A surface cyclone system tracking algorithm

Abstract

The present invention provides a ground cyclone system tracking algorithm, which relates to the technical field of ground cyclone system tracking. The algorithm specifically includes the following steps: S1. Cyclone identification algorithm. The algorithm starts with the value and number of contour lines to be analyzed in the analysis area, and then analyzes the low-pressure blocks corresponding to each contour line one by one starting from the minimum contour line, and then determines the cyclone system. Compared with most cyclone identification methods, it is more in line with the habit of subjective analysis and more consistent with the results of subjective analysis. S2. Cyclone tracking. Based on the low-pressure system data of each time calculated in the previous step, the Lagrangian method is used, and the nearest neighbor rule and cross-correlation method are applied to judge the temporal similarity of the low-pressure system, and a similarity discrimination factor of the cross-correlation method is constructed. This is different from most existing tracking algorithms in that a mixed method based on the cyclone centroid position and the cross-correlation method is used for tracking, and a similarity discrimination operator is constructed, so that the tracking result is more reasonable.

Description

Technical Field

[0001] The present invention relates to the technical field of ground cyclone system tracking, and in particular to a ground cyclone system tracking algorithm. Background Art

[0002] Automatic cyclone identification and tracking has always been an important technology of concern to meteorological scientists. As early as 1985, Akyildiz automatically identified and tracked cyclone centers when evaluating the performance of numerical models. To this day, many meteorological scientists are still continuously improving and developing automatic cyclone identification and tracking technology, such as Lu and Chang in 2017, Lakkis et al., and Valsangkar et al. in 2019.

[0003] Many cyclone identification methods begin by searching for the center of minimum pressure or the center of maximum vorticity (Hoskins and Hodges, 2002; Feser et al., 2015; Yang et al., 2015; Massey et al., 2017; Chang, 2017; Lu, 2017). Furthermore, many automatic identification algorithms perform spatial filtering before starting identification to eliminate the influence of large or small noise scales (Anderson et al., 2003; Zappa et al., 2013; Feser et al., 2015; Massey, 2016).

[0004] Most automatic identification algorithms (Hanley and Caballero, 2012; Lu, 2017; Valsangkar et al., 2019) use the method of determining the value of a grid point in the pressure field to be less than the values ​​of the surrounding eight points. They first determine the candidate center of the cyclone and then use the center to determine the range of the cyclone's various contour lines. This method has the potential and feasibility to further improve the accuracy of cyclone system tracking.

[0005] References:

[0006] Akyildiz,V.,1984:Systematic errors in the behavior of cyclones in the ECMWF operational models.Tellus,37A,297-308.

[0007] Hoskins B,Hodges K.2002.New perspectives on the northern hemispherewinter storm tracks.J.Atmos.Sci.59:1041–1061.

[0008] Anderson DK,Hodges KI,Hoskins BJ.2003.Sensitivity of feature-basedanalysis methods of storm tracks to the form of background fieldremoval.Mon.Weather Rev.131:565–573.

[0009] Inatsu M.2009.The neighbor enclosed area tracking algorithm forextratropical wintertime cyclones.Atmos.Sci.Lett.10:267-272.

[0010] Hodges,K.I.1994.A general-method for tracking analysis and itsapplication to meteorological data.Monthly Weather Review 122:2573-2586.

[0011] Hewson TD.2009.Diminutive frontal waves─a link between fronts andcyclones.J.Atmos.Sci.66:116–132.

[0012] Lakkis SG,Canziani P,Yuchechen A et al.2019.A 4D feature-trackingalgorithm:a multidimensional view of cyclonesystems.Q.J.R.Meteorol.Soc.145:1–23.

[0013] Zappa G,Shaffrey LC,Hodges KI et al.2013.A multimodel assessment offuture projections of North Atlantic and European extratropical cyclones inthe CMIP5 climate models.J.Cl imate 26:5846–5862.

[0014] Feser F,Barcikowska M,Krueger O et al.2015.Storminess over the NorthAtlantic and northwestern Europe-a review.Q.J.R.Meteorol.Soc.141:350–382.

[0015] Massey NR.2016.Feature tracking in high-resolution regional climatedata.Comput.Geosci.93:36–44.

[0016] Hanley,J.,and R.Caballero,2012.Objective identification and trackingof multicentre cyclones in the ERA-Interim reanalysis dataset.Quart.J.Roy.Meteor.Soc.138:612–625.

[0017] Yang X,Vecchi GA,Gudgel RG et al.2015.Seasonal predictability ofextratropical storm tracks in GFDL’s high-resolution climate predictionmodel.J.Cl imate 28:3592–3611.

[0018] Chang EKM.2017.Projected significant increase in the number ofextreme extratropical cyclones in the Southern Hemisphere.J.Climate 30:4915–4935.

[0019] Lu,C.H.2017.A modified algorithm for identifying and trackingextratropical cyclones.Adv.Atmos.Sci.34(7):909–924.

