Tropical cyclone automatic tracking method based on relative threshold value

Abstract

The invention relates to the field of tropical cyclone tracking, in particular to a tropical cyclone automatic tracking method based on a relative threshold value. The method comprises the following steps: acquiring meteorological historical data through a numerical weather forecast model, and obtaining the relative vorticity and warm heart strength of a historical tropical cyclone according to the meteorological historical data; historical tropical cyclone data are collected, and the relative threshold value of the relative vorticity and the relative threshold value of the warm heart strength are calculated according to the historical tropical cyclone data; acquiring meteorological data of different times according to the numerical weather forecast model, and obtaining candidate areas of different times according to the relative threshold and the meteorological data; obtaining latitude and longitude data and intensity data of a plurality of groups of tropical cyclone centers at different times according to the candidate areas at different times; and according to the latitude and longitude data and the intensity data, using a clustering algorithm to obtain time-varying path and intensity information of each tropical cyclone. According to the invention, the problems of low tracking efficiency and poor precision of a plurality of tropical cyclones in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the field of tropical cyclone tracking, and specifically to an automatic tropical cyclone tracking method based on relative thresholds. Background Art

[0002] Tropical cyclones are major meteorological disasters that affect many regions globally. Their formation, development, and dissipation processes are complex, and they exhibit different characteristics under different climate conditions. Using high-resolution numerical weather prediction models and meteorological data, and locating and tracking tropical cyclones therein are one of the main basic tools for forecasting and studying tropical cyclones.

[0003] Main related content of the prior art: The existing mainstream tropical cyclone tracking algorithms are divided into two parts. One part is tropical cyclone positioning: based on the main characteristics of tropical cyclones, and by setting fixed thresholds to locate the spatial positions of tropical cyclones at different times; the other part is tracking. Based on the moving average speed and the sequence of positions of tropical cyclones, the positions of tropical cyclones within a certain range are connected, and finally the path of the tropical cyclone is formed.

[0004] Problems and disadvantages of the prior art: First, the existing mainstream algorithms have insufficient tracking integrity for high-resolution weather prediction models (such as spatial resolutions of 3.5 km and 1 km), which is reflected in a large deviation between the number of tracked tropical cyclones of different intensity levels and the observations. Most of the existing mainstream tracking algorithms were developed relatively early and were developed based on coarser-resolution atmospheric numerical models, without considering the characteristics of tropical cyclones at high resolutions in high-resolution and even ultra-high-resolution models. Therefore, serious omission problems occur in the latest ultra-high-resolution prediction models, so that the number of tracked tropical cyclones is significantly less than the actual number; on the other hand, the mainstream tropical cyclone tracking scheme has low tracking efficiency and takes too long for ultra-high-resolution weather prediction models. This is because the mainstream tracking scheme tracks tropical cyclones one by one according to time steps and spatial positions. In high-resolution models, when there are multiple potential tropical cyclone paths, the mainstream tracking scheme takes a long time. When multiple forecast results need to be processed, the mainstream tracking scheme has low operating efficiency and requires a large amount of running time. Summary of the Invention

[0005] Aiming at the defects in the prior art, the present invention provides an automatic tropical cyclone tracking method based on relative thresholds, which solves the problems of low tracking efficiency and poor accuracy for multiple tropical cyclones in the prior art.

[0006] To achieve the above object, the automatic tracking method of tropical cyclones based on relative thresholds provided by the present invention includes: obtaining meteorological historical data through a numerical weather prediction model, and obtaining the relative vorticity and warm core intensity of historical tropical cyclones based on the meteorological historical data; collecting historical tropical cyclone data, and calculating the relative thresholds of the relative vorticity and the warm core intensity based on the historical tropical cyclone data; obtaining meteorological data at different times through the numerical weather prediction model, and obtaining candidate regions at different times based on the relative thresholds and the meteorological data; obtaining longitude and latitude data and intensity data of the centers of tropical cyclones in multiple groups at different times based on the candidate regions at different times; and obtaining the path and intensity information of each tropical cyclone changing with time by using a clustering algorithm based on the longitude and latitude data and the intensity data.

[0007] The present invention obtains meteorological data through a numerical weather prediction model and combines historical tropical cyclone data to comprehensively and integrally explore the data value. Calculating the relative vorticity, warm core intensity, and relative threshold using meteorological historical data provides a basis for accurately screening candidate regions in the subsequent process, giving full play to the guiding role of the data. Using a clustering algorithm to obtain the path and intensity information changing with time helps to accurately analyze the development dynamics of tropical cyclones, and overall improves the accuracy of tracking the paths and intensities of tropical cyclones.

[0008] Optionally, the obtaining the relative vorticity and warm core intensity of historical tropical cyclones based on the meteorological historical data includes: extracting the meridional wind component and zonal wind component at a distance of 10 m from the ground in the meteorological historical data; calculating the relative vorticity of the tropical cyclone based on the meridional wind component and the zonal wind component; respectively extracting the first average temperature and the second average temperature of the 200 hPa isobaric surface and the 500 hPa isobaric surface in the meteorological historical data; and calculating the warm core intensity of the tropical cyclone based on the first average temperature and the second average temperature.

[0009] The present invention accurately obtains key meteorological elements by extracting the meridional and zonal wind components at specific heights and the temperatures of different isobaric surfaces. When calculating the relative vorticity, using the meridional and zonal wind components can effectively reflect the rotational characteristics of the air near the tropical cyclone and provide a basis for judging its dynamic structure. Calculating the warm core intensity based on the temperatures of different isobaric surfaces can measure the thermal characteristics of the tropical cyclone. These calculation methods cooperate with each other to comprehensively present the thermal and dynamic properties of the tropical cyclone, laying a solid data foundation for subsequent calculation of relative thresholds, determination of candidate regions, and tracking of the path and intensity changes of the tropical cyclone, and improving the accuracy and reliability of tropical cyclone tracking.

[0010] Optionally, the calculating of the relative vorticity and the relative threshold of the warm core intensity based on the historical tropical cyclone data includes: based on the relative vorticity and the warm core intensity, using statistical analysis method to extract the historical relative vorticity values and historical warm core structure values within the 2° spherical neighborhood of the historical tropical cyclone center from the historical tropical cyclone data; calculating the historical relative vorticity maximum value and historical warm core structure maximum value of each historical tropical cyclone according to the historical relative vorticity values and the historical warm core structure values; respectively calculating the 95% quantiles of the historical relative vorticity maximum value and the historical warm core structure maximum value to obtain the relative threshold.

