A storm tracking analysis method based on multi-source rainfall monitoring data cooperation

Abstract

This invention discloses a rainstorm tracking and analysis method based on multi-source rainfall monitoring data collaboration. It constructs a dual-source input channel for rainfall in the cloud from a rain-measuring radar and for rainfall measured at ground rain gauges, collecting radar reflectivity, its inverted rainfall, and ground-measured rainfall data, and extracting key parameters. A time window of 1-2 hours is set to trace back from the current point, with the first appearance of the rainstorm center as the tracking starting point. The 1-hour window is iteratively updated to track the evolution of the rainstorm center's location, coverage area, and other dimensions. An adaptive weighted intelligent algorithm is used to predict the movement direction of the rainstorm center based on historical data. Three types of cloud physical characteristics—resident, mobile, and train-like multi-center mobile—are integrated to construct a complete analysis model and output results. This invention achieves full-link rainfall monitoring, improves the comprehensiveness of rainstorm tracking and prediction accuracy, and solves the problems of blind spots in monitoring from single data sources, poor collaboration among multi-source data, and insufficient accuracy in rainstorm tracking and prediction in existing technologies.

Description

Technical Field

[0001] This invention relates to the fields of rain-measuring radar, rainstorm monitoring, meteorological data analysis, and disaster early warning technology. Specifically, it relates to a rainstorm tracking and analysis method based on the collaboration of multi-source rainfall monitoring data, which is an intelligent rainstorm tracking and analysis method based on the collaboration of rain-measuring radar cloud rain and ground station rainfall. Background Technology

[0002] Heavy rain, as a frequent meteorological disaster, is characterized by its sudden onset, rapid evolution, and concentrated impact. Accurate and real-time tracking of the center location, evolution process, and movement direction of heavy rain is crucial for early warning and reducing disaster losses. Currently, heavy rain monitoring mainly relies on two types of data sources. One type is cloud-based rainfall data collected by rain-measuring radar. This data can characterize the formation state of aerial rainfall through physical quantities such as radar reflectivity. Some technologies can calculate cloud-based rainfall based on reflectivity inversion, enabling early detection of rainfall. However, this type of data only reflects the potential for rainfall in the clouds and cannot accurately correspond to the actual rainfall on the ground. It is prone to monitoring deviations due to factors such as cloud dissipation and changes in water vapor transport. The other type is measured data from ground-based rain gauges, which can accurately record the actual rainfall on the ground. However, there are limitations in the density of station deployment, making it difficult to cover the entire area. Furthermore, there is a lag in the response to the dynamic movement of the heavy rain center. Relying solely on this type of data cannot achieve early tracking and trend prediction of heavy rain.

[0003] In existing rainstorm tracking technologies, some solutions rely on a single data source for analysis, resulting in monitoring blind spots or insufficient accuracy. A few multi-source data collaboration solutions simply overlay radar and ground station data without establishing targeted collaboration logic to address the differences in the characteristics of rain in clouds and rain falling to the ground, nor do they optimize tracking strategies based on the physical characteristics of rainstorm clouds themselves. Furthermore, traditional rainstorm tracking often employs a fixed time interval analysis model, lacking the ability to dynamically capture the entire process of a rainstorm center's evolution from its emergence. It also lacks differentiated prediction logic for different types of rainstorms, such as stationary, mobile, and multi-center mobile rainstorms, leading to insufficient accuracy in predicting the direction of rainstorm center movement to meet practical disaster prevention and mitigation needs. Summary of the Invention

[0004] The purpose of this invention is to provide a rainstorm tracking and analysis method based on the collaboration of multi-source rainfall monitoring data. It addresses the technical problems of existing rainstorm tracking technologies, such as monitoring blind spots due to single data sources, poor collaboration of multi-source data, and insufficient tracking and prediction accuracy due to the lack of integration with the physical characteristics of rainstorm clouds. The proposed method is an intelligent rainstorm tracking and analysis method based on the collaboration of rain-measuring radar cloud rain and ground station rainfall. It enables accurate and dynamic tracking of the rainstorm center and scientific prediction of its movement direction, providing reliable technical support for disaster prevention, mitigation and early warning.

