Hail recognition early warning method based on storm space-time channel

Abstract

The invention provides a hail identification and early warning method based on a storm space-time channel, and the method comprises the steps: constructing a complete system from the tracking of a storm to the early warning of a disaster-bearing body, storing the information of a storm carrier in all directions through the space-time channel, and precisely identifying hail through multi-feature fusion and an SVM algorithm. According to the method, radar data inversion and a semi-Lagrange extrapolation method are used for realizing hail extrapolation forecasting, the hail forecasting and disaster-bearing body data are deeply fused, accurate early warning is provided for different disaster-bearing bodies, and the loss caused by hail disasters is reduced to the greatest extent. According to the invention, the historical motion trail and the coverage range of the storm carrier can be stored in all directions, the states of the storm at different moments can be known by analyzing data in the space-time channel, and the evolution trend of the storm on the time axis can be clearly seen; through combination of multi-source data, the accuracy and reliability of hail early warning are remarkably improved, hail falling area forecast and disaster-bearing body data covering population, building and agricultural distribution are subjected to overlay analysis, and the risk is accurately evaluated.

Description

Technical Field

[0001] The present invention relates to the technical field of hail identification and early warning, and in particular to a hail identification and early warning method based on a storm space-time channel. Background Art

[0002] As global climate change continues, extreme weather events are becoming more frequent. Hail, a highly destructive meteorological disaster, poses a serious threat to agriculture, industry, transportation, and the safety of people's lives and property. Accurate hail warnings are crucial for proactively implementing preventive measures and minimizing disaster losses. The meteorological community is committed to developing more accurate and efficient hail warning technologies to meet the urgent needs of all sectors of society for meteorological disaster prevention.

[0003] In the development of hail warning technology, how to comprehensively obtain storm information, accurately identify hail, and provide effective warnings for different disaster-prone bodies has always been a key research focus and difficulty. Traditional meteorological monitoring methods and warning methods have exposed many limitations when dealing with complex and changing weather conditions, and are unable to meet the current requirements for high-precision, timely, and targeted hail warnings. For example:

[0004] Radar-based hail warning schemes primarily utilize meteorological radar to monitor atmospheric echoes, analyzing parameters such as echo intensity and velocity to determine areas where hail is likely to occur. This approach relies on identifying echo characteristics potentially associated with hail based on the differences in radar wave reflection characteristics of different meteorological targets. For example, when echo intensity reaches a certain threshold and exhibits a specific morphology, hail is likely to occur. However, this approach has significant shortcomings. First, it typically relies on single-volume scan data, capturing only the instantaneous state of the storm at a specific moment, failing to capture the storm's evolution over time or its continuous trajectory in space. This makes it difficult to accurately assess the storm's development trajectory, potentially missing the optimal warning opportunity. Second, while storm clouds can be detected, critical information about the specific hailstones within them, such as their size, location, and timing, remains difficult to accurately forecast.

[0005] Numerical forecasting schemes based on atmospheric physics rely on physical principles such as atmospheric dynamics and thermodynamics to simulate and predict atmospheric motion and changes through the development of complex mathematical models. These models account for the interplay of various physical processes in the atmosphere, such as heat transfer, water vapor phase transitions, and airflow, and can theoretically provide a relatively comprehensive forecast for hail. In practice, initial meteorological conditions, such as temperature, humidity, air pressure, and wind speed, are input. The models then use numerical calculations to simulate future atmospheric conditions, thereby predicting the likely location and time of hail. However, this approach also has limitations. Due to the extreme complexity of the atmospheric system, numerical models cannot fully and accurately describe all atmospheric phenomena and processes. Furthermore, they lack effective calibration with real-time observational data. Although the models can simulate according to physical laws, actual atmospheric conditions are subject to change. Without real-time, accurate observational data to calibrate the models, forecasts can deviate significantly from actual conditions, leading to omissions or misreporting of hail forecasts and reducing the reliability and practicality of early warnings. Summary of the Invention

[0006] In view of this, the present invention proposes a hail identification and early warning method based on storm spatiotemporal channels to solve the problems existing in the above-mentioned prior art.