[0011] The present invention focuses on selecting data in the 2° spherical neighborhood of the historical tropical cyclone center, accurately locates the area closely related to the tropical cyclone, avoids interference from other irrelevant areas, ensures the effectiveness and pertinence of the extracted data, and lays a foundation for accurately calculating the relative threshold. Using the statistical analysis method to calculate the historical relative vorticity maximum value, historical warm core structure maximum value and their 95% quantiles as the relative threshold can effectively reflect the distribution of tropical cyclone-related characteristics in the historical data. This statistical method based on a large amount of historical data can minimize the influence of accidental factors and improve the scientificity and accuracy of the threshold.

[0012] Optionally, the relative threshold satisfies the following formula: is the relative threshold, is the candidate value of the variable, is the indicator function, is the historical relative vorticity maximum value or historical warm core structure maximum value of the th historical tropical cyclone,

[0013] The relative threshold formula of the present invention constructs an empirical distribution based on a large amount of historical tropical cyclone data, can fully exploit historical information, reduce accidental errors, make the threshold more stable and representative, and further judges whether the candidate value condition of the variable is satisfied through the indicator function, and accurately defines the relative threshold in a strict mathematical form, improving the calculation accuracy of the relative threshold.

[0014] Optionally, the obtaining of candidate regions at different times according to the relative threshold and the meteorological data includes: setting constraint conditions using the relative threshold, and screening the meteorological data using the constraint conditions to obtain candidate regions at different times; The candidate region satisfies the following formula: where, is the relative vorticity relative threshold, is the relative vorticity in meteorological data, is the relative threshold of warm core intensity, is the warm core intensity in meteorological data.

[0015] By setting constraint conditions in two dimensions of relative vorticity and warm core intensity, the present invention screens meteorological data from two key perspectives of dynamics and thermodynamics, comprehensively and accurately capturing tropical cyclone-related features. This multi-condition joint screening method avoids the limitations of a single index and improves the accuracy and reliability of screening.

[0016] Optionally, the intensity data includes the maximum 10m wind speed and the lowest sea level pressure. The obtaining of the longitude and latitude data and intensity data of the centers of tropical cyclones in multiple groups at different times according to the candidate regions at different times includes: calculating the 10m wind speed at different times based on the candidate regions at different times using the meteorological data; determining the vorticity maximum point within the candidate region, and using the 10m wind speed to determine the wind speed minimum point at different times within a 2° spherical neighborhood centered on the vorticity maximum point; obtaining the longitude and latitude data of the centers of tropical cyclones at different times according to the wind speed minimum points at different times; calculating the maximum 10m wind speed and the lowest sea level pressure within a 2° spherical neighborhood of the centers of tropical cyclones at different times using the meteorological data.

[0017] The present invention calculates the 10m wind speed based on the candidate region, and accurately locks the longitude and latitude of the center by determining the wind speed minimum point with the vorticity maximum point, fully considering the rotation characteristics of tropical cyclones, effectively excluding interference, and improving the positioning accuracy. In terms of obtaining intensity data, the maximum 10m wind speed and the lowest sea level pressure are calculated within a specific spherical neighborhood around the determined center to ensure that the obtained data truly reflects the core intensity of tropical cyclones.

[0018] Optionally, the obtaining of the path and intensity information of each tropical cyclone changing with time according to the longitude and latitude data and the intensity data using a clustering algorithm includes: constructing a three-dimensional feature vector of each candidate region according to the longitude and latitude data and the intensity data; constructing an adjacency structure of the candidate regions according to the three-dimensional feature vector; classifying the central nodes of the candidate regions according to the adjacency structure; constructing a density connectivity clustering according to the classification result, and obtaining the path and intensity information of each tropical cyclone changing with time according to the density connectivity clustering.

[0019] The present invention constructs a three-dimensional feature vector through longitude, latitude and intensity data, effectively integrating key information of tropical cyclones and providing a structured data basis for subsequent analysis. Based on the three-dimensional feature vector, an adjacency structure is constructed, which can accurately measure the association between candidate regions, taking into account spatial positions and intensity differences, and conforming to the actual characteristics of tropical cyclones. By classifying central nodes and constructing density connectivity clustering, effective identification and division of tropical cyclones are achieved. This clustering method fully explores the hidden connections between data and can accurately improve the accuracy of tropical cyclone tracking.

[0020] Optionally, constructing the adjacency structure of the candidate region according to the three-dimensional feature vector includes: calculating the spatial distance and intensity difference between the centers of tropical cyclones at two different times according to the three-dimensional feature vector; constructing a comprehensive distance metric function between the centers of tropical cyclones at two different times according to the spatial distance and the intensity difference; constructing the adjacency structure of the candidate region according to the comprehensive distance metric function.

[0021] By calculating the spatial distance and intensity difference between the centers of tropical cyclones at different times, the present invention comprehensively considers the change characteristics of tropical cyclones in terms of time, space and intensity. On this basis, the constructed comprehensive distance metric function organically combines spatial and intensity factors, more truly reflects the similarity degree and correlation tightness between tropical cyclones at different times, and the adjacency structure constructed accordingly can accurately divide the mutual relationship of tropical cyclones at different moments, improving the scientificity and accuracy of adjacency structure construction.

[0022] Optionally, constructing the adjacency structure of the candidate region according to the comprehensive distance metric function includes: obtaining the historical path data of the tropical cyclone, and obtaining the historical spatial data and historical intensity data of the historical centers of the tropical cyclones at two different times according to the historical path data; respectively calculating the 95% quantiles of the historical spatial data and the historical intensity data to obtain a historical spatial threshold and a historical intensity threshold; calculating a clustering radius according to the historical spatial threshold and the historical intensity threshold; constructing the adjacency structure of the candidate region according to the clustering radius and the comprehensive distance metric function.

[0023] Calculating the spatial threshold and intensity threshold using historical path data can fully exploit the value of historical data, and these thresholds represent the general characteristics of tropical cyclones in terms of spatial and intensity changes. The clustering radius calculated through the spatial threshold and intensity threshold provides a scientific criterion for defining the association between the centers of tropical cyclones at different times. Constructing the adjacency structure in combination with the comprehensive distance metric function further improves the scientificity and accuracy of adjacency structure construction.