[0005] To achieve the above objectives, the solution of the present invention is as follows: A method for tracking and analyzing rainstorms based on the collaborative analysis of multi-source rainfall monitoring data includes the following steps: Step 1: Construct a dual-source rainfall data input channel: including a rain-measuring radar cloud rainfall data channel and a ground-based rain gauge measured rainfall data channel, wherein: The cloud rain data channel of the rain measuring radar collects radar reflectivity and inverted rainfall calculated by a high-precision inversion model based on the radar reflectivity. The landing area parameters and rainfall threshold parameters corresponding to the radar reflectivity and rainfall are extracted as cloud rainfall characterization data. Ground-based rain gauges collect measured rainfall data from the ground. The measured rainfall data is then processed by spatial interpolation and planar gridding based on the landing area to generate a planar rainfall model covering the monitoring area. During the modeling process, abnormal data is identified and filtered out simultaneously. This planar rainfall model is then used as the representation data of the ground-based rainfall. Cloud-based precipitation data and ground-based precipitation data are referred to as dual-source precipitation data. Step 2: Set the rainstorm tracking time window: For dual-source rainfall data, take the current time as the cutoff point and trace back 1-2 hours as the rainstorm tracking analysis time range. The moment when the rainstorm center first appears within the time range is taken as the tracking start node. Step 3: Start dynamic cyclic iterative tracking of the rainstorm center: For each new set of dual-source rainfall data, update the 1-hour time window for iterative calculation and continuously track the evolution of the rainstorm center in the past 1 hour; Step 4: Prediction of the movement direction of the rainstorm center: Based on the dual-source rainfall data and rainstorm center evolution data collected in multiple historical iterations, the movement direction of the rainstorm center at the next moment is predicted through intelligent algorithm fitting analysis. Step 5: Multi-element fusion to construct an intelligent analysis model: Introduce the physical occurrence characteristics of rainstorm clouds as auxiliary analysis elements, and fuse the cloud physical occurrence characteristics with dual-source rainfall data, rainstorm center evolution data, and rainstorm center movement direction prediction results in multiple dimensions to construct a complete intelligent rainstorm tracking and analysis model, and output rainstorm tracking and prediction results.

[0006] The solution further states that the physical characteristics of the rainstorm clouds include at least three types: stationary, mobile, and train-like multi-center mobile.

[0007] The scheme is further described as follows: The operation logic of the high-precision inversion model is as follows: with radar reflectivity intensity data as the core input, after feature denoising, parameter calibration and error correction, the output is the inverted rainfall that matches the actual rainfall in the cloud, and the calculation accuracy of the inverted rainfall is positively correlated with the radar reflectivity acquisition resolution and the inversion algorithm.

[0008] The solution further involves: performing time synchronization calibration on the dual-source rainfall data: aligning the cloud rainfall data from the rain-measuring radar and the measured rainfall data from the ground rain monitoring station according to a preset time granularity to ensure the spatiotemporal consistency of the two types of data within the same monitoring period and improve the accuracy of collaborative analysis.

[0009] The scheme further states that the criteria for determining the rainstorm center are as follows: combining the reflectivity threshold, inverted rainfall threshold, and corresponding coverage area threshold in the rainstorm data from the rain-measuring radar cloud, and matching the rainstorm level threshold of the actual rainfall data from the ground rain gauge station, and simultaneously meeting the preset collaborative determination conditions, it is determined to be a rainstorm center.

[0010] The solution further includes: in step 4, the intelligent algorithm has the ability to adaptively adjust weights; based on the physical characteristics of different types of rainstorm clouds introduced in step 5, the fitting weights of historical data are adjusted, and differentiated prediction logic is adopted for stationary, mobile, and train-type multi-center mobile rainstorms to improve the accuracy of predicting the direction of movement of the rainstorm center.

[0011] The solution further includes: in step 3, the tracking dimensions of the evolution process of the rainstorm center include at least location coordinates, coverage area, rainfall intensity level, and core area diffusion rate. Each iteration updates the data of each dimension to form a rainstorm evolution time series archive to support subsequent prediction and analysis.

[0012] The solution further includes: the method further includes preprocessing the dual-source rainfall data, the preprocessing including outlier removal, missing value completion and data standardization, wherein outlier removal is determined based on a preset reasonable range of rainfall physical quantities, and missing value completion is achieved by interpolation of adjacent time periods in the same region.

[0013] The solution further states that the preset collaborative judgment conditions are dynamically adjusted based on the climate characteristics and terrain conditions of the monitoring area to adapt to the rainstorm formation patterns of different regions, thereby improving the regional adaptability of the rainstorm center judgment.

