Dual-polarization radar multi-parameter fusion rainfall estimation system and method

Abstract

The invention discloses a dual-polarization radar multi-parameter fusion rainfall estimation system and method, and relates to the technical field of meteorological radars, and the system comprises a data collection module which is used for starting a dual-polarization radar to carry out long-term observation when an observation scene with uniformly distributed light rain weather appears, recording radar polarization parameters at different elevation angles, and carrying out the data collection module; transmitting to the deviation analysis module; the deviation analysis module is used for receiving the radar polarization parameters and calculating ZDR statistical characteristics according to the reference uniform scene; according to the invention, by constructing the antenna isolation elevation angle dependence model, dynamic reverse correction is carried out on ZDR and rhohv measurement values under different elevation angles, the influence of dynamic degradation of the antenna cross polarization isolation along with the scanning elevation angle on the measurement values is effectively eliminated, and real-time sensing and compensation of the isolation degradation along with the dynamic change of the elevation angle are realized; the radar can keep high measurement precision in the whole scanning space, and the conditions of precipitation phase state misjudgment and missing report are reduced, so that the accuracy of precipitation estimation is improved.

Description

Technical Field

[0001] This invention relates to the field of meteorological radar technology, specifically to a dual-polarization radar multi-parameter fusion precipitation estimation system and method. Background Technology

[0002] Dual-polarization radar plays a crucial role in meteorological monitoring. It can transmit and receive horizontally and vertically polarized waves, acquiring more meteorological information by measuring multiple parameters such as differential reflectivity (ZDR) and differential propagation phase shift (KDP), thus enabling accurate precipitation estimation. Compared to traditional radar, it significantly improves the accuracy of precipitation type and intensity identification and estimation. However, the measurement accuracy of dual-polarization radar is highly dependent on the polarization purity of the radar system itself, specifically the antenna's cross-polarization isolation. This reflects the degree of signal leakage received by the V channel when transmitting H-waves. Ideally, the antenna should have extremely high isolation to ensure minimal signal crosstalk between the transmitting and receiving channels.

[0003] Currently, through long-term observation and analysis, it has been found that the cross-polarization isolation of large parabolic antennas will dynamically and nonlinearly deteriorate with the change of scanning elevation angle during actual operation. This is mainly because the antenna feed and its supporting structure are subjected to different gravitational loads at different elevation angles, resulting in micron-level mechanical deformation, which in turn changes the polarization characteristics of the antenna. This dynamic degradation may be particularly significant in a certain elevation angle range (such as between 0.5° and 2.0°), forming an "isolation concave point".

[0004] For example, application number CN202411782937.9 discloses an S-band dual-polarization radar precipitation estimation method based on historical optimal coefficients. By estimating precipitation relationships by referring to dual polarization quantities, it proposes to add observational information differential reflectivity ZDR and differential phase shift rate KDP to a three-dimensional reflectivity Z for mosaicking. By splitting the processing unit, the processing range is narrowed, improving the accuracy of radar precipitation estimation. In the processing unit, a scheme for calculating combined reflectivity feature codes and matching combined reflectivity feature codes with fitting coefficients is used to match historical fitting relationships. According to the processing unit at different geographical locations, the optimal fitting coefficient of the corresponding processing unit is quickly obtained and adjusted through a deviation algorithm to improve the speed of radar precipitation estimation in real-time operations. However, it lacks consideration of the problems of differential reflectivity measurement deviation and co-polarization correlation coefficient distortion, resulting in the observation information not eliminating deviations and affecting the judgment of precipitation type and rainfall intensity estimation.

[0005] The existing technology has the following shortcomings: The cross-polarization isolation value of the antenna is usually fixed in the laboratory or at a specific elevation angle. However, the dynamic changes in actual scanning are ignored. The effect of this dynamic degradation is not a simple increase in noise. Instead, at a specific azimuth and elevation angle, it causes systematic distortion in the differential reflectivity (ZDR) and co-polarization correlation coefficient (ρhv) measurements of non-spherical particles such as hail and melt layer particles. Especially in strong convective storms, low isolation will cause the ZDR measurement of the target to deviate significantly from the true value, which will lead to misjudgment of precipitation phase and subsequent incorrect selection of rainfall intensity estimation algorithm. In addition, low isolation will cause the ρhv value to drop abnormally, which may be misjudged as a large number of non-meteorological clutter or biological particles in the radar beam. This will cause the quality control algorithm to erroneously reject effective precipitation echoes, resulting in missed precipitation estimates.

[0006] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] The purpose of this invention is to provide a dual-polarization radar multi-parameter fusion precipitation estimation system and method. This invention solves the problems in the background art by constructing an antenna isolation elevation angle dependent model and dynamically reverse correcting the ZDR and ρhv measurements at different elevation angles.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a dual-polarization radar multi-parameter fusion precipitation estimation system, including a data collection module: when a uniformly distributed light rain weather observation scenario occurs, the dual-polarization radar is activated for long-term observation, and the radar polarization parameters at different elevation angles are recorded and transmitted to the deviation analysis module.

[0009] Deviation Analysis Module: Receives radar polarization parameters, calculates ZDR statistical characteristics based on a reference uniform scene, presets a ZDR threshold range based on the theoretical value of 0dB, triggers anomaly detection when the measured ZDR value exceeds the set ZDR threshold range, calculates the deviation data between the actual ZDR measurement value exceeding the set ZDR threshold range and the theoretical value of 0dB at different elevation angles, combines ρhv analysis to determine the abnormal elevation angle, and transmits it to the compensation model construction module.

[0010] Compensation model construction module: Combining the analysis results of deviation and abnormal elevation angle, the deformation-elevation angle relationship curve under the antenna's own weight is obtained through finite element analysis. Then, data fitting and regression analysis techniques are used to establish an isolation compensation model for the elevation angle. Taking the scanning elevation angle as input, the corresponding isolation compensation values ​​of ZDR and ρhv are predicted and output, and transmitted to the real-time correction module.

[0011] Real-time correction module: Receives real-time radar polarization parameters and isolation compensation model, designs real-time correction algorithm, performs reverse correction on the real-time measured ZDR and ρhv values ​​based on the isolation compensation value obtained from the isolation compensation model according to the current scanning elevation angle, and verifies the rationality of the corrected values ​​and evaluates the quality of the values ​​to obtain corrected, high-quality radar polarization parameters.

[0012] Multi-parameter fusion precipitation estimation module: It has a built-in multi-parameter fusion precipitation estimation algorithm, which identifies precipitation type and estimates rainfall intensity based on the finally obtained corrected, high-quality radar polarization parameters.

[0013] Optionally, the steps for recording the radar polarization parameters are as follows:

[0014] A1. When meteorological staff identify a “widespread, evenly distributed light rain” observation scene at a selected meteorological monitoring station based on weather forecast information or real-time radar panoramic scanning, they activate the dual-polarization radar to conduct long-term observations according to the preset volume scan mode, and collect radar base data, including complex signals of the horizontal and vertical channels on each range database.

[0015] A2. Stable base data is generated by smoothing the radar base data of multiple consecutive transmitted pulses by performing coherent integration and incoherent averaging in the time dimension.

[0016] A3. Based on the electromagnetic wave scattering and polarization theory, radar polarization parameters are calculated on the stable base data to obtain the reflectivity factor Zh, differential reflectivity ZDR, co-polarization correlation coefficient ρhv, and the corresponding scanning elevation angle information θ.

