Thunderstorm identification method based on radar reflectivity data

Abstract

The invention aims to provide a thunderstorm identification method based on radar reflectivity data, which creatively introduces a connected domain region marking algorithm for binary image analysis in computer vision into the field of radar data processing, expands the connected domain region marking algorithm from two-dimensional radar data processing to three-dimensional radar data, and applies an improved region growing algorithm to identify the thunderstorm. The internal structure of the thunderstorm is accurately identified, the false merging problem existing in a traditional method is solved, and the accuracy and practicability of thunderstorm identification are improved.

Description

Technical Field

[0001] This invention relates to the field of meteorological monitoring and forecasting technology, and in particular to a thunderstorm identification method based on radar reflectivity data. Background Technology

[0002] Thunderstorms are a type of severe convective weather phenomenon characterized by their sudden onset, short lifespan, strong locality, and high risk of causing disasters. Accurately identifying thunderstorms and analyzing their internal structure is of great significance for the early warning and forecasting of severe convective weather.

[0003] Traditional thunderstorm identification methods often employ a single threshold method, such as classifying areas with radar reflectivity greater than 35 dBZ as strong convection areas. This method suffers from problems such as a single threshold and unclear definition, making it difficult to effectively distinguish the complex internal structure of thunderstorms. Furthermore, conventional methods have limited effectiveness in handling the splitting and merging processes during thunderstorm formation and dissipation, and are prone to producing false merging phenomena.

[0004] In the prior art, CN110942111B discloses a method for identifying strong convective cloud clusters based on radar and satellite data fusion, but this method relies on satellite data, and its identification effect is limited when satellite data is unavailable. CN107229084B discloses a method for identifying convective systems based on a tree structure, but this method has high computational complexity and poor real-time performance.

[0005] Therefore, a thunderstorm identification method is needed that can accurately identify the internal structure of thunderstorms while ensuring computational efficiency. Summary of the Invention

[0006] The purpose of this invention is to provide a thunderstorm identification method based on radar reflectivity data. It creatively introduces the connected component region labeling algorithm used in binary image analysis in computer vision into the field of radar data processing, and extends it from two-dimensional radar data processing to three-dimensional radar data. By applying an improved region growing algorithm, it accurately identifies the internal structure of thunderstorms, solves the false merging problem in traditional methods, and improves the accuracy and practicality of thunderstorm identification.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a thunderstorm identification method based on radar reflectivity data, comprising the following steps: Step S1: Acquire three-dimensional radar reflectivity data and perform preprocessing; Step S2: Set a first reflectivity threshold and a second reflectivity threshold, wherein the second reflectivity threshold is higher than the first reflectivity threshold; Step S3: Using a connected component labeling algorithm, perform connected component analysis on radar grid data that are above the first reflectivity threshold to identify large thunderstorm clusters; Step S4: Using the same method as in step S3, perform connected component analysis on radar grid data that are above the second reflectivity threshold to identify thunderstorm cells and calculate the centroid and size of each thunderstorm cell; Step S5: Based on the size ratio between thunderstorm cells and their respective large thunderstorm clusters, select the centroids of thunderstorm cells that meet the conditions as seed points for regional growth. Step S6: Using a region growing algorithm, with the selected seed point as the center, perform region growing based on neighborhood connectivity and reflectivity threshold conditions to achieve separation and identification of the internal structure of thunderstorms.

[0008] Preferably, the preprocessing in step S1 includes: if two-dimensional recognition is performed, the three-dimensional radar data is converted into a two-dimensional combined reflectivity, that is, the maximum value of all vertical layers at each horizontal point is calculated; if three-dimensional recognition is performed, the radar data is interpolated to a three-dimensional grid with a vertical resolution of 0.5 km.

[0009] Preferably, the preprocessing in step S1 includes: quality control of the radar data; if the radar reflectivity is higher than the maximum reflectivity threshold, the grid data is considered invalid.

[0010] Preferably, the connected component labeling algorithm in step S3 includes: if processing two-dimensional radar data, the current grid neighborhood can be defined as a 4-neighborhood, including (x-1, y), (x+1, y), (x, y+1), (x, y-1) or an 8-neighborhood, including (x-1, y-1), (x-1, y), (x-1, y+1), (x, y-1), (x, y+1), (x+1, y-1), (x+1, y), (x+1, y+1); if processing three-dimensional radar data, the current grid neighborhood needs to be extended to three dimensions, including but not limited to (x-1, y, z), (x, y-1, z), (x, y, z-1), (x+1, y, z), (x, y+1, z), (x, y, z-1), (x+1, y, z), (x, y, z-1 ... z+1), and its diagonal neighborhood voxels, such as (x+1, y+1, z+1), to ensure 6-connectivity, 18-connectivity, and 26-connectivity in 3D space.

