A tornado grading early warning method based on multi-dimensional variable characteristics

Abstract

The application discloses a tornado grading early warning method based on multi-dimensional variable characteristics, which comprises the following steps: 1, radar weather scanning; 2, identifying strong convection area; 3, calculating radar inversion vorticity; 4, identifying tornado characteristics: specifically comprising constructing a Bayesian classification model, identifying tornado characteristics in the lowest layer of elevation angle, and identifying tornado characteristics in the second-lowest layer of elevation angle; 5, tornado prediction: when tornado characteristics of more than 7 distance bins are identified in any one of the lowest two elevation angles, and the maximum vorticity intensity exceeds 5*10 ‑3 s ‑1 , a tornado is early warned; when tornado characteristics of more than 7 distance bins are identified in both of the lowest two elevation angles, and the vorticity intensity exceeds 10*10 ‑3 s ‑1 , a strong tornado is early warned; otherwise, no tornado is predicted. The application combines the vorticity intensity and the number of distance bins with tornado characteristics in the two layers of low elevation angles to grade and early warn the tornado, thereby providing an important reference for forecasters to timely issue accurate early warning signals and improving the tornado early warning capability.

Description

Technical Field

[0001] This invention relates to the field of weather forecasting technology, and in particular to a tornado classification and early warning method based on multidimensional variable characteristics. Background Technology

[0002] Among various meteorological disasters, tornadoes are the most intense atmospheric vortex phenomenon. Tornadoes typically originate from small- to medium-scale convective weather systems, with a funnel-shaped cloud extending from the base of a thunderstorm cloud to the ground exhibiting intense rotation. While their impact area is small, their destructive power is extremely strong, capable of causing significant casualties and severe property damage. Due to their small spatial and temporal scale, localized and sudden nature, and rapid development, current numerical models have very limited ability to predict tornadoes, and traditional observation methods, due to the sparse number of observation stations, also struggle to capture their intensity.

[0003] Prior to the deployment of the new generation of Doppler weather radar networks, the analysis and research on tornadoes mainly focused on disaster descriptions, the weather and climate characteristics of tornadoes, and the weather background analysis of some individual tornado cases. With the development of detection technology, significant progress has been made in tornado detection and early warning since the deployment of the new generation of weather radar networks. Because Doppler weather radar can measure not only the intensity of precipitation echoes but also the velocity of precipitation particles along the radar's radial direction, it enables the detection and early warning of tornadoes. Simultaneously, scholars have conducted a series of research studies using Doppler weather radar, including the analysis of tornado storm radar echo structure and the evolution characteristics of mesocyclones, thereby improving the timeliness of near-term tornado monitoring and early warning. Automatic tornado identification has also been implemented, providing crucial assistance for tornado monitoring and early warning. Unlike ground lightning warnings and precipitation estimates, which rely on microphysical information observed by dual-polarization radar, tornado identification is currently mainly based on Doppler velocity. The identification work is carried out based on the positive and negative radial velocities generated by the rotation of the tornado observed by radar. However, since tornadoes develop at low heights, and low-level echoes and radial velocities are often affected by non-meteorological targets such as ground objects, the identification success rate is seriously affected.

[0004] In recent years, my country has constructed and upgraded various dual-polarization radars across multiple bands, including S-band, C-band, and X-band. These radars not only provide reflectivity factors (Z...) H ), radial velocity and spectral width, and also provide differential reflectivity (Z). DR ), differential phase shift rate (K) DP ) and cross-correlation coefficient (ρ) hvThese variables can be used to analyze information such as the phase state, concentration, and particle size of water condensate particles observed by radar, as well as ground object and biological echoes. During the formation of tornadoes, due to strong rotation and updrafts, objects on the ground are often carried into the air, resulting in radar detecting signals with distinct characteristics that differ from meteorological echoes. How to further develop a noise-resistant eddy current inversion identification algorithm based on radar radial velocity, and combine it with dual-polarization radar observation information as an aid, to ultimately achieve high-precision automatic identification and early warning of tornadoes, is an urgent problem to be solved. Summary of the Invention

[0005] The technical problem to be solved by this invention is to address the shortcomings of the existing technology by providing a tornado classification and early warning method based on multidimensional variable features. This method can classify and warn of tornadoes into three levels: no tornadoes, tornadoes, and severe tornadoes. Severe tornadoes represent tornadoes with distinct observational characteristics and strong destructive potential. Tornadoes represent tornadoes with very similar characteristics. The tornado classification and early warning provides important reference information for forecasters to issue accurate early warning signals in a timely manner, thereby improving the early warning capability for tornadoes.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0007] A tornado classification and early warning method based on multidimensional variable features includes the following steps.