[0017] A4. Based on long-term observations by dual-polarization radar, a "uniform reference target area" that meets the spatial conditions is selected through preset physical criteria, including the following steps:

[0018] B401. At each scanning elevation angle θ, perform gridding or partitioning analysis on the data within the radar scanning range to generate several analysis areas;

[0019] B402. For each analysis region, calculate the spatial gradient of the reflectivity factor Zh for the corresponding analysis region, and select regions where the spatial gradient value of Zh is less than the preset Zh threshold.

[0020] B403. Based on the above-mentioned same analysis region, it is further required that the average copolarization correlation coefficient ρhv is greater than the first preset threshold.

[0021] B404. At the same time, it is required that the average reflectance factor Zh of the analysis area be within the preset range corresponding to light rain weather.

[0022] B405. A spatial region is marked as a valid "uniform reference target area" only when it meets all three of the above conditions.

[0023] A5. For each identified uniform reference target region, perform the following steps:

[0024] B501. Extract the original ZDR measurement values ​​of all distance library sampling points within the corresponding analysis area;

[0025] B501. Calculate the statistical mean and statistical variance of these ZDR measurements;

[0026] B501. Package the scanning elevation angle θ, the statistical mean and statistical variance obtained from the ZDR measurement, and the average ρhv value of the corresponding analysis area into a standardized data record;

[0027] A6. Record all data of the identified uniform reference target areas, along with the corresponding timestamps to form a data packet of radar polarization parameters, and store it in a long-term database.

[0028] Optionally, the calculation steps for the ZDR statistical features are as follows:

[0029] Based on the data packets of the pre-screened "uniform reference target area" transmitted from the data collection module, all data are grouped according to the scanning elevation angle θ.

[0030] For all data packets at each scanning elevation angle θ, the average ρhv and Zh values ​​of the region associated with the scanning elevation angle θ are effectively verified to obtain all valid "uniform reference target area" data packets;

[0031] For each scanning elevation angle θ, the ZDR sampling points in all valid "uniform reference target area" data packets are collected to form a ZDR sample set under the corresponding scanning elevation angle θ, and the statistical characteristics of the arithmetic mean and variance of the ZDR sample set are calculated.

[0032] Optionally, the triggering steps for the anomaly detection are as follows:

[0033] Based on the theoretical value of 0dB, a ZDR threshold range is preset, where the preset value of the ZDR threshold range is the theoretical value of 0dB ± 0.1dB. Therefore, ZDR... th = [-0.1dB, +0.1dB], where ZDR th This represents the preset ZDR threshold range;

[0034] If the arithmetic mean of the measured ZDR values ​​at a certain scanning elevation angle θ falls within the preset ZDR threshold range, the corresponding scanning elevation angle is considered normal and no further analysis is required. Otherwise, if the measured ZDR value at a certain scanning elevation angle θ exceeds the preset ZDR threshold range, anomaly detection is triggered.

[0035] For the scanning elevation angle θ that triggers anomaly detection, the mean ZDR data of the scanning elevation angle θ that triggers anomaly detection during the historical normal period is retrieved from the long-term database, and the historical long-term arithmetic mean and standard deviation are calculated as the historical statistical benchmark of the scanning elevation angle θ that triggers anomaly detection under normal conditions.

[0036] A one-sample T-test is performed between the ZDR mean calculated within the current observation period and the arithmetic mean in the historical statistical benchmark to calculate the T-statistic value.

[0037] Based on the T-statistic, a corresponding probability value is obtained by looking up the T-distribution table with the degrees of freedom. A significance level is preset and compared with the corresponding probability value. If the corresponding probability value is less than the preset significance level, the null hypothesis is rejected and the ZDR deviation of the current scanning elevation angle θ is considered statistically significant. At this time, the abnormal state of the scanning elevation angle is officially "triggered".

[0038] Optionally, the steps for obtaining the deviation data and abnormal elevation angle are as follows:

[0039] For each scan elevation angle θ that triggers anomaly detection, calculate the absolute deviation between the actual measured mean ZDR and the theoretical value of 0dB at the corresponding scan elevation angle θ.

[0040] The average ρhv value corresponding to the scanning elevation angle θ that triggers anomaly detection is retrieved within the current observation period. A threshold is preset based on the historical normal benchmark value of ρhv, and compared with the average ρhv value. When the average ρhv value corresponding to the scanning elevation angle that triggers anomaly detection is lower than the preset threshold, it is considered that the ρhv of the current scanning elevation angle θ has decreased significantly.

[0041] Only when the scanning elevation angle θ that triggers anomaly detection simultaneously meets both the conditions of "statistically significant ZDR deviation" and "significant synchronous decrease of ρhv" will the scanning elevation angle θ that triggers anomaly detection be determined as a reliable abnormal elevation angle.

[0042] Optionally, the steps for establishing the isolation compensation model are as follows:

[0043] The analysis results of the receiving deviation and abnormal elevation angle include all the identified abnormal elevation angles, as well as the ZDR and ρhv deviations of the corresponding abnormal elevation angles, and the deformation of the corresponding abnormal elevation angle is obtained from the deformation-elevation angle relationship curve as physical prior knowledge.

[0044] The deformation of the abnormal elevation angle and the corresponding abnormal elevation angle constitute the input feature vector of the isolation compensation model, and the ZDR and ρhv deviation of the corresponding abnormal elevation angle constitute the output feature vector of the isolation compensation model.

[0045] An isolation compensation model is constructed using Gaussian process regression (GPR) algorithm. The input feature vector is fed into the GPR model for training and learning to obtain the output feature vector. Given the data and kernel function, the posterior distribution of the function is solved, and the hyperparameters of the kernel function are optimized by maximizing the marginal likelihood function.

[0046] After training, a complete GPR model is obtained. At this point, an arbitrary real-time scanning elevation angle can be input to obtain the deformation of the corresponding scanning elevation angle. Based on the new input feature vector formed by the new deformation, the GPR model is used for prediction, and the best estimated ZDR and ρhv isolation compensation value corresponding to the input real-time scanning elevation angle are output.

[0047] Optionally, the design steps of the real-time correction algorithm are as follows:

[0048] At the start of the radar volume scan cycle, the trained isolation compensation model is loaded and initialized from the storage medium, while the physical reasonable ranges of ZDR and ρhv are preset, as well as the flag bits for quality control.

[0049] Real-time monitoring and parsing of radar base data; for each radar sampling point, extracting the current scanning elevation angle and the original polarization parameter values ​​of ZDR and ρhv measured in real time from the data packet of radar polarization parameters;

[0050] Call the pre-trained isolation compensation model, input the extracted current scanning elevation angle and perform forward prediction, and output the optimal ZDR and ρhv isolation compensation values ​​corresponding to the current scanning elevation angle;

[0051] The isolation compensation based on the isolation compensation model output is applied to the original ZDR and ρhv measurements to perform reverse correction operations, resulting in the preliminary corrected polarization parameter values ​​of ZDR and ρhv.

[0052] Optionally, the steps for obtaining the corrected, high-quality radar polarization parameters are as follows:

[0053] Perform a boundary check on the physical rationality of the pre-corrected polarization parameter values ​​to determine whether the pre-corrected polarization parameter values ​​fall within the preset physical rationality range;

[0054] Based on the rationality test results, a quality flag bit is assigned to each parameter of each radar sampling point. When both parameters pass the rationality test, they are marked as high-quality data, and the preliminary correction value is output as the final value.

