heavy precipitation while reducing the overestimation of light precipitation.

[0076] Comparing the effects of different correction methods, the PDF algorithm, while performing well in densely networked radar systems, performs poorly in the complex terrain of Southwest China. The RFC method, however, effectively corrects biases caused by terrain obstruction. However, the RFC method tends to overestimate small precipitation events, potentially leading to a large positive bias in the overall merged precipitation data. To leverage the strengths of different methods, the radar precipitation bias correction scheme was adjusted: the RFC method is primarily used in Southwest China to address biases caused by terrain obstruction, while the PDF method is primarily used in areas with higher radar quality to address biases caused by different weather and precipitation systems.

[0077] This invention utilizes a ground station grid analysis field and a ground-radar climatological ratio analysis field to correct biases in radar network-estimated precipitation products. The feasibility of PDF bias correction and the effectiveness of different correction methods are compared and analyzed. The results from different correction methods show that both PDF and climatological bias correction can improve the quality of radar precipitation products to some extent. However, the experimental results indicate that both methods have certain limitations. While the PDF algorithm can effectively correct radar biases in densely networked conditions, its performance is poor in the complex terrain of southwestern China. The RFC method, on the other hand, effectively corrects biases caused by terrain obstruction. However, the RFC method tends to overestimate small amounts of precipitation, potentially leading to a large positive bias in the overall merged precipitation data.

[0078] The correction results were independently verified by retaining precipitation observation data from over 2400 national stations. Statistical analysis shows that the comprehensively optimized correction method demonstrates superior correction performance in areas with complex terrain in Southwest China and regions with low radar quality. As shown in Table 2, in areas with complex terrain in Southwest China, the relative deviation of the comprehensively optimized correction results is -1.8%, which is better than PDF deviation correction and climatological deviation correction (-18.6% and -3.8%, respectively). In areas with low radar quality, the comprehensively optimized correction results improved the relative deviation (PDF deviation correction and climatological deviation correction were -77.1% and -49.4%, respectively) to -46.5%, and the relative coefficients (PDF deviation correction and climatological deviation correction were 0.269 and 0.375, respectively) to 0.419. From the spatial distribution of precipitation after correction and the statistical analysis of the assessment, the comprehensively optimized correction effect provides a greater improvement in the accuracy of radar products, is more consistent with ground observations, and enhances the reliability and operational applicability of radar precipitation products.

[0079] Table 2. Test results of different bias correction schemes in the southwest region and areas with poor radar quality.

[0080]

[0081] This invention achieves effective utilization of precipitation observation data from ground meteorological stations. By leveraging various analysis fields, such as the ground station grid analysis field, it corrects the biases in radar network-based precipitation estimation products, thereby rectifying estimation errors and realizing multi-scale, multi-type bias collaborative optimization and correction, thus improving product reliability and applicability. On the other hand, the corrected data can improve the accuracy of precipitation forecasts, provide precise information for multi-source fusion real-time products, optimize weather monitoring, provide strong support for multiple meteorological services, and promote the overall improvement of meteorological services.

[0082] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of the claims of this patent application.

Claims

1. A radar precipitation bias correction algorithm based on the combination of probability density matching and climatological ratio method, characterized in that, include: S1. The collection and preprocessing of ground station observation data yields the ground precipitation grid analysis field; S2, PDF bias correction for radar precipitation S21. Perform a consistency matching analysis between ground-observed precipitation and radar-estimated precipitation; S22. Construct a correction model and adjust the spatiotemporal matching window between ground and radar precipitation PDF samples; S23. Use the ground precipitation grid analysis field to perform PDF bias correction on radar network-estimated precipitation data; S3. Climatic bias correction of radar precipitation S31. For areas where radar precipitation is severely obstructed by terrain and has poor quality, the INCA algorithm is used to correct the precipitation by calculating the climate proportion correction factor (RFC). The calculation formula is as follows: Among them, P i,j For ground-observed precipitation values ​​at specified grid locations, P Radar,i,j Precipitation values ​​are estimated by radar at specified grid locations; S32. Construct a correction model, use the ground-radar climatological ratio analysis field to correct the bias in the radar network-estimated precipitation data, apply the RFC to the original radar-estimated precipitation in the grid, and obtain the climatological bias-corrected radar-estimated precipitation. The calculation formula is as follows: P * Radar,i,j =RFC i,j *P Radar,i,j (2)。 2. The radar precipitation bias correction algorithm based on the combination of the probability density matching and the climatological ratio method according to claim 1, characterized in that, In step S1, precipitation data from ground meteorological observation stations are collected, and quality control processing is performed to obtain quality-controlled ground observation precipitation data. Using the optimal interpolation method, the quality-controlled ground observation precipitation data is processed to generate a 1km ground grid analysis precipitation product, namely the ground precipitation grid analysis field.

