Method for early hail cloud recognition based on polarization radar

Abstract

The application discloses a hail cloud early identification method based on polarization radar, comprising the following steps: establishing a sample library; extracting multiple determination factors of hail cloud monomer radar data 20 minutes before hail occurrence in the sample library; establishing a logistic regression model for multiple determination factors in the sample library; selecting an optimal weight algorithm by using a cross-validation mechanism; calculating the weight of multiple determination factors; determining a hail early warning index threshold; comparing the early warning index with the hail early warning index threshold; if the early warning index is greater than or equal to the hail early warning index threshold, the early hail cloud monomer is determined; otherwise, the general thunderstorm cloud is determined. The application can effectively improve the early hail cloud identification accuracy, accurately distinguish the hail cloud from the general thunderstorm cloud, and assist weather forecasting and artificial hail prevention operation based on polarization radar data and three key determination factors.

Description

Technical Field

[0001] This invention relates to the field of early meteorological identification technology, and in particular to an early identification method for hail clouds based on polarization radar. Background Technology

[0002] Hail clouds and other thunderstorm clouds (thunderclouds or weak convective clouds) are all forms of severe convective weather. Compared to severe convective clouds, weak convective clouds are easier to identify because they develop less vigorously, generally have shorter lifespans, smaller cloud areas, weaker radar echo intensity, and rarely produce lightning or precipitation. Hail clouds and thunderclouds share many similarities in their development, making early differentiation more challenging. In the past, limitations in detection methods made it difficult to understand the microphysical changes within clouds in real time, thus hindering the early identification of hail clouds. However, with the development and use of new detection instruments such as dual-polarization weather radar, solving the problem of early hail cloud identification has become possible.

[0003] Early hail cloud development, before large hailstones have formed, can be effectively mitigated by disrupting the environmental conditions that lead to large hailstone formation. Hail clouds are often accompanied by strong lightning and winds during their development, leading to hail, localized heavy rain, and thunderstorms. Furthermore, early identification of hail clouds allows for more preparation time for hail prevention and mitigation efforts. Therefore, a polarization radar-based method for early hail cloud identification is needed. Summary of the Invention

[0004] The purpose of this invention is to provide an early identification method for hail clouds based on polarization radar, so as to solve the problems existing in the prior art.

[0005] To achieve the above objectives, the present invention is implemented according to the following technical solution:

[0006] On one hand, the present invention includes the following steps:

[0007] Step 1: Based on historical polarization radar data, screen hail events verified by ground observations and establish a sample database;

[0008] Step 2: Extract multiple decision factors from the single-cell radar data of hail clouds 20 minutes before the hail event in the sample library, and standardize the decision factors; the decision factors include the vertical integral liquid water VIL factor FVIL and the differential reflectivity column height factor F. ZDRc The difference factor F between the top height of the 45dBZ reflectivity echo and the height of the 0℃ layer. ΔH45 ;

[0009] Step 3: Establish a logistic regression model for multiple decision factors in the sample library, use cross-validation mechanism to select the optimal weight algorithm, calculate the weights of multiple decision factors, and determine the hail warning index threshold.

[0010] Step 4: Extract the determination factors from the strong convective cloud detection data and perform standardization processing. Input the processed determination factors into the logistic regression model to calculate the warning index. Compare the warning index with the hail warning index threshold. If the warning index ≥ the hail warning index threshold, it is determined to be an early hail cloud; otherwise, it is determined to be a general thunderstorm cloud.

[0011] The logistic regression model calculates the linear weighted sum of each decision factor, Fhail, and then substitutes Fhail into the sigmoid function to obtain the warning index P. 冰雹 The linear weighted sum F 冰雹 The calculation formula is:

[0012] F 冰雹 =W VIL ×F VIL +W ZDRc ×F ZDRc +W ΔH45 ×F ΔH45

[0013] In the formula W VIL For F VIL weights, W ZDRc For F ZDRc weights, W ΔH45 For F ΔH4 The weights are determined by the number of data points, and the minimum value of F_hail is 0, while the maximum value changes dynamically based on the amount of data input. The formula for calculating the sigmoid function is as follows:

[0014]

[0015] Where F 冰雹 The linear weighted sum of the decision factors is given, where e is the natural constant.