[0055] If any parameter exceeds the reasonable range, it is marked as data that needs to be processed, and the backup correction strategy is activated. The backup correction strategy includes elevation angle interpolation, neighborhood substitution, or conservative backoff.

[0056] The final correction value of each radar sampling point is packaged with the corresponding quality flag to form a standardized, high-quality radar polarization parameter data packet containing quality information.

[0057] A dual-polarization radar multi-parameter fusion precipitation estimation method includes the following steps: S1, select a meteorological monitoring station and dynamically identify uniformly distributed light rain weather as the observation scenario of the natural calibration field, monitor radar echo data at different elevation angles in real time during the volume scan of the dual-polarization radar, and record radar polarization parameters at different elevation angles through long-term observation, including differential reflectivity ZDR, co-polarization correlation coefficient ρhv and corresponding scanning elevation angle information;

[0058] S2. Based on the theoretical value of 0dB, a ZDR threshold range is preset. When the measured ZDR value exceeds the preset ZDR threshold range, an anomaly detection mechanism is triggered. The deviation analysis of the actual ZDR measurement value exceeding the preset ZDR threshold range at different elevation angles and the theoretical value of 0dB is performed to generate ZDR deviation data at different elevation angles. At the same time, the ρhv at different elevation angles is compared with the preset threshold to determine and mark the abnormal elevation angle.

[0059] S3. Based on the deviation data and abnormal elevation angle after the anomaly is triggered, the physical prior knowledge of the radar antenna structure is integrated, and data fitting and regression analysis techniques are used to construct an isolation compensation model for the elevation angle. The output isolation compensation value is predicted by inputting the scanning elevation angle.

[0060] S4. Design a real-time correction algorithm. Based on the current scanning elevation angle, obtain the isolation compensation value from the isolation compensation model, perform reverse correction on the real-time measured ZDR and ρhv values, and monitor and evaluate the quality of the corrected results in real time to continuously optimize the real-time correction algorithm. At the same time, input the finally corrected and high-quality radar polarization parameters into the multi-parameter fusion precipitation estimation algorithm for accurate precipitation type identification and rainfall intensity estimation.

[0061] Optionally, the processing steps of the multi-parameter fusion precipitation estimation algorithm are as follows:

[0062] Based on the final obtained high-quality radar polarization parameters data packet, including reflectivity factor Zh, differential reflectivity ZDR, co-polarization correlation coefficient ρhv, scanning elevation angle information θ, and quality flag bit for each data point, the data is filtered for high quality according to the quality flag bit.

[0063] Introducing the cross-correlation coefficient between the differential phase KDP and zero hysteresis;

[0064] Fuzzy logic technology is used to classify precipitation particles to identify precipitation types, and an adaptive rainfall intensity inversion algorithm is combined to calculate instantaneous rainfall intensity.

[0065] The estimated instantaneous rainfall intensity value is integrated along the radar ray path and then interpolated in two-dimensional / three-dimensional grids on the Geographic Information System (GIS) platform to generate the final rainfall intensity distribution map or hourly cumulative precipitation map covering the radar detection range.

[0066] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0067] This invention constructs an antenna isolation elevation angle-dependent model to dynamically reverse-correct ZDR and ρhv measurements at different elevation angles. This effectively eliminates the impact of dynamic degradation of antenna cross-polarization isolation with scanning elevation angle on the measured values, achieving real-time sensing and compensation for isolation degradation that dynamically changes with elevation angle. This ensures the radar maintains high measurement accuracy throughout the scanning space, reducing misjudgments and missed detections of precipitation phases, thereby improving the accuracy of precipitation estimation. By utilizing natural uniform precipitation as a "natural calibration source," cost-effective and uninterrupted online calibration is achieved without expensive additional hardware or interruption of normal radar observation. Furthermore, the use of a joint criterion of ZDR and ρhv to trigger anomalies effectively avoids weather interference that may affect a single parameter, improving the reliability and accuracy of anomaly detection. A Gaussian process regression algorithm is introduced to construct an isolation compensation model, combined with a real-time correction algorithm. This approach considers improving the data quality of ZDR and ρhv, the efficiency and accuracy of real-time data processing, and enables rapid and accurate correction within the radar's real-time data stream. This improves the real-time performance of meteorological monitoring, the accuracy and reliability of estimation results, and ensures the system's robustness under complex weather conditions. Attached Figure Description

[0068] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0069] Figure 1 This is a block diagram of the dual-polarization radar multi-parameter fusion precipitation estimation system of the present invention.

[0070] Figure 2 This is a flowchart of the dual-polarization radar multi-parameter fusion precipitation estimation method of the present invention. Detailed Implementation

[0071] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0072] Example 1

[0073] This invention provides, for example Figure 1 The dual-polarization radar multi-parameter fusion precipitation estimation system shown includes a data collection module: when a uniformly distributed light rain weather observation scenario occurs, the dual-polarization radar is activated for long-term observation, and the differential reflectivity ZDR, co-polarization correlation coefficient ρhv and radar polarization parameters with corresponding scanning elevation angle information at different elevation angles are recorded and transmitted to the deviation analysis module.

[0074] Specifically, the steps for recording radar polarization parameters are as follows:

[0075] A1. When meteorological staff identify a “widespread, evenly distributed light rain” observation scene at a selected meteorological monitoring station based on weather forecast information or real-time radar panoramic scanning, they activate the dual-polarization radar to conduct long-term observations according to the preset volume scan mode, and collect radar base data, including complex signals of the horizontal and vertical channels on each range database.

[0076] A2. By smoothing the radar base data of multiple consecutive transmitted pulses by coherent integration and incoherent averaging in the time dimension, stable base data is generated to improve the data signal-to-noise ratio and suppress random fluctuations.

[0077] A3. Based on electromagnetic wave scattering and polarization theory, radar polarization parameters are calculated from the stable base data to obtain the reflectivity factor Zh, differential reflectivity ZDR, co-polarization correlation coefficient ρhv, and the corresponding scanning elevation angle information θ. The formula for calculating the reflectivity factor Zh is: In the formula, Zh represents the horizontal polarization reflectivity factor in dBZ, C0 represents a comprehensive calibration constant including radar constant and range correction, and S hh Expressed as the backscattering coefficient for horizontal transmission and horizontal reception, It represents the average of the squares of the backscattering coefficient modes for horizontal transmission and horizontal reception, refers to the average echo power of horizontal polarization, and ε0 represents the dielectric constant factor related to the complex refractive index of water.