[0007] To achieve the above objectives, the present invention proposes a hail identification and early warning method based on a storm spatiotemporal channel, comprising:

[0008] (a) Storm spatiotemporal channel construction: Three-dimensional echo data, including the current moment and n historical volume scans, is acquired through meteorological radar. Storm cells are identified based on a preset echo intensity threshold, and morphological processing methods are used to remove noise interference. Main sections are cut vertically along the storm's moving direction, and the motion trajectory of the storm cells is tracked based on adjacent volume scan data. The spatiotemporal channel model is constructed by connecting the main sections at consecutive moments.

[0009] (b) Hail feature fusion identification: In the four-dimensional space of the spatiotemporal channel, radar reflectivity data, dual-polarization radar data, numerical forecast field data, sounding data, and ground meteorological data are integrated to extract a multidimensional feature set including vertically integrated liquid water content, three-body scattering characteristics, circulation dynamic parameters, atmospheric stability index, and terrain influence factors. After normalization, the feature set is input into the support vector machine model for hail probability classification;

[0010] (c) Dynamic extrapolation forecast: Based on the inversion of the three-dimensional wind field from the radar radial velocity field, the semi-Lagrangian method is used to track the motion trajectory of the storm carrier. The hail drop location is corrected in combination with the hail particle settling velocity, and a hail drop area forecast grid with minute-level time resolution is generated.

[0011] (d) Warning issuance: Through GIS spatial analysis, the forecast grid is matched with the vector data of the disaster-prone body in time and space, and graded warning information is generated based on hail size and impact time. The information is then pushed to the affected areas through a multi-channel warning system.

[0012] Extract hail identification vectors based on weather radar echo data;

[0013] identifying hail characteristics based on the hail identification carrier;

[0014] Based on the hail characteristics, a support vector machine is used to estimate the hailfall location;

[0015] The semi-Lagrangian method is used to extrapolate the hailfall area to construct the hailfall forecast area.

[0016] The data overlay method is used to conduct hail risk assessment and generate early warning information.

[0017] Furthermore, the process of identifying a storm cell based on a preset echo intensity threshold includes:

[0018] Setting an echo intensity threshold, when the echo intensity of the weather radar echo data is greater than the echo intensity threshold, it is marked as a potential storm cell area;

[0019] Acquire a spatiotemporal channel of storm movement based on the storm cell area;

[0020] Calculating storm geometry based on the spatiotemporal channel;

[0021] Based on the storm geometric characteristics, storm cells with obvious convective characteristics and a probability of producing hail are selected as the hail identification carriers.

[0022] Furthermore, the hail characteristics include reflectivity characteristics, dual-polarization data characteristics, circulation form characteristics, sounding data characteristics, and ground element characteristics;

[0023] The reflectivity characteristics include maximum reflectivity, weak echo area, storm core height, vertically integrated liquid water content, echo top height, and three-body scattering parameters;

[0024] The dual-polarization data features include differential reflectivity, differential phase ratio, and reflectivity correlation coefficient;

[0025] The circulation features include high-pressure centers, low-pressure centers, pressure troughs, pressure ridges, convergence and divergence areas of airflow, and vertical rising and sinking movement areas of the air pressure field;

[0026] The sounding data characteristics include the temperature, humidity, air pressure, wind speed, wind direction change data with altitude at the nearby high-altitude station and atmospheric vertical stability parameters;

[0027] The ground element characteristics include ground air pressure, temperature, relative humidity, pressure change and terrain data.

[0028] Furthermore, the process of inputting the normalized data into the support vector machine model for hail probability classification includes:

[0029] constructing a training data set based on the hail characteristics, and training a support vector machine model based on the training data set;

[0030] The normalized hail features are input into the trained support vector machine model to output the probability of hail.

[0031] Furthermore, the process of inputting the normalized data into the support vector machine model for hail probability classification further includes:

[0032] A relationship model between hail size and characteristic parameters is constructed based on reflectivity characteristics and dual-polarization data features.

[0033] Based on the relationship model, the location of the hailfall area is estimated according to the hail characteristics and spatial structure of each storm carrier in the space-time channel, as well as the evolution of the storm carrier over time and space, combined with the characteristics of sounding data and ground elements.

[0034] Furthermore, the relationship model is shown as follows:

[0035]

[0036] Among them, a 1-9 is the coefficient, Z max is the maximum horizontal reflectivity of the storm carrier, Z dr is the differential reflectivity, K dp is the differential phase ratio, ρ hv is the correlation coefficient of radar reflectivity, is the mean horizontal reflectivity of the storm carrier, H core is the core height of the storm carrier, A wer is the area of ​​weak echo area, H top is the echo top height, and VIL is the vertical liquid water content.