[0020] Valsangkar AA,Monteiro JM,Narayanan V.2019.An ExploratoryFramework for Cyclone Identification and Tracking.IEEETrans.Vis.Comput.Graph.25(3):1460-1473. Summary of the Invention

[0021] (1) Technical problems solved

[0022] In response to the shortcomings of the existing technology, the present invention provides a ground cyclone system tracking algorithm. When identifying a ground cyclone, the algorithm starts with the value and number of contour lines that need to be analyzed in the analysis area, and then analyzes the low-pressure blocks corresponding to each contour line one by one starting from the minimum contour line, and then determines the cyclone system. Compared with most cyclone identification methods, the algorithm is more in line with the habit of subjective analysis and is more consistent with the results of subjective analysis. After cyclone identification, tracking is performed. Based on the low-pressure system data of each time calculated in the previous step, the Lagrangian method is used, and the nearest neighbor rule and cross-correlation method are applied to judge the temporal similarity of the low-pressure system, and a similarity discrimination factor of the cross-correlation method is constructed. The difference between this and most existing tracking algorithms is that a mixed tracking method based on the cyclone centroid position and the cross-correlation method is used, and a shape similarity discrimination operator is constructed, so that the tracking result is more reasonable.

[0023] (2) Technical solution

[0024] To achieve the above objectives, the present invention is implemented through the following technical solutions: a ground cyclone system tracking algorithm, specifically comprising the following steps:

[0025] S1. Cyclone Identification Algorithm

[0026] A. Set the parameters of the equal longitude and latitude grid within a limited area

[0027] Set the number of longitude grid points to SNnum, the number of latitude grid points to EWnum, the starting longitude to blon, the starting latitude to blat, the ending longitude to elon, the ending latitude to elat, the longitude distance of the grid points to detSN, the latitude distance of the grid points to detEW, where detSN equals detEW, and SNnum = (elat-blat) / detSN+1, EWnum = (elon-blon) / detEW+1. Then the grid points are numbered 1, 2, ..., EWnum from west to east, and 1, 2, ..., SNnum from south to north.

[0028] B. Determine the value of the contour line

[0029] Calculate the minimum and maximum values ​​of the sea level pressure field in the region, with 2hPa as the contour interval, and take the value between the minimum and maximum values ​​that is divisible by 2 as the contour value. The contour values ​​are arranged from small to large and are recorded as Lslp[i], i=1,2,...n;

[0030] C. Identify contour lines;

[0031] a. Determine whether the contour line is a closed contour line

[0032] If the point G[i] on the contour line is located at the edge of the finite region, it is an open contour line, denoted by O. If any point G[i] on the contour line is no longer at the edge of the finite region, it is a closed contour line, denoted by C.

[0033] b. Determine whether the contour line is the contour line of the low-pressure system

[0034] If it is a closed contour line, calculate whether the values ​​of all grid points in the contour line are less than or equal to the value of the contour line. If they are less than or equal to the value of the contour line, it is a low-pressure contour line, otherwise it is a high-pressure contour line and is removed;

[0035] If it is an open contour line, the least squares method is used to fit a circle to all points, and the center position of the fitted circle is calculated. The grid point values ​​in the sector surrounded by the center of the circle and the contour line are judged. If all grid point values ​​are less than or equal to the value of the contour line, it is a low-pressure contour line, otherwise it is a high-pressure contour line and is removed.

[0036] D. Identification of low-pressure blocks;

[0037] 1) Block identification

[0038] Assume that any point in a block is less than or equal to the value of the contour line corresponding to the block, and any grid point in the block can always find at least one adjacent grid point. The adjacent condition is that the grid points are separated by a latitudinal or longitudinal grid distance. Then the set of these points is called a block.

[0039] 2) Determine whether the block is a low-voltage block

[0040] The contour line corresponding to the low-pressure block is judged to be the contour line of the block. The judgment condition is that all points on the contour line can find a point in the block that satisfies the condition, and the points on the line are adjacent to the points in the block. The adjacent condition is the same as step 1). If the condition is met, it is a low-pressure block and is retained. Otherwise, it is discarded.

[0041] 3) Determine whether the low-pressure block is a closed low-pressure block

[0042] If a grid point in a low-pressure block is located at the edge of a limited area, it is an open low-pressure block, otherwise it is a closed low-pressure block;

[0043] E. Identification of low-pressure systems;

[0044] ①. Find Block[j] that contains Block[i], that is, all grid points on Block[i] are on Block[j], i≠j, and so on, and mark the blocks that meet the conditions until all low-pressure blocks are found, recorded as low-pressure system C[1], and record the value of the contour line of the low-pressure block, whether the corresponding low-pressure block is a closed isobaric block and equal to the grid points contained in the block;

[0045] ②. Find out whether there is an unmarked block included in the low-pressure system C[1]. If so, mark the low-pressure block and record the value of the corresponding low-pressure block's contour line, whether the corresponding low-pressure block is a closed isobaric block and the grid points included in the block;

[0046] ③. Repeat ① and ② until all low-voltage blocks are marked, and finally obtain n low-voltage systems, i.e., C[i], i = 1, 2, ..., n;

[0047] ④. Calculate the centroid position and lowest grid point value for the grid points of the low-pressure block corresponding to the minimum contour value of each low-pressure system, or for the low-pressure block that does not include other low-pressure blocks. Record the centroid position and lowest grid point value as the center position and central lowest pressure of the low-pressure system.