[0008] Step 1, Radar Weather Scan: The radar performs a radar scan of the local forecast area according to the set sampling frequency; during each radar scan, it scans at least n set elevation angles from bottom to top or from top to bottom; after each elevation angle scan is completed, a radar scan image is obtained; each radar scan image has a uniform arrangement of a rows and b columns. There are several range databases; each range database corresponds to a set of radar observation parameters; each set of radar observation parameters includes the reflectivity factor Z. H Differential reflectivity Z DR Zero-lag correlation coefficient ρ hv Differential phase standard deviation SD(Φ) DP ) and radar radial velocity u; where; n≥3, n=1 represents the lowest elevation angle, n=2 represents the second-to-last elevation angle; a>10, b>10; the length of each range station is The center of the local forecast area is weather forecast point A.

[0009] Step 2: Identify strong convection regions: Combining objective convection identification methods with strong convection echo top height and vertical cumulative liquid water content (VIL), strong convection regions are identified from the local forecast area in Step 1; the distance reservoir located in the strong convection region is called the strong convection distance reservoir.

[0010] Step 3: Calculate radar inversion vorticity For each strong convection range library in the lowest elevation angle and the penultimate elevation angle, based on their respective radar radial velocity u and range library length... The corresponding radar inversion vorticity is obtained through inversion calculation. .

[0011] Step 4: Identify tornado features, specifically including the following steps:

[0012] Step 4-1: Construct a Bayesian classification model Bayesian classification model In For input sample data, and ,in, , , , , Bayesian classification model In It belongs to the tornado category, and ,in, Indicates no tornadoes. It indicates the presence of a tornado.

[0013] Step 4-2: Identify tornado features at the lowest elevation angle: For each strong convection distance in the lowest elevation angle database, the Bayesian classification model constructed in Step 4-1 is used. Tornado feature identification is performed; assuming that the lowest elevation angle identifies M strong convective distances with tornado features.

[0014] Step 4-3, Identifying Tornado Features at the Penultimate Elevation Angle: Using the tornado feature identification method from Step 4-2, N strong convective distances with tornado features were identified at the penultimate elevation angle.

[0015] Step 5, Tornado Forecasting: Tornado forecasts are categorized into three types: no tornadoes, tornadoes, and severe tornadoes. The specific forecasting method is as follows:

[0016] A. Strong tornado: When M≥7, N≥7 and When all three conditions are met simultaneously, the weather forecast point A is likely to experience a strong tornado; among them... Radar inversion vorticity for all strong convective range libraries in the lowest and penultimate elevation angles. The maximum value; This represents the maximum threshold for vorticity intensity.

[0017] B. Tornado: When M≥7 or N≥7 and At that time, a tornado was observed at forecast point A; among them, This is the minimum threshold for vorticity intensity, and .

[0018] C. No tornadoes: When the conditions for strong tornadoes and tornadoes are not met, there are no tornadoes at the forecast point A.

[0019] Step 2, the method for identifying areas of strong convection, specifically includes the following steps:

[0020] Step 2-1: Identify convection regions: Use objective convection identification methods to identify the convection regions and their locations.

[0021] Step 2-2, Calculate VIL: Calculate the strong convection echo top height and vertical cumulative liquid water content (VIL) for the identified convection regions.

[0022] Step 2-3: Identify strong convection areas: Compare the VIL calculated in Step 2-2 with the set VIL threshold. Convection areas that exceed the set VIL threshold are called strong convection areas.

[0023] In step 3, radar inversion vorticity The calculation formula is:

[0024] ;

[0025] In the formula, The distance in row i and column j The radar radial velocity of the ku; where 1≤i≤a, 1≤j≤b.

[0026] The weighting coefficient for the distance database in the i-th row and j-th column is a set value that increases as the distance database moves further away from the center of the weather forecast point A.

[0027] The position of the distance from the library in the i-th row and j-th column can be calculated by interpolation.

[0028] In step 3, radar inversion vorticity The method for obtaining the calculation formula includes the following steps:

[0029] Step 3-1, Linear Interpolation: Using bilinear interpolation, the radar radial velocity u is interpolated to the Cartesian coordinate system.

