CN110687618A - Automatic nowcasting method for short-time strong rainfall event of multi-monomer convection system - Google Patents
Automatic nowcasting method for short-time strong rainfall event of multi-monomer convection system Download PDFInfo
- Publication number
- CN110687618A CN110687618A CN201910913517.2A CN201910913517A CN110687618A CN 110687618 A CN110687618 A CN 110687618A CN 201910913517 A CN201910913517 A CN 201910913517A CN 110687618 A CN110687618 A CN 110687618A
- Authority
- CN
- China
- Prior art keywords
- monomer
- convection
- monomers
- short
- point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Environmental & Geological Engineering (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Atmospheric Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an automatic nowcasting method for a short-time strong precipitation event of a multi-monomer convection system, which comprises the following steps of: and identifying and tracking the convection single body. The method comprises the steps of performing convection monomer identification by using a multi-threshold self-adaptive algorithm, obtaining a convection monomer identification result which simultaneously retains core and peripheral related information of a convection monomer, and performing convection monomer tracking by using an optical flow algorithm to obtain the speed of the convection monomer; and identifying the multi-monomer convection system. According to the space-time correlation and the monomer speed between each pair of convection monomers in the multi-monomer convection system, the position prediction of the convection monomers at the next moment is given, the superposition coefficient between the convection monomers is calculated, a correlation matrix is established according to the superposition coefficient, and the identification result of the multi-monomer convection system is obtained by using a transfer closure clustering method; and constructing a multi-monomer convection system graph model and identifying a short-time strong precipitation event. The method realizes the close forecast of the short-time strong rainfall event of the automatic multi-monomer convection system, carries out early warning on disasters in time, and reduces economic loss and casualties.
Description
Technical Field
The invention relates to the field of meteorology, in particular to an automatic nowcasting method for a short-time strong rainfall event of a multi-monomer convection system.
Background
Short-term heavy precipitation (the term "short-term heavy precipitation" as used herein refers to the weather of China nationsPrecipitation events with hourly precipitation greater than 20mm as specified by the central business standards) are one of the most prominent strong convection disasters in china, and the short-term strong precipitation emphasizes the convection property and the short-term duration of precipitation more than the ordinary heavy precipitation[1]. Because the large precipitation is accumulated in a short time, torrential flood is often formed, and urban waterlogging, torrential flood, debris flow and other hazards are caused. The nowcasting of the short-time strong rainfall has important application value for disaster prevention and disaster control.
The existing approach forecast method for short-time heavy rainfall can be mainly divided into two types: method based on numerical weather forecast[2]And methods based on radar extrapolation techniques. The method based on the numerical weather forecast describes the change of atmospheric environment through a group of mathematical physics equations, and can solve the atmospheric physical state at any moment under the condition of giving initial conditions through various atmospheric observation technologies, thereby carrying out short-time strong precipitation proximity forecast. Method based on radar extrapolation technology uses Doppler weather radar to perform approach prediction, and mainly comprises centroid tracking method[3-4]And cross correlation method[5-6]The methods obtain the change trend of the convection system from the radar data at two adjacent moments, so as to carry out the nowcasting of the future short-time strong precipitation event.
In the process of implementing the invention, the inventor finds that at least the following disadvantages and shortcomings exist in the prior art:
in the methods described in the above documents, a method based on numerical weather prediction can simulate the atmospheric physical change process, but requires more computing resources, is slow in operation speed, is limited by resolution, and cannot describe a convection system with a small scale and a short life cycle; the method based on the radar extrapolation technology operates at a high speed, but cannot simulate complex meteorological processes due to lack of description of an atmospheric physical system. Therefore, the existing method has low recognition rate on the short-time strong precipitation event of the multi-monomer convection system, and cannot accurately predict disasters, thereby causing economic loss and casualties.
[ reference documents ]
[1] Poplar, Sunyun pine, Mausxu, et al. Beijing area short-time heavy precipitation process Multi-Scale circulation characteristics [ J ] Meteorological report, 2016,74(6): 919-.
[2]Sokol Z,V,Zacharov P,et al.Nowcasting of hailstormssimulated by the NWP model COSMO for the area of the Czech Republic[J].Atmospheric research,2016,171:66-76.
[3]Dixon M,Wiener G.TITAN:Thunderstorm identification,tracking,analysis,and nowcasting—A radar-based methodology[J].Journal of atmosphericand oceanic technology,1993,10(6):785-797.
[4]Rossi P J,Chandrasekar V,Hasu V,et al.Kalman filtering–basedprobabilistic nowcasting of object-oriented tracked convective storms[J].Journal of Atmospheric and Oceanic Technology,2015,32(3):461-477.
[5]Johnson J T,MacKeen P L,Witt A,et al.The storm cell identificationand tracking algorithm:An enhanced WSR-88D algorithm[J].Weather andforecasting,1998,13(2):263-276.
[6]Liu Y,Xi D G,Li Z L,et al.A new methodology for pixel-quantitativeprecipitation nowcasting using a pyramid Lucas Kanade optical flow approach[J].Journal of Hydrology,2015,529:354-364.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides an automatic close prediction method for a short-time strong rainfall event of a multi-monomer convection system, and solves the problems that in the prior art, the recognition rate of the short-time strong rainfall event of the multi-monomer convection system is low, disaster prediction cannot be accurately carried out, and economic loss and casualties are caused.