[0037] Furthermore, the process of tracking the trajectory of the storm carrier using the semi-Lagrangian method includes:

[0038] The radar radial velocity data is transformed into a certain form, and the relationship between the spatial derivative and the time derivative is used to solve the divergence and vorticity of the horizontal wind field;

[0039] Based on the divergence and vorticity, the horizontal component of the transient wind field is inverted using a constraint equation;

[0040] Based on the horizontal component and the three-dimensional radar echo characteristics, a semi-Lagrangian method is used to treat the storm carrier and the hailfall area within it as Lagrangian objects moving with the atmosphere. Based on the atmospheric motion equations, the position of the storm carrier at future times is tracked, and an extrapolated forecast of the hailfall area is achieved through the particle trajectory method of fluid mechanics.

[0041] Using wind field data and the settling velocity of hail particles, the horizontal displacement of the hailfall area is corrected to generate the hailfall area distribution at different forecast times.

[0042] The wind field predicted by the numerical model is used to further correct the position of the hailfall area to obtain the predicted hailfall area.

[0043] Furthermore, the process of correcting the falling area location based on the hail particle settling velocity includes:

[0044] Calculating the hail settling velocity and hail horizontal displacement based on the predicted falling area of ​​the hail body;

[0045] The longitude and latitude of the hail fall area outline are updated based on the hail settling velocity and the hail horizontal offset, and the position of the hail fall area is updated based on the longitude and latitude.

[0046] Furthermore, the process of generating graded warning information based on hail size and impact time includes:

[0047] GIS is used to integrate the disaster-prone body information, and combined with latitude and longitude positioning and time dimension, the disaster-prone body information is matched with the hail body forecast area in real time;

[0048] For vulnerable hazard-bearing bodies, a separate superposition analysis method is used for risk assessment;

[0049] Based on the risk assessment results, yellow warning information, orange warning information and red warning information are generated respectively.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] The proposed spatiotemporal channel identification solution utilizes multiple volume scans to identify storm paths, establish a spatiotemporal channel of the storm's trajectory, and then identify hail within this channel. This method comprehensively preserves the historical trajectory and coverage of storm carriers. By analyzing the data within the spatiotemporal channel, not only can the storm's state at different moments be understood, but its evolution over time can also be clearly visualized, such as its gradual intensification, changes in its movement direction, and the expansion or contraction of its coverage.

[0052] The application comprehensively uses radar data and numerical weather prediction data, and ground meteorological station and sounding data. Hail is identified based on radar, hail bodies are accurately judged by analyzing echo data, and future characteristics thereof are predicted. Meanwhile, high altitude station, ground station and circulation situation data are used for correction. The high altitude station provides vertical atmospheric information, the ground station reflects ground meteorological elements, and the circulation situation shows the macro background. The combination of multiple data sources significantly improves the accuracy and reliability of hail warning.

[0053] The application superimposes and analyzes hail falling area prediction and disaster-bearing body data covering population, buildings and agricultural distribution, accurately assesses the risk, and generates different levels of warning through time, time and space subdivision, and delivers it through various channels to perform hierarchical warning on different risk areas. BRIEF DESCRIPTION OF DRAWINGS

[0054] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The accompanying drawings are included to provide a description of the preferred embodiments and are not intended to limit the scope of the application. Moreover, the same reference numerals are used throughout the same figures. In the drawings:

[0055] Figure 1 A flowchart of the method of the application is shown;

[0056] Figure 2 A schematic diagram of the storm carrier and the space-time channel formed by its movement in the embodiment of the application is shown;

[0057] Figure 3 A schematic diagram of hail falling area extrapolation in the embodiment of the application is shown;

[0058] Figure 4 A schematic diagram of hail falling area correction in the embodiment of the application is shown. DETAILED DESCRIPTION

[0059] Exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0060] The present embodiment proposes a hail identification and warning method based on storm space-time channel, as shown in Figure 1 , comprising:

[0061] 1.1 Extracting the storm spatiotemporal channel as a carrier for hail identification

[0062] Using the echo data of the weather radar, set the echo intensity threshold Z th , for the echo intensity Z(x,y), if Z(x,y)>Z th , then the area is marked as a potential storm cell area. Then, morphological operations such as dilation and erosion are applied to remove noise and small interference areas.