[0030] Step 3-2: Calculate the position of the distance library: In the Cartesian coordinate system, calculate the position of the distance library in the i-th row and j-th column. The specific calculation formula is as follows:

[0031] (3-1);

[0032] In the formula, The distance from the distance library in the i-th row and j-th column to the radar center; Let be the radar azimuth angle of the range library in the i-th row and j-th column.

[0033] Step 3-3: Construct the radar radial velocity estimation of the range library in the i-th row and j-th column. The expression is as follows:

[0034] (3-2);

[0035] In the formula, Let be the radar radial velocity at weather forecast point A, as observed by radar.

[0036] Let be the radial shear of the distance library in the i-th row and j-th column, an unknown variable.

[0037] This is the distance from weather forecast point A to the radar center.

[0038] Steps 3-4: Define the value function Specifically:

[0039] (3-3);

[0040] In the formula, The actual radar radial observation value of the range library in the i-th row and j-th column is given.

[0041] Steps 3-5: Solve for radar inversion vorticity Substitute equation (3-1) into equation (3-2), then substitute equation (3-2) into equation (3-3), and solve equation (3-3) to obtain the radar-derived vorticity. .

[0042] Step 3-6, Vortex Coordinate Transformation: Transform the radar-inverted vorticity obtained in Step 3-5. Interpolate back to the original polar coordinates.

[0043] In step 1, the local forecast area is a 2×2 km area centered on the weather forecast point A.

[0044] In step 1, n=9.

[0045] In step 5, the wind force of the strong tornado is no less than EF2.

[0046] In step 5, =10×10 -3 s -1 , =5×10 -3 s -1 .

[0047] The radar in step 1 is a dual-polarization radar.

[0048] The present invention has the following beneficial effects:

[0049] 1. This invention combines vortex intensity and the number of distance databases with tornado characteristics in two low elevation angles to enable graded early warning of tornadoes, including no tornadoes, tornadoes, and strong tornadoes. This provides important reference information for forecasters to issue accurate early warning signals in a timely manner, thereby improving the early warning capability for tornadoes.

[0050] 2. This invention utilizes local functional theory and the method of minimizing the value function to invert and calculate the radial velocity u of precipitation particle radar to obtain the corresponding radar inversion vorticity. This significantly reduces the impact of non-meteorological targets such as ground features on the radar's radial velocity, thereby improving the success rate of tornado identification.

[0051] 3. The reflectivity factor Z is used in this invention. H Differential reflectivity Z DR Zero-lag correlation coefficient ρ hv Differential phase standard deviation SD(Φ) DP ) and radar inversion vorticity Five variables are used as input sample data for the Bayesian classification model to identify tornado features, further reducing the influence of non-meteorological targets such as ground objects on the radial velocity of the radar and improving the success rate of tornado identification.

[0052] 4. Radar inversion vorticity in this invention During the calculation, radial velocity information over the entire 2×2 km range is utilized. Missing measurements at single or multiple points, as well as interference from noise, have little impact on the results. Attached Figure Description

[0053] Figure 1 The flowchart of a tornado classification and early warning method based on multidimensional variable features according to the present invention is shown.

[0054] Figure 2 The radar radial velocity map of the tornado in Shengze Town, Wujiang District, Suzhou is displayed.

[0055] Figure 3 The image shows the tornado reflectance factor and identification results for Shengze Town, Wujiang District, Suzhou. Detailed Implementation

[0056] The present invention will now be described in further detail with reference to the accompanying drawings and specific preferred embodiments.

[0057] This embodiment uses the observation of a tornado in Shengze, Suzhou on May 14, 2021, by a Shanghai dual-polarization radar as an example for illustration.

[0058] like Figure 1 As shown, a tornado classification and early warning method based on multidimensional variable features includes the following steps.

[0059] Step 1: Radar Weather Scan

[0060] The radar (preferably a dual-polarization radar) performs a radar scan of the local forecast area at a set sampling frequency. The local forecast area is preferably a 2×2 km region centered on the weather forecast point A.

[0061] During each radar scan, the scan is performed in n predetermined elevation angles, either from bottom to top or from top to bottom. After each elevation angle scan is completed, a radar scan image is obtained. Wherein, n≥3, and in this embodiment, n=9 is preferred; n=1 represents the lowest elevation angle, and n=2 represents the second to last elevation angle.