In order to solve the technical problem, the invention provides an automatic nowcasting method for a short-time strong precipitation event of a multi-monomer convection system, which comprises the following steps:
1) and identifying and tracking the convection single body. The method comprises the steps of performing convection monomer identification by using a multi-threshold self-adaptive algorithm, obtaining a convection monomer identification result which simultaneously retains core and peripheral related information of a convection monomer, and performing convection monomer tracking by using an optical flow algorithm to obtain the speed of the convection monomer; 2) and identifying the multi-monomer convection system. According to the space-time correlation and the monomer speed between each pair of convection monomers in the multi-monomer convection system, the position prediction of the convection monomers at the next moment is given, the superposition coefficient between the convection monomers is calculated, a correlation matrix is established according to the superposition coefficient, and the identification result of the multi-monomer convection system is obtained by using a transfer closure clustering method; 3) and constructing a multi-monomer convection system graph model and identifying a short-time strong precipitation event. And establishing a graph model for each multi-monomer convection system, obtaining graph characteristics, and performing the close prediction of the short-time strong precipitation event of the multi-monomer convection system by using a random forest model.
The method comprises the following steps: identifying and tracking a convection cell;
1-1) setting the threshold TlowAnd determining a convection cell preliminary candidate region. For each point in the radar image, if the reflectivity value of the point is greater than or equal to TlowSetting the point as a candidate point of the convection monomer; if the reflectivity value of the point is less than TlowSetting the point as a non-convective monomer point;
1-2) setting a plurality of reflectivity threshold values T from large to smallhigh,Thigh-5,Thigh-10...Tlow]The shell is used for determining the core of the convection current monomer and the maximum area containing the single core. For an area, if the reflectivity of each point is greater than the current threshold and no other area determined by the greater threshold is included, the area is called a convection cell core. If one area is larger than the current threshold and only contains one convection monomer core, and the next-level threshold and other areas are jointly contained in one area, the area is called as a convection monomer mesochite;
1-3) alternately using an expansion algorithm in image morphology for each mesochite region, and sequentially filling peripheral regions except the mesochite region in the monomer candidate region to obtain a final convective monomer identification result;
1-4) for each identified convection monomer, carrying out convection monomer tracking by using an optical flow algorithm, and calculating to obtain the speed of each convection monomer;
step two: identifying a multi-monomer convection system;
2-1) based on the convection single body identification result obtained in the step one, for each convection single body, assuming that the speed at the current moment is vtThen:
vt+1=α0vt+α1vt-1+α2vt-2,(0≤α≤1) (1)
using vt+1Giving the monomer position at the next moment;
2-2) the extrapolation results at a plurality of moments are superposed together, and the superposition coefficient eta between the monomers is calculated. Suppose there are two monomers at time tAndthe extrapolation results at their 9 time points are respectivelyAndthen calculate:
superposition coefficient etaABCan be used as a monomerAndthe correlation metric of (a) is applied in the subsequent multi-monomer convection system identification;
2-3) suppose that in a certain volume sweep result of the radar, there are n monomers O1,O2,...,On. Establishing an n x n Boolean relation matrix Rn×n=[rij]n×nR is a symmetric matrix, RijRepresents OiAnd OjCorrelation between, given a correlation metric threshold Tre,rijSatisfies the relationship:
2-4) performing transitive closure clustering on the Boolean matrix Rn×n=[rij]n×nPerforming p (p is an integer, p is more than or equal to 2) power-Boolean operation to obtainIf it isIs a full 1 matrix orAnd finishing clustering. Otherwise, repeating the step 2-3;
2-5) relationship matrixAll the elements of 1 in each row in the system are a multi-monomer convection system, and repeated results in the system are removed, so that a multi-monomer convection system identification result taking the superposition relationship as monomer similarity measurement can be obtained (each result may contain a plurality of convection monomers or only one convection monomer);
step three, constructing a multi-monomer convection system graph model and identifying a short-time strong precipitation event;
3-1) one multimonomer comprising a plurality of convective monomersA convection system for calculating a convection cell characteristic EOC for each convection cellL,EOCS,HOR30,LOR30,VILMAX,VILAVERAnd RPS. Wherein EOCLThe length of a long axis for fitting the outline ellipse of the convection monomer; EOCSThe length of a short shaft for fitting the outline ellipse of the convection monomer; HOR30The echo peak height is 30dBZ of the convection monomer; LOR30The echo bottom is 30dBZ for the convection monomer; VILMAXCalculating the maximum value of VIL for convection monomers point by point, wherein VIL refers to the vertically accumulated liquid water content and is an estimated value of the total precipitation of each point in the convection monomers; VILAVERCalculating the average value after VIL point by point for the convection monomer; RPS is the ratio of the projection of the monomer in the direction of the velocity vector to the monomer velocity;
3-2) for a multi-monomer convection system, a graph model was built, denoted G ═ U, E. Where U is the set of nodes and E is the set of edges. Creating a node u for each convection cell in the systemi,uiIs the characteristic of the corresponding monomer, will uiAnd adding the data into the set U. For any two points U in UiAnd ujAssuming that their velocity vectors are respectively viV and vj. Calculating uiTo ujDistance vector l ofijAnd is used to represent uiAnd ujParameter t of relative distance-to-velocity ratioij:
And sets a threshold value TtIf | tij|≤TtThen a connection u is creatediAnd ujEdge e ofijAnd useAs eij(ii) an attribute of (d);
3-3) assume that there is a single graph G ═ (U, E), G containing n nodes, each node having attributes that are a 7-dimensional vector. The adjacency matrix of the cell graph is denoted by a,a is an n × n symmetric array containing information of all sides of the monomer map if uiAnd ujThere is an edge e betweenijThen Aij=eijOtherwise Aij0. Use ofAn initial node attribute matrix representing a node simplex graph, each row representing a node. Defining a node fusion operation:
Xi+1=D(A+I)Xi,(i=0,1,2,...) (7)
where D is a row normalization coefficient matrix. And updating the node attribute into the attribute weighted sum of the node and the adjacent nodes every time the fusion operation is executed, wherein the weight is the spatial distribution relation between the nodes. Performing two times of fusion operation, and taking the two times of fusion result together with the original attribute as the characteristic X of the monomer mapfeature=[X0,X1,X2]。
3-4) to XfeatureAnd (6) sorting. XfeatureIs an n x 21 matrix, and specifies the characteristics VIL which can most directly express local water content information in the monomer characteristicsMAXI.e. XfeatureAs a sorting index with X in column 19featureArranged from large to small, thus ensuring that among the characteristics of each monomer map, VILMAXThe monomer subgraphs with the highest content are arranged foremost in total. At the same time in order to ensure XfeatureThe number of lines of (A) is fixed, and the ordered X is takenfeatureThe first two rows as final features of the simplex graphIf it isIf the amount is less than two lines, the amount is 0.