[0016] Furthermore, using 30 dBZ as the identification threshold, strong echo cells are identified to confirm cell information. The maximum and minimum vertical integral liquid water VIL values ​​within a cell are calculated using the cell information and VIL grid data, ultimately determining the VIL factor FVIL. The formula for calculating the vertical integral liquid water VIL factor FVIL is as follows:

[0017]

[0018] In the formula, VIL max The maximum value of VIL within a hail cloud cell. max L represents the minimum VIL value in the hail case sample library.max H represents the maximum VIL value in the hail case sample library.

[0019] Furthermore, the formula for calculating the differential reflectivity column height factor FZDRc is as follows:

[0020]

[0021] In the formula H ZDRc H represents the maximum height of the differential reflectivity column within a hail cloud cell, extending vertically above the 0°C layer and with a ZDR ≥ 1 dB. ZDRc L is the minimum differential reflectance column height of hail cases in the sample library, and H is... ZDRc H represents the height of the maximum differential reflectance column for hail cases in the sample library.

[0022] Furthermore, the formula for calculating the difference factor FΔH45 between the 45dBZ reflectivity echo top height and the 0℃ layer height is as follows:

[0023]

[0024] In the formula H ΔH45 H represents the difference between the top height of the 45 dBZ reflectivity echo within a single hail cloud cell and the height of the 0°C layer. ΔH45 L is the minimum 45 dBZ reflectance height difference of hail cases in the sample library, H ΔH45 H represents the maximum 45 dBZ reflectance height difference of hail cases in the sample library.

[0025] Furthermore, the standardization process is min-max standardization; and before the standardization process, it also includes a step of using the 3σ method to remove extreme outliers from the decision factor.

[0026] Furthermore, the polarization radar data needs to be gridded; the gridding process includes: (1) assigning values ​​to grid points based on the longitude, latitude, altitude and radar position of grid points in the Cartesian coordinate system, combined with the elevation angle, azimuth and slant range of radar echo data in the polar coordinate system, using the nearest neighbor assignment method, filtering out clutter during the assignment process, and determining clutter and removing it when the number of valid data in the 8 azimuths around a reflectivity data is ≤3;

[0027] (2) Using data from adjacent elevation layers in polar coordinate system, the eight-point Barnes interpolation algorithm is used for processing.

[0028] Furthermore, in step three, the optimal weighting algorithm adopts one of the following: L1 regularized logistic regression algorithm, genetic algorithm, and grid algorithm, and the algorithm with the highest AUC score is selected as the optimal weighting algorithm through the AUC score of the cross-validation mechanism.

[0029] Furthermore, in step three, the method for determining the hail warning index threshold G hail includes: extracting the warning index calculated by the judgment factor 20 minutes before the hail occurs from the sample library, and simultaneously extracting the warning index increment calculated by the judgment factor 30 minutes before the hail occurs and the previous 20 minutes, and combining the warning index and the warning index increment to divide the threshold.

[0030] Furthermore, the method for classifying the warning level corresponding to the hail warning index threshold G includes: when the warning index increment is >0.06 or the warning index is >0.6, it is a red warning, indicating that hail may occur in the next 20 minutes; when the warning index increment is between 0.05 and 0.06 or the warning index is between 0.5 and 0.6, it is an orange warning; the rest are low-risk levels.

[0031] The beneficial effects of this invention are:

[0032] This invention is an early identification method for hail clouds based on polarization radar. Compared with existing technologies, this invention has the following technical advantages:

[0033] Based on polarization radar data and three key decision factors, this invention effectively improves the accuracy of early hail cloud identification and can accurately distinguish hail clouds from general thunderstorm clouds, thus assisting in weather forecasting and artificial hail suppression operations. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the adjustment system framework for an early hail cloud identification method based on polarization radar according to the present invention.

[0035] Figure 2 The image shows a broken line plot of the VIL time-series change during a hailstorm in Area A on July 24, 2022, as an implementation example of an early hail cloud identification method based on polarization radar according to the present invention.