[0048] F. Obtain low-pressure system data of sea level pressure field

[0049] Low-pressure system data includes: the number of low-pressure systems, the values ​​of the contour lines of the low-pressure systems, the grid points and grid values ​​of the low-pressure systems, the center position of the pressure system, the lowest pressure at the center, and the grid points and grid values ​​of the low-pressure block where the center is located;

[0050] S2. Cyclone tracking algorithm

[0051] (1) Setting of relevant parameters of trajectory tracking algorithm

[0052] Nearest neighbor method: The threshold value of the distance Dc between the centers of two cyclonic systems at adjacent times is 250.0;

[0053] Cross-correlation method: If the number of grid points of the contour pressure blocks corresponding to the central pressure of the cyclone system at adjacent times is recorded as p1 and p2 respectively, then the discriminant function of the cyclone system similarity of the cross-correlation method is set to: shr = 0.5*corr + 0.5*(p / (p1+p2)), where corr is the maximum value that the correlation coefficient of the P1 pressure block can reach within the rectangular area formed by the P2 pressure block through spatial translation, and p is the number of valid points of the P1 pressure block within the rectangular interval formed by the P2 pressure block at this time, and the distance threshold (Dcc) between the centers of two similar cyclones is 1000 km;

[0054] (2) Cyclone system number

[0055] Take the center point of all cyclone systems at the first time as the starting point of the trajectory. Suppose there are x cyclone systems in total, and the cyclone systems are numbered 1, 2, ..., n one by one. Suppose each cyclone system has y centers, and they are numbered 1, 2, ..., n one by one. Then there are x trajectories in total, recorded as Trk[x], x = 1, 2, ..., n. Record whether the contour line corresponding to the center is closed, the longitude and latitude of the center, the lowest pressure of the center, the cyclone system number corresponding to the center, the center number, and the position of the corresponding pressure block point.

[0056] (3) Trajectory Trk[x] comparison

[0057] Take the center point of the cyclone system at the second time, calculate the distance between the last center in each trajectory Trk[x] and the cyclone center at this time, and take the center with the smallest distance <= 250.0 km, and add it to the trajectory Trk[x];

[0058] If the trajectory Trk[x] does not have a cyclone system center point that meets the conditions, then calculate the similarity coefficient shr between the pressure block corresponding to the last center point of Trk[x] and the pressure blocks corresponding to the center points of all cyclone systems at this time. If shr>0.5 and the distance Dcc between the two pressure blocks is less than 1000km, then determine that the center point of this pressure system is the latest center point of Trk[x], and record whether the contour line corresponding to the center is closed, the center longitude and latitude, the center minimum pressure, the cyclone system number corresponding to the center, the center number and the position of the corresponding pressure block point;

[0059] If the center point of the cyclone system at the second time cannot find a matching trajectory Trk[x], then add a new trajectory until all the center points of the cyclone system at the second time are exhausted;

[0060] Assume that there are n newly added trajectories, denoted as Trk[x+i], i=1,2,...,n, and record whether the contour line corresponding to the center is closed, the longitude and latitude of the center, the lowest pressure of the center, the cyclone system number corresponding to the center, the center number and the position of the corresponding pressure block point;

[0061] (4) Obtain the full-time cyclone activity trajectory Trk[i]

[0062] Repeat step 3 for the cyclone system center points from the third time to the last time, and finally obtain the trajectories of all cyclone activities in all data sets, Trk[i], i = 1, 2, ..., n;

[0063] (5) Operational analysis

[0064] α. Calculate the length of the trajectory Trk[i], where i = 1, 2,..., n. If the length is less than 2, then eliminate this trajectory;

[0065] β. Take any trajectory Trk[x] as the final trajectory Track[1] after operation analysis. If it meets the conditions of operation analysis, then mark that the operation analysis has been performed, and record the parameters of this trajectory: central longitude and latitude, central lowest air pressure, whether the isopleth corresponding to the center is closed, the cyclone system number corresponding to the center, the center number, and the position of the air pressure block point corresponding to it;

[0066] γ. Take the Track[1] trajectory as the mother trajectory, and perform operation analysis on the trajectory Trk[y], where x ≠ y. If it meets the conditions of operation analysis, then mark that the operation analysis has been performed, record the main trajectory obtained each time as Track[1], and record the time points of trajectory merging, separation, or reconnection. Analyze the secondary trajectory after the operation, and record the central longitude and latitude, central lowest air pressure, whether the isopleth corresponding to the center is closed, the cyclone system number corresponding to the center, the center number, and the position of the air pressure block point corresponding to it;

[0067] δ. Still take Track[1] as the mother trajectory, and perform operation analysis on all unmarked trajectories Trk exhaustively, and repeat step γ;

[0068] ε. For the remaining unmarked trajectories Trk, repeat steps β, γ, and δ to obtain the Track trajectories. Suppose there are n in total, then denote them as Track[i], where i = 1, 2,..., n.