[0063] Using the current moment volume scan data and combining it with the previous n radar volume scan data (n=3 by default), the storm cells are identified one by one, and the main cross section of the volume scan storm cell at a single moment is intercepted along the vertical plane in the direction of the storm movement. The tracking trajectory is calculated, and the main cross sections of the storm cells at multiple moments are connected along the trajectory to form the space-time channel of the storm movement, such as Figure 2 shown.

[0064] For each storm space-time channel segmented, calculate its geometric characteristics (main cross-sectional area A = ∑ x,y I(x,y), where I(x,y) is a binary image of the storm cell area, which is 1 within the cell area and 0 otherwise; the perimeter P is calculated using a boundary tracking algorithm; the shape factor ), and screen out storm cells with obvious convective characteristics and a probability of producing hail as the carriers (cell_body) for hail identification.

[0065] Figure 2 The area enclosed by the middle dotted line is the space-time channel proposed by the present invention, which can preserve the historical movement trajectory and coverage of the storm carrier.

[0066] Within the four-dimensional space of the spatiotemporal channel, numerical weather forecast data, as well as ground-based meteorological station and sounding data, were incorporated. The specific method for extracting hail-related characteristic factors was as follows.

[0067] 1.2 Hail Feature Recognition

[0068] 1.2.1 Hail-related feature extraction

[0069] Reflectivity feature extraction

[0070] Calculate the maximum reflectivity: Use a sliding window algorithm with a window size of m×n to calculate the maximum reflectivity Z point by point within the storm cell. max ,Right now Where W(x,y) is a sliding window centered at (x,y).

[0071] Identify weak echo areas: by setting the echo intensity lower limit Z low And regional connectivity conditions, mark the weak echo area. Calculate its area Awer and other characteristic parameters.

[0072] Determine the height of the storm core: By analyzing the echo intensity distribution Z(h) at different altitudes h, find the altitude h with the strongest echo intensity core ,Right now

[0073] Calculate the vertically integrated liquid water content (VIL): Based on the relationship between radar reflectivity and liquid water content, the liquid water content of the storm cell in the vertical direction is integrated and calculated:

[0074]

[0075] Among them H bot and H top are the bottom and top heights of the storm cell, respectively.

[0076] Determine the echo top height: Scan the radar echo data upward to find the echo intensity below the threshold Z. echo-th Height h core-top , as the echo top height.

[0077] Identify three-body scattering: By analyzing the multipath propagation characteristics and specific echo morphology of radar echoes, detect whether three-body scattering exists and record parameters such as its location and intensity.

[0078] Dual-polarization data feature extraction

[0079] Extracting differential reflectivity Z from dual-polarization radar data dr , differential phase ratio K dp , the correlation coefficient of radar reflectivity ρ hv . Z dr Focus on the data in the range of [-0.3,4.5], ρ hv The interest interval is [0.75, 1.0], K dp is [-2.0,10.0].

[0080] Circulation form feature extraction

[0081] Obtain pressure and wind data at 500hPa, 700hPa, and 850hPa from meteorological satellites and high-altitude sounding equipment. Analyze pressure data to identify weather system features such as high-pressure centers, low-pressure centers, troughs, and ridges. Analyze wind data to determine areas of convergence and divergence, as well as vertical ascent and descent.

[0082] Comprehensively analyze the pressure and wind field characteristics at different altitudes to determine the impact of circulation forms on the dynamic and thermal conditions for the occurrence of severe convective weather.

[0083] Radiosonde data feature extraction

[0084] Collect the radiosonde data of nearby upper air stations, including temperature T, humidity rh, pressure P, wind speed V, wind direction θ, and other data that change with height h. Calculate the vertical stability parameters of the atmosphere, including: convective available potential energy (CAPE), lifting condensation level (LCL), Showalter index, and K index, etc.

[0085] Ground element feature extraction

[0086] Using the ground meteorological station network, real-time collection of ground pressure P s , air temperature T s , relative humidity RH, pressure change ΔP, and other meteorological element data. Combined with terrain data, analyze the distribution characteristics of ground elements under different terrain conditions, and consider the influence of terrain lifting and blocking on the formation of hail.