[0078] The formula for calculating differential reflectance (ZDR) is as follows: In the formula, ZDR represents differential reflectivity, with units of dB, which reflects the difference in scattering intensity of the target in the horizontal and vertical directions. This is expressed as the average echo power transmitted vertically and received vertically, and for a spherical raindrop, then... At this point, ZDR ≈ 0 dB;

[0079] The formula for calculating the copolarization correlation coefficient ρhv is: In the formula, ρhv represents the copolarization correlation coefficient, and the value of ρhv ranges from 0 to 1. Represented as the modulus of the cross-correlation value. This is expressed as the cross-correlation value between the horizontal and vertical channel signals. Represented as S vv The complex conjugate of , where, in a uniform light rain region, the raindrop particles have the same shape and phase, and the two signals are highly correlated, at which point ρhv≈1;

[0080] A4. Based on long-term observations by dual-polarization radar, a "uniform reference target area" that meets the spatial conditions is selected through preset physical criteria, including the following steps:

[0081] B401. At each specific scanning elevation angle θ, perform gridding or partitioning analysis on the data within the radar scanning range to generate several analysis areas;

[0082] B402. For each analysis region, calculate the spatial gradient of the reflectivity factor Zh for the corresponding analysis region, and select regions where the spatial gradient value of Zh is less than the preset Zh threshold. The preset Zh threshold indicates that the echo intensity distribution within the analysis region is uniform.

[0083] B403. Based on the same analysis area mentioned above, it is further required that the average co-polarization correlation coefficient ρhv is greater than the first preset threshold. The first preset threshold indicates that the meteorological particle properties are uniform and consistent within the analysis area, which is used to ensure that there is no non-meteorological clutter interference within the analysis area. The value of the first preset threshold is 0.98.

[0084] B404. At the same time, it is required that the average reflectance factor Zh of the analysis area be within the preset range corresponding to light rain weather, wherein the preset range for light rain weather is within 15dBZ-30dBZ.

[0085] B405. A spatial region is marked as a valid "uniform reference target area" only when it meets all three of the above conditions.

[0086] A5. For each identified uniform reference target region, perform the following steps:

[0087] B501. Extract the original ZDR measurement values ​​of all distance library sampling points within the corresponding analysis area;

[0088] B501. Calculate the statistical mean and statistical variance of these ZDR measurements, where the mean is used to estimate the systematic bias and the variance is used to assess the reliability of the measurement at the corresponding scanning elevation angle.

[0089] B501. Package the scanning elevation angle θ, the statistical mean and statistical variance obtained from the ZDR measurement, and the average ρhv value of the corresponding analysis area into a standardized data record;

[0090] A6. Record all data of the identified uniform reference target areas, along with the corresponding timestamps to form a data packet of radar polarization parameters, and store it in a long-term database.

[0091] Deviation Analysis Module: Receives radar polarization parameters, calculates ZDR statistical characteristics based on a reference uniform scene, presets a ZDR threshold range based on the theoretical value of 0dB, triggers anomaly detection when the measured ZDR value exceeds the set ZDR threshold range, calculates the deviation data between the actual ZDR measurement value exceeding the set ZDR threshold range and the theoretical value of 0dB at different elevation angles, combines ρhv analysis to determine the abnormal elevation angle, and transmits it to the compensation model construction module.

[0092] Specifically, the calculation steps for ZDR statistical characteristics are as follows:

[0093] Based on the data packets of the pre-screened "uniform reference target area" transmitted from the data collection module, all data are grouped according to the scanning elevation angle θ to ensure that the data in the same group originate from the same elevation angle of the radar.

[0094] For all data packets at each specific scanning elevation angle θ, the average ρhv and Zh values ​​of the region associated with the scanning elevation angle θ are effectively verified to ensure that they still meet the physical definition of "uniformly distributed light rain" and to remove data that may have failed during transmission or storage, so as to obtain all valid "uniform reference target area" data packets.

[0095] For each specific scanning elevation angle θ, the ZDR sampling points in all valid "uniform reference target area" data packets are aggregated to form a ZDR sample set at the corresponding scanning elevation angle θ. The statistical characteristics of the arithmetic mean and variance of the ZDR sample set are calculated. The formula for calculating the arithmetic mean is as follows: In the formula, μ ZDR (θ) represents the statistical mean of all ZDR sampling points at the scanning elevation angle θ, i.e., the ZDR deviation estimate, representing the central tendency of the systematic deviation of the ZDR measurement value relative to the theoretical value of 0 dB at that scanning elevation angle θ. N represents the total number of ZDR sampling points in all effective "uniform reference target areas" at the scanning elevation angle θ. n(θ) represents the measured value of the nth ZDR sampling point at the scanning elevation angle θ, where θ represents the specific radar scanning elevation angle;

[0096] The formula for calculating variance is: In the formula, It is expressed as the statistical variance of all ZDR sampling points at the scanning elevation angle θ, representing the dispersion or fluctuation range of the ZDR measurement values ​​at that scanning elevation angle θ.

[0097] Specifically, the steps for triggering anomaly detection are as follows:

[0098] Based on the theoretical value of 0dB, a ZDR threshold range is preset, where the preset value of the ZDR threshold range is the theoretical value of 0dB ± 0.1dB. Therefore, ZDR... th = [-0.1dB, +0.1dB], where ZDR th This represents the preset ZDR threshold range;

[0099] If the arithmetic mean of the measured ZDR values ​​at a certain scanning elevation angle θ falls within the preset ZDR threshold range, the corresponding scanning elevation angle is considered normal and no further analysis is required. Otherwise, if the measured ZDR value at a certain scanning elevation angle θ exceeds the preset ZDR threshold range, anomaly detection is triggered.

[0100] For the scanning elevation angle θ that triggers anomaly detection, the mean ZDR data of the scanning elevation angle θ that triggers anomaly detection during the historical normal period is retrieved from the long-term database, and the historical long-term arithmetic mean and standard deviation are calculated as the historical statistical benchmark of the scanning elevation angle θ that triggers anomaly detection under normal conditions.

[0101] A one-sample t-test is performed between the ZDR mean calculated for the current observation period and the arithmetic mean of the historical statistical benchmark to calculate the t-statistic. The formula for calculating the t-statistic is as follows: In the formula, This is expressed as the T-statistic value. This represents the long-term average of the ZDR mean during historical normal periods, expressed as the scanning elevation angle θ that triggers anomaly detection. It represents the long-term standard deviation of the ZDR mean during historical normal periods, where the scanning elevation angle θ that triggers anomaly detection is represented. m represents the sample size used to calculate the ZDR mean in the current observation period, i.e., the number of target areas, rather than the total number of ZDR sampling points.

[0102] Based on the T-statistic, a corresponding probability value is obtained by looking up the T-distribution table with the degrees of freedom. A significance level is preset and compared with the corresponding probability value. If the corresponding probability value is less than the preset significance level, the null hypothesis is rejected, and it is considered that the ZDR deviation of the current scanning elevation angle θ is statistically significant and not caused by random fluctuations. At this time, the abnormal state of the scanning elevation angle is officially "triggered". The null hypothesis is that "the current ZDR mean is not significantly different from the arithmetic mean in the historical statistical benchmark".

[0103] The preset significance level is usually set to 0.01, which represents the risk tolerance for false alarms;

[0104] The expression for statistical significance of ZDR bias is p≤α, and α=0.01, where p represents the probability value obtained from the T-distribution table, and α represents the preset significance level.