[0069] Preferably, the automatic recognition algorithm uses the sea - level air pressure field of grid points with equal longitude and latitude in a limited area for automatic recognition of low - pressure systems.

[0070] Preferably, each grid point number record in step A of S1 consists of two parameters, namely the east - west grid point number and the north - south grid point number.

[0071] Preferably, for two adjacent grid point values (denoted as v1, v2) in step C of S1, if v1 < Lslp[i] <= v2, then take the position of grid point v2 as the position of the isopleth, and record the isopleth grid point position G[i], where i = 1, 2,..., n, and the set of all points adjacent to the positions of G[i] and G[i + 1].

[0072] Preferably, step D of S1 needs to calculate the low - pressure blocks corresponding to the values of all isopleths in the region.

[0073] Preferably, step E of S1 arranges the low - pressure block pairs in ascending order according to the values of the corresponding isopleths, and denotes them as Block[i], where i = 1, 2,..., n.

[0074] Preferably, the step S2 (5) performs fusion, separation, and reconnection analysis on all trajectories. The fusion, separation, and reconnection conditions are detailed in Hanley and Caballero (2012). In particular, the length of the past and future trajectory branches in the algorithm is required to be 1.

[0075] (3) Beneficial effects

[0076] The present invention provides a ground cyclone system tracking algorithm. It has the following beneficial effects:

[0077] 1. The present invention provides a ground cyclone system tracking algorithm. When identifying ground cyclones, the system starts with the value and number of contour lines to be analyzed in the analysis area, then analyzes the low-pressure blocks corresponding to each contour line one by one starting from the minimum contour line, and then determines the cyclone system. Compared with most cyclone identification methods, this algorithm is more in line with the habit of subjective analysis and is more consistent with the results of subjective analysis.

[0078] 2. The present invention provides a ground cyclone system tracking algorithm. After identifying the ground cyclone, the ground cyclone system tracking algorithm continues to track. Based on the low-pressure system data of each time calculated in the previous step, the Lagrangian method is used, and the nearest neighbor rule and cross-correlation method are applied to judge the temporal similarity of the low-pressure system, and a similarity discrimination factor of the cross-correlation method is constructed. The difference between this algorithm and most existing tracking algorithms is that the algorithm uses a mixed tracking method based on the cyclone centroid position and the cross-correlation method, and constructs a shape similarity discrimination operator, so the tracking results are more reasonable. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 It is a schematic diagram of the process of the present invention;

[0080] Figure 2 This is a schematic diagram of the trajectory of the Jianghuai cyclone center from 11:00 on June 27, 2020 to 6:00 on July 1, 2020. DETAILED DESCRIPTION

[0081] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0082] Example:

[0083] like Figure 1-2 As shown, an embodiment of the present invention provides a ground cyclone system tracking algorithm, which specifically includes the following steps:

[0084] S1. Cyclone Identification Algorithm

[0085] A. Set the parameters of the equal longitude and latitude grid within a limited area

[0086] Set the number of longitude grid points to SNnum, the number of latitude grid points to EWnum, the starting longitude to blon, the starting latitude to blat, the ending longitude to elon, the ending latitude to elat, the longitude distance of the grid points to detSN, the latitude distance of the grid points to detEW, where detSN equals detEW, and SNnum = (elat-blat) / detSN+1, EWnum = (elon-blon) / detEW+1. Then the grid points are numbered 1, 2, ..., EWnum from west to east, and 1, 2, ..., SNnum from south to north.

[0087] B. Determine the value of the contour line

[0088] Calculate the minimum and maximum values ​​of the sea level pressure field in the region, with 2hPa as the contour interval, and take the value between the minimum and maximum values ​​that is divisible by 2 as the contour value. The contour values ​​are arranged from small to large and are recorded as Lslp[i], i=1,2,...n;

[0089] C. Identify contour lines;

[0090] a. Determine whether the contour line is a closed contour line

[0091] If the point G[i] on the contour line is located at the edge of the finite region, it is an open contour line, denoted by O. If any point G[i] on the contour line is no longer at the edge of the finite region, it is a closed contour line, denoted by C.

[0092] b. Determine whether the contour line is the contour line of the low-pressure system

[0093] If it is a closed contour line, calculate whether the values ​​of all grid points in the contour line are less than or equal to the value of the contour line. If they are less than or equal to the value of the contour line, it is a low-pressure contour line, otherwise it is a high-pressure contour line and is removed;

[0094] If it is an open contour line, the least squares method is used to fit a circle to all points, and the center position of the fitted circle is calculated. The grid point values ​​in the sector surrounded by the center of the circle and the contour line are judged. If all grid point values ​​are less than or equal to the value of the contour line, it is a low-pressure contour line, otherwise it is a high-pressure contour line and is removed.