3-5) training a binary model of the random forest by using a large amount of historical meteorological data. For each multi-monomer convection system, one graph model G ═ (U, E) can be obtained. For the graph model, a graph feature representation can be obtainedWill be provided withAnd inputting a random forest model, and carrying out the nowcasting of the short-time strong rainfall event of the multi-monomer convection system.
Compared with the prior art, the invention has the beneficial effects that:
the technical scheme provided by the invention has the beneficial effects that: the method comprises the steps of modeling a multi-monomer convection system by using a graph model, converting physical characteristics and spatial distribution of the multi-monomer convection system into the graph model and graph characteristics, and carrying out short-time strong precipitation event close prediction on the graph characteristics obtained by the multi-monomer convection system by using a random forest model. The invention realizes high-quality automatic nowcasting of short-time strong rainfall of the multi-monomer convection system, and is beneficial to timely forecasting of weather disasters so as to reduce economic loss and casualties; and the effectiveness of the method is verified through experiments.
Drawings
FIG. 1 is a flow chart of an automatic nowcasting method for a short-term heavy precipitation event in a multi-cell convection system according to the present invention;
FIG. 2 is a schematic diagram of a convective cell identification algorithm;
fig. 3 is a schematic view of a multi-monomer convection system identification. Obtaining the speed of convection monomers by a convection monomer tracking method, predicting the positions of the convection monomers, calculating the superposition coefficient between the convection monomers, establishing a relation Boolean matrix, and obtaining the identification result of the multi-monomer convection system by transfer closure clustering;
FIG. 4 is a schematic diagram of a calculation method of physical characteristics RPS of a convection current monomer;
FIG. 5 is a schematic diagram of a multi-monomer convection system chart model construction method.
Detailed Description
The technical solutions of the present invention are further described in detail with reference to the accompanying drawings and specific embodiments, which are only illustrative of the present invention and are not intended to limit the present invention.
The invention provides an automatic nowcasting method for a short-time strong rainfall event of a multi-monomer convection system, which is designed according to the following design concept: and performing convection monomer identification by using a multi-threshold mathematical morphology-based method to obtain a convection monomer area. Tracking the single convection by using an optical flow algorithm to obtain the speed of the single convection; predicting the positions of the convection monomers by using the obtained convection monomer speed, calculating the correlation among the convection monomers by using the time-space superposition of the convection monomers in the future, and identifying a multi-monomer convection system by using transfer closure clustering; according to the space-time characteristics of the multi-monomer convection system, a graph model is constructed, monomers in the multi-monomer convection system are mapped to be nodes in the graph model, and the physical characteristics of the convection monomers are set to be attributes of the corresponding nodes. The edges in the graph model are used to describe the spatial distribution of the multi-monomer convection system. And generating a graph characteristic which is not influenced by the node sequence for each multi-monomer convection system corresponding to a graph model. 1079 cases of historical data were used to train a random forest model for the imminent prediction of short-term heavy precipitation events in a multi-modal convection system.