[0036] Figure 3 The following is an implementation example of the early identification method for hail clouds based on polarization radar of the present invention: a broken line showing the time-series change of differential reflectivity column height during a hail process in area B on June 5, 2024.

[0037] Figure 4 This is an implementation example of the early identification method for hail clouds based on polarization radar of the present invention, using differential reflectivity radar echo profiles of a hailfall process in a certain area on August 9, 2022. Detailed Implementation

[0038] The present invention will be further described below through specific embodiments. The illustrative embodiments and descriptions herein are used to explain the present invention, but are not intended to limit the present invention.

[0039] like Figure 1As shown, an early identification method for hail clouds based on polarization radar includes the following steps:

[0040] The development of hail clouds requires a sufficient supply of water vapor. To represent the water vapor content information of a three-dimensional spatial grid using two-dimensional data, the more vigorous the convection, the larger the VIL value. A large VIL value is also strongly correlated with ground-based hail. For example... Figure 2 The figure shows the time-series variation of VIL during the hailstorm in Area A on July 24, 2022.

[0041] Using individual cell information and VIL grid data, the maximum and minimum values ​​of VIL within an individual cell are calculated, and the VIL factor is finally determined. This paper defines the VIL factor as F. VIL It means using W VIL Determine the influence weight of the VIL factor in the early hail cloud identification system, and use the maximum VIL value within a single area. max As a parameter of the influence of the factor.

[0042]

[0043] VIL max L represents the minimum VIL value in the hail case sample library. max H represents the maximum VIL value in the hail case sample library.

[0044] Differential reflectivity column height H ZDRc factor

[0045] Dual-polarization radar, by emitting horizontally and vertically polarized electromagnetic waves, can obtain not only the reflectivity factor (ZH), but also dual-polarization parameters such as the differential reflectivity factor (ZDR), differential propagation phase shift rate (KDP), and correlation coefficient (CC). These parameters are closely related to the phase state, shape, spatial orientation, and distribution of precipitation particles, and can be used to identify specific microphysical and dynamic processes. Among these, the ZDR column, as one of the most prominent dual-polarization radar features of convective storms, is closely related to the dynamics and microphysical structure within the convective storm and has important indicative significance for the development and evolution of convective storms.

[0046] The core logic behind using ZDR column height as a criterion for hail cloud identification is that its height directly reflects the intensity of strong updrafts within the cloud, the vertical transport capacity of large water droplets, and whether it crosses the 0°C layer to enter the freezing zone. A higher ZDR column indicates that the cloud possesses sufficient liquid water supply, strong updraft support, and freezing conditions necessary for hail formation, thus becoming one of the important indicators for radar monitoring of hail clouds. In practical operations, meteorologists can quickly determine whether a cloud cluster has the potential for hail by analyzing the relative position of the ZDR column height and the 0°C layer, providing crucial information for hail warnings. Figure 3The figure shows the time-series variation of differential reflectance column height during the hail event in the region on June 5, 2024. Figure 4 The image shows the differential reflectivity radar echo profile of a hailstorm event in a certain area on August 9, 2022.

[0047] Using individual cell information and differential reflectivity grid data, the maximum height of the columnar differential reflectivity factor enhancement region extending vertically above the 0°C layer within the individual cell, with a ZDR ≥ 1 dB, is calculated, and the final ZDRc factor F is determined. ZDRc :

[0048]

[0049] H ZDRc L is the height of the minimum differential reflectance column in the hail sample library, and H is... ZDRc H represents the height of the maximum differential reflectance column in the hail sample library.

[0050] 45dBZ reflectivity height difference factor

[0051] Hail formation requires two core conditions: sufficient liquid water being transported to the freezing zone and strong updrafts supporting particle growth. The altitude difference, through a combination of radar observations and atmospheric thermodynamic parameters, directly quantifies these two conditions: 45 dBZ is a key threshold for radar reflectivity, typically corresponding to relatively strong precipitation particles (such as large water droplets, hail embryos, or small hailstones). Its presence indicates a sufficient concentration of large particles within the cloud, the "material basis" for hail formation. The 0°C layer is the boundary between liquid water (warm zone) and the freezing zone (cold zone) in the atmosphere; particles can only freeze and grow into hailstones when transported above the 0°C layer.