[0011] Preferably, the size of the large thunderstorm cluster identified in step S3 should not be smaller than the minimum thunderstorm size.

[0012] Preferably, the size of the thunderstorm cell identified in step S4 should not be smaller than the minimum cell size.

[0013] Preferably, the size calculation of the large thunderstorm cluster and thunderstorm cell includes: For two-dimensional recognition, the area of ​​each connected component is calculated, the number of grid cells in the two-dimensional connected components is counted, and then multiplied by the horizontal spatial resolution of the grid data. The calculation formula is as follows:

[0014] For 3D recognition, the volume of each connected component is calculated, the number of grid points in the 3D connected component is counted, and then multiplied by the horizontal and vertical spatial resolutions of the grid data. The calculation formula is as follows:

[0015] Where grid_num is the number of grid points in the connected domain, reso is the horizontal spatial resolution of the grid data, and h is the vertical spatial resolution.

[0016] Preferably, the centroid calculation of the thunderstorm cell in step S4 includes: For connected components in two-dimensional recognition, each connected component is considered to be an irregular polygon with uniform density. This is achieved by dividing it into n finite simple shapes A1, A2, ..., A n We can use the following formula to calculate the centroid coordinates G(x, y) of the connected component:

[0017] The centroid of these simple figures is represented as (G ix G iy ), with an area of ​​S i (i=1, 2, ..., n). The centroid of the entire connected region is the area-weighted average of the centroids of these simple graphs.

[0018] For the connected regions in 3D recognition, they are considered to be irregular cubes with uniform density, and their centroids G(x) are... c ,y c , z c The calculation formula is as follows:

[0019]

[0020]

[0021] In the calculation, the irregular cube is divided into multiple regular solid shapes with easily computed centroids (e.g., small cubes, triangular prisms, tetrahedrons, etc.), and the centroid coordinates (x, y, y) of each regular part are calculated. i , y i , z i ) and volume V i Finally, the centroid coordinates of the entire three-dimensional connected domain are calculated by weighted averaging.

[0022] Preferably, the seed point selection conditions in step S5 include: A large thunderstorm cluster contains at least two thunderstorm cells. The size of each thunderstorm cell is more than a first proportion threshold relative to the sum of the sizes of all thunderstorm cells in the large thunderstorm cluster, and the sum of the sizes of all thunderstorm cells is more than a second proportion threshold relative to the total size of the large thunderstorm cluster.

[0023] Preferably, the region growing algorithm includes: optimizing the selection of the initial seed point. As described in step 5, if an unvisited pixel or voxel in the centroid neighborhood of each thunderstorm cell meets the region growing conditions, it is considered that the pixel or voxel is similar to it, and it is included in the region of the thunderstorm cell as the seed point for the next growth. This process continues to expand outward until the growth stopping condition is met, and no new pixels or voxels are included in the region, thus completing the entire thunderstorm cell region growing process.

[0024] Preferably, the region growth stopping condition in step S6 is: there are no unvisited pixels around the current growth region that meet the reflectivity threshold condition.

[0025] This invention provides a thunderstorm identification method based on radar reflectivity data. The identification method includes: setting two radar reflectivity thresholds, using a connected component labeling algorithm to identify large thunderstorm clusters and individual thunderstorm cells within large thunderstorm clusters, and using a region growing algorithm to separate and identify the internal structure of thunderstorms. This method can accurately identify the internal structure of thunderstorms, solve the false merging problem in traditional methods, and improve the accuracy and practicality of thunderstorm identification.

[0026] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a flowchart illustrating the thunderstorm identification method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the internal structure analysis of a two-dimensional thunderstorm cluster according to an embodiment of the present invention; Detailed Implementation

[0029] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0030] The terms "comprising" and "having," and any variations thereof, in the specification, embodiments, claims, and drawings of this invention are intended to cover non-exclusive inclusion, such as including a series of steps or units.