[0024] The adjacency structure satisfies the following formula: Wherein, is the adjacency structure constructed according to the three-dimensional feature vector ; is the other three-dimensional feature vector that is adjacent to the three-dimensional feature vector and the comprehensive distance metric function is less than the clustering radius value ; is the and spatial distance value calculated using the spherical cosine formula is the intensity weight is the calculated and intensity difference value is the clustering radius value

[0025] The adjacency structure formula of the present invention comprehensively considers spatial distance and intensity difference, calculates the spatial distance value using the spherical cosine formula, and combines the intensity weight to measure the intensity difference value, comprehensively and accurately depicting the relationship between three-dimensional feature vectors. The clustering radius value is set as the judgment criterion in the formula, which can effectively screen out other vectors that are adjacent to the target three-dimensional feature vector and the comprehensive distance metric function meets the requirements, constructing a more reasonable adjacency structure and improving the accuracy of adjacency structure calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is the flowchart of the tropical cyclone automatic tracking method based on relative threshold according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described here are only for illustrative purposes and are not used to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it will be apparent to those of ordinary skill in the art that: the present invention does not have to be practiced with these specific details. In other instances, well-known circuits, software, or methods have not been described in detail in order to avoid obscuring the present invention.

[0028] Throughout the specification, the reference to "one embodiment", "an embodiment", "one example" or "an example" means that the specific features, structures, or characteristics described in connection with the embodiment or example are included in at least one embodiment of the present invention. Thus, the phrases "in one embodiment", "in an embodiment", "one example" or "an example" appearing throughout the specification do not necessarily all refer to the same embodiment or example. In addition, the specific features, structures, or characteristics may be combined in any suitable combination and / or sub-combination in one or more embodiments or examples. In addition, those of ordinary skill in the art should understand that the drawings provided herein are for illustrative purposes only and are not necessarily drawn to scale.

[0029] Please refer to Figure 1 , Figure 1 which is a flowchart of the automatic tracking method for tropical cyclones based on relative thresholds according to an embodiment of the present invention. As Figure 1 shown, the method includes the following steps: Step S1: Obtain meteorological historical data through a numerical weather prediction model, and obtain the relative vorticity and warm core intensity of historical tropical cyclones based on the meteorological historical data.

[0030] Among them, obtaining the relative vorticity and warm core intensity of historical tropical cyclones based on the meteorological historical data specifically includes the following sub-steps: Step S101: Extract the meridional wind component and zonal wind component at a distance of 10 m from the ground in the meteorological historical data.

[0031] In this embodiment, the numerical weather prediction model is a scientific means of making weather forecasts based on a mathematical model of the atmosphere and using the current weather conditions as input data. It is based on the basic laws of fluid mechanics and thermodynamics, describes the laws of atmospheric motion with a set of mathematical equations, and numerically solves these mathematical equations through a high-performance computer. Before the operation, a large amount of current meteorological observation data needs to be collected, such as temperature, pressure, humidity, precipitation, wind speed, etc. obtained from meteorological stations, meteorological satellites, meteorological balloons, etc., as the initial conditions of the model. The numerical weather prediction model is divided into global models and regional models, etc. The global model aims to accurately predict the global weather; the regional model focuses on the weather conditions in a specific region. In practical applications, multiple models are often used in combination to reduce the uncertainty of weather forecasts and extend the predictable time range.

[0032] First, relying on an advanced numerical weather prediction model, with the help of a high-performance computing cluster, using a vast amount of meteorological observation data as the initial field, through complex fluid mechanics and thermodynamics equation operations, generate meteorological historical data covering many years and the global scope. Then, use the professional meteorological data processing software GrADS (Grid Analysis and Display System). In the GrADS operation interface, by writing specific control statements and script codes, call its rich data reading and analysis function library. According to the precise height instruction (10 m from the ground), conduct a detailed screening from the huge meteorological historical dataset, and extract the two key elements of the meridional wind component and the zonal wind component.

[0033] The wind field data at a height of 10 m above the ground is closely related to the surface friction effect and has a crucial impact on the formation and maintenance of the near-surface circulation structure of tropical cyclones. The meridional wind component represents the air flow transportation in the north-south direction, and the zonal wind component reflects the air flow variation in the east-west direction. Their combined effect is of great significance for characterizing the rotation, movement, and energy exchange of tropical cyclones, and also lays a solid data foundation for subsequent in-depth calculation of important parameters such as relative vorticity.

[0034] Step S102, calculate the relative vorticity of the tropical cyclone according to the meridional wind component and the zonal wind component.

[0035] In this embodiment, relative vorticity is a physical quantity that measures the degree of rotation of an air parcel around the vertical axis. For a strongly rotating weather system such as a tropical cyclone, calculating its relative vorticity can help to deeply understand its dynamic characteristics and development mechanism.

[0036] First, the study area is carefully divided into small grids, and each grid corresponds to specific values of the meridional wind component and the zonal wind component. Then, the zonal wind components of the adjacent grids on the east and west sides of a certain grid point can be taken, their differences can be analyzed, and combined with the distance information of the grid to obtain the variation of the zonal wind component in the east-west direction. Similarly, the meridional wind components of the adjacent grids on the east and west sides of the same grid point are taken, their differences are analyzed, and combined with the distance information of the grid to obtain the variation of the meridional wind component in the north-south direction. Immediately afterwards, the relative vorticity of the tropical cyclone is calculated by combining the variations of the wind components in these two directions.

[0037] The relative vorticity satisfies the following formula: where is the relative vorticity at longitude , latitude and time , is the meridional wind component at 10 m above the ground, is the zonal wind component at 10 m above the ground, is the distance coordinate in the east-west direction, is the distance coordinate in the north-south direction.

[0038] Step S103, extract the first average temperature and the second average temperature of the 200 hPa isobaric surface and the 500 hPa isobaric surface respectively from the meteorological historical data.

[0039] In this embodiment, using the GrADS software and based on the standard settings of the pressure layer, the 200 hPa isobaric surface and the 500 hPa isobaric surface are accurately located. Using the temperature data screening function of the software, the temperature data corresponding to the isobaric surface is screened out from the massive meteorological historical data, and these temperature data are discretely distributed at different spatial positions. Immediately afterwards, for each point where the average temperature is to be calculated, with this point as the center, within a circular neighborhood with a radius of 2°, all temperature data points are collected, and then a suitable mathematical statistical method, such as the arithmetic mean method, is used to process the temperature values of these data points to calculate the spatial average temperature within this neighborhood. After completing the above operations for all points on the 200 hPa isobaric surface, the spatial average temperatures of the obtained points are averaged again, and the result obtained is the average temperature of the 200 hPa isobaric surface, which is the first average temperature. Similarly, the average temperature of the 500 hPa isobaric surface is the second average temperature.