As shown in fig. 1, the method mainly includes convection single body identification and tracking, multi-single body convection system identification, multi-single body convection system graph model construction and short-time strong precipitation event identification; the specific contents are as follows:
the method comprises the following steps: identifying and tracking a convection cell;
1-1) setting the threshold TlowAnd determining a convection cell preliminary candidate region. For each point in the radar image, if the reflectivity value of the point is greater than or equal to TlowSetting the point as a candidate point of the convection monomer; if the reflectivity value of the point is less than TlowSetting the point as a non-convective monomer point; as shown in step 1 in FIG. 2, for an area in the Doppler weather radar reflectivity image, let T belowEach point in the traversal area is set to be a convective monomer candidate area for points having a reflectivity value greater than or equal to 35dBZ, and the other points are set to be non-convective monomer candidate areas. For theIn step 1 of fig. 2, a region with a minimum reflectance value of 30dBZ may be obtained, and 2 convection monomer candidate regions with a minimum reflectance value of 35dBZ included therein may be obtained;
1-2) setting a plurality of reflectivity threshold values T from large to smallhigh,Thigh-5,Thigh-10...Tlow]The shell is used for determining the core of the convection current monomer and the maximum area containing the single core. For an area, if the reflectivity of each point is greater than the current threshold and no other area determined by the greater threshold is included, the area is called a convection cell core. If an area is larger than the current threshold and only contains one convection monomer core, and the next level threshold and other areas are jointly contained in an area, the area is called a convection monomer mesochite. As shown in step 2 of fig. 2, let the threshold T be used for two convection current monomer candidate regionshighThe lower core in the first candidate region may be determined at 50dBZ, and the upper core in the first candidate region and the two cores in the second candidate region may not be determined at this threshold. When the threshold is ThighAt-5-45 dBZ, the kernel above the first candidate region and the two kernels in the second candidate region may be determined. Since at the next level of threshold, the two core regions in the second candidate region will be encompassed by the same region, the two cores in the second candidate region are also two mesochites at the same time. When the threshold is ThighAt-10-40 dBZ, the mesochites of the two convective cell cores in the first candidate region may be determined;
1-3) alternately using an expansion algorithm in image morphology for each mesochite region, and sequentially filling peripheral regions except the mesochite region in the monomer candidate region to obtain a final convection monomer identification result. As in step 3 of fig. 2, the two candidate regions are each divided into two convection monomer regions according to the mesochite thereof;
1-4) for each identified convection monomer, carrying out convection monomer tracking by using an optical flow algorithm, and calculating to obtain the speed of each convection monomer;
step two: identifying a multi-monomer convection system;
2-1) based on the convection single body identification result obtained in the step one, for each convection single body, assuming that the speed at the current moment is vtThen:
vt+1=α0vt+α1vt-1+α2vt-2,(0≤α≤1) (1)
using vt+1The monomer position at the next moment is given. As in the step of extrapolating the positions of the convection monomers in fig. 3, the current speed of the convection monomer is used to calculate the speed of the convection monomer at the next moment, so as to obtain the position prediction of the convection monomer at the next moment, which is represented by a dashed outline, and so on to obtain the position prediction results of the convection monomer at the subsequent 9 moments;
2-2) the extrapolation results at a plurality of moments are superposed together, and the superposition coefficient eta between the monomers is calculated. Suppose there are two monomers at time tAndthe extrapolation results at their 9 time points are respectivelyAndthen calculate:
superposition coefficient etaABCan be used as a monomerAndthe correlation metric of (a) is applied in the subsequent multi-monomer convection system identification. As shown in fig. 3, in the step of calculating the correlation between the convection monomers, the prediction result of the convection monomer position obtained in step (2-2) can obtain the overlapping description between the monomers at the next 9 times, which is shown as the colored filling area at the intersection of the dashed outlines in the figure, and this overlapping description can be used to calculate the overlap coefficient η;
2-3) suppose that in a certain volume sweep result of the radar, there are n monomers O1,O2,...,On. Establishing an n x n Boolean relation matrix Rn×n=[rij]n×nR is a symmetric matrix, RijRepresents OiAnd OjCorrelation between, given a correlation metric threshold Tre,rijSatisfies the relationship:
as shown in the correlation calculation step of convection monomers in fig. 3, 4 convection monomers form a 4 × 4 boolean relationship matrix, and the correlation coefficient η between the monomers is calculated and compared with the threshold TreAs can be seen by comparison, there is a correlation between the convection monomers 1 and 2, and a correlation between 2 and 3. So r in the Boolean relationship matrix12=r21=r23=r321, others are 0;
2-4) performing transitive closure clustering on the Boolean matrix Rn×n=[rij]n×nPerforming p (p is an integer, p is more than or equal to 2) power-Boolean operation to obtainIf it isIs a full 1 matrix orAnd finishing clustering. Otherwise, repeating the step (2-4) until p is equal to p + 1;
2-5) relationship matrixAll the elements of 1 in each row in the multi-monomer convection system are a multi-monomer convection system, and repeated results in the multi-monomer convection system are removed, so that a multi-monomer convection system identification result taking the superposition relationship as monomer similarity measurement can be obtained (each result may contain a plurality of convection monomers or only one convection monomer). As shown in the convection system identification step in fig. 3, after the boolean relationship matrix is subjected to multiple exponentiation operations of the transitive closure clustering, the 1 st, 2 nd, and 3 rd rows in the relationship matrix are the same, which indicates that the convection monomers 1, 2, and 3 are a convection system. The 4 th row in the relation matrix indicates that the convection monomer 4 is a convection system;
step three, constructing a multi-monomer convection system graph model and identifying a short-time strong precipitation event;
3-1) a multi-monomer convection system comprising a plurality of convection monomers, wherein for each convection monomer, a convection monomer characteristic EOC is calculatedL,EOCS,HOR30,LOR30,VILMAX,VILAVERAnd RPS. Wherein EOCLThe length of a long axis for fitting the outline ellipse of the convection monomer; EOCSThe length of a short shaft for fitting the outline ellipse of the convection monomer; HOR30The echo peak height is 30dBZ of the convection monomer; LOR30The echo bottom is 30dBZ for the convection monomer; VILMAXCalculating the maximum value of VIL for convection monomers point by point, wherein VIL refers to the vertically accumulated liquid water content and is an estimated value of the total precipitation of each point in the convection monomers; VILAVERCalculating the average value after VIL point by point for the convection monomer; the RPS is the ratio of the projection of the monomer in the direction of the velocity vector to the size of the monomer velocity. As shown in FIG. 4, for the calculation method of RPS, for a convective monomer, the projected length C in the velocity direction is calculatedpro,CproThe ratio of the velocity of the convection monomer and the velocity | v | is RPS;
3-2) for a multi-monomer convection system, a graph model was built, denoted G ═ U, E. Where U is the set of nodes and E is the set of edges. Creating a node u for each convection cell in the systemi,uiIs the characteristic of the corresponding monomer, will uiAnd adding the data into the set U. For any two points U in UiAnd ujAssuming that their velocity vectors are respectively viV and vj. Calculating uiTo ujDistance vector l ofijAnd is used to represent uiAnd ujParameter t of relative distance-to-velocity ratioij:
And sets a threshold value TtIf | tij|≤TtThen a connection u is creatediAnd ujEdge e ofijAnd useAs eijThe attribute of (2). As shown in fig. 5, for a multi-monomer convection system composed of 4 convection monomers, a graph model is created, each convection monomer corresponds to a node in the graph model, and the physical characteristics of each convection monomer are calculated as the attributes of the corresponding node. Calculate t between monomers, as shown, t12,t14,t13>TtSo edges are established between nodes 1 and 2, 1 and 4, 1 and 3, respectively, and will beAs an attribute of the edge connecting nodes i and j. t is t23,t34,t24<TtSo no edges are established between nodes 2 and 3, 3 and 4, and 2 and 4.