[0014] Compared with the prior art, the present invention has the following advantages: (1) The dual-source data collection of rain in the cloud by rain measuring radar and rain falling to the ground by ground station is adopted to make up for the blind spots of monitoring by a single data source, realize the full-link characterization of rainfall from cloud formation to ground landing, and improve the comprehensiveness of rainstorm monitoring; (2) Dynamic tracking is carried out with a base time window of 1-2 hours and a sliding iteration granularity of 1 hour. Combined with multi-dimensional evolution data recording, the entire process of the rainstorm center is accurately captured without missing any key evolution stages. (3) By integrating the physical characteristics of rainstorm clouds and combining them with intelligent algorithms that have adaptive weight adjustment capabilities, differentiated analysis and prediction logic is adopted for different types of rainstorms, which greatly improves the accuracy of predicting the direction of movement of the rainstorm center. (4) The criteria for determining the center of rainstorm can be dynamically adapted to regional climate and topographic features. The preprocessing and synchronous calibration steps ensure data quality, making the method widely applicable and meeting the rainstorm monitoring and early warning needs of multiple fields such as meteorology, hydrology, and disaster prevention and mitigation.

[0015] The invention will be further explained in detail below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the dual-source rainfall data input channel architecture; Figure 2 This is a schematic diagram of the dynamic tracking and iteration process for rainstorms. Figure 3 This is a schematic diagram of a multi-factor fusion analysis model. Detailed Implementation

[0017] A rainstorm tracking and analysis method based on multi-source rainfall monitoring data collaboration is an intelligent rainstorm tracking and analysis method based on the collaboration of cloud-based rainfall from rainfall measuring radar and ground-based rainfall, including the following steps: Step 1: Construct a dual-source rainfall data input channel: including a rain-measuring radar cloud rainfall data channel and a ground-based rain gauge measured rainfall data channel, wherein: The cloud rain data channel of the rain measuring radar collects radar reflectivity and inverted rainfall calculated by a high-precision inversion model based on the radar reflectivity. The landing area parameters and rainfall threshold parameters corresponding to the radar reflectivity and rainfall are extracted as cloud rainfall characterization data. Ground-based rain gauges collect measured rainfall data from the ground. The measured rainfall data is then processed by spatial interpolation and planar gridding based on the landing area to generate a planar rainfall model covering the monitoring area. During the modeling process, abnormal data is identified and filtered out simultaneously. This planar rainfall model is then used as the representation data of the ground-based rainfall. Cloud-based precipitation data and ground-based precipitation data are referred to as dual-source precipitation data. See appendix Figure 1 The diagram shows the architecture of the dual-source rainfall data input channel. It marks the composition and data flow of the rain-measuring radar cloud rain channel and the ground rain gauge actual measurement channel, and clarifies the core parameter extraction process.

[0018] One specific embodiment is as follows: Four rain-measuring radars and 1,000 ground rain gauges were deployed in the monitoring area, covering the urban area and surrounding towns. The monitoring target is sudden summer rainstorms, and it can track the entire process of the rainstorm center from its formation to its landing and predict its movement direction.

[0019] 1.1 Establishment of Cloud Rainfall Data Channel for Rainfall Measurement Radar: The rainfall measurement radar is started, and raw radar reflectivity (unit: dBZ) data of the monitoring area is acquired with a 5-minute acquisition cycle. The reflectivity data is input into a preset high-precision inversion rainfall model. This model first performs feature denoising on the reflectivity data to remove atmospheric clutter interference, then completes parameter calibration through a relational model calibrated by regional measurements, and finally outputs inversion rainfall (unit: mm / h) after error correction. Simultaneously, threshold parameters of reflectivity (rainstorm-level reflectivity threshold ≥ 40 dBZ), point parameters (coordinates of high reflectivity areas), and area parameters (area covered by reflectivity ≥ 40 dBZ), as well as threshold parameters of inversion rainfall (rainstorm-level derived rainfall threshold ≥ 8 mm / h), point parameters (coordinates of high inversion rainfall areas), and area parameters (area covered by inversion rainfall ≥ 8 mm / h), are extracted and integrated as data for cloud rainfall characterization.

[0020] 1.2 Construction of Ground Rain Gauge Data Channel: 1000 ground rain gauges collected point-based data of groundfall rainfall in real time with a 10-minute acquisition cycle. Based on the pre-defined rainfall area division scheme for the region, the Kriging interpolation algorithm was used to perform areal grid modeling on the collected point data to generate a global areal rainfall model. During the modeling process, anomaly data filtering was performed simultaneously. Based on the pre-defined reasonable fluctuation range of rainfall data, instantaneous abnormal data caused by equipment failure or external interference from individual stations were identified and removed. After modeling and filtering were completed, the rainfall magnitude of each grid within the model coverage area was labeled as data for storing groundfall rainfall characteristics.