[0105] Specifically, the steps for obtaining deviation data and abnormal elevation angles are as follows:

[0106] For each scan elevation angle θ that triggers anomaly detection, calculate the absolute deviation between the measured mean ZDR at the corresponding scan elevation angle θ and the theoretical value of 0 dB. The formula for calculating the absolute deviation is Δ. ZDR (θ)=|μ ZDR (θ)-0dB|, where, Δ ZDR (θ) represents the absolute deviation data of the ZDR measurement at the abnormal scanning elevation angle θ;

[0107] The average ρhv value within the current observation period is retrieved corresponding to the scanning elevation angle θ that triggers anomaly detection. A threshold is preset based on the historical normal benchmark value of ρhv, and compared with the average ρhv value. When the average ρhv value within the current observation period corresponding to the scanning elevation angle that triggers anomaly detection is lower than the preset threshold, it is considered that the ρhv of the current scanning elevation angle θ has decreased significantly. The expression for a significant decrease in ρhv is ρhv history (θ)-ρhv(θ)>δ ρhv , and δ ρhv =0.03, where ρhv history ρhv(θ) represents the average ρhv value of the scan elevation angle θ during historical normal periods, and ρhv(θ) represents the average co-polarization correlation coefficient of the current observation period at the abnormal scan elevation angle θ. ρhv This represents the preset threshold for determining the decrease in ρhv;

[0108] Only when the scanning elevation angle θ that triggers anomaly detection simultaneously meets both the conditions of "statistically significant ZDR deviation" and "significant synchronous decrease of ρhv" will the scanning elevation angle θ that triggers anomaly detection be determined as a reliable abnormal elevation angle.

[0109] Compensation model construction module: Combining the analysis results of deviation and abnormal elevation angle, the deformation-elevation angle relationship curve under the antenna's own weight is obtained through finite element analysis. Then, data fitting and regression analysis techniques are used to establish an isolation compensation model for the elevation angle. Taking the scanning elevation angle as input, the corresponding isolation compensation values ​​of ZDR and ρhv are predicted and output, and transmitted to the real-time correction module.

[0110] Specifically, the steps for obtaining the deformation-elevation angle relationship curve are as follows:

[0111] Based on the actual engineering design drawings of the dual-polarization radar antenna, an accurate three-dimensional geometric model is established in the finite element analysis software, including the main reflector, sub-reflector, feed source and support structure. In the three-dimensional geometric model, each component is given real material properties such as elastic modulus, Poisson's ratio and density.

[0112] The three-dimensional geometric model is finely meshed to ensure that the mesh is dense enough in key stress areas such as the feed and support arm to capture small deformations. Boundary conditions are set, and the connection surface between the antenna base and the tower is set as a fixed constraint to simulate the state in actual installation.

[0113] Gravity load is applied to the three-dimensional geometric model, and the direction varies with the antenna elevation angle. Discrete elevation angle conditions are defined from 0° to 90° in 5° intervals. For each condition, the direction of the gravity vector in the coordinate system of the three-dimensional geometric model is calculated and used as the load condition for static analysis.

[0114] For each elevation angle condition, run the static analysis solver to solve the equilibrium equations of the entire antenna structure under gravity load and calculate the displacement vector of each node;

[0115] Analyze the solution results and extract the deformation of the feed phase center node, especially the displacement component perpendicular to the reflector, because it directly affects the radiation field distribution and polarization purity of the electromagnetic wave. Summarize the displacement of the feed phase center for each elevation angle condition and construct a continuous deformation-elevation angle relationship curve by interpolation.

[0116] Specifically, the steps for establishing the isolation compensation model are as follows:

[0117] The analysis results of the receiving deviation and abnormal elevation angle include all the identified abnormal elevation angles, as well as the ZDR and ρhv deviations of the corresponding abnormal elevation angles, and the deformation of the corresponding abnormal elevation angle is obtained from the deformation-elevation angle relationship curve as physical prior knowledge.

[0118] The deformation of the abnormal elevation angle and the corresponding abnormal elevation angle constitute the input feature vector of the isolation compensation model, and the ZDR and ρhv deviation of the corresponding abnormal elevation angle constitute the output feature vector of the isolation compensation model. The expression for the input feature vector is as follows: In the formula, Let θ be the i-th input feature vector. i Let ζ(θ) be the i-th abnormal elevation angle. i Let ) represent the deformation of the i-th abnormal elevation angle;

[0119] The expression for the output feature vector is: And Δ ρhv (θ)=ρhv(θ)-δ ρhv In the formula, Let Δ be the i-th output feature vector. ZDR (θ i ) represents the absolute deviation data of the ZDR measurement value at the i-th abnormal elevation angle, Δ ρhv (θ i ) represents the deviation value of the average co-polarization correlation coefficient observed at the i-th anomalous elevation angle, Δ ρhv (θ) represents the deviation of the average copolarization correlation coefficient for the current observation period at the abnormal scan elevation angle θ;

[0120] An isolation compensation model is constructed using the Gaussian Process Regression (GPR) algorithm. The input feature vector is fed into the GPR model for training, yielding the output feature vector. Given the data and kernel function, the posterior distribution of the function is calculated, and the hyperparameters of the kernel function are optimized by maximizing the marginal likelihood function. The prior definition of the GPR model is as follows: And K In the formula, GPR() represents the relationship between the isolation compensation value and the input feature vector in the GPR model, where GPR() stands for Gaussian process function. It is represented as the mean function of the input feature vector. It is represented as a kernel function for two input feature vectors. Let denot be the signal variance representing the range of variation of the control function output, exp() represent the natural exponential function, T represent the transpose, and Λ represent the diagonal matrix representing the range of influence of the control input eigenvectors on the output. ζ is represented as the noise variance in the observed data. ij It is represented as the shape variable of two input feature vectors;

[0121] After training, a complete GPR model is obtained. At this point, inputting any real-time scanning elevation angle yields the corresponding deformation variables. Based on the new input feature vector formed by these deformation variables, the GPR model performs predictions, outputting the optimal estimated ZDR and ρhv isolation compensation value for the corresponding input real-time scanning elevation angle. The posterior prediction distribution expression of the GPR model is as follows: and In the formula, This is represented as a new input feature vector The predicted compensation value, Represented as the new input feature vector Let Q be the covariance vector among all training data points, and I be the identity matrix. This is expressed as prediction variance, used to measure the uncertainty of predicted values. It is represented as a kernel function for two new input feature vectors.

[0122] Real-time correction module: Receives real-time radar polarization parameters and isolation compensation model, designs real-time correction algorithm, performs reverse correction on the real-time measured ZDR and ρhv values ​​based on the isolation compensation value obtained from the isolation compensation model according to the current scanning elevation angle, and verifies the rationality of the corrected values ​​and evaluates the quality of the values ​​to obtain corrected, high-quality radar polarization parameters.

[0123] Specifically, the design steps of the real-time correction algorithm are as follows:

[0124] At the start of the radar volume scan cycle, the trained isolation compensation model is loaded and initialized from the storage medium, while the physical reasonable ranges of ZDR and ρhv are preset, as well as the flag bits for quality control.