[0052] When the top of the 45 dBZ echo exceeds the 0°C layer (i.e., the height difference is positive), it means that large particles (liquid water droplets or embryos) corresponding to strong reflectivity are transported to the freezing zone by updrafts, creating conditions for freezing and growth. If the height difference is negative (the top of the 45 dBZ echo is below the 0°C layer), the particles cannot enter the freezing zone and can only form rain or graupel, unlikely to develop into hail. Through statistical analysis of a large amount of observational data, meteorologists can determine the critical height difference thresholds for different regions and seasons. When the actual height difference exceeds the threshold, the probability of hail increases significantly, providing a clear quantitative indicator for forecasting.

[0053] Using individual cell information and reflectivity grid data, the difference between the 45 dBZ reflectivity height and the 0°C layer height was calculated, and the 45 dBZ reflectivity height difference factor F was finally determined. ΔH45 .

[0054]

[0055] H ΔH45L represents the minimum 45 dBZ reflectance height difference in the hail sample library, and H represents... ΔH45 H represents the maximum 45 dBZ reflectance height difference in the hail case sample library.

[0056] After calculating the function values ​​of each of the above sensitivity factors, calculate the linear weighted sum F of each decision factor. 冰雹 The warning index P is calculated by substituting it into the sigmoid function. 冰雹 P 冰雹 Hail cloud warning index threshold G 冰雹 The comparison is performed, and those cells exceeding the threshold are identified as early hail clouds, while others are identified as general thunderstorm clouds.

[0057] F 冰雹 =W VIL ×F VIL +W ZDRc ×F ZDRc +W ΔH45 ×F ΔH45

[0058] F 冰雹 The maximum value changes dynamically depending on the amount of data input, and the minimum value is 0.

[0059]

[0060] If P 冰雹 ≥G 冰雹 If it is, then it is identified as a hail cloud; otherwise, it is another type of thunderstorm cloud.

[0061] Radar data gridding

[0062] The gridding process for radar data begins by combining the longitude, latitude, and altitude of grid points in the Cartesian coordinate system with the radar's location. Then, using the elevation, azimuth, and slant range of the radar echo data in the polar coordinate system, neighboring points are used to assign values ​​to each grid point, resulting in the detected value at that point. During this assignment process, clutter removal is performed on the base data; if fewer than three data points exist in the eight azimuth directions surrounding a reflectivity data point, it is considered a single clutter element and is filtered out to ensure the continuity of the data image. To further ensure data continuity and integrity, an eight-point Barnes interpolation algorithm is used with neighboring elevation layer data in the polar coordinate system.

[0063] Extraction of hail cloud warning index threshold

[0064] Based on historical polarization radar data, hail events verified by ground observations were screened to establish a sample library; each judgment factor of the hail cloud individual radar data 20 minutes before the hail occurred was extracted and min-max standardized; a logistic regression model was established for each judgment factor in the sample library, and the optimal weight algorithm was selected using a cross-validation mechanism to calculate the weight of each judgment factor and determine the hail warning index threshold.

[0065] Verification and Analysis of Early Hail Cloud Identification Effectiveness

[0066] In this embodiment, using a dual-polarization radar hail dataset from a certain city from 2022 to 2024 and a dual-polarization radar hail dataset from other regions from 2018 to 2023, suitable hail cases were selected, and key time points for each case were marked: the time of the first hailfall, 30 minutes before the hailfall, and 20 minutes before the hailfall. Judgment factors were extracted from the radar detection data of hail clouds at key time points, including the maximum value of vertically integrated liquid water VIL, the height of the differential reflectivity column, and the top height of the 45dBZ threshold echo. The 0°C height was determined using radiosonde data, and finally the difference between the 45dBZ top height and the 0°C height was determined.