[0021] 1.3 Dual-source data preprocessing and synchronous calibration: First, the two types of data are preprocessed. Outliers are removed based on the reasonable range of rainfall physical quantities (reflectivity 0-70dBZ, rainfall 0-100mm / h). Missing values ​​are filled using linear interpolation of data from adjacent time periods in the same region. Then, the data is standardized. Subsequently, the time synchronization calibration of the dual-source data is completed at a 10-minute time granularity to ensure that the cloud rain data and the ground rain data in the same time period are consistent in time and space.

[0022] Step 2: Set the rainstorm tracking time window: For dual-source rainfall data, take the current time as the cutoff point and trace back 1-2 hours as the rainstorm tracking analysis time range. The moment when the rainstorm center first appears within the time range is taken as the tracking start node. See appendix Figure 2 The flowchart shows the entire process of dynamic tracking and iteration of rainstorms, including setting time windows, iterative update logic, and recording evolution data.

[0023] One specific embodiment is as follows: Using the current time monitored by the system in real time as the cutoff point, trace back 2 hours as the basic analysis time range for rainstorm tracking, retrieve the pre-processed dual-source rainfall data within this period, identify the rainstorm center according to the judgment criteria in step 4, mark the moment when the rainstorm center is first identified within the time range as the rainstorm tracking start node, and start the subsequent tracking process.

[0024] Step 3: Start dynamic cyclic iterative tracking of the rainstorm center: For each new set of dual-source rainfall data, update the 1-hour time window for iterative calculation and continuously track the evolution of the rainstorm center in the past 1 hour; One specific embodiment is as follows: The cyclic iterative analysis mode is adopted, with the iteration cycle consistent with the synchronization cycle of the dual-source data, set to 10 minutes. Each time a new set of synchronized dual-source data is accessed, the 1-hour sliding time window is updated to focus on the evolution of the rainstorm center in the past hour.

[0025] During the tracking process, multi-dimensional evolution data of the rainstorm center are recorded simultaneously, including location coordinates (precisely marked with latitude and longitude), coverage area (statistical projection area of ​​high-value areas), rainfall intensity level (classified by rainstorm and torrential rain level), and core area diffusion rate (calculated as the ratio of the movement distance of the core area boundary to the time in adjacent time periods). The data of each dimension are archived in time series to form a rainstorm evolution time series archive, providing data support for the prediction of movement direction.

[0026] Step 4: Prediction of the movement direction of the rainstorm center: Based on the dual-source rainfall data and rainstorm center evolution data collected in multiple historical iterations, the movement direction of the rainstorm center at the next moment is predicted through intelligent algorithm fitting analysis. One specific embodiment is as follows: 4.1 Determination of Rainstorm Center: A dual-source data collaborative determination standard is adopted. The preset collaborative determination conditions are: "Rainfall data in radar clouds meet the following requirements: reflectivity ≥ 40 dBZ and coverage area ≥ 5 km², inverted rainfall ≥ 8 mm / h and coverage area ≥ 3 km², and the actual rainfall measured by the corresponding ground rain gauge in the area is ≥ 8 mm / h". When all conditions are met, it is determined to be a rainstorm center. The reflectivity and rainfall thresholds can be adjusted according to the topographic differences between mountainous and plain areas in the region to improve regional adaptability.

[0027] 4.2 Prediction of the movement direction of the rainstorm center: Retrieve historical dual-source rainfall data and rainstorm evolution time series archives collected in multiple iterations, and input them into an intelligent fitting algorithm with adaptive weight adjustment capabilities; The algorithm first identifies the cloud physical characteristics corresponding to the current rainstorm, distinguishing between three types: stationary, mobile, and train-like multi-center mobile. Then, it adjusts the fitting weights of historical data in a targeted manner—increasing the weight of the intensity dimension for stationary rainstorms, increasing the weight of the position evolution dimension for mobile rainstorms, and increasing the weight of the multi-center collaborative evolution dimension for train-like multi-center mobile rainstorms. Differentiated prediction logic fitting analysis is used to output the movement direction of the rainstorm center and the prediction confidence at the next moment.

[0028] Step 5: Multi-element fusion to construct an intelligent analysis model: Introduce the physical occurrence characteristics of rainstorm clouds as auxiliary analysis elements, and fuse the cloud physical occurrence characteristics with dual-source rainfall data, rainstorm center evolution data, and rainstorm center movement direction prediction results in multiple dimensions to construct a complete intelligent rainstorm tracking and analysis model, and output rainstorm tracking and prediction results.