[0040] Step S104, calculate the warm core intensity of the tropical cyclone according to the first average temperature and the second average temperature.

[0041] In this embodiment, the warm core intensity is an index that measures the degree to which the central region of a tropical cyclone is warmer than the surrounding environment. Its value is often obtained by calculating the temperature difference between the center of the system and a certain range around it, or the temperature difference between different isobaric surfaces. The larger the difference, the stronger the warm core intensity.

[0042] The warm core intensity satisfies the following formula: Wherein, is the warm core intensity at longitude , latitude and time , is the first average temperature, is the second average temperature.

[0043] Step S2, collect historical tropical cyclone data, and calculate the relative vorticity and the relative threshold of the warm core intensity according to the historical tropical cyclone data.

[0044] Among them, calculating the relative vorticity and the relative threshold of the warm core intensity according to the historical tropical cyclone data specifically includes the following sub-steps: Step S201, based on the relative vorticity and the warm core intensity, use statistical analysis to extract the historical relative vorticity value and the historical warm core structure value within the 2° spherical neighborhood of the historical tropical cyclone center from the historical tropical cyclone data.

[0045] In this embodiment, through professional meteorological data platforms, such as the International Best Track Archive for Climate Stewardship (IBTrACS) of the World Meteorological Organization, the official database of the National Hurricane Center (NHC) of the United States, and the relevant data resource libraries of the China Meteorological Administration, historical tropical cyclone data are comprehensively collected. These data contain rich information such as the generation time, movement trajectory, central longitude and latitude, and intensity change of tropical cyclones. At the same time, it should be understood that the relative vorticity and warm core intensity data calculated from historical meteorological data in the early stage are global data calculated based on meteorological elements at different grid points. According to the central longitude and latitude in the historical tropical cyclone data, with the center of each tropical cyclone as the center of the circle, the 2° spherical neighborhood range is determined using geographic information technology and related algorithms. Within this neighborhood range, the corresponding relative vorticity and warm core intensity data are screened out. Then, statistical analysis methods, such as calculating the maximum value, average value, median, etc., are used to process the relative vorticity data within the neighborhood to obtain the historical relative vorticity value that can represent the characteristics of this neighborhood; similarly, similar statistical processing is performed on the warm core intensity data within the neighborhood to obtain the historical warm core structure value.

[0046] Step S202, calculate the historical relative vorticity maximum value and the historical warm core structure maximum value of each historical tropical cyclone according to the historical relative vorticity value and the historical warm core structure value.

[0047] In this embodiment, for each historical tropical cyclone, first, a 2° spherical neighborhood range centered on the center is delimited according to its central longitude and latitude. Then, within this neighborhood range, the relative vorticity values at different position points at a specific moment are checked one by one, and through mutual comparison, the largest value among them is found. This value is the historical relative vorticity maximum value of the historical tropical cyclone within the 2° spherical neighborhood centered at the center at this moment. Similarly, the warm core structure values at different position points within the neighborhood at a specific moment are comprehensively compared, and the largest value is selected. This is the historical warm core structure maximum value of the historical tropical cyclone within the 2° spherical neighborhood centered at the center at this moment.

[0048] The historical relative vorticity maximum value and the historical warm core structure maximum value satisfy the following formula: where is the historical relative vorticity maximum value or the historical warm core structure maximum value, is the function for obtaining the maximum value of the historical relative vorticity or the historical warm core structure within the spherical neighborhood with a distance not exceeding 2° centered on the tropical cyclone center at longitude , latitude and time .

[0049] Step S203: Calculate the 95% quantiles of the historical relative vorticity maximum and the historical warm core structure maximum respectively to obtain the relative thresholds.

[0050] In this embodiment, first, determine the number of historical tropical cyclones involved in the calculation and the historical relative vorticity maximum or the historical warm core structure maximum corresponding to each historical tropical cyclone. Then, within a reasonable value range, continuously select the value of the variable x. For each selected x, check all the relative vorticity maxima or historical warm core structure maxima one by one. If the relative vorticity maximum or the historical warm core structure maximum is less than or equal to x, record it as 1; otherwise, record it as 0. Accumulate these 1s and 0s and divide by N to calculate the ratio. Continue this process until an x is found such that the ratio reaches or exceeds 95%. Among all the eligible x values, find the largest lower bound, which is the relative threshold we want. Perform the above operations on the historical relative vorticity maximum and the historical warm core structure maximum respectively to obtain their corresponding relative thresholds.

[0051] The 95% quantile is a common quantile index, indicating that 95% of the data values in the dataset are less than or equal to this quantile, and only 5% of the data values are greater than this quantile. It reflects the positional characteristics of the data distribution and is often used to exclude outliers and determine the "general level" or "critical value" of the data.

[0052] The relative threshold satisfies the following formula: is the relative threshold, is the candidate value of the variable, is the indicator function, is the historical relative vorticity maximum or the historical warm core structure maximum of the th historical tropical cyclone,

[0053] Step S3: Obtain meteorological data at different times according to the numerical weather prediction model, and obtain candidate regions at different times according to the relative threshold and the meteorological data.

[0054] Among them, obtaining candidate regions at different times according to the relative threshold and the meteorological data specifically includes the following sub-steps: Set constraint conditions using the relative threshold, and use the constraint conditions to screen the meteorological data to obtain candidate regions at different times.

[0055] In this embodiment, the relative thresholds of relative vorticity and warm core intensity are used as screening criteria to check the relative vorticity in the meteorological data. If it is not lower than the relative threshold of relative vorticity, this part of the data is retained first. Then, the warm core intensity is examined. If it is not lower than the relative threshold of warm core intensity, it is also retained. Only the regions that satisfy both conditions simultaneously are selected. In this way, the meteorological data at different times are processed in this manner to obtain the regions that meet the requirements at different times, that is, the candidate regions at different times.

[0056] The candidate regions satisfy the following formula: Wherein, is the relative threshold of relative vorticity, is the relative vorticity in the meteorological data, is the relative threshold of warm core intensity, is the warm core intensity in the meteorological data.

[0057] Step S4, obtain the longitude and latitude data and intensity data of the centers of tropical cyclones at different times according to the candidate regions at different times.