[0125] Real-time monitoring and parsing of radar base data; for each radar sampling point, extracting the current scanning elevation angle and the original polarization parameter values ​​of ZDR and ρhv measured in real time from the data packet of radar polarization parameters;

[0126] The pre-trained isolation compensation model is invoked, the extracted current scan elevation angle is input, and forward prediction is performed. The output is the optimal ZDR and ρhv isolation compensation value corresponding to the current scan elevation angle. The expression for the isolation compensation value prediction is as follows: In the formula, This is represented as the isolation compensation model for the current real-time scanning elevation angle θ. current Predicted ZDR compensation value, This is represented as the isolation compensation model for the current real-time scanning elevation angle θ. current The predicted ρhv compensation value, θ currentThis is expressed as the current real-time scanning elevation angle of the radar antenna;

[0127] The isolation compensation output from the isolation compensation model is applied to perform a reverse correction operation on the original ZDR and ρhv measurements. Specifically, the ZDR compensation value is subtracted from the original ZDR measurements, and the original ρhv measurements are additively compensated to obtain the polarization parameter values ​​of ZDR and ρhv after preliminary correction. The expression for the reverse correction operation is as follows: and In the formula, ZDR′ current This represents the ZDR parameter values ​​after preliminary inverse correction calculations. raw Represented as the uncorrected raw ZDR measurement extracted from real-time radar base data, ρhv′ current This represents the ρhv parameter value after preliminary inverse correction calculation. raw This is represented as the uncorrected raw ρhv measurement extracted from real-time radar base data.

[0128] Specifically, the steps for obtaining the corrected, high-quality radar polarization parameters are as follows:

[0129] A boundary check for the physical rationality of the initially corrected polarization parameter values ​​is performed to determine whether the initially corrected polarization parameter values ​​fall within a preset physical rationality range. The expression for the physical rationality boundary check is ZDR′. current ∈[ZDR min ZDR max ], and ρhv′ current ∈[ρhv min ,ρhv max In the formula, [ZDR] min ZDR max ] represents the physical reasonable range of the preset ZDR, [ρhv min ,ρhv max ] represents the preset physical reasonable range of ρhv;

[0130] Based on the rationality test results, a quality flag bit is assigned to each parameter of each radar sampling point. When both parameters pass the rationality test, they are marked as high-quality data, and the preliminary correction value is output as the final value.

[0131] If any parameter exceeds the reasonable range, it is marked as data that needs to be processed, and the backup correction strategy is activated. The backup correction strategy includes elevation angle interpolation, neighborhood substitution, or conservative backoff. Among them, elevation angle interpolation is to use the compensation value of the elevation angle adjacent to the current elevation angle that has passed the reasonableness test to perform linear interpolation, obtain a new compensation value, and recalculate the correction value.

[0132] Neighborhood substitution directly uses the correction values ​​of spatially adjacent sampling points labeled as high-quality data at the same elevation angle for substitution;

[0133] Conservative rollback involves reverting to using the original measurements, but downgrading the quality indicators of the original measurements to low-quality data.

[0134] The final correction value of each radar sampling point is packaged with the corresponding quality flag to form a standardized, high-quality radar polarization parameter data packet containing quality information.

[0135] Multi-parameter fusion precipitation estimation module: It has a built-in multi-parameter fusion precipitation estimation algorithm, which identifies precipitation type and estimates rainfall intensity based on the finally obtained corrected, high-quality radar polarization parameters.

[0136] Example 2

[0137] This invention provides, for example Figure 2 The method for multi-parameter fusion precipitation estimation using dual-polarization radar, as shown, includes the following steps:

[0138] S1. Select a meteorological monitoring station and dynamically identify uniformly distributed light rain weather as the observation scenario of the natural calibration field. During the volume scan of the dual polarization radar, monitor the radar echo data at different elevation angles in real time. Record the radar polarization parameters at different elevation angles through long-term observation, including differential reflectivity ZDR, co-polarization correlation coefficient ρhv and corresponding scanning elevation angle information.

[0139] S2. Based on the theoretical value of 0dB, a ZDR threshold range is preset. When the measured ZDR value exceeds the preset ZDR threshold range, an anomaly detection mechanism is triggered. The deviation analysis of the actual ZDR measurement value exceeding the preset ZDR threshold range at different elevation angles and the theoretical value of 0dB is performed to generate ZDR deviation data at different elevation angles. At the same time, the ρhv at different elevation angles is compared with the preset threshold to determine and mark the abnormal elevation angle.

[0140] S3. Based on the deviation data and abnormal elevation angle after the anomaly is triggered, the physical prior knowledge of the radar antenna structure is integrated, and data fitting and regression analysis techniques are used to construct an isolation compensation model for the elevation angle. The output isolation compensation value is predicted by inputting the scanning elevation angle.

[0141] S4. Design a real-time correction algorithm. Based on the current scanning elevation angle, obtain the isolation compensation value from the isolation compensation model, perform reverse correction on the real-time measured ZDR and ρhv values, and monitor and evaluate the quality of the corrected results in real time to continuously optimize the real-time correction algorithm. At the same time, input the finally corrected and high-quality radar polarization parameters into the multi-parameter fusion precipitation estimation algorithm for accurate precipitation type identification and rainfall intensity estimation.

[0142] Specifically, the processing steps of the multi-parameter fusion precipitation estimation algorithm are as follows:

[0143] Based on the final data packet containing high-quality radar polarization parameters after correction, including reflectivity factor Zh, differential reflectivity ZDR, co-polarization correlation coefficient ρhv, scanning elevation angle information θ, and quality flag bits for each data point, the data is filtered for high-quality data according to the quality flag bits. Data points marked as low-quality data or invalid are excluded from the current precipitation estimation calculation or given extremely low weights to ensure that the estimation process is based on reliable data.

[0144] To achieve precipitation type identification and rainfall intensity estimation, the specific differential phase KPD and zero-hysteresis cross-correlation coefficient are calculated. The formula for calculating the specific differential phase KPD is as follows: In the formula, KDP represents the differential phase, and φ DP The differential propagation phase is expressed in degrees (°), and r represents the distance in kilometers (km). It is expressed as the differential propagation phase with respect to distance, and in practice, it is calculated using sliding window linear regression or smooth difference method;

[0145] Fuzzy logic technology is used to classify precipitation particles to identify precipitation types, and an adaptive rainfall intensity inversion algorithm is combined to calculate instantaneous rainfall intensity. The steps of fuzzy logic technology for classifying precipitation particles are as follows:

[0146] The clear input parameters Zh, ZDR, ρhv, and KDP are converted into membership values ​​belonging to different raindrop particle types through preset membership functions, including drizzle, rain, heavy rain, dry snow, wet snow, hail / graupel, etc., and each parameter has a membership function for each particle type.

[0147] By applying a pre-defined fuzzy rule base, all input membership values ​​are logically combined through the rule base to generate a rule strength for each particle type.

[0148] The maximum membership degree processing is performed on the rule intensity of all particle types to assign a primary precipitation type to each sampling point;

[0149] The adaptive rainfall intensity inversion algorithm dynamically selects the optimal instantaneous rainfall intensity for each sampling point. It is applicable to uniform light rain, heavy rainfall, hail, most rainfall events, and can invert drop spectrum. The expression for the adaptive rainfall intensity inversion algorithm for uniform light rain is E(Zh)=a·(Zh) b In the formula, E(Zh) represents the rainfall intensity of uniform light rain, in mm / h; a and b represent empirical coefficients adjusted according to regional and climatic characteristics; and Zh represents the horizontal reflectance factor, converted from dBZ to mm. 6 / m 3 ;

[0150] The expression for the adaptive rainfall intensity inversion algorithm applicable to heavy rainfall and hail scenarios is E(KDP) = a·(KDP). b In the formula, E(KDP) represents the intensity of heavy rainfall, including hail;

[0151] The expression for the adaptive rainfall intensity inversion algorithm, applicable to most rainfall events and capable of retrieving droplet spectra, is E(Zh,ZDR)=a·(Zh) b ·(10 0.1ZDR ) c In the formula, E(Zh,ZDR) represents the rainfall intensity of most rainfall events, which can be used to retrieve the droplet spectrum, and c represents the empirical coefficient of the droplet spectrum type.