3-3) assume that there is a single graph G ═ (U, E), G containing n nodes, each node having attributes that are a 7-dimensional vector. A is used for representing the adjacent matrix of the single map, and A is an n multiplied by n symmetric matrix and contains the information of all sides of the single map if uiAnd ujThere is an edge e betweenijThen Aij=eijOtherwise Aij0. Use ofAn initial node attribute matrix representing a node simplex graph, each row representing a node. Defining a node fusion operation:
Xi+1=D(A+I)Xi,(i=0,1,2,...) (7)
where D is a row normalization coefficient matrix. And updating the node attribute into the attribute weighted sum of the node and the adjacent nodes every time the fusion operation is executed, wherein the weight is the spatial distribution relation between the nodes. Performing two times of fusion operation, and taking the two times of fusion result together with the original attribute as the characteristic X of the monomer mapfeature=[X0,X1,X2]。
3-4) to XfeatureAnd (6) sorting. XfeatureIs an n x 21 matrix, and specifies the characteristics VIL which can most directly express local water content information in the monomer characteristicsMAXI.e. XfeatureAs a sorting index with X in column 19featureArranged from large to small, thus ensuring that among the characteristics of each monomer map, VILMAXThe monomer subgraphs with the highest content are arranged foremost in total. At the same time in order to ensure XfeatureThe number of lines of (A) is fixed, and the ordered X is takenfeatureThe first two rows as final features of the simplex graphIf it isIf the amount is less than two lines, the amount is 0.
3-5) training a binary model of the random forest by using a large amount of historical meteorological data. For each multi-monomer convection system, one graph model G ═ (U, E) can be obtained. For the graph model, a graph feature representation can be obtainedWill be provided withAnd inputting a random forest model, and carrying out the nowcasting of the short-time strong rainfall event of the multi-monomer convection system.
The feasibility of the automatic nowcasting method for the short-time strong precipitation event of the multi-monomer convection system provided by the embodiment of the invention is verified by the following specific tests, which are described in detail as follows:
the data of the effectiveness of the test method is provided by the atmospheric detection center of the China weather service bureau. Specifically, the data are 5S-band Doppler weather radar data from Binzhou, Jinan, Qingdao, Shijiazhuang and Weifang, and weather automatic station data covered by the detection ranges of the radar. The time ranges were 6,7, 8, 9 months of 2015 and 2016. These data were collated to give 1349 weather samples, of which 590 short-time heavy precipitation events and 759 non-short-time heavy precipitation events. 1079 of these samples were used for the adjustment of various parameter parameters in the method and 270 were used for the final performance evaluation of the method. As shown in table 1, the final performance of the algorithm is described using three indexes commonly used in meteorology, namely, hit rate (POD), false positive rate (FAR), and Critical Success Index (CSI).
TABLE 1 evaluation results of examples of the present invention
While the present invention has been described with reference to the accompanying drawings, the present invention is not limited to the above-described embodiments, which are illustrative only and not restrictive, and various modifications which do not depart from the spirit of the present invention and which are intended to be covered by the claims of the present invention may be made by those skilled in the art.
Claims (4)
1. An automatic nowcasting method for a short-time strong precipitation event of a multi-monomer convection system is characterized by comprising the following steps:
identifying and tracking a convection cell; the method comprises the steps of performing convection monomer identification by using a multi-threshold self-adaptive algorithm, obtaining a convection monomer identification result which simultaneously retains core and peripheral related information of a convection monomer, and performing convection monomer tracking by using an optical flow algorithm to obtain the speed of the convection monomer;
identifying a multi-monomer convection system; according to the space-time correlation and the monomer speed between each pair of convection monomers in the multi-monomer convection system, the position prediction of the convection monomers at the next moment is given, the superposition coefficient between the convection monomers is calculated, a correlation matrix is established according to the superposition coefficient, and the identification result of the multi-monomer convection system is obtained by using a transfer closure clustering method;
constructing a multi-monomer convection system graph model and identifying a short-time strong precipitation event; and establishing a graph model for each multi-monomer convection system, obtaining graph characteristics, and performing the close prediction of the short-time strong precipitation event of the multi-monomer convection system by using a random forest model.