[0058] Among them, obtaining the longitude and latitude data and intensity data of the centers of tropical cyclones at different times according to the candidate regions at different times specifically includes the following sub-steps: Step S401, calculate the 10m wind speed at different times based on the candidate regions at different times using the meteorological data.

[0059] In this embodiment, the 10m wind speed refers to the horizontal air movement speed at a height of 10 meters above the ground. For the candidate region at each time, the corresponding meteorological data within this region are collected, and these data include various elements such as air pressure, temperature, humidity, and wind direction. Then, professional meteorological algorithms and models are used, such as numerical models based on the principles of atmospheric dynamics, and calculations are carried out in combination with the mutual relationships among the various meteorological elements within the candidate region. During the calculation process, the impacts of the horizontal and vertical movements of the atmosphere on the wind speed, as well as the effects of factors such as terrain and underlying surface, are fully considered. Through comprehensive analysis and operation of these meteorological data, the 10m wind speed values within the candidate regions at different times are finally obtained.

[0060] Step S402, determine the point of maximum vorticity within the candidate region, and use the 10m wind speed to determine the point of minimum wind speed at different times within a 2° spherical neighborhood centered on the point of maximum vorticity.

[0061] In this embodiment, for the candidate regions at each time instance, the meteorological data therein is carefully sorted out, with a particular focus on the important meteorological element of vorticity. The vorticity values at each position point within the candidate region are accurately obtained. Through point-by-point comparison of the magnitudes of the vorticity values, a comprehensive and detailed analysis is carried out. Finally, the point with the largest vorticity value is found and determined as the vorticity maximum point.

[0062] After successfully determining the vorticity maximum point, taking this point as the precise central reference, according to the geospatial calculation method, a 2° spherical neighborhood range centered on it is accurately delimited. Within this specific neighborhood, for different time instances, the corresponding 10m wind speed data is systematically collected. The 10m wind speed data for each time instance contains the wind speed information of each position point within the neighborhood.

[0063] Then, a detailed comparative analysis is carried out on the 10m wind speed values of each position point within the neighborhood for each time instance. In this process, through rigorous comparison and screening, the point with the smallest wind speed value at that time instance is found, and this point is identified as the wind speed minimum point for that time instance. According to such an operation process, for different time instances, the corresponding wind speed minimum points are respectively determined within the 2° spherical neighborhood. The determination of these wind speed minimum points at different time instances provides key and precisely located basic data information for subsequent in-depth research on the structural characteristics and variation laws of tropical cyclones at different times.

[0064] Step S403, obtaining the longitude and latitude data of the tropical cyclone center at different time instances based on the wind speed minimum points at different time instances.

[0065] Among them, the obtaining of the longitude and latitude data of the tropical cyclone center at different time instances based on the wind speed minimum points at different time instances includes: Extracting the distribution characteristics of the meteorological elements in the meteorological data; Correcting the wind speed minimum values according to the distribution characteristics to obtain the corrected wind speed minimum values at different time instances; Obtaining the longitude and latitude data of the tropical cyclone center at different time instances based on the corrected wind speed minimum values at different time instances.

[0066] In this embodiment, the obtained meteorological data is comprehensively and meticulously analyzed to extract various meteorological elements, such as the distribution characteristics of 10m wind speed, relative vorticity, warm core intensity, air pressure, temperature, humidity, wind direction, etc. at different times and spatial positions, including the positions of high-value areas and low-value areas of each element, change trends, gradient directions and intensities, etc. These characteristics are clearly presented by drawing contour maps, analyzing statistical data, etc. Then, based on the distribution characteristics of the extracted meteorological elements, in-depth evaluation is carried out on the determined minimum wind speed points at different times. Considering that the minimum wind speed may be affected by factors such as the surrounding air pressure gradient, topography, and the synergistic influence of other meteorological elements, if the minimum wind speed point is in an area with abnormal air pressure gradient, or in a complex terrain (such as valleys, leeward slopes of mountains), and there are obvious contradictions with the distribution of other meteorological elements, it is appropriately adjusted according to relevant meteorological principles and empirical models to obtain more accurate minimum corrected wind speeds at different times that can reflect the actual situation. Finally, based on these minimum corrected wind speeds at different times, combined with geographic information system technology, as well as the structural characteristics and formation and development laws of tropical cyclones, the spatial position relationship of the minimum corrected wind speed points and their correlation with the distribution of other meteorological elements are comprehensively analyzed, and their coordinates are slightly adjusted to further determine relatively accurate longitude and latitude data of the tropical cyclone center at different times.

[0067] In an alternative embodiment, the minimum corrected wind speed satisfies the following formula: Wherein, is the minimum corrected wind speed at the th time, is the minimum wind speed at the th time, and are both weights, is the relative vorticity eigenvalue extracted from the distribution characteristics of meteorological elements, is the air pressure gradient eigenvalue extracted from the distribution characteristics of meteorological elements.

[0068] The above and can be confirmed and calibrated through the statistical relationship between historical tropical cyclone data and the corresponding wind speed, vorticity, and air pressure gradient.

[0069] Step S404, calculate the maximum 10m wind speed and the lowest sea level air pressure within the 2° spherical neighborhood of the tropical cyclone center at different times using the meteorological data.

[0070] In this embodiment, the intensity data includes the maximum 10m wind speed and the lowest sea level air pressure.

[0071] Based on the known longitude and latitude data of the tropical cyclone center at different times, for each time, with the tropical cyclone center as the center of the circle, a 2° spherical neighborhood is delimited according to the geographical information system technology. Using the professional meteorological data processing software GrADS, according to the delimited regional scope and time, the corresponding data subset is accurately extracted from the acquired meteorological data. Focus on extracting the 10m wind speed and sea level pressure data of each grid point within this neighborhood, and at the same time extract auxiliary meteorological element data such as temperature, humidity, and wind direction for subsequent comprehensive analysis. The extracted data is sorted in an orderly manner according to the longitude and latitude coordinates of the grid points to construct a two-dimensional array structure that is convenient for calculation and analysis, ensuring that each data point can accurately correspond to its geographical location, providing a clear data basis for calculating the maximum 10m wind speed and the lowest sea level pressure in the follow-up.

[0072] For the 10m wind speed data within the neighborhood of each sorted time, it is processed using programming languages such as Python. With the help of the powerful array operation function of the NumPy library, the wind speed data is converted into a NumPy array. By calling the maximum value function np.max() of the array, each wind speed value in the array is quickly traversed and compared to determine the maximum 10m wind speed value within the 2° spherical neighborhood at this time.