3. The radar precipitation bias correction algorithm based on the combination of the probability density matching and the climatological ratio method according to claim 2, characterized in that, The method for processing ground-observed precipitation data after quality control involves using a single data source control module, a multi-source data collaborative quality control module, and a dynamic blacklist module for data quality control.

4. The radar precipitation bias correction algorithm based on the combination of probability density matching and climatological ratio method according to claim 3, characterized in that, The single data source control module includes metadata checking, feature value checking, boundary value checking, and dead value checking, which can eliminate gross errors and dead values ​​that have not changed for a long time in real-time precipitation data; the multi-source data collaborative quality control module is a QC algorithm developed based on the consistency of meteorological observation data related to various precipitation from radar and weather phenomena, which can accurately identify false clear-sky precipitation that is difficult to identify by rapid quality control, as well as false 0 value data in typical precipitation areas. The data, after being controlled by the single data source control module and the multi-source data collaborative quality control module, is further processed by the dynamic blacklist module to remove observation stations with high error rates and long error durations. At the same time, a flexible dynamic evaluation mechanism is adopted to remove the data from the blacklist after the data quality is restored. Data with quality control codes of 0, 1, 3, and 4 are selected to participate in subsequent fusion analysis.

5. The radar precipitation bias correction algorithm based on the combination of the probability density matching and the climatological ratio method according to claim 3, characterized in that, Step S1, which uses optimal interpolation to generate precipitation products for 1km ground grid analysis, includes: 1) Establish a grid background field for precipitation climate values; 2) Calculate the precipitation ratio data of each station and interpolate to generate the corresponding grid field. The ratio data is a new element defined with the help of the climate background field: Precipitation ratio = Station observed precipitation / Corresponding grid precipitation climate value; 3) The precipitation grid field is generated by multiplying the precipitation ratio grid field with the corresponding climate background field. The interpolation method used in step 2) to generate the grid field is the optimal interpolation method, and the calculation formula (3) is as follows: That is, the analysis value A of the grid. k The initial estimate F at that point k Including the deviation between the observed value and the initial estimate at that point, the deviation is calculated from n known initial estimates F within the specified analysis range. i With the observed value O i The bias-weighted estimate is obtained.

6. The radar precipitation bias correction algorithm based on a combination of probability density matching and climatological ratio method according to claim 1, characterized in that, In step S21, the consistency matching analysis includes: (1) data preparation and time alignment, (2) consistency index calculation, and (3) determination of the optimal lag time. The analysis analyzes the difference between the cumulative precipitation observed by the ground automatic station in the first 10 minutes and the radar QPE at the lag time of 0 minutes during the precipitation process, and uses the correlation coefficient, root mean square error, and relative deviation index to determine the optimal lag time.

7. The radar precipitation bias correction algorithm based on a combination of probability density matching and climatological ratio method according to claim 1, characterized in that, In step S22, the spatiotemporal matching window parameters are set to 1 hour and 35 km, and the minimum number of valid sample pairs participating in PDF matching is 120.

8. The radar precipitation bias correction algorithm based on a combination of probability density matching and climatological ratio method according to claim 1, characterized in that, In step S1, the ground precipitation grid analysis field is an interpolated precipitation field with equal latitude and longitude grids generated using precipitation data from ground stations.

9. The radar precipitation bias correction algorithm based on a combination of probability density matching and climatological ratio method according to claim 1, characterized in that, In step S32, the radar network-estimated precipitation data is a radar network-estimated precipitation product developed by the National Meteorological Information Center using basic data from more than 200 operational weather radars across the country.

10. The radar precipitation bias correction algorithm based on a combination of probability density matching and climatological ratio method according to claim 1, characterized in that, In step S32, the ground-radar climate ratio analysis field is an analysis grid climate ratio correction factor calculated using the ratio of historical ground observations and radar-estimated cumulative precipitation over the past three months.

Images