[0087] 1.2.2 Hail identification

[0088] Data normalization: Normalize the extracted reflectivity-related features, dual-polarization-related features, circulation pattern features, radiosonde data features, and ground element features. Use the formula:

[0089]

[0090] Where x is the original feature value, x min and x max are the minimum and maximum values of the feature, respectively.

[0091] Construct the training data set: Use historical hail and non-hail sample data to construct the training data set. For each sample, extract the above-mentioned fused feature vector and label it as a hail sample or not.

[0092] Support vector machine (SVM) training: Use the support vector machine (SVM) algorithm, select the radial basis function (RBF) as the kernel function K(x i ,x j ) = exp (-γ‖x i -x j ‖ 2 ), where γ is the kernel function parameter, x i and x j are the sample feature vectors. By adjusting the parameters of SVM (such as penalty parameter C and kernel function parameter γ), use the training data set for training, and optimize the performance of the model.

[0093] Hail prediction: Input the real-time collected and processed fusion feature vector of the storm cell into the trained SVM model, and the model will judge whether the storm carrier will produce hail according to the learned classification rule, and output the hail probability P(hail).

[0094] 1.2.3 Hail Parameter Assessment

[0095] Hail size estimation: Based on the reflectivity characteristics and dual-polarization data characteristics, a relationship model between hail size and characteristic parameters is established.

[0096]

[0097] where a 1-9 is the coefficient, Z max is the maximum horizontal reflectivity of the storm carrier, Z dr is the differential reflectivity, K dp is the differential phase ratio, ρ hv is the correlation coefficient of radar reflectivity, is the mean horizontal reflectivity of the storm carrier, H core is the core height of the storm carrier, A wer is the area of ​​weak echo area, H top is the echo top height, and VIL is the vertical liquid water content.

[0098] Hailfall area estimation: Based on the hail characteristics and spatial structure of each storm carrier in the space-time channel, as well as the evolution of the storm carrier over time and space, combined with the detection data of ground stations and high-altitude stations, comprehensive calculation and analysis are performed to estimate the location of the hailfall area.

[0099] 1.3 Hail Extrapolation Forecast

[0100] Radar radial velocity data is processed and converted from polar coordinates to Cartesian coordinates (with the x-axis pointing eastward, parallel to the equator, and the y-axis pointing northward, perpendicular to the equator). The relationship between spatial and temporal derivatives is used to determine the divergence and vorticity of the horizontal wind field. Furthermore, the linear optical flow method (LK method) is used to combine the divergence and vorticity fields to invert the horizontal components (u and v) of the transient wind field using constraint equations.

[0101] Extrapolation employs a semi-Lagrangian approach, based on the three-dimensional echo characteristics of radar observations and wind field inversion results. The storm carrier and the hailfall zone within it are considered Lagrangian objects moving with the atmosphere. Based on the atmospheric equations of motion, the storm carrier's future position is tracked along wind field streamlines. Using the particle trajectory method from fluid dynamics, an extrapolated forecast of the hailfall zone is achieved.

[0102] At the hourly time resolution (S-band radar data is saved every 6 minutes or so), gridded wind field data is used, combined with the hail particle settling velocity to correct the horizontal displacement of the hailfall area, and the hailfall area distribution at different forecast times is generated, such as Figure 3 shown.

[0103] The wind field predicted by the numerical model is further used to modify the position of the hail fall area. The numerical weather model provides wind field prediction data at different heights, which is used to modify the hail body fall area at different heights, as shown in Figure 4 The specific method is as follows:

[0104] 1. Obtain the hail extrapolation fall area result.

[0105] 2. Calculate the hail sinking speed. The hail sinking speed is denoted as V t :

[0106]

[0107] In the formula, g is the acceleration of gravity (unit: m / s 2 ), p h is the hail density (unit: kg / m 3 ), p a is the air density (unit: kg / m 3 ), C d is the drag coefficient, and D is the hail diameter (unit: m)

[0108] For the value of the drag coefficient C d , the present application defines:

[0109]

[0110] 3. Calculate the horizontal offset. According to the hail sinking speed V t and the core height Core_height of the storm carrier, the hail sinking time t f is calculated,

[0111] t f = Core_height / V t

[0112] When the hail falls from high altitude, it will be affected by the horizontal wind. It is assumed that the hail passes through n height layers when falling, and the offsets in the x direction and the y direction are denoted as Deltax and Deltay, respectively:

[0113]

[0114] Where u i and v i are the numerical prediction wind field results at different height layers.