[0095] D. Identification of low-pressure blocks;

[0096] 1) Block identification

[0097] Assume that any point in a block is less than or equal to the value of the contour line corresponding to the block, and any grid point in the block can always find at least one adjacent grid point. The adjacent condition is that the grid points are separated by a latitudinal or longitudinal grid distance. Then the set of these points is called a block.

[0098] 2) Determine whether the block is a low-voltage block

[0099] The contour line corresponding to the low-pressure block is judged to be the contour line of the block. The judgment condition is that all points on the contour line can find a point in the block that satisfies the condition, and the points on the line are adjacent to the points in the block. The adjacent condition is the same as step 1). If the condition is met, it is a low-pressure block and is retained. Otherwise, it is discarded.

[0100] 3) Determine whether the low-pressure block is a closed low-pressure block

[0101] If a grid point in a low-pressure block is located at the edge of a limited area, it is an open low-pressure block, otherwise it is a closed low-pressure block;

[0102] E. Identification of low-pressure systems;

[0103] ①. Find Block[j] that contains Block[i], that is, all grid points on Block[i] are on Block[j], i≠j, and so on, and mark the blocks that meet the conditions until all low-pressure blocks are found, recorded as low-pressure system C[1], and record the value of the contour line of the low-pressure block, whether the corresponding low-pressure block is a closed isobaric block and equal to the grid points contained in the block;

[0104] ②. Find out whether there is an unmarked block included in the low-pressure system C[1]. If so, mark the low-pressure block and record the value of the corresponding low-pressure block's contour line, whether the corresponding low-pressure block is a closed isobaric block and the grid points included in the block;

[0105] ③. Repeat ① and ② until all low-voltage blocks are marked, and finally obtain n low-voltage systems, i.e., C[i], i = 1, 2, ..., n;

[0106] ④. Calculate the centroid position and lowest grid point value for the grid points of the low-pressure block corresponding to the minimum contour value of each low-pressure system, or for the low-pressure block that does not include other low-pressure blocks. Record the centroid position and lowest grid point value as the center position and central lowest pressure of the low-pressure system.

[0107] F. Obtain low-pressure system data of sea level pressure field

[0108] Low-pressure system data includes: the number of low-pressure systems, the values ​​of the contour lines of the low-pressure systems, the grid points and grid values ​​of the low-pressure systems, the center position of the pressure system, the lowest pressure at the center, and the grid points and grid values ​​of the low-pressure block where the center is located;

[0109] S2. Cyclone tracking algorithm

[0110] (1) Setting of relevant parameters of trajectory tracking algorithm

[0111] Nearest neighbor method: The threshold value of the distance Dc between the centers of two cyclonic systems at adjacent times is 250.0;

[0112] Cross-correlation method: If the number of grid points of the contour pressure blocks corresponding to the central pressure of the cyclone system at adjacent times is recorded as p1 and p2 respectively, then the discriminant function of the cyclone system similarity of the cross-correlation method is set to: shr = 0.5*corr + 0.5*(p / (p1+p2)), where corr is the maximum value that the correlation coefficient of the P1 pressure block can reach within the rectangular area formed by the P2 pressure block through spatial translation, and p is the number of valid points of the P1 pressure block within the rectangular interval formed by the P2 pressure block at this time, and the distance threshold (Dcc) between the centers of two similar cyclones is 1000 km;

[0113] (2) Cyclone system number

[0114] Take the center point of all cyclone systems at the first time as the starting point of the trajectory. Suppose there are x cyclone systems in total, and the cyclone systems are numbered 1, 2, ..., n one by one. Suppose each cyclone system has y centers, and they are numbered 1, 2, ..., n one by one. Then there are x trajectories in total, recorded as Trk[x], x = 1, 2, ..., n. Record whether the contour line corresponding to the center is closed, the longitude and latitude of the center, the lowest pressure of the center, the cyclone system number corresponding to the center, the center number, and the position of the corresponding pressure block point.

[0115] (3) Trajectory Trk[x] comparison

[0116] Take the center point of the cyclone system at the second time, calculate the distance between the last center in each trajectory Trk[x] and the cyclone center at this time, and take the center with the smallest distance <= 250.0 km, and add it to the trajectory Trk[x];

[0117] If the trajectory Trk[x] does not have a cyclone system center point that meets the conditions, then calculate the similarity coefficient shr between the pressure block corresponding to the last center point of Trk[x] and the pressure blocks corresponding to the center points of all cyclone systems at this time. If shr>0.5 and the distance Dcc between the two pressure blocks is less than 1000km, then determine that the center point of this pressure system is the latest center point of Trk[x], and record whether the contour line corresponding to the center is closed, the center longitude and latitude, the center minimum pressure, the cyclone system number corresponding to the center, the center number and the position of the corresponding pressure block point;