[0031] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0032] like Figure 1 The flowchart shown illustrates the thunderstorm identification method based on radar reflectivity data in this embodiment of the invention. The identification method includes the following steps: Step S1: Data acquisition and preprocessing.

[0033] We obtained a nationwide 3D radar reflectivity factor mosaic product, covering the entire Chinese region, with a time resolution of 6 minutes and a horizontal resolution of 0.01°×0.01°. The vertical detection altitude ranges from 0.5km to 16km, with intervals of 0.5km for altitudes below 8km and 1km for altitudes above 8km.

[0034] If two-dimensional thunderstorm identification is used, the three-dimensional radar data is preprocessed into a combined reflectivity; if three-dimensional thunderstorm identification is used, the height layer of the data is interpolated to three-dimensional radar data with uniform height intervals, i.e., a vertical resolution of 0.5km.

[0035] Step S2: Based on the physical characteristics of thunderstorm clouds, such as their horizontal scale and the intensity of radar reflectivity factors, set the parameters required for thunderstorm identification, including: Recognition type (two-dimensional or three-dimensional) First reflectivity threshold: 40 dBZ Second reflectivity threshold: 45dBZ Maximum reflectivity threshold: 80 dBZ Minimum thunderstorm size: 50km 2 (Two-dimensional) or 50km 3 (3D) Maximum thunderstorm size: 10 5 km 2 (Two-dimensional) or 10 5 km 3 (3D) Minimum unit size: 20km 2(Two-dimensional) or 20km 3 (3D) First proportional threshold: 0.05 Second proportional threshold: 0.5.

[0036] Step S3: Identification of large thunderstorm clusters.

[0037] Figure 2 A schematic diagram of the internal structure analysis of a two-dimensional thunderstorm cluster provided by an embodiment of the present invention is shown.

[0038] In stage 1, based on a first reflectivity threshold (40 dBZ), a connected component labeling algorithm is used to identify multiple large thunderstorm clusters in space. In this step, regions with an area or volume smaller than the minimum thunderstorm size and larger than the maximum thunderstorm size are filtered out, and the valid thunderstorm regions are labeled, such as Cell 1.

[0039] For two-dimensional radar data, the connected component labeling algorithm scans the data from top to bottom and from left to right. During the first traversal, it checks the non-empty pixels above and to the left of (x, y) that are defined as neighbors, and labels the connected components of the radar data. Then, it uses a union-find data structure to store the relationships between the connected components and connects pixels with the same label. After the second traversal, the final connected component labeling result of the radar data is formed.

[0040] For 3D radar data, a connected component labeling algorithm is applied to sequentially check the 3D neighborhood of each non-empty voxel along the X, Y, and Z axes. Specifically, non-empty voxels in (x-1, y, z), (x, y-1, z), and (x, y, z-1) in the 6-neighborhood are scanned, or non-empty voxels in the 26-neighborhood whose X, Y, or Z coordinate values ​​are less than or equal to the current (x, y, z) voxel are scanned. At the same time, the label equivalence relation is recorded, and the final 3D connected component is output after the second traversal.

[0041] Step S4: Thunderstorm cell identification. Based on the second reflectivity threshold (45dBZ), thunderstorm cells are identified. A valid thunderstorm cell should have a size no smaller than the minimum cell size defined in the thunderstorm identification parameters of step S2. Valid thunderstorm cells are marked as follows: Figure 2 In stage 2, Cell 1-1 and Cell 1-2 are located in Cell 1, and the centroid coordinates and size of each thunderstorm cell are calculated.

[0042] Step S5: Seed point selection. For each thunderstorm cell, find the large thunderstorm cluster to which it belongs. When the following conditions are met, use the centroid of the thunderstorm cell as the initial seed point for region growth: (1) A large thunderstorm cluster contains at least two individual thunderstorms; (2) The ratio of the size of each thunderstorm cell to the sum of the sizes of all thunderstorm cells belonging to the same storm cluster exceeds the first ratio threshold (0.05). (3) The ratio of the sum of the sizes of all thunderstorm cells belonging to the same storm cluster to the total size of the storm cluster exceeds the second proportional threshold (0.5).