[0115] 4. Hail fall area correction. Obtain the longitude and latitude of the hail extrapolation fall area contour, and collect them as {Lon_pre, Lat_pre}. Traverse Lon_pre and Lat_pre respectively, and update the position of the fall area:

[0116]

[0117] where lon new ,lat new is the latitude and longitude of the contour points of the hailstone extrapolated falling area after correction, lon pre_i and lat pre_i It is the longitude and latitude of the original landing area.

[0118] 1.4 Disaster-prone body warning

[0119] After hailstone identification, the system of the present invention further realizes all-round and accurate tiered early warning of hailstone disasters.

[0120] At the data overlay level, advanced GIS technology is used to precisely integrate disaster-prone information such as population distribution, building distribution, and agricultural planting areas in a shape format. This information is then combined with latitude and longitude positioning and time dimensions to match this with hail forecast data in real time. Furthermore, overlaying is performed by administrative division, down to the township and village level, enabling minute-by-minute forecasts of hail impact times and estimated hail sizes. This not only enables grassroots communities to prepare for emergencies in advance but also allows for the rational allocation of protective resources based on forecast information.

[0121] This invention uses a separate overlay analysis approach for vulnerable areas such as schools, chemical parks, power facilities, and military facilities. By deeply analyzing the unique attributes and potential risks of these key areas, a focused early warning and prevention mechanism is established, ensuring that these high-risk areas receive the most timely and accurate warning information before hail disasters strike.

[0122] Based on a thorough risk assessment, warning information is generated at different levels. A yellow alert is issued in low-risk areas, reminding residents to be aware of weather changes; an orange alert is issued in medium-risk areas, advising residents to take comprehensive precautions; and a red alert is issued in high-risk areas, urging residents to take immediate emergency measures. Ultimately, through diverse communication channels such as text messages and radio broadcasts, warning information is quickly and accurately delivered to residents and relevant departments in affected areas, effectively raising awareness and response capabilities across society and minimizing casualties and property losses caused by hailstorms.

[0123] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A hail identification and early warning method based on storm spatiotemporal channels, characterized in that: include: (a) Storm spatiotemporal channel construction: 3D echo data including the current time and n historical volume scans are acquired through weather radar. Storm cells are identified based on a preset echo intensity threshold, and morphological processing methods are used to remove noise interference. The main section is cut vertically along the storm's moving direction, and the motion trajectory of the storm cell is tracked based on the adjacent volume scan data. The space-time channel model is constructed by connecting the main sections at consecutive moments. (b) Hail feature fusion identification: In the four-dimensional space of the spatiotemporal channel, radar reflectivity data, dual-polarization radar data, numerical forecast field data, sounding data, and ground meteorological data are integrated to extract a multidimensional feature set including vertically integrated liquid water content, three-body scattering characteristics, circulation dynamic parameters, atmospheric stability index, and terrain influence factors. After normalization, the feature set is input into the support vector machine model for hail probability classification; (c) Dynamic extrapolation forecast: Based on the inversion of the three-dimensional wind field from the radar radial velocity field, the semi-Lagrangian method is used to track the motion trajectory of the storm carrier. The hail drop location is corrected in combination with the hail particle settling velocity, and a hail drop area forecast grid with minute-level time resolution is generated. (d) Warning issuance: Through GIS spatial analysis, the forecast grid is matched with the vector data of the disaster-prone body in time and space, and graded warning information is generated based on hail size and impact time. The information is then pushed to the affected areas through a multi-channel warning system.

2. The hail identification and early warning method based on storm spatiotemporal channels according to claim 1 is characterized in that: The process of identifying storm cells based on a preset echo intensity threshold includes: Setting an echo intensity threshold, when the echo intensity of the weather radar echo data is greater than the echo intensity threshold, it is marked as a potential storm cell area; Acquire a spatiotemporal channel of storm movement based on the storm cell area; Calculating storm geometry based on the spatiotemporal channel; Based on the storm geometric characteristics, storm cells with obvious convective characteristics and a probability of producing hail are selected as the hail identification carriers.