[0029] See appendix Figure 3 Multi-factor fusion analysis model diagram; the diagram shows the fusion relationship between dual-source data, cloud physical characteristics, intelligent algorithms, and prediction results.

[0030] One specific embodiment is as follows: The physical characteristics of rainstorm clouds are introduced as core auxiliary elements, and the criteria for determining three types of characteristics are clarified: stationary rainstorms are characterized by a fixed rainstorm center and a continuous increase or maintenance of intensity; mobile rainstorms are characterized by a rainstorm center moving at a constant speed in a single direction; and train-type multi-center mobile rainstorms are characterized by multiple rainstorm centers moving sequentially in the same direction and having superimposed effects.

[0031] The physical characteristics of the cloud layer are integrated with dual-source rainfall data, rainstorm center evolution data, and rainstorm center movement direction prediction results in multiple dimensions to construct a complete intelligent rainstorm tracking and analysis model. The model output generates a visualized rainstorm tracking trajectory map, movement direction prediction map, and data report, providing an intuitive and reliable decision-making basis for disaster prevention, mitigation, and early warning.

[0032] Among them, the physical characteristics of the rainstorm clouds include at least three types: stationary, mobile, and train-like multi-center mobile.

[0033] The operation logic of the high-precision inversion model is as follows: with radar reflectivity intensity data as the core input, after feature denoising, parameter calibration and error correction, the output is the inverted rainfall that matches the actual rainfall in the cloud, and the calculation accuracy of the inverted rainfall is positively correlated with the radar reflectivity acquisition resolution and the inversion algorithm.

[0034] Time synchronization calibration is performed on the dual-source rainfall data: the cloud rainfall data from the rain measuring radar and the measured rainfall data from the ground rain monitoring station are aligned according to a preset time granularity to ensure the spatiotemporal consistency of the two types of data within the same monitoring period and improve the accuracy of collaborative analysis.

[0035] The criteria for determining the center of a rainstorm are as follows: when the parameters of reflectivity threshold T1, inverted rainfall threshold T2, and corresponding coverage area threshold T3 in the rainstorm data from the rain-measuring radar cloud rain data are combined, and the rainstorm level threshold T4 of the actual rainfall data from the ground rain gauge is matched, and the preset collaborative determination conditions are met, it is determined to be the center of a rainstorm.

[0036] In step 4, the intelligent algorithm has the ability to adaptively adjust weights; based on the physical characteristics of different types of rainstorm clouds introduced in step 5, it adjusts the fitting weights of historical data, and adopts differentiated prediction logic for stationary, mobile, and train-like multi-center mobile rainstorms to improve the accuracy of predicting the direction of movement of the rainstorm center.

[0037] In step 3, the tracking dimensions of the evolution process of the rainstorm center include at least location coordinates, coverage area, rainfall intensity level, and core area diffusion rate. Each iteration updates the data of each dimension to form a rainstorm evolution time series archive to support subsequent prediction and analysis.

[0038] The method further includes preprocessing the dual-source rainfall data. The preprocessing includes outlier removal, missing value completion, and data standardization. Outlier removal is determined based on a preset reasonable range of rainfall physical quantities, and missing value completion is achieved by interpolating data from adjacent time periods in the same region.

[0039] The preset collaborative judgment conditions are dynamically adjusted based on the climate characteristics and terrain conditions of the monitoring area to adapt to the rainstorm formation patterns of different regions, thereby improving the regional adaptability of the rainstorm center judgment.

[0040] Through the above-described rainstorm tracking and analysis method based on multi-source rainfall monitoring data collaboration, the accurate tracking of the rainstorm center and the scientific prediction of its movement direction were achieved. According to actual measurements, the accuracy of rainstorm center identification was improved by 35% compared with a single data source, and the prediction error of movement direction was controlled within 10km, which can effectively support regional rainstorm disaster early warning work.

Claims

1. A method for tracking and analyzing rainstorms based on the collaborative analysis of multi-source rainfall monitoring data, characterized in that, The rainstorm tracking and analysis method includes the following steps: Step 1: Construct a dual-source rainfall data input channel: including a rain-measuring radar cloud rainfall data channel and a ground-based rain gauge measured rainfall data channel, wherein: The cloud rain data channel of the rain measuring radar collects radar reflectivity and inverted rainfall calculated by a high-precision inversion model based on the radar reflectivity. The landing area parameters and rainfall threshold parameters corresponding to the radar reflectivity and rainfall are extracted as cloud rainfall characterization data. Ground-based rain gauges collect measured rainfall data from the ground. The measured rainfall data is then processed by spatial interpolation and planar gridding based on the landing area to generate a planar rainfall model covering the monitoring area. During the modeling process, abnormal data is identified and filtered out simultaneously. This planar rainfall model is then used as the representation data of the ground-based rainfall. Cloud-based precipitation data and ground-based precipitation data are referred to as dual-source precipitation data. Step 2: Set the rainstorm tracking time window: For dual-source rainfall data, take the current time as the cutoff point and trace back 1-2 hours as the rainstorm tracking analysis time range. The moment when the rainstorm center first appears within the time range is taken as the tracking start node. Step 3: Start dynamic cyclic iterative tracking of the rainstorm center: For each new set of dual-source rainfall data, update the 1-hour time window for iterative calculation and continuously track the evolution of the rainstorm center in the past 1 hour; Step 4: Prediction of the movement direction of the rainstorm center: Based on the dual-source rainfall data and rainstorm center evolution data collected in multiple historical iterations, the movement direction of the rainstorm center at the next moment is predicted through intelligent algorithm fitting analysis. Step 5: Multi-element fusion to construct an intelligent analysis model: Introduce the physical occurrence characteristics of rainstorm clouds as auxiliary analysis elements, and fuse the cloud physical occurrence characteristics with dual-source rainfall data, rainstorm center evolution data, and rainstorm center movement direction prediction results in multiple dimensions to construct a complete intelligent rainstorm tracking and analysis model, and output rainstorm tracking and prediction results.

2. The rainstorm tracking and analysis method according to claim 1, characterized in that, The physical characteristics of the rainstorm clouds include at least three types: stationary, mobile, and train-like multi-center mobile.

3. The rainstorm tracking and analysis method according to claim 1, characterized in that, The operation logic of the high-precision inversion model is as follows: with radar reflectivity intensity data as the core input, after feature denoising, parameter calibration and error correction, the output is the inverted rainfall that matches the actual rainfall in the cloud, and the calculation accuracy of the inverted rainfall is positively correlated with the radar reflectivity acquisition resolution and the inversion algorithm.

4. The rainstorm tracking and analysis method according to claim 1, characterized in that, Time synchronization calibration is performed on the dual-source rainfall data: the cloud rainfall data from the rain measuring radar and the measured rainfall data from the ground rain monitoring station are aligned according to a preset time granularity to ensure the spatiotemporal consistency of the two types of data within the same monitoring period and improve the accuracy of collaborative analysis.

5. The rainstorm tracking and analysis method according to claim 1, characterized in that, The criteria for determining the center of a rainstorm are as follows: when the reflectivity threshold, inverted rainfall threshold, and corresponding coverage area threshold in the rainstorm data from the rain-measuring radar cloud rain data are combined with the rainstorm level threshold of the actual rainfall data from the ground rain gauge, and when the preset collaborative determination conditions are met, it is determined to be the center of a rainstorm.

6. The rainstorm tracking and analysis method according to claim 1, characterized in that, In step 4, the intelligent algorithm has the ability to adaptively adjust weights; based on the physical characteristics of different types of rainstorm clouds introduced in step 5, it adjusts the fitting weights of historical data, and adopts differentiated prediction logic for stationary, mobile, and train-like multi-center mobile rainstorms to improve the accuracy of predicting the direction of movement of the rainstorm center.

7. The rainstorm tracking and analysis method according to claim 1, characterized in that, In step 3, the tracking dimensions of the evolution process of the rainstorm center include at least location coordinates, coverage area, rainfall intensity level, and core area diffusion rate. Each iteration updates the data of each dimension to form a rainstorm evolution time series archive to support subsequent prediction and analysis.

8. The rainstorm tracking and analysis method according to claim 1, characterized in that, The method further includes preprocessing the dual-source rainfall data. The preprocessing includes outlier removal, missing value completion, and data standardization. Outlier removal is determined based on a preset reasonable range of rainfall physical quantities, and missing value completion is achieved by interpolating data from adjacent time periods in the same region.

9. The rainstorm tracking and analysis method according to claim 5, characterized in that, The preset collaborative judgment conditions are dynamically adjusted based on the climate characteristics and terrain conditions of the monitoring area to adapt to the rainstorm formation patterns of different regions, thereby improving the regional adaptability of the rainstorm center judgment.

Images