[0118] If the center point of the cyclone system at the second time cannot find a matching trajectory Trk[x], then add a new trajectory until all the center points of the cyclone system at the second time are exhausted;

[0119] Assume that there are n newly added trajectories, denoted as Trk[x+i], i=1,2,...,n, and record whether the contour line corresponding to the center is closed, the longitude and latitude of the center, the lowest pressure of the center, the cyclone system number corresponding to the center, the center number and the position of the corresponding pressure block point;

[0120] (4) Obtain the full-time cyclone activity trajectory Trk[i]

[0121] Repeat step 3 for the cyclone system center points from the third time to the last time, and finally obtain the trajectories of all cyclone activities in all data sets, Trk[i], i = 1, 2, ..., n;

[0122] (5) Operational analysis

[0123] α. Calculate the length of the trajectory Trk[i], i = 1, 2, ..., n. If the length is less than 2, remove the trajectory;

[0124] β. Take any trajectory Trk[x] as the final trajectory Track[1] after the operation analysis. If it meets the conditions of the operation analysis, it is marked as having been analyzed. The parameters of the trajectory are recorded: the longitude and latitude of the center, the lowest pressure of the center, whether the contour line corresponding to the center is closed, the cyclone system number corresponding to the center, the center number and the position of the corresponding pressure block point;

[0125] γ. Take Track[1] as the parent track and perform an operational analysis on track Trk[y], where x≠y. If the conditions for operational analysis are met, the analysis is marked as completed. The main track obtained from each analysis is recorded as Track[1], and the time points of track merging, separation, or reconnection are recorded. The secondary track after the analysis operation is recorded, along with the center longitude and latitude, the lowest center pressure, whether the contour line corresponding to the center is closed, the cyclone system number corresponding to the center, the center number, and the position of the corresponding pressure block point.

[0126] δ. Still taking Track[1] as the parent track, perform operation analysis on all unlabeled tracks Trk and repeat step γ;

[0127] ε. Repeat steps β, γ, and δ for the remaining unmarked tracks Trk to obtain Track. Suppose there are n tracks in total, which are recorded as Track[i], where i = 1, 2, ..., n.

[0128] There are usually two types of methods for tracking cyclone trajectories. One is the Euler method, and the other is the Lagrangian method. This algorithm is the Lagrangian method, and its principle is based on the data of the low-pressure system at each time step calculated in the previous step. The nearest neighbor rule and cross-correlation method are applied to judge the similarity of the low-pressure system in time, which is different from the algorithms that solely apply the nearest neighbor method or feature similarity (Inatsu, 2009; Hodges, 1994; Hanley and Caballero, 2012, Hewson, 2009), and is also different from the comprehensive application of the nearest neighbor method and feature similarity method by Lu (2017). The cross-correlation method is applied in this algorithm, and a discriminant factor for system similarity is constructed, making the tracking results more accurate.

[0129] The automatic recognition algorithm uses the sea-level pressure field of grid points with equal longitude and latitude in a limited area for the automatic recognition of low-pressure systems. Each record of the grid point number in step A of S1 consists of two parameters, namely the east-west grid point number and the north-south grid point number. In step C of S1, for two adjacent grid point values (assumed to be v1 and v2), if v1 < Lslp[i] <= v2, then the position of grid point v2 is taken as the position of the isoline, and the isoline grid point position G[i] is recorded, where i = 1, 2,..., n, and the set of all points adjacent to the positions of G[i] and G[i + 1]. In step D of S1, it is necessary to calculate the low-pressure blocks corresponding to the values of all isolines in the region. In step E of S1, the low-pressure block pairs are arranged in ascending order according to the values of the corresponding isolines, denoted as Block[i], where i = 1, 2,..., n. In step (5) of S2, fusion, separation, and reconnection operations and analyses are performed on all trajectories. The conditions for fusion, separation, and reconnection are detailed in Hanley and Caballero (2012). In particular, the length requirements for the past and future trajectory branches in the algorithm are 1.

[0130] This algorithm is applied to the ERA5 dataset (European Centre for Medium-Range Weather Forecasts reanalysis data) from June 27 to July 1, 2020. The time resolution of this reanalysis data is 1 hour, and the spatial resolution is an equal longitude-latitude grid of 0.25° × 0.25°. The intercepted spatial range is: 26°N - 36°N in the north-south direction and 110°E - 126°E in the east-west direction. As a result, the whole process of a Jianghuai cyclone from brewing and formation to moving eastward and entering the sea is obtained (as Figure 1 ) As can be seen from the figure, from 11:00 on June 27, 2020, when the low-pressure system started to move, it strengthened to form a cyclone center with a central pressure of 998 hPa at 15:00 on June 27, and then started to move eastward. During this period, the cyclone center merged once, split three times, and reconnected twice, lasting for 92 hours.