[0067] The maximum vertical integral liquid water volume (VIL), differential reflectivity column height, and 45 dBZ echo top height difference were min-max standardized for the values ​​30 minutes and 20 minutes before hailfall. Extreme outliers were then removed using the 3σ method to create a training sample library for the logistic regression model for hail cloud determination.

[0068] The standardized formula for this example is as follows (based on data from the most recent 20 minutes):

[0069] F VIL =(VIL) max –6.48) / (45.42-6.48)

[0070] F ZDRc =(H ZDRc -0.00) / (11.68-0.00)

[0071] F ΔH45 =(H ΔH45 --1.47) / (6.82--1.47)

[0072] The logistic regression coefficient method was used to calculate the weights of the decision factors extracted from the sample database 20 minutes before hail. The optimal weight optimization algorithm was selected based on the AUC score of the cross-validation mechanism. This example used L1 regularized logistic regression, genetic algorithm, and grid algorithm. The following are the decision factor weights and their AUC scores calculated by the three algorithms:

[0073] L1 regularized logistic regression algorithm (AUC: 0.7882)

[0074] WVIL : 0.1153; W ZDRC : 0.2383; W ΔH45 0.6464

[0075] Genetic Algorithm (AUC: 0.8118)

[0076] W VIL : 0.1947; W ZDRC : 0.1888; W ΔH45 0.6164

[0077] Grid algorithm (AUC: 0.8750)

[0078] W VIL : 0.3401; W ZDRC : 0.0993; W ΔH45 0.5607

[0079] Therefore, this example uses the grid algorithm with the highest AUC score to determine the weights of the decision factors, resulting in the following logistic regression model:

[0080] F 冰雹 =0.3401×F VIL +0.0993×F ZDRc +0.5607×F ΔH45

[0081] The weighted sum of the decision factors obtained from the logistic regression model is then substituted into the sigmoid function to convert it into a hail warning index.

[0082]

[0083] Extracting G from individual samples in the hail dataset 冰雹 For the hail warning threshold, this example uses the decision factor 20 minutes before hail to extract the hail warning index threshold, and uses the difference between the decision factors 30 minutes and 20 minutes before hail to extract the hail warning increment index threshold: a red warning is issued when the index increment is >0.06 or the index is >0.6, indicating that hail may occur in the next 20 minutes; an orange warning is issued when the index increment is between 0.05 and 0.06 or the index is between 0.5 and 0.6. Other values ​​indicate low risk; please pay attention to weather changes.

[0084] like Figure 2-4 As shown, 16 hail clouds occurring in different regions outside the training samples were selected. Assessment factors were extracted 30 minutes and 20 minutes before the peak of hail or thunderstorm cloud echo intensity, and warning indices were calculated, resulting in the table below. Of the 16 hail cases, 15 were correctly identified as red warnings (meaning hail was expected within the next 20 minutes), and 1 was falsely reported as an orange warning, resulting in a hail identification accuracy rate of 93.75%.

[0085] Table 1. Test results of the model in the embodiment.

[0086]

[0087]