2. The method of automatic nowcasting for short-term heavy precipitation events in multi-cell convection systems as claimed in claim 1, wherein said convection cell identification and tracking specifically comprises the steps of:
setting a threshold TlowDetermining a convection cell preliminary candidate region; for each point in the radar image, if the reflectivity value of the point is greater than or equal to TlowSetting the point as a candidate point of the convection monomer; if the reflectivity value of the point is less than TlowSetting the point as a non-convective monomer point;
setting a plurality of reflectivity threshold values T from large to smallhigh,Thigh-5,Thigh-10...Tlow]A core for defining a convection cell and a maximum area mesochite containing a single core; for an area, if the reflectivity of each point in the area is greater than the current threshold and the area determined by other larger thresholds is not included in the area, the area is called as a convection monomer core; if an area is larger than the current threshold and only contains one convection single core, and the next level of threshold andthe other regions are contained in one region together, and the region is called a convection monomer mesochite;
for each mesochite area, alternately using an expansion algorithm in image morphology, and sequentially filling peripheral areas except the mesochite area in the monomer candidate area to obtain a final convection monomer identification result;
and for each identified convection monomer, carrying out convection monomer tracking by using an optical flow algorithm, and calculating to obtain the speed of each convection monomer.
3. The method of automatic nowcasting for transient heavy precipitation events in a multi-monomer convection system as claimed in claim 1, wherein said identification of the multi-monomer convection system specifically comprises the steps of:
based on the obtained identification result of the convection single body, for each convection single body, the speed at the current moment is assumed to be vtThen:
vt+1=α0vt+α1vt-1+α2vt-2,(0≤α≤1) (1)
using vt+1Giving the monomer position at the next moment;
stacking the extrapolation results at a plurality of moments together, and calculating the stacking coefficient eta between the monomers; suppose there are two monomers at time tAndthe extrapolation results at their 9 time points are respectivelyAndthen calculate:
superposition coefficient etaABCan be used as a monomerAndthe correlation metric of (a) is applied in the subsequent multi-monomer convection system identification;
suppose that in a certain volume sweep result of the radar, n single bodies O exist1,O2,...,On(ii) a Establishing an n x n Boolean relation matrix Rn×n=[rij]n×nR is a symmetric matrix, RijRepresents OiAnd OjCorrelation between, given a correlation metric threshold Tre,rijSatisfies the relationship:
performing transitive closure clustering on a Boolean matrix Rn×n=[rij]n×nPerforming p (p is an integer, p is more than or equal to 2) power-Boolean operation to obtainIf it isIs a full 1 matrix orFinishing clustering; otherwise, repeating the step 2-3; relationship matrixAll the elements of 1 in each row are a multi-monomer convection system, and repeated results in the multi-monomer convection system are removed, so that a multi-monomer convection system identification result taking the superposition relationship as monomer similarity measurement can be obtained.
4. The method of claim 1, wherein the building of the multi-monomer convection system graph model and the identifying of the short-time heavy precipitation event comprise the following steps:
a multi-monomer convection system comprising a plurality of convection monomers, wherein for each convection monomer, a convection monomer characteristic EOC is calculatedL,EOCS,HOR30,LOR30,VILMAX,VILAVERRPS; wherein EOCLThe length of a long axis for fitting the outline ellipse of the convection monomer; EOCSThe length of a short shaft for fitting the outline ellipse of the convection monomer; HOR30The echo peak height is 30dBZ of the convection monomer; LOR30The echo bottom is 30dBZ for the convection monomer; VILMAXCalculating the maximum value of VIL for convection monomers point by point, wherein VIL refers to the vertically accumulated liquid water content and is an estimated value of the total precipitation of each point in the convection monomers; VILAVERCalculating the average value after VIL point by point for the convection monomer; RPS is the ratio of the projection of the monomer in the direction of the velocity vector to the monomer velocity;
for a multi-monomer convection system, a graph model is established, and is expressed as G ═ U, E; where U is the set of nodes and E is the set of edges; creating a node u for each convection cell in the systemi,uiIs the characteristic of the corresponding monomer, will uiAdding the obtained solution into a set U; for any two points U in UiAnd ujAssuming that their velocity vectors are respectively viV and vj(ii) a Calculating uiTo ujDistance vector l ofijAnd is used forRepresents uiAnd ujParameter t of relative distance-to-velocity ratioij:
And sets a threshold value TtIf | tij|≤TtThen a connection u is creatediAnd ujEdge e ofijAnd useAs eij(ii) an attribute of (d); assuming that a single graph G ═ (U, E) exists, G contains n nodes, and the attribute of each node is a 7-dimensional vector; a is used for representing the adjacent matrix of the single map, and A is an n multiplied by n symmetric matrix and contains the information of all sides of the single map if uiAnd ujThere is an edge e betweenijThen Aij=eijOtherwise Aij0; use ofAn initial node attribute matrix representing a node simplex graph, each row representing a node. Defining a node fusion operation:
Xi+1=D(A+I)Xi,(i=0,1,2,...) (7)