[0131] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A ground cyclone system tracking algorithm, characterized by: The specific steps include: S1. Cyclone Identification Algorithm A. Set the parameters of the equal longitude and latitude grid within a limited area Set the number of longitude grid points to SNnum, the number of latitude grid points to EWnum, the starting longitude to blon, the starting latitude to blat, the ending longitude to elon, the ending latitude to elat, the longitude distance of the grid points to detSN, the latitude distance of the grid points to detEW, where detSN equals detEW, and SNnum = (elat-blat) / detSN+1, EWnum = (elon-blon) / detEW+1. Then the grid points are numbered 1, 2, ..., EWnum from west to east, and 1, 2, ..., SNnum from south to north. B. Determine the value of the contour line Calculate the minimum and maximum values ​​of the sea level pressure field in the region, with 2hPa as the contour interval, and take the value between the minimum and maximum values ​​that is divisible by 2 as the contour value. The contour values ​​are arranged from small to large and are recorded as Lslp[i], i=1,2,...n; C. Identify contour lines; a. Determine whether the contour line is a closed contour line If the point G[i] on the contour line is located at the edge of the finite region, it is an open contour line, denoted by O. If any point G[i] on the contour line is no longer at the edge of the finite region, it is a closed contour line, denoted by C. b. Determine whether the contour line is the contour line of the low-pressure system If it is a closed contour line, calculate whether the values ​​of all grid points in the contour line are less than or equal to the value of the contour line. If they are less than or equal to the value of the contour line, it is a low-pressure contour line, otherwise it is a high-pressure contour line and is removed; If it is an open contour line, the least squares method is used to fit a circle to all points, and the center position of the fitted circle is calculated. The grid point values ​​in the sector surrounded by the center of the circle and the contour line are judged. If all grid point values ​​are less than or equal to the value of the contour line, it is a low-pressure contour line, otherwise it is a high-pressure contour line and is removed. D. Identification of low-pressure blocks; 1) Block identification Assume that any point in a block is less than or equal to the value of the contour line corresponding to the block, and any grid point in the block can always find at least one adjacent grid point. The adjacent condition is that the grid points are separated by a latitudinal or longitudinal grid distance. Then the set of these points is called a block. 2) Determine whether the block is a low-voltage block The contour line corresponding to the low-pressure block is judged to be the contour line of the block. The judgment condition is that all points on the contour line can find a point in the block that satisfies the condition, and the points on the line are adjacent to the points in the block. The adjacent condition is the same as step 1). If the condition is met, it is a low-pressure block and is retained. Otherwise, it is discarded. 3) Determine whether the low-pressure block is a closed low-pressure block If a grid point in a low-pressure block is located at the edge of a limited area, it is an open low-pressure block, otherwise it is a closed low-pressure block; E. Identification of low-pressure systems; ①. Find Block[j] that contains Block[i], that is, all grid points on Block[i] are on Block[j], i≠j, and so on, and mark the blocks that meet the conditions until all low-pressure blocks are found, recorded as low-pressure system C[1], and record the value of the contour line of the low-pressure block, whether the corresponding low-pressure block is a closed isobaric block and equal to the grid points contained in the block; ②. Find out whether there is an unmarked block included in the low-pressure system C[1]. If so, mark the low-pressure block and record the value of the corresponding low-pressure block's contour line, whether the corresponding low-pressure block is a closed isobaric block and the grid points included in the block; ③. Repeat ① and ② until all low-voltage blocks are marked, and finally obtain n low-voltage systems, i.e., C[i], i = 1, 2, ..., n; ④. Calculate the centroid position and lowest grid point value for the grid points of the low-pressure block corresponding to the minimum contour value of each low-pressure system, or for the low-pressure block that does not include other low-pressure blocks. Record the centroid position and lowest grid point value as the center position and central lowest pressure of the low-pressure system. F. Obtain low-pressure system data of sea level pressure field Low-pressure system data includes: the number of low-pressure systems, the values ​​of the contour lines of the low-pressure systems, the grid points and grid values ​​of the low-pressure systems, the center position of the pressure system, the lowest pressure at the center, and the grid points and grid values ​​of the low-pressure block where the center is located; S2. Cyclone tracking algorithm (1) Setting of relevant parameters of trajectory tracking algorithm Nearest neighbor method: The threshold value of the distance Dc between the centers of two cyclonic systems at adjacent times is 250.0; Cross-correlation method: If the number of grid points of the contour pressure blocks corresponding to the central pressure of the cyclone system at adjacent times is recorded as p1 and p2 respectively, then the discriminant function of the cyclone system similarity of the cross-correlation method is set to: shr = 0.5*corr + 0.5*(p / (p1+p2)), where corr is the maximum value that the correlation coefficient of the P1 pressure block can reach within the rectangular area formed by the P2 pressure block through spatial translation, and p is the number of valid points of the P1 pressure block within the rectangular interval formed by the P2 pressure block at this time, and the distance threshold (Dcc) between the centers of two similar cyclones is 1000 km; (2) Cyclone system number Take the center point of all cyclone systems at the first time as the starting point of the trajectory. Suppose there are x cyclone systems in total, and the cyclone systems are numbered 1, 2, ..., n one by one. Suppose each cyclone system has y centers, and they are numbered 1, 2, ..., n one by one. Then there are x trajectories in total, recorded as Trk[x], x = 1, 2, ..., n. Record whether the contour line corresponding to the center is closed, the longitude and latitude of the center, the lowest pressure of the center, the cyclone system number corresponding to the center, the center number, and the position of the corresponding pressure block point. (3) Trajectory Trk[x] comparison Take the center point of the cyclone system at the second time, calculate the distance between the last center in each trajectory Trk[x] and the cyclone center at this time, and take the center with the smallest distance <= 250.0 km, and add it to the trajectory Trk[x]; If the trajectory Trk[x] does not have a cyclone system center point that meets the conditions, then calculate the similarity coefficient shr between the pressure block corresponding to the last center point of Trk[x] and the pressure blocks corresponding to the center points of all cyclone systems at this time. If shr>0.5 and the distance Dcc between the two pressure blocks is less than 1000km, then determine that the center point of this pressure system is the latest center point of Trk[x], and record whether the contour line corresponding to the center is closed, the center longitude and latitude, the center minimum pressure, the cyclone system number corresponding to the center, the center number and the position of the corresponding pressure block point; If the center point of the cyclone system at the second time cannot find a matching trajectory Trk[x], then add a new trajectory until all the center points of the cyclone system at the second time are exhausted; Assume that there are n newly added trajectories, denoted as Trk[x+i], i=1,2,...,n, and record whether the contour line corresponding to the center is closed, the longitude and latitude of the center, the lowest pressure of the center, the cyclone system number corresponding to the center, the center number and the position of the corresponding pressure block point; (4) Obtain the full-time cyclone activity trajectory Trk[i] Repeat step 3 for the cyclone system center points from the third time to the last time, and finally obtain the trajectories of all cyclone activities in all data sets, Trk[i], i = 1, 2, ..., n; (5) Operational analysis α. Calculate the length of the trajectory Trk[i], i = 1, 2, ..., n. If the length is less than 2, remove the trajectory; β. Take any trajectory Trk[x] as the final trajectory Track[1] after the operation analysis. If it meets the conditions of the operation analysis, it is marked as having been analyzed. The parameters of the trajectory are recorded: the longitude and latitude of the center, the lowest pressure of the center, whether the contour line corresponding to the center is closed, the cyclone system number corresponding to the center, the center number and the position of the corresponding pressure block point; γ. Take Track[1] as the parent track and perform an operational analysis on track Trk[y], where x≠y. If the conditions for operational analysis are met, the analysis is marked as completed. The main track obtained from each analysis is recorded as Track[1], and the time points of track merging, separation, or reconnection are recorded. The secondary track after the analysis operation is recorded, along with the center longitude and latitude, the lowest center pressure, whether the contour line corresponding to the center is closed, the cyclone system number corresponding to the center, the center number, and the position of the corresponding pressure block point. δ. Still taking Track[1] as the parent track, perform operation analysis on all unlabeled tracks Trk and repeat step γ; ε. Repeat steps β, γ, and δ for the remaining unmarked tracks Trk to obtain Track. Suppose there are n tracks in total, which are recorded as Track[i], where i = 1, 2, ..., n.