3. The hail identification and early warning method based on storm spatiotemporal channels according to claim 1 is characterized in that: The hail characteristics include reflectivity characteristics, dual-polarization data characteristics, circulation form characteristics, sounding data characteristics, and ground element characteristics; The reflectivity characteristics include maximum reflectivity, weak echo area, storm core height, vertically integrated liquid water content, echo top height, and three-body scattering parameters; The dual-polarization data features include differential reflectivity, differential phase ratio, and reflectivity correlation coefficient; The circulation features include high-pressure centers, low-pressure centers, pressure troughs, pressure ridges, convergence and divergence areas of airflow, and vertical rising and sinking movement areas of the air pressure field; The sounding data characteristics include the temperature, humidity, air pressure, wind speed, wind direction change data with altitude at the nearby high-altitude station and atmospheric vertical stability parameters; The ground element characteristics include ground air pressure, temperature, relative humidity, pressure change and terrain data.

4. The hail identification and early warning method based on storm spatiotemporal channels according to claim 1, characterized in that: The process of inputting the normalized data into the support vector machine model for hail probability classification includes: constructing a training data set based on the hail characteristics, and training a support vector machine model based on the training data set; The normalized hail features are input into the trained support vector machine model to output the probability of hail.

5. The hail identification and early warning method based on storm spatiotemporal channels according to claim 4 is characterized in that: The process of inputting the normalized data into the support vector machine model for hail probability classification also includes: A relationship model between hail size and characteristic parameters is constructed based on reflectivity characteristics and dual-polarization data features. Based on the relationship model, the location of the hailfall area is estimated according to the hail characteristics and spatial structure of each storm carrier in the space-time channel, as well as the evolution of the storm carrier over time and space, combined with the characteristics of sounding data and ground elements.

6. The hail identification and early warning method based on storm spatiotemporal channels according to claim 5, characterized in that: The relationship model is shown below: Among them, a 1-9 is the coefficient, Z max is the maximum horizontal reflectivity of the storm carrier, Z dr is the differential reflectivity, K dp is the differential phase ratio, ρ hv is the correlation coefficient of radar reflectivity, is the mean horizontal reflectivity of the storm carrier, H core is the core height of the storm carrier, A wer is the area of ​​weak echo area, H top is the echo top height, and VIL is the vertical liquid water content.

7. The hail identification and early warning method based on storm spatiotemporal channels according to claim 1, characterized in that: The process of tracking the trajectory of storm carriers using the semi-Lagrangian method includes: The radar radial velocity data is transformed into a certain form, and the relationship between the spatial derivative and the time derivative is used to solve the divergence and vorticity of the horizontal wind field; Based on the divergence and vorticity, the horizontal component of the transient wind field is inverted using a constraint equation; Based on the horizontal component and the three-dimensional radar echo characteristics, a semi-Lagrangian method is used to treat the storm carrier and the hailfall area within it as Lagrangian objects moving with the atmosphere. Based on the atmospheric motion equations, the position of the storm carrier at future times is tracked, and an extrapolated forecast of the hailfall area is achieved through the particle trajectory method of fluid mechanics. Using wind field data and the hail particle settling velocity, the horizontal displacement of the hailfall area is corrected to generate the hailfall area distribution at different forecast times. The wind field predicted by the numerical model is used to further correct the position of the hailfall area to obtain the predicted hailfall area.

8. The hail identification and early warning method based on storm spatiotemporal channels according to claim 7, characterized in that: The process of correcting the location of the hail drop zone based on the hail particle settling velocity includes: Calculating the hail settling velocity and hail horizontal displacement based on the predicted falling area of ​​the hail body; The longitude and latitude of the hail fall area contour are updated based on the hail settling velocity and the hail horizontal offset, and the position of the hail fall area is updated based on the longitude and latitude.

9. The hail identification and early warning method based on storm spatiotemporal channels according to claim 1, characterized in that: The process of generating graded warning information based on hail size and impact time includes: GIS is used to integrate the disaster-prone body information, and combined with latitude and longitude positioning and time dimension, the disaster-prone body information is matched with the hail body forecast area in real time; For vulnerable hazard-bearing bodies, a separate superposition analysis method is used for risk assessment; Based on the risk assessment results, yellow warning information, orange warning information and red warning information are generated respectively.

Images