wherein D is a row normalization coefficient matrix; each time the fusion operation is executed, the node attribute is updated to be the attribute weighted sum of the node and the adjacent nodes, and the weight is the spatial distribution relation between the nodes; performing two times of fusion operation, and taking the two times of fusion result together with the original attribute as the characteristic X of the monomer mapfeature=[X0,X1,X2];
To XfeatureAnd (6) sorting. XfeatureIs an n x 21 matrix, and specifies the characteristics VIL which can most directly express local water content information in the monomer characteristicsMAXI.e. XfeatureAs a sorting index with X in column 19featureArranged from large to small, so as to ensure the characteristics of each single graphMiddle, VILMAXThe monomer subgraphs with the largest content are arranged at the forefront in total; at the same time in order to ensure XfeatureThe number of lines of (A) is fixed, and the ordered X is takenfeatureThe first two rows as final features of the simplex graphIf it isIf the number is less than two, the number is 0;
training a binary classification model of a random forest by using a large amount of historical meteorological data; for each multi-monomer convection system, one graph model G ═ (U, E) may be obtained; for the graph model, a graph feature representation can be obtainedWill be provided withAnd inputting a random forest model, and carrying out the nowcasting of the short-time strong rainfall event of the multi-monomer convection system.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910913517.2A CN110687618B (en) | 2019-09-25 | 2019-09-25 | Automatic nowcasting method for short-time strong rainfall event of multi-monomer convection system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910913517.2A CN110687618B (en) | 2019-09-25 | 2019-09-25 | Automatic nowcasting method for short-time strong rainfall event of multi-monomer convection system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110687618A true CN110687618A (en) | 2020-01-14 |
CN110687618B CN110687618B (en) | 2021-10-01 |
Family
ID=69110093
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910913517.2A Active CN110687618B (en) | 2019-09-25 | 2019-09-25 | Automatic nowcasting method for short-time strong rainfall event of multi-monomer convection system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110687618B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111427101A (en) * | 2020-04-07 | 2020-07-17 | 江苏省气象科学研究所 | Thunderstorm strong wind grading early warning method, system, storage medium and computer program |
CN112926664A (en) * | 2021-03-01 | 2021-06-08 | 南京信息工程大学 | Feature selection and CART forest short-time strong rainfall forecasting method based on evolutionary algorithm |
CN113447931A (en) * | 2021-06-10 | 2021-09-28 | 天津大学 | Short-time strong precipitation identification method based on Doppler radar data |
Citations (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070005249A1 (en) * | 2005-06-30 | 2007-01-04 | Massachusetts Institute Of Technology | Weather radar echo tops forecast generation |
CN102721987A (en) * | 2012-06-12 | 2012-10-10 | 中国海洋大学 | Method for prewarning Doppler radar remote sensing strong storm |
CN103529492A (en) * | 2013-09-22 | 2014-01-22 | 天津大学 | Storm body position and form prediction method based on Doppler radar reflectivity image |
US20140300510A1 (en) * | 2008-08-26 | 2014-10-09 | Viasat, Inc. | Weather detection using satellite communication signals |
CN104237890A (en) * | 2014-09-03 | 2014-12-24 | 天津大学 | Recognition and forecast method for rainstorm caused by train effect |
CN104463196A (en) * | 2014-11-11 | 2015-03-25 | 中国人民解放军理工大学 | Video-based weather phenomenon recognition method |
CN104977584A (en) * | 2015-06-29 | 2015-10-14 | 深圳市气象台 | Convective weather approach prediction method and system |
CN105158762A (en) * | 2014-03-31 | 2015-12-16 | 霍尼韦尔国际公司 | Identifying and tracking convective weather cells |
US9507022B1 (en) * | 2008-03-07 | 2016-11-29 | Rockwell Collins, Inc. | Weather radar system and method for estimating vertically integrated liquid content |
CN107229084A (en) * | 2017-06-08 | 2017-10-03 | 天津大学 | A kind of automatic identification, tracks and predicts contracurrent system mesh calibration method |
CN107301272A (en) * | 2017-05-24 | 2017-10-27 | 天津大学 | Strong convection monomer structure and architectural feature visualization space-time method for decomposing |
US20170336533A1 (en) * | 2016-05-20 | 2017-11-23 | The Climate Corporation | Radar based precipitation estimates using spatiotemporal interpolation |
CN108562903A (en) * | 2017-12-25 | 2018-09-21 | 天津大学 | Strong convection system dynamic field structure recognition methods based on Doppler radar |
CN108665486A (en) * | 2018-03-31 | 2018-10-16 | 天津大学 | A kind of labeling method of strong convection monomer sample |
CN109300174A (en) * | 2018-11-27 | 2019-02-01 | 杨波 | A kind of Severe Convective Weather Forecasting analysis system |
US10345483B2 (en) * | 2015-06-10 | 2019-07-09 | Escaype | Observer-based meteorology and image identification |
-
2019
- 2019-09-25 CN CN201910913517.2A patent/CN110687618B/en active Active
Patent Citations (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070005249A1 (en) * | 2005-06-30 | 2007-01-04 | Massachusetts Institute Of Technology | Weather radar echo tops forecast generation |
US9507022B1 (en) * | 2008-03-07 | 2016-11-29 | Rockwell Collins, Inc. | Weather radar system and method for estimating vertically integrated liquid content |
US20140300510A1 (en) * | 2008-08-26 | 2014-10-09 | Viasat, Inc. | Weather detection using satellite communication signals |
CN102721987A (en) * | 2012-06-12 | 2012-10-10 | 中国海洋大学 | Method for prewarning Doppler radar remote sensing strong storm |
CN103529492A (en) * | 2013-09-22 | 2014-01-22 | 天津大学 | Storm body position and form prediction method based on Doppler radar reflectivity image |
CN105158762A (en) * | 2014-03-31 | 2015-12-16 | 霍尼韦尔国际公司 | Identifying and tracking convective weather cells |