2. A ground cyclone system tracking algorithm according to claim 1, characterized in that: The automatic identification algorithm uses the grid sea level pressure field of equal longitude and latitude in a limited area to automatically identify low-pressure systems.

3. The surface cyclone system tracking algorithm according to claim 1, characterized in that: The record of each grid point number in step S1 A consists of two parameters, namely the east-west grid point number and the north-south grid point number.

4. The surface cyclone system tracking algorithm according to claim 1, characterized in that: In step C of S1, for two adjacent grid point values (denoted as v1, v2), if v1 < Lslp[i] <= v2, then take the position of grid point v2 as the position of the contour line, and record the contour line grid point position G[i], where i = 1, 2,..., n, and G[i] is the set of all points adjacent to the position of G[i + 1].

5. The surface cyclone system tracking algorithm according to claim 1, characterized in that: In step D of S1, it is necessary to calculate the low-pressure blocks corresponding to the values of all contour lines within the region.

6. A surface cyclone system tracking algorithm according to claim 1, characterized in that: In step E of S1, the low-pressure block pairs are arranged in ascending order according to the values of the corresponding contour lines, denoted as Block[i], where i = 1, 2,..., n.

7. The surface cyclone system tracking algorithm according to claim 1, characterized in that: In step (5) of S2, fusion, separation, and reconnection operation analysis are performed on all trajectories. The conditions for fusion, separation, and reconnection are described in detail in Hanley and Caballero (2012), but the lengths of the past and future trajectory branches in this algorithm are required to be 1.

Images