[0073] For the sea level pressure data, a similar process to calculating the maximum 10m wind speed is adopted. Similarly, the sea level pressure data is sorted into a NumPy array, and the np.min() function is used to obtain the minimum value in the array, that is, the lowest sea level pressure value within the 2° spherical neighborhood of the tropical cyclone center at this time. Since the sea level pressure may be affected by factors such as terrain and tides, before calculation, the pressure data is corrected in combination with high-precision terrain data and tide models. For the terrain factor, according to the formula of the change of air pressure with height, the air pressure measurement values at different altitudes are corrected; for the tide influence, according to the tide forecast data, the air pressure data in the coastal area is adjusted accordingly. Through these correction steps, the interference of external factors on the pressure data is eliminated to ensure that the calculated lowest sea level pressure is accurate and reliable.

[0074] Step S5, using the clustering algorithm based on the longitude and latitude data and the intensity data to obtain the path and intensity information of each tropical cyclone changing with time.

[0075] Among them, using the clustering algorithm based on the longitude and latitude data and the intensity data to obtain the path and intensity information of each tropical cyclone changing with time specifically includes the following sub-steps: Step S501, constructing a three-dimensional feature vector for each of the candidate regions according to the longitude and latitude data and the intensity data.

[0076] In this embodiment, the longitude and latitude data includes longitude data and latitude data, and the intensity data includes the maximum 10m wind speed and the lowest sea level pressure. In order to make these different types and magnitudes of data comparable and operable, the longitude and latitude data, the maximum 10m wind speed, and the lowest sea level pressure are respectively standardized. For the longitude and latitude, normalization mapping is performed according to the longitude and latitude span of the research area. For the maximum 10m wind speed and the lowest sea level pressure, the Z-score standardization method is used to eliminate the influence of dimensions. After standardization, the standardized longitude value is used as the first component of the three-dimensional feature vector, the standardized latitude value is used as the second component, and the standardized maximum 10m wind speed and the lowest sea level pressure are linearly combined according to a certain weight (for example, the weight is determined according to the correlation analysis of historical data to make the contributions of the two to the feature vector more reasonable), and the obtained value is used as the third component, thereby constructing a unique three-dimensional feature vector for each candidate area.

[0077] Step S502, construct the adjacency structure of the candidate area according to the three-dimensional feature vector.

[0078] Among them, constructing the adjacency structure of the candidate area according to the three-dimensional feature vector specifically includes the following sub-steps: Step S50201, calculate the spatial distance and intensity difference between the centers of two tropical cyclones at different times according to the three-dimensional feature vector.

[0079] In this embodiment, for any two vectors at different times, calculations are carried out. The first two dimensions in the three-dimensional feature vector respectively represent the longitude and latitude information of the center of the tropical cyclone. With the help of the spherical cosine formula, by inputting the longitude and latitude values corresponding to the centers at two times, the spatial distance between them on the earth's sphere is accurately calculated. This method takes into account the curvature of the earth and can effectively reflect the actual spatial interval. The third dimension of the vector integrates the maximum 10m wind speed and the lowest sea level pressure information. Subtract the numerical values of this dimension of the two-time vectors and take the absolute value to obtain the preliminary value of the intensity difference. Considering that the wind speed and air pressure have different degrees of influence on the intensity, according to historical data and meteorological principles, their respective weights are determined, and the preliminary intensity difference value is weighted and corrected, so as to obtain a value that can more accurately reflect the intensity difference between the centers of two tropical cyclones at different times.

[0080] In an alternative embodiment, the spatial distance satisfies the following formula: Wherein, is the three-dimensional feature vector calculated using the spherical cosine formula and the three-dimensional feature vector adjacent and the comprehensive distance metric function is less than the clustering radius value of other three-dimensional feature vectors The spatial distance value between is the average radius of the Earth, and are the longitudes of the centers of two tropical cyclones at different times respectively, and are the latitudes of the centers of two tropical cyclones at different times respectively.

[0081] In an optional embodiment, the intensity difference satisfies the following formula: Wherein, is and is the intensity difference value of and are the intensity indexes of the centers of two tropical cyclones at different times respectively.

[0082] Step S50202, construct a comprehensive distance metric function between the centers of two tropical cyclones at different times according to the spatial distance and the intensity difference.

[0083] In this embodiment, the comprehensive distance metric function is a mathematical tool that combines the spatial distance and the intensity difference. Through a specific calculation method, a value is obtained that can comprehensively reflect the degree of difference between the three-dimensional feature vectors corresponding to the centers of tropical cyclones at different times. This value can be used to judge the similarity or relevance between objects. In the study of tropical cyclones, it can assist in analyzing the changes of the centers of tropical cyclones at different times.

[0084] When studying the centers of tropical cyclones at different times, in order to accurately measure the comprehensive difference between them, a comprehensive distance metric function is constructed. First, determine two three-dimensional feature vectors, calculate the spatial distance value between the two three-dimensional feature vectors using the spherical cosine formula to reflect their spatial position differences; then calculate the intensity difference value between the two three-dimensional feature vectors to show the differences in intensity. Then, set weights to reflect the importance of the intensity difference in the comprehensive measurement. Finally, square the spatial distance value and the intensity difference value respectively, combine them according to the weights and take the square root to obtain the comprehensive distance metric value, thereby quantitatively describing the comprehensive difference situation in space and intensity between the centers of two tropical cyclones at different times.

[0085] The comprehensive distance metric function satisfies the following formula: Wherein, is the comprehensive distance metric function, is the three-dimensional feature vector, is adjacent to the three-dimensional feature vector and the value of the comprehensive distance metric function is less than the clustering radius value Other three-dimensional eigenvectors calculated using the spherical cosine formula and the spatial distance value between is the calculated and intensity difference value is the intensity weight

[0086] Among them is the historical intensity difference standard deviation

[0087] Step S50203, construct the adjacency structure of the candidate region according to the comprehensive distance metric function

[0088] Among them, constructing the adjacency structure of the candidate region according to the comprehensive distance metric function specifically includes the following sub-steps Step S5020301, obtain the historical path data of the tropical cyclone, and obtain the historical spatial data and historical intensity data of the historical tropical cyclone centers at two different times according to the historical path data

[0089] In this embodiment, these data detail the central longitude and latitude coordinates (spatial positions) of the tropical cyclone at different time points and the corresponding maximum 10m wind speed and minimum sea level pressure (intensity indicators). For two different times of the research target, first locate the records of these two times in the historical path data, extract the corresponding two central longitudes and latitudes, calculate the spherical distance between them using the spherical cosine formula as the historical spatial data; at the same time, extract the maximum wind speed and minimum pressure at the corresponding time and calculate the intensity difference as the historical intensity data. During this process, it is necessary to carefully check the accuracy and integrity of the data. For possible missing or incorrect data, methods such as data interpolation and cross-verification with other reliable data sources are used for processing to ensure that the obtained spatial data and intensity data can truly and accurately reflect the characteristics of the historical tropical cyclone centers at two different times

[0090] Step S5020302, calculate the 95th percentiles of the historical spatial data and the historical intensity data respectively to obtain the historical spatial threshold and the historical intensity threshold

[0091] Collect the spherical distance data between the centers of tropical cyclones at different times. Carefully arrange these data in ascending order to form an ordered data series. Then, count how many such data there are in total. Suppose there are n in total. Next, calculate 95% of n, that is, multiply n by 0.95 to get a value. If this value is an integer, directly find the data at the corresponding position in the sorted data series. This data is the 95th percentile of the historical spatial data, that is, the historical spatial threshold. If this value is not an integer, take the closest integer smaller than it. Suppose this integer is m. First, find the mth data in the data series, then find the (m + 1)th data, and subtract the mth data from the (m + 1)th data to get a difference. Subtract m from the non-integer value calculated just now to get a decimal. Finally, add the product of this decimal and the difference to the mth data, and the result obtained is the historical spatial threshold.

[0092] Perform the same process on the historical intensity data. Collect the data reflecting the intensity changes of tropical cyclones, such as the maximum wind speed difference, the minimum pressure difference, etc., and arrange them in ascending order. Use the same method as for processing the historical spatial data to calculate 95% of the number of data, and determine the 95th percentile according to whether the result is an integer. This percentile is the historical intensity threshold, which represents a boundary value for the intensity change of tropical cyclones.

[0093] Step S5020303, calculate the clustering radius according to the historical spatial threshold and the historical intensity threshold.

[0094] In this embodiment, the calculation of the clustering radius is based on the in-depth mining of the historical characteristics of tropical cyclones, aiming to provide a reasonable quantitative standard for judging the relevance between the centers of tropical cyclones at different times. The historical spatial threshold reflects the general upper limit of the spatial distance between adjacent-time centers during the historical movement of tropical cyclones, and the distance between the centers of tropical cyclones at adjacent times will not exceed the historical spatial threshold. Similarly, the historical intensity threshold represents the common boundary in terms of the intensity change of tropical cyclones, and its intensity change will not exceed the historical intensity threshold.

[0095] When calculating the clustering radius, the historical spatial threshold and the historical intensity threshold are comprehensively considered. Since the spatial movement and intensity change of tropical cyclones are both crucial for determining their identity and development path, a specific calculation method is used to combine the two. Usually, an intensity weight is determined according to historical data and meteorological principles, and this weight is used to reflect the importance of intensity change in the overall judgment. Add the square of the historical spatial threshold to the square of the product of the intensity weight and the historical intensity threshold, and then take the square root of the sum of the two. The result obtained is the clustering radius.

[0096] The clustering radius plays a crucial role in subsequent tropical cyclone tracking and analysis. When constructing the adjacency structure of candidate regions, it serves as a measurement criterion to determine whether the comprehensive distance between the three-dimensional feature vectors of tropical cyclone centers at different times is within a reasonable range. If the comprehensive distance between two times is less than the clustering radius, they are considered to have a strong correlation and are likely to represent the states of the same tropical cyclone at different times, thus laying the foundation for accurately depicting the path and intensity information of the tropical cyclone over time.

[0097] The clustering radius satisfies the following formula: where, is the clustering radius, is the historical spatial threshold, is the intensity weight, is the historical intensity threshold.

[0098] Step S5020304: Construct the adjacency structure of the candidate region according to the clustering radius and the comprehensive distance metric function.

[0099] The adjacency structure satisfies the following formula: where is the adjacency structure constructed based on the three-dimensional feature vector ; are other three-dimensional feature vectors that are adjacent to the three-dimensional feature vector and whose comprehensive distance metric function is less than the clustering radius value ; is the spatial distance value between and calculated using the spherical cosine formula; is the intensity weight; is the intensity difference value calculated for and ; is the clustering radius value.

[0100] Step S503: Classify the central nodes of the candidate region according to the adjacency structure.

[0101] In this embodiment, the adjacency structure defines the relationship between three-dimensional feature vectors. By integrating the distance metric function and the clustering radius, it is determined which nodes are adjacent to each other and may belong to the same tropical cyclone. Based on this, when classifying the central nodes, the central nodes of all candidate regions are traversed. For each node, other nodes that are adjacent to it and have a comprehensive distance less than the clustering radius are judged according to the adjacency structure, and these mutually adjacent nodes are grouped into the same class. During the classification process, specific algorithms and logics are adopted. For example, starting from an unclassified node, it is marked as the starting node of a new class, and then according to the adjacency relationship, all associated adjacent nodes are included in this class and marked as classified. Then continue to find the next unclassified node and repeat the above process until all nodes are classified. Through such a classification method, the central nodes of tropical cyclones with similar spatio-temporal characteristics and intensity characteristics can be grouped together, clearly distinguishing different tropical cyclone individuals. This not only helps to analyze the independent development path and intensity change of each tropical cyclone over time, but also avoids misassociating the nodes of different cyclones, thus greatly improving the accuracy and reliability of tropical cyclone tracking.

[0102] Step S504: Construct a density-connected clustering according to the classification result, and obtain the path and intensity information of each tropical cyclone changing with time according to the density-connected clustering.

[0103] In this embodiment, the density-connected clustering is a clustering method based on the density relationship between data points. In the present invention, the central nodes in each class can be regarded as a set of data points with similar spatial and intensity characteristics. For these sets, the core points are first determined, that is, those nodes with a sufficient number of adjacent points in their neighborhoods (the adjacency relationship is determined by the adjacency structure and the clustering radius). The core points represent the relatively stable and significant states of the tropical cyclone at a certain moment.

[0104] Starting from the core points, the nodes directly or indirectly connected to them are connected through the density-reachable relationship. If a node can reach another node through a series of density-connected core points, then these two nodes belong to the same density-connected clustering. During the construction process, the clustering range is continuously expanded, and the eligible nodes are gradually included to form complete clustering clusters. Each clustering cluster corresponds to a set of states of a tropical cyclone at different times.

[0105] Through such a clustering method, different tropical cyclone individuals can be clearly distinguished. For each clustering cluster, according to the chronological order of the nodes within the cluster, the corresponding longitude and latitude data are extracted in sequence, and these longitude and latitude data are connected in chronological order to form the movement path of the tropical cyclone. At the same time, based on the intensity data in the nodes, including the maximum 10m wind speed and the lowest sea level pressure, the intensity change situation at different times can be intuitively presented. This process not only effectively integrates various information obtained in the previous steps, but also deeply explores the internal connections between the data, providing solid data support for comprehensively and accurately studying the development and evolution laws of tropical cyclones.

[0106] Automatically track the tropical cyclone based on the said path and the said intensity information.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered by the scope of the claims and the description of the present invention.

Claims

1. An automatic tracking method for tropical cyclones based on relative thresholds, characterized in that, The method includes: Obtaining meteorological historical data through a numerical weather prediction model, and obtaining the relative vorticity and warm core intensity of historical tropical cyclones based on the meteorological historical data; Collecting historical tropical cyclone data, and calculating the relative thresholds of the relative vorticity and the warm core intensity based on the historical tropical cyclone data; Obtaining meteorological data at different times according to the numerical weather prediction model, and obtaining candidate regions at different times according to the relative thresholds and the meteorological data; Obtaining latitude and longitude data and intensity data of multiple groups of tropical cyclone centers at different times based on the candidate regions at different times; Obtaining the path and intensity information of each tropical cyclone changing with time by using a clustering algorithm based on the latitude and longitude data and the intensity data.

2. The automatic tracking method of tropical cyclones based on relative thresholds according to claim 1, characterized in that The obtaining the relative vorticity and the warm core intensity of historical tropical cyclones based on the meteorological historical data includes: Extracting the meridional wind component and the zonal wind component at a distance of 10 m from the ground in the meteorological historical data; Calculating the relative vorticity of the tropical cyclone according to the meridional wind component and the zonal wind component; Respectively extracting the first average temperature and the second average temperature of the 200 hPa isobaric surface and the 500 hPa isobaric surface in the meteorological historical data; Calculating the warm core intensity of the tropical cyclone according to the first average temperature and the second average temperature.

3. The automatic tracking method of tropical cyclones based on relative thresholds according to claim 1, characterized in that The calculating the relative thresholds of the relative vorticity and the warm core intensity based on the historical tropical cyclone data includes: Based on the relative vorticity and the warm core intensity, using a statistical analysis method to extract the historical relative vorticity values and historical warm core structure values within a 2° spherical neighborhood of the historical tropical cyclone center in the historical tropical cyclone data; Calculating the historical relative vorticity maximum value and the historical warm core structure maximum value of each historical tropical cyclone according to the historical relative vorticity values and the historical warm core structure values; Respectively calculating the 95% quantiles of the historical relative vorticity maximum value and the historical warm core structure maximum value to obtain the relative thresholds.

4. The automatic tracking method of tropical cyclones based on relative thresholds according to claim 3, characterized in that The relative thresholds satisfy the following formula: is the relative threshold, is the candidate value of the variable, is the indicator function, is the maximum historical relative vorticity or maximum historical warm core structure of the th historical tropical cyclone, and is the number of all historical tropical cyclones used to construct the empirical distribution.

5. The automatic tracking method of tropical cyclones based on relative thresholds according to claim 1, characterized in that, The obtaining candidate regions at different times according to the relative thresholds and the meteorological data includes: Setting constraint conditions by using the relative thresholds, and screening the meteorological data by using the constraint conditions to obtain candidate regions at different times; The candidate regions satisfy the following formula: Among them, is the relative threshold of relative vorticity, is the relative vorticity in meteorological data, is the relative threshold of warm core intensity, is the warm core intensity in meteorological data.

6. The automatic tracking method of tropical cyclones based on relative thresholds according to claim 1, characterized in that, The intensity data includes the maximum 10 m wind speed and the lowest sea level pressure. The obtaining latitude and longitude data and intensity data of multiple groups of tropical cyclone centers at different times based on the candidate regions at different times includes: Calculating the 10 m wind speed at different times by using the meteorological data based on the candidate regions at different times; [[ID= ​ ​ 7. The automatic tracking method of tropical cyclones based on relative thresholds according to claim 1, characterized in that The path and intensity information of each tropical cyclone over time obtained by using a clustering algorithm based on the latitude and longitude data and the intensity data includes: Construct a three-dimensional feature vector for each of the candidate regions according to the latitude and longitude data and the intensity data; Construct an adjacency structure for the candidate regions according to the three-dimensional feature vector; Classify the central nodes of the candidate regions according to the adjacency structure; Construct a density connectivity clustering according to the result of the classification, and obtain the path and intensity information of each tropical cyclone over time according to the density connectivity clustering.

8. The automatic tracking method of tropical cyclones based on relative thresholds according to claim 7, characterized in that, The constructing of the adjacency structure of the candidate regions according to the three-dimensional feature vector includes: Calculate the spatial distance and intensity difference between the centers of two tropical cyclones at different times according to the three-dimensional feature vector; Construct a comprehensive distance metric function between the centers of two tropical cyclones at different times according to the spatial distance and the intensity difference; Construct the adjacency structure of the candidate regions according to the comprehensive distance metric function.

9. The automatic tracking method for tropical cyclones based on relative thresholds according to claim 8, characterized in that, The constructing of the adjacency structure of the candidate regions according to the comprehensive distance metric function includes: Obtain the historical path data of the tropical cyclone, and obtain the historical spatial data and historical intensity data of the historical tropical cyclone centers at two different times according to the historical path data; Calculate the 95th percentiles of the historical spatial data and the historical intensity data respectively to obtain a historical spatial threshold and a historical intensity threshold; Calculate the clustering radius according to the historical spatial threshold and the historical intensity threshold; Construct the adjacency structure of the candidate regions according to the clustering radius and the comprehensive distance metric function.

10. The automatic tracking method for tropical cyclones based on relative thresholds according to claim 9, wherein The adjacency structure satisfies the following formula: Among them, is the adjacency structure constructed according to the three-dimensional feature vector . is another three-dimensional feature vector that is adjacent to the three-dimensional feature vector and the comprehensive distance metric function is less than the clustering radius value . is the spatial distance value between and calculated using the spherical cosine formula. is the intensity weight. is and 's intensity difference value. is the clustering radius value.

Images