[0152] The estimated instantaneous rainfall intensity value is integrated along the radar ray path and then interpolated in two-dimensional / three-dimensional grids on the Geographic Information System (GIS) platform to generate the final rainfall intensity distribution map or hourly cumulative precipitation map covering the radar detection range.

[0153] It should be noted that the theoretical value of 0dB is because the particles in the raindrops can be approximated as spherical. Theoretically, the differential reflectivity (ZDR) should be close to 0dB. The uniformly distributed light rain weather can effectively eliminate the interference of other factors on ZDR, making it easier to accurately analyze the change of antenna isolation with elevation angle.

[0154] The present invention provides a dual-polarization radar multi-parameter fusion precipitation estimation method, which is implemented through the aforementioned dual-polarization radar multi-parameter fusion precipitation estimation system. For details of the specific method and process of the dual-polarization radar multi-parameter fusion precipitation estimation method, please refer to the aforementioned embodiment of the dual-polarization radar multi-parameter fusion precipitation estimation system, which will not be repeated here.

[0155] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0156] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0157] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0158] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0159] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A dual-polarization radar multi-parameter fusion precipitation estimation system, characterized in that, Includes a data collection module: When a uniformly distributed light rain occurs during the observation scenario, the dual-polarization radar is activated for long-term observation, and the radar polarization parameters at different elevation angles are recorded and transmitted to the deviation analysis module. Deviation Analysis Module: Receives radar polarization parameters, calculates ZDR statistical characteristics based on a reference uniform scene, presets a ZDR threshold range based on the theoretical value of 0dB, triggers anomaly detection when the measured ZDR value exceeds the set ZDR threshold range, calculates the deviation data between the actual ZDR measurement value exceeding the set ZDR threshold range and the theoretical value of 0dB at different elevation angles, combines ρhv analysis to determine the abnormal elevation angle, and transmits it to the compensation model construction module. Compensation model construction module: Combining the analysis results of deviation and abnormal elevation angle, the deformation-elevation angle relationship curve under the antenna's own weight is obtained through finite element analysis. Then, data fitting and regression analysis techniques are used to establish an isolation compensation model for the elevation angle. Taking the scanning elevation angle as input, the corresponding isolation compensation values ​​of ZDR and ρhv are predicted and output, and transmitted to the real-time correction module. Real-time correction module: Receives real-time radar polarization parameters and isolation compensation model, designs real-time correction algorithm, performs reverse correction on the real-time measured ZDR and ρhv values ​​based on the isolation compensation value obtained from the isolation compensation model according to the current scanning elevation angle, and verifies the rationality of the corrected values ​​and evaluates the quality of the values ​​to obtain corrected, high-quality radar polarization parameters. Multi-parameter fusion precipitation estimation module: It has a built-in multi-parameter fusion precipitation estimation algorithm, which identifies precipitation type and estimates rainfall intensity based on the finally obtained corrected, high-quality radar polarization parameters.

2. The dual-polarization radar multi-parameter fusion precipitation estimation system according to claim 1, characterized in that, The steps for recording the radar polarization parameters are as follows: A1. When meteorological staff identify a "widespread, evenly distributed light rain" observation scenario at a selected meteorological monitoring station based on weather forecast information or real-time radar panoramic scanning, they activate the dual-polarization radar to conduct long-term observations according to the preset volume scan mode, collecting radar base data, including complex signals of the horizontal and vertical channels on each distance database. A2. Stable base data is generated by smoothing the radar base data of multiple consecutive transmitted pulses by performing coherent integration and incoherent averaging in the time dimension. A3. Based on the electromagnetic wave scattering and polarization theory, radar polarization parameters are calculated on the stable base data to obtain the reflectivity factor Zh, differential reflectivity ZDR, co-polarization correlation coefficient ρhv, and the corresponding scanning elevation angle information θ. A4. Based on long-term observations by dual-polarization radar, a "uniform reference target area" that meets the spatial conditions is selected through preset physical criteria, including the following steps: B401. At each scanning elevation angle θ, perform gridding or partitioning analysis on the data within the radar scanning range to generate several analysis areas; B402. For each analysis region, calculate the spatial gradient of the reflectivity factor Zh for the corresponding analysis region, and select regions where the spatial gradient value of Zh is less than the preset Zh threshold. B403. Based on the above-mentioned same analysis region, it is further required that the average copolarization correlation coefficient ρhv is greater than the first preset threshold. B404. At the same time, it is required that the average reflectance factor Zh of the analysis area be within the preset range corresponding to light rain weather. B405. A spatial region is marked as a valid "uniform reference target area" only when it meets all three of the above conditions. A5. For each identified uniform reference target region, perform the following steps: B501. Extract the original ZDR measurement values ​​of all distance library sampling points within the corresponding analysis area; B501. Calculate the statistical mean and statistical variance of these ZDR measurements; B501. Package the scanning elevation angle θ, the statistical mean and statistical variance obtained from the ZDR measurement, and the average ρhv value of the corresponding analysis area into a standardized data record; A6. Record all data of the identified uniform reference target areas, along with the corresponding timestamps to form a data packet of radar polarization parameters, and store it in a long-term database.

3. The dual-polarization radar multi-parameter fusion precipitation estimation system according to claim 2, characterized in that, The calculation steps for the ZDR statistical characteristics are as follows: Based on the data packets of the pre-screened "uniform reference target area" transmitted from the data collection module, all data are grouped according to the scanning elevation angle θ. For all data packets at each scanning elevation angle θ, the average ρhv and Zh values ​​of the region associated with the scanning elevation angle θ are effectively verified to obtain all valid "uniform reference target area" data packets; For each scanning elevation angle θ, the ZDR sampling points in all valid "uniform reference target area" data packets are collected to form a ZDR sample set under the corresponding scanning elevation angle θ, and the statistical characteristics of the arithmetic mean and variance of the ZDR sample set are calculated.

4. The dual-polarization radar multi-parameter fusion precipitation estimation system according to claim 3, characterized in that, The triggering steps for the anomaly detection are as follows: Based on the theoretical value of 0dB, a ZDR threshold range is preset, where the preset value of the ZDR threshold range is the theoretical value of 0dB ± 0.1dB. Therefore, ZDR... th = [-0.1dB, +0.1dB], where ZDR th This represents the preset ZDR threshold range; If the arithmetic mean of the measured ZDR values ​​at a certain scanning elevation angle θ falls within the preset ZDR threshold range, the corresponding scanning elevation angle is considered normal and no further analysis is required. Otherwise, if the measured ZDR value at a certain scanning elevation angle θ exceeds the preset ZDR threshold range, anomaly detection is triggered. For the scanning elevation angle θ that triggers anomaly detection, the mean ZDR data of the scanning elevation angle θ that triggers anomaly detection during the historical normal period is retrieved from the long-term database, and the historical long-term arithmetic mean and standard deviation are calculated as the historical statistical benchmark of the scanning elevation angle θ that triggers anomaly detection under normal conditions. A one-sample T-test is performed between the ZDR mean calculated within the current observation period and the arithmetic mean in the historical statistical benchmark to calculate the T-statistic value. Based on the T-statistic, a corresponding probability value is obtained by looking up the T-distribution table with the degrees of freedom. A significance level is preset and compared with the corresponding probability value. If the corresponding probability value is less than the preset significance level, the null hypothesis is rejected and the ZDR deviation of the current scanning elevation angle θ is considered statistically significant. At this time, the abnormal state of the scanning elevation angle is officially "triggered".

5. A dual-polarization radar multi-parameter fusion precipitation estimation system according to claim 4, characterized in that, The steps for obtaining the deviation data and abnormal elevation angle are as follows: For each scan elevation angle θ that triggers anomaly detection, calculate the absolute deviation between the actual measured mean ZDR and the theoretical value of 0dB at the corresponding scan elevation angle θ. The average ρhv value corresponding to the scanning elevation angle θ that triggers anomaly detection is retrieved within the current observation period. A threshold is preset based on the historical normal benchmark value of ρhv, and compared with the average ρhv value. When the average ρhv value corresponding to the scanning elevation angle that triggers anomaly detection is lower than the preset threshold, it is considered that the ρhv of the current scanning elevation angle θ has decreased significantly. Only when the scanning elevation angle θ that triggers anomaly detection simultaneously meets both the conditions of "significant ZDR deviation" and "significant synchronous decrease of ρhv" will the scanning elevation angle θ that triggers anomaly detection be determined as a reliable abnormal elevation angle.

6. A dual-polarization radar multi-parameter fusion precipitation estimation system according to claim 5, characterized in that, The steps for establishing the isolation compensation model are as follows: The analysis results of the receiving deviation and abnormal elevation angle include all the identified abnormal elevation angles, as well as the ZDR and ρhv deviations of the corresponding abnormal elevation angles, and the deformation of the corresponding abnormal elevation angle is obtained from the deformation-elevation angle relationship curve as physical prior knowledge. The deformation of the abnormal elevation angle and the corresponding abnormal elevation angle constitute the input feature vector of the isolation compensation model, and the ZDR and ρhv deviation of the corresponding abnormal elevation angle constitute the output feature vector of the isolation compensation model. An isolation compensation model is constructed using Gaussian process regression (GPR) algorithm. The input feature vector is fed into the GPR model for training and learning to obtain the output feature vector. Given the data and kernel function, the posterior distribution of the function is solved, and the hyperparameters of the kernel function are optimized by maximizing the marginal likelihood function. After training, a complete GPR model is obtained. At this point, an arbitrary real-time scanning elevation angle can be input to obtain the deformation of the corresponding scanning elevation angle. Based on the new input feature vector formed by the new deformation, the GPR model is used for prediction, and the best estimated ZDR and ρhv isolation compensation value corresponding to the input real-time scanning elevation angle are output.

7. A dual-polarization radar multi-parameter fusion precipitation estimation system according to claim 6, characterized in that, The design steps of the real-time correction algorithm are as follows: At the start of the radar volume scan cycle, the trained isolation compensation model is loaded and initialized from the storage medium, while the physical reasonable ranges of ZDR and ρhv are preset, as well as the flag bits for quality control. Real-time monitoring and parsing of radar base data; for each radar sampling point, extracting the current scanning elevation angle and the original polarization parameter values ​​of ZDR and ρhv measured in real time from the data packet of radar polarization parameters; Call the pre-trained isolation compensation model, input the extracted current scan elevation angle and perform forward prediction, and output the optimal ZDR and ρhv isolation compensation values ​​corresponding to the current scan elevation angle; The isolation compensation based on the isolation compensation model output is applied to the original ZDR and ρhv measurements to perform reverse correction operations, resulting in preliminary corrected polarization parameter values ​​for ZDR and ρhv.

8. A dual-polarization radar multi-parameter fusion precipitation estimation system according to claim 7, characterized in that, The steps for obtaining the corrected, high-quality radar polarization parameters are as follows: Perform a boundary check on the physical rationality of the pre-corrected polarization parameter values ​​to determine whether the pre-corrected polarization parameter values ​​fall within the preset physical rationality range; Based on the rationality test results, a quality flag bit is assigned to each parameter of each radar sampling point. When both parameters pass the rationality test, they are marked as high-quality data, and the preliminary correction value is output as the final value. If any parameter exceeds the reasonable range, it is marked as data that needs to be processed, and the backup correction strategy is activated. The backup correction strategy includes elevation angle interpolation, neighborhood substitution, or conservative backoff. The final correction value of each radar sampling point is packaged with the corresponding quality flag to form a standardized, high-quality radar polarization parameter data packet containing quality information.

9. A method for multi-parameter fusion precipitation estimation using dual-polarization radar, implemented using a dual-polarization radar multi-parameter fusion precipitation estimation system as described in any one of claims 1-8, characterized in that... The process includes the following steps: S1. Select a meteorological monitoring station and dynamically identify evenly distributed light rain weather as the observation scenario of the natural calibration field. During the volume scan of the dual-polarization radar, monitor the radar echo data at different elevation angles in real time. Record the radar polarization parameters at different elevation angles through long-term observation, including differential reflectivity ZDR, co-polarization correlation coefficient ρhv and the corresponding scanning elevation angle information. S2. Based on the theoretical value of 0dB, a ZDR threshold range is preset. When the measured ZDR value exceeds the preset ZDR threshold range, an anomaly detection mechanism is triggered. The deviation analysis of the actual ZDR measurement value exceeding the preset ZDR threshold range at different elevation angles and the theoretical value of 0dB is performed to generate ZDR deviation data at different elevation angles. At the same time, the ρhv at different elevation angles is compared with the preset threshold to determine and mark the abnormal elevation angle. S3. Based on the deviation data and abnormal elevation angle after the anomaly is triggered, the physical prior knowledge of the radar antenna structure is integrated, and data fitting and regression analysis techniques are used to construct an isolation compensation model for the elevation angle. The output isolation compensation value is predicted by inputting the scanning elevation angle. S4. Design a real-time correction algorithm. Based on the current scanning elevation angle, obtain the isolation compensation value from the isolation compensation model, perform reverse correction on the real-time measured ZDR and ρhv values, and monitor and evaluate the quality of the corrected results in real time to continuously optimize the real-time correction algorithm. At the same time, input the finally corrected and high-quality radar polarization parameters into the multi-parameter fusion precipitation estimation algorithm for accurate precipitation type identification and rainfall intensity estimation.

10. The method for multi-parameter fusion precipitation estimation using dual-polarization radar according to claim 9, characterized in that, The processing steps of the multi-parameter fusion precipitation estimation algorithm are as follows: Based on the final obtained high-quality radar polarization parameters data packet, including reflectivity factor Zh, differential reflectivity ZDR, co-polarization correlation coefficient ρhv, scanning elevation angle information θ, and quality flag bit for each data point, the data is filtered for high quality according to the quality flag bit. Introducing the cross-correlation coefficient between the differential phase KDP and zero hysteresis; Fuzzy logic technology is used to classify precipitation particles to identify precipitation types, and an adaptive rainfall intensity inversion algorithm is combined to calculate instantaneous rainfall intensity. The estimated instantaneous rainfall intensity value is integrated along the radar ray path and then interpolated in two-dimensional / three-dimensional grids on the Geographic Information System (GIS) platform to generate the final rainfall intensity distribution map or hourly cumulative precipitation map covering the radar detection range.

Images