[0043] Step S6: Region growing. A region growing algorithm is used, with the selected seed point as the center (e.g., ...). Figure 2 The region growing algorithm involves growing regions marked with "+" in stage 3. The algorithm includes the following steps: (1) Using the initial seed point as the center, check its neighboring points; (2) Incorporate unvisited points in the neighborhood with reflectivity greater than the first reflectivity threshold (40dBZ) into the growth region and mark them as visited; (3) Use the newly included points as new seed points and repeat the above process; (4) When no new points can be incorporated, the growth process stops, that is, the separation of the internal structure of the thunderstorm cluster is completed (e.g., Figure 2 Stage 4).

[0044] Beneficial effects: Compared with existing technologies, this invention effectively identifies the internal structure of thunderstorms and avoids false merging by using a dual threshold design and region growing algorithm; combined with a connected region labeling algorithm, it improves the accuracy and efficiency of thunderstorm identification; by selecting seed points based on size ratio conditions, it improves the traditional region growing algorithm, optimizes the starting position of region growing, and improves the accuracy of internal structure separation; it supports two-dimensional and three-dimensional identification modes to adapt to different application needs; and it can achieve accurate identification with a single radar data source without relying on satellite data.

[0045] The above specific embodiments further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A thunderstorm identification method based on radar reflectivity data, characterized in that, The identification method includes the following steps: Step S1: Acquire three-dimensional radar reflectivity data and perform preprocessing, including height layer interpolation or combined reflectivity calculation; Step S2: Set a first reflectivity threshold and a second reflectivity threshold, wherein the second reflectivity threshold is higher than the first reflectivity threshold; Step S3: Using a connected component labeling algorithm, perform connected component analysis on radar grid data that are above the first reflectivity threshold to identify large thunderstorm clusters; Step S4: Using the same method as in step S3, perform connected component analysis on radar grid data that are above the second reflectivity threshold to identify thunderstorm cells and calculate the centroid and size of each thunderstorm cell; Step S5: Based on the size ratio between thunderstorm cells and their respective large thunderstorm clusters, select the centroids of thunderstorm cells that meet the conditions as seed points for regional growth. Step S6: Using a region growing algorithm, with the selected seed point as the center, perform region growing based on neighborhood connectivity and reflectivity threshold conditions to separate and identify the internal structure of thunderstorms.

2. The method according to claim 1, characterized in that, The preprocessing steps include: if two-dimensional recognition is to be performed, the three-dimensional radar data is converted into a two-dimensional combined reflectivity by calculating the maximum vertical layer value of each horizontal point; if three-dimensional recognition is to be performed, the radar data is interpolated to a three-dimensional grid with a vertical resolution of 0.5 km.

3. The method according to claim 1, characterized in that, The reflectivity of the radar grid data marked with connected regions should not exceed the maximum reflectivity threshold.

4. The method according to claim 1, characterized in that, The size of the identified large thunderstorm cluster should not be smaller than the smallest thunderstorm size.

5. The method according to claim 1, characterized in that, The size of the identified thunderstorm cells should not be smaller than the minimum cell size.

6. The method according to claim 1, characterized in that, The neighborhood definitions used in the connected component labeling algorithm and the region growing algorithm include: for two-dimensional data, 4-neighborhood or 8-neighborhood is used; for three-dimensional data, 6-neighborhood, 18-neighborhood or 26-neighborhood is used.

7. The method according to claim 1, characterized in that, The method for calculating the size of large thunderstorm clusters and individual thunderstorm cells is achieved by multiplying the number of grids in the two-dimensional connected domain by the spatial resolution of the grid data.

8. The method according to claim 1, characterized in that, The centroid calculation of the large thunderstorm cluster and individual thunderstorm cells is achieved by treating the connected domain as an irregular polygon or irregular cube with uniform density, dividing the irregular two-dimensional or three-dimensional space into regular two-dimensional or three-dimensional shapes, calculating the centroid coordinates and area or volume of the regular two-dimensional or three-dimensional shapes respectively, and finally performing a weighted average of the centroid coordinates with area or volume as the weight.

9. The method according to claim 1, characterized in that, The selection criteria for the regional growth seed points include: the large thunderstorm cluster contains at least two thunderstorm cells; the size of each thunderstorm cell accounts for more than a first proportion threshold of the sum of the sizes of all thunderstorm cells in the large thunderstorm cluster; and the sum of the sizes of all thunderstorm cells accounts for more than a second proportion threshold of the total size of the large thunderstorm cluster.

10. The method according to claim 1, characterized in that, The growth in the region is stopped when there are no unvisited pixels in the vicinity of the current growth region that meet the reflectivity threshold condition.

Images