[0088] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for early identification of hail clouds based on polarization radar, characterized in that, Includes the following steps: Step 1: Based on historical polarization radar data, screen hail events verified by ground observations and establish a sample database; Step 2: Extract multiple decision factors from the radar data of individual hail clouds 20 minutes before the hailstorm in the sample library, and standardize the decision factors; the decision factors include the vertical integral liquid water VIL factor (FVIL) and the differential reflectivity column height factor. 45dBZ reflectivity echo top height difference factor with 0℃ layer height difference factor ; Step 3: Establish a logistic regression model for multiple decision factors in the sample library, use cross-validation mechanism to select the optimal weight algorithm, calculate the weights of multiple decision factors, and determine the hail warning index threshold. Step 4: Extract the determination factors from the strong convective cloud detection data and perform standardization processing. Input the processed determination factors into the logistic regression model to calculate the warning index. Compare the warning index with the hail warning index threshold. If the warning index ≥ the hail warning index threshold, it is determined to be an early hail cloud; otherwise, it is determined to be a general thunderstorm cloud. The logistic regression model calculates the linear weighted sum of each decision factor, Fhail, and then substitutes Fhail into the sigmoid function to obtain the warning index P. 冰雹 The linear weighted sum The calculation formula is: F 冰雹 =W VIL ×F VIL +W ZDRc ×F ZDRc +W ΔH45 ×F ΔH45 In the formula W VIL For F VIL weights, W ZDRc For F ZDRc weights, W ΔH45 For F ΔH4 The weights are determined by the number of data points, and the minimum value of F_hail is 0, while the maximum value changes dynamically based on the amount of data input. The formula for calculating the sigmoid function is as follows: P 冰雹 = ； in The sum of linear weighted factors is used as the criterion, where e is the natural constant. Strong echo cells were identified using 30 dBZ as the identification threshold. Cell information was confirmed, and the maximum / minimum vertical integral liquid water VIL value within each cell was calculated using the cell information and VIL grid data. Finally, the VIL factor FVIL was determined. The formula for calculating the vertical integral liquid water VIL factor FVIL is as follows: ； In the formula, VIL max The maximum value of VIL within a single hail cloud cell. max L represents the minimum VIL value in the hail case sample library. max H represents the maximum VIL value in the hail case sample library; In step three, the method for determining the hail warning index threshold G for hail includes: extracting the warning index calculated from the judgment factors 20 minutes before hail occurrence in the sample library, and simultaneously extracting the increments of the warning index calculated from the judgment factors 30 minutes and 20 minutes before hail occurrence, and combining the warning index and the increments of the warning index to define the threshold; the method for classifying the warning level corresponding to the hail warning index threshold G for hail includes: when the increment of the warning index > 0.06 or the warning index > 0.6, it is a red warning, indicating that hail may occur in the next 20 minutes; when the increment of the warning index is between 0.05 and 0.06 or the warning index is between 0.5 and 0.6, it is an orange warning; the rest are low-risk levels.

2. The method for early identification of hail clouds based on polarization radar according to claim 1, characterized in that, The formula for calculating the differential reflectivity column height factor FZDRc is as follows: ； In the formula The maximum height of the differential reflectivity column extending vertically above the 0°C layer within a single hail cloud cell, with a ZDR ≥ 1 dB. The minimum differential reflectance column height for hail cases in the sample library. The maximum differential reflectance column height is given for hail cases in the sample library.

3. The method for early identification of hail clouds based on polarization radar according to claim 1, characterized in that, The formula for calculating the difference factor FΔH45 between the top height of the 45dBZ reflectivity echo and the height of the 0℃ layer is as follows: ； In the formula This represents the difference between the top height of the 45 dBZ reflectivity echo and the height of the 0°C layer within a single hail cloud cell. The minimum 45 dBZ reflectance height difference for hail cases in the sample library. The maximum 45 dBZ reflectance height difference for hail cases in the sample library.

4. The method for early identification of hail clouds based on polarization radar according to claim 1, characterized in that, The standardization process is min-max standardization; and before the standardization process, it also includes a step of using the 3σ method to remove extreme outliers from the decision factor.

5. The method for early identification of hail clouds based on polarization radar according to claim 1, characterized in that, The polarization radar data needs to be gridded; the gridding process includes: (1) Based on the longitude, latitude, altitude and radar position of the grid points in the Cartesian coordinate system, combined with the elevation angle, azimuth and slant range of the radar echo data in the polar coordinate system, the nearest neighbor assignment method is used to assign values ​​to the grid points. Clutter is filtered out during the assignment process. When the number of valid data in the 8 azimuths around a reflectivity data is ≤3, it is judged as clutter and removed. (2) Using data from adjacent elevation layers in polar coordinate system, the eight-point Barnes interpolation algorithm is used for processing.

6. The method for early identification of hail clouds based on polarization radar according to claim 1, characterized in that, In step three, the optimal weighting algorithm adopts one of the following: L1 regularized logistic regression algorithm, genetic algorithm, and grid algorithm. The algorithm with the highest AUC score is selected as the optimal weighting algorithm through cross-validation mechanism.

Images