[0062] Each radar scan image has a uniform arrangement of a rows and b columns. There are several distance libraries; where a > 10 and b > 10.

[0063] Each range database corresponds to a set of radar observation parameters; each set of radar observation parameters includes the reflectivity factor Z. H Differential reflectivity Z DR Zero-lag correlation coefficient ρ hv Differential phase standard deviation SD(Φ) DP ) and radar radial velocity u; the length of each range is .

[0064] In this embodiment, the radar radial velocity u-map of the tornado in Shengze, Suzhou is shown below. Figure 2 As shown.

[0065] Step 2: Identify strong convection regions: Combining objective convection identification methods with strong convection echo top height and vertical cumulative liquid water content (VIL), strong convection regions are identified from the local forecast area in Step 1.

[0066] The method for identifying strong convection regions described above preferably includes the following steps:

[0067] Step 2-1: Identify convection regions: Use objective convection identification methods to identify the convection regions and their locations.

[0068] Step 2-2, Calculate VIL: For each distance reservoir in the identified convection region, calculate the strong convection echo top height and the vertical cumulative liquid water content (VIL). The specific calculation formula is as follows:

[0069] ;

[0070] Among them, Z k Z represents the reflectivity factor Z at the corresponding distance in the k-th elevation angle. H Where 1≤k≤n-1.

[0071] Z k+1 Z represents the reflectivity factor Z of the distance library at the elevation angle of the (k+1)th layer. H .

[0072] This represents the height difference between the elevation angles of two adjacent floors.

[0073] Step 2-3: Identify Severe Convection Regions: Compare the VIL calculated in Step 2-2 with the set VIL threshold. Convection regions exceeding the set VIL threshold are called severe convection regions. The distance database located within severe convection regions is called the severe convection distance database.

[0074] Step 3: Calculate radar inversion vorticity For each strong convection range library in the lowest elevation angle and the penultimate elevation angle, based on their respective radar radial velocity u and range library length... The corresponding radar inversion vorticity is obtained through inversion calculation. .

[0075] The above radar inversion vorticity The preferred method for obtaining the [method] includes the following steps.

[0076] Step 3-1, Linear Interpolation: Using bilinear interpolation, the radar radial velocity u is interpolated to the Cartesian coordinate system.

[0077] The preferred horizontal resolution for the above interpolation is 200 m.

[0078] Step 3-2: Calculate the position of the distance library: In the Cartesian coordinate system, calculate the position of the distance library in the i-th row and j-th column. The specific calculation formula is as follows:

[0079] (3-1);

[0080] In the formula, The distance from the distance library in the i-th row and j-th column to the radar center; Let be the radar azimuth angle of the range library in the i-th row and j-th column.

[0081] Step 3-3: Construct the radar radial velocity estimation of the range library in the i-th row and j-th column. The expression is as follows:

[0082] (3-2);

[0083] In the formula, Let be the radar radial velocity at weather forecast point A, as observed by radar.

[0084] Let be the radial shear of the distance library in the i-th row and j-th column, an unknown variable.

[0085] This is the distance from weather forecast point A to the radar center.

[0086] Steps 3-4: Define the value function Specifically:

[0087] (3-3);

[0088] In the formula, The actual radar radial observation value of the range library in the i-th row and j-th column is given.

[0089] Steps 3-5: Solve for radar inversion vorticity Substituting equation (3-1) into equation (3-2), and then substituting (3-2) into equation (3-3), we can solve equation (3-3) by minimizing the value function to obtain the partial derivative. Thus, the radar inversion vorticity is obtained. The calculation formula is:

[0090] ;

[0091] In the formula, Let be the radar radial velocity of the range ku in the i-th row and j-th column; where 1≤i≤a, 1≤j≤b.

[0092] The weighting coefficient for the distance database in the i-th row and j-th column is a set value that increases as the distance database moves further away from the center of the weather forecast point A.

[0093] The position of the distance from the library in the i-th row and j-th column can be calculated by interpolation.

[0094] This embodiment solves... At the same time, it can also be solved. .

[0095] Step 3-6, Vortex Coordinate Transformation: Transform the radar-inverted vorticity obtained in Step 3-5. Interpolate back to the original polar coordinates.

[0096] In traditional methods, azimuth shear is usually estimated by using the velocity variation of adjacent distance reservoirs. The advantage of this invention in estimating vorticity using local functional theory is that it utilizes radial velocity information over the entire 2×2 km range, and the results are not significantly affected by missing measurements at single or multiple points or by noise and other interference.

[0097] Step 4: Identify tornado features, specifically including the following steps:

[0098] Step 4-1: Construct a Bayesian classification model Bayesian classification model In For input sample data, and ,in, , , , , Bayesian classification model In It belongs to the tornado category, and ,in, Indicates no tornadoes. It indicates the presence of a tornado.

[0099] The above Bayesian classification model The preferred expression is:

[0100] ;

[0101] In the formula, The normalization parameter represents the classification category of the l-th tornado; where l = 1 or 2.

[0102] The prior probability density function represents the classification category of the l-th tornado.

[0103] The conditional similarity probability density function represents the l-th type of tornado classification.

[0104] Step 4-2: Identify tornado features at the lowest elevation angle: For each strong convection distance in the lowest elevation angle database, the Bayesian classification model constructed in Step 4-1 is used. Tornado feature identification is performed; assuming that the lowest elevation angle identifies M strong convective distances with tornado features.

[0105] Step 4-3, Identifying Tornado Features at the Penultimate Elevation Angle: Using the tornado feature identification method from Step 4-2, N strong convective distances with tornado features were identified at the penultimate elevation angle.

[0106] Step 5, Tornado Forecast: Tornado forecast types are divided into three categories: no tornadoes, tornadoes, and severe tornadoes. Among them, the wind level of severe tornadoes is preferably not less than EF2.

[0107] The preferred method for tornado forecasting described above is:

[0108] A. Strong tornado: When M≥7, N≥7 and When all three conditions are met simultaneously, a strong tornado is predicted at weather forecast point A, and a warning is issued; among them... Radar inversion vorticity for all strong convective range libraries in the lowest and penultimate elevation angles. The maximum value; The maximum threshold for vorticity intensity is preferably [value missing]. =10×10 -3 s -1 .

[0109] B. Tornado: When M≥7 or N≥7 (that is, there are at least 7 strong convective distances with tornado characteristics in either the lowest elevation angle or the penultimate elevation angle) and At that time, a tornado was predicted at forecast point A and a warning was issued; among them, This is the minimum threshold for vorticity intensity, and In this embodiment, preferably... =5×10 -3 s -1 .

[0110] C. No tornadoes: When the conditions for strong tornadoes and tornadoes are not met, there are no tornadoes at the forecast point A.

[0111] In this embodiment, the tornado forecast results for Shengze, Suzhou are as follows: Figure 3 As shown, the red triangle represents the location of the identified powerful tornado.

[0112] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A method for tornado classification and early warning based on multi-dimensional variable characteristics, characterized in that: Includes the following steps: Step 1, Radar Weather Scan: The radar performs a radar scan of the local forecast area according to the set sampling frequency; during each radar scan, the data is processed from bottom to top or from top to bottom. n The radar scan is performed at a set elevation angle; after each elevation angle scan, a radar scan image is obtained; each radar scan image has a [missing information - likely a specific feature or characteristic]. a OK b Columns are evenly arranged There are several range databases; each range database corresponds to a set of radar observation parameters; each set of radar observation parameters includes the reflectivity factor Z. H Differential reflectivity Z DR Zero-lag correlation coefficient ρ hv Differential phase standard deviation SD(Φ) DP and radar radial velocity u ;in ； n ≥3, n =1 indicates the lowest elevation angle. n =2 indicates the second-to-last elevation angle; a >10, b >10; the length of each distance library is The center of the local forecast area is weather forecast point A; Step 2: Identify strong convection regions: Combining objective convection identification methods with strong convection echo top height and vertical cumulative liquid water content (VIL), identify strong convection regions from the local forecast area in Step 1. The distance reservoir located in a region of strong convection is called the strong convection distance reservoir; Step 3: Calculate radar inversion vorticity For each strong convection range library in the lowest elevation angle and the penultimate elevation angle, based on their respective radar radial velocity... u and distance to the library length The corresponding radar inversion vorticity is obtained through inversion calculation. ; Step 4: Identify tornado features, specifically including the following steps: Step 4-1: Construct a Bayesian classification model Bayesian classification model In For input sample data, and Bayesian classification model middle The category is tornado, and ,in, Indicates no tornadoes. This indicates the presence of a tornado; Step 4-2: Identify tornado features at the lowest elevation angle: For each strong convection distance in the lowest elevation angle database, the Bayesian classification model constructed in Step 4-1 is used. Perform tornado feature identification; assume that M strong convective distances with tornado features are identified at the lowest elevation angle. Step 4-3, Identifying Tornado Features at the Penultimate Elevation Angle: Using the tornado feature identification method from Step 4-2, N strong convection distances with tornado features were identified at the penultimate elevation angle. Step 5, Tornado Forecasting: Tornado forecasts are categorized into three types: no tornadoes, tornadoes, and severe tornadoes. The specific forecasting method is as follows: A. Strong tornado: When M≥7, N≥7 and When all three conditions are met simultaneously, the weather forecast point A is likely to experience a strong tornado; among them... Radar inversion vorticity for all strong convective range libraries in the lowest and penultimate elevation angles. The maximum value; This represents the maximum threshold for vorticity intensity. B. Tornado: When M≥7 or N≥7 and At that time, a tornado was observed at forecast point A; among them, This is the minimum threshold for vorticity intensity, and ; C. No tornadoes: When the conditions for strong tornadoes and tornadoes are not met, there are no tornadoes at the forecast point A.

2. The tornado classification and early warning method based on multidimensional variable features according to claim 1, characterized in that: Step 2, the method for identifying areas of strong convection, specifically includes the following steps: Step 2-1: Identify convection regions: Use objective convection identification methods to identify the convection regions and their locations; Step 2-2, Calculate VIL: Calculate the strong convection echo top height and vertical cumulative liquid water content (VIL) for the identified convection regions; Step 2-3: Identify strong convection areas: Compare the VIL calculated in Step 2-2 with the set VIL threshold. Convection areas that exceed the set VIL threshold are called strong convection areas.

3. The tornado classification and early warning method based on multidimensional variable features according to claim 1, characterized in that: In step 3, the formula for calculating radar-inverted vorticity is: ； In the formula, For the first i OK j The radar radial velocity of the column range library; where 1≤ i ≤ a ,1≤ j ≤ b ; For the first i OK j The weighting coefficient of the distance database is set to increase as the distance database moves further away from the center of weather forecast point A. For the first i OK j The position of a column relative to the library can be calculated through interpolation.

4. The tornado classification and early warning method based on multidimensional variable features according to claim 3, characterized in that: Step 3, the method for obtaining the radar inversion vorticity calculation formula, includes the following steps: Step 3-1, Linear Interpolation: Using bilinear interpolation, the radar radial velocity is... u Interpolate to the Cartesian coordinate system; Step 3-2: Calculate the distance to the library position: In the Cartesian coordinate system, calculate the distance to the library position. i OK j Column distance from the library The specific calculation formula is as follows: （3-1）； In the formula, For the first i OK j The distance from the column to the radar center; For the first i OK j Radar azimuth of the range database; Step 3-3, Construct the first i OK j Radar radial velocity estimation of the range library The expression is as follows: （3-2）； In the formula, The radar radial velocity at weather forecast point A is the radar observation value. For the first i OK j Radial shear of column distance library, unknown variable; The distance from weather forecast point A to the radar center; Steps 3-4: Define the value function Specifically: （3-3）； In the formula, For the first i OK j Actual radar radial observations of the range library; Steps 3-5: Solve for radar inversion vorticity Substitute equation (3-1) into equation (3-2), then substitute equation (3-2) into equation (3-3), and solve equation (3-3) to obtain the radar-derived vorticity. ; Step 3-6, Vortex Coordinate Transformation: Transform the radar-inverted vortex obtained in Step 3-5. Interpolate back to the original polar coordinates.

5. The tornado classification and early warning method based on multidimensional variable features according to claim 1, characterized in that: In step 1, the local forecast area is centered on weather forecast point A. km area.

6. The tornado classification and early warning method based on multidimensional variable features according to claim 1, characterized in that: In step 1, n =9.

7. The tornado classification and early warning method based on multidimensional variable features according to claim 1, characterized in that: In step 5, the wind force of the strong tornado is no less than EF2.

8. The tornado classification and early warning method based on multidimensional variable features according to claim 1, characterized in that: In step 5, .

9. The tornado classification and early warning method based on multidimensional variable features according to claim 1, characterized in that: The radar in step 1 is a dual-polarization radar.

Images