CN104237890A (en) * | 2014-09-03 | 2014-12-24 | 天津大学 | Recognition and forecast method for rainstorm caused by train effect |
CN104463196A (en) * | 2014-11-11 | 2015-03-25 | 中国人民解放军理工大学 | Video-based weather phenomenon recognition method |
US10345483B2 (en) * | 2015-06-10 | 2019-07-09 | Escaype | Observer-based meteorology and image identification |
CN104977584A (en) * | 2015-06-29 | 2015-10-14 | 深圳市气象台 | Convective weather approach prediction method and system |
US20170336533A1 (en) * | 2016-05-20 | 2017-11-23 | The Climate Corporation | Radar based precipitation estimates using spatiotemporal interpolation |
CN107301272A (en) * | 2017-05-24 | 2017-10-27 | 天津大学 | Strong convection monomer structure and architectural feature visualization space-time method for decomposing |
CN107229084A (en) * | 2017-06-08 | 2017-10-03 | 天津大学 | A kind of automatic identification, tracks and predicts contracurrent system mesh calibration method |
CN108562903A (en) * | 2017-12-25 | 2018-09-21 | 天津大学 | Strong convection system dynamic field structure recognition methods based on Doppler radar |
CN108665486A (en) * | 2018-03-31 | 2018-10-16 | 天津大学 | A kind of labeling method of strong convection monomer sample |
CN109300174A (en) * | 2018-11-27 | 2019-02-01 | 杨波 | A kind of Severe Convective Weather Forecasting analysis system |
Non-Patent Citations (9)
Title |
---|
J HOU, WANG P: "Storm Tracking via Tree Structure Representation of Radar Data", 《JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY》 * |
MORAL A D, RIGO T, LLASAT M C.: "A radar-based centroid tracking algorithm for severe weather surveillance: Identifying split/merge processes in convective systems", 《ATMOSPHERIC RESEARCH》 * |
ROSSI P J,CHANDRASEKAR V,HASU V,ET AL.: "Kalman filtering-based probabilistic nowcasting of object-oriented tracked convective storms", 《JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY》 * |
STALKER, JAMES, R, ET AL.: "A Method to Identify Convective Cells within Multicell Thunderstorms from Multiple Doppler Radar Data", 《MONTHLY WEATHER REVIEW》 * |
侯谨毅: "基于模式识别理论方法的强对流天气分类临近预报研究", 《中国博士学位论文全文数据库信息科技辑》 * |
张春桂,周乐照,林炳青: "基于卫星和雷达资料的对流云团识别跟踪", 《气象科技》 * |
王萍,潘跃: "基于显著性特征的大冰雹识别模型", 《物理学报》 * |
钟海燕, 李玲等: "基于随机森林的短时临近降雨预报方法", 《广西师范学院学报(自然科学版)》 * |
高毅: "小尺度对流单体的分类识别方法研究", 《中国优秀硕士学位论文全文数据库基础科学辑》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111427101A (en) * | 2020-04-07 | 2020-07-17 | 江苏省气象科学研究所 | Thunderstorm strong wind grading early warning method, system, storage medium and computer program |
CN111427101B (en) * | 2020-04-07 | 2022-04-26 | 南京气象科技创新研究院 | Thunderstorm strong wind grading early warning method, system and storage medium |
CN112926664A (en) * | 2021-03-01 | 2021-06-08 | 南京信息工程大学 | Feature selection and CART forest short-time strong rainfall forecasting method based on evolutionary algorithm |
CN112926664B (en) * | 2021-03-01 | 2023-11-24 | 南京信息工程大学 | Feature selection and CART forest short-time strong precipitation prediction method based on evolutionary algorithm |
CN113447931A (en) * | 2021-06-10 | 2021-09-28 | 天津大学 | Short-time strong precipitation identification method based on Doppler radar data |
Also Published As
Publication number | Publication date |
---|---|
CN110687618B (en) | 2021-10-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110059554B (en) | Multi-branch target detection method based on traffic scene | |
WO2021139069A1 (en) | General target detection method for adaptive attention guidance mechanism | |
CN111091105B (en) | Remote sensing image target detection method based on new frame regression loss function | |
CN110687618B (en) | Automatic nowcasting method for short-time strong rainfall event of multi-monomer convection system | |
CN109932730B (en) | Laser radar target detection method based on multi-scale monopole three-dimensional detection network | |
CN113496104B (en) | Precipitation prediction correction method and system based on deep learning | |
CN110223302A (en) | A kind of naval vessel multi-target detection method extracted based on rotary area | |
CN104931934B (en) | A kind of radar plot condensing method based on PAM cluster analyses | |
CN110197215A (en) | A kind of ground perception point cloud semantic segmentation method of autonomous driving | |
CN107016677A (en) | A kind of cloud atlas dividing method based on FCN and CNN | |
CN109765559A (en) | A kind of mountain area disastrous rainstorm recognition methods based on Doppler radar and deep learning | |
CN112836713A (en) | Image anchor-frame-free detection-based mesoscale convection system identification and tracking method | |
CN106651865B (en) | A kind of optimum segmentation scale automatic selecting method of new high-resolution remote sensing image | |
CN110378297A (en) | A kind of Remote Sensing Target detection method based on deep learning | |
CN111709285A (en) | Epidemic situation protection monitoring method and device based on unmanned aerial vehicle and storage medium | |
CN104237890B (en) | The heavy rain identification that one is caused by " train effect " and forecasting procedure | |
CN109242019A (en) | A kind of water surface optics Small object quickly detects and tracking | |
CN110942111A (en) | Method and device for identifying strong convection cloud cluster | |
CN105046259A (en) | Coronal mass ejection (CME) detection method based on multi-feature fusion | |
CN108254750A (en) | A kind of downburst intelligent recognition method for early warning based on Radar Data | |
CN113627440A (en) | Large-scale point cloud semantic segmentation method based on lightweight neural network | |
Zhao et al. | Vehicle detection based on improved yolov3 algorithm | |
CN115728841A (en) | Strong convection weather early warning method based on inspection and evaluation | |
CN103150566A (en) | Automatic detecting method of remote sensing ground object target based on random geometric model | |
Zhang et al. | Learnable optical flow network for radar echo extrapolation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |