CN115902812B - Automatic discriminating method, system, equipment and terminal for short-time heavy rain weather background - Google Patents

Abstract

The invention belongs to the technical field of short-time storm recognition, and discloses a method, a system, equipment and a terminal for automatically distinguishing a short-time storm weather background, wherein the historical process example data of short-time storm is utilized to construct MQVP corresponding to each double-polarization radar time and time, so as to construct an MQVP data set of the short-time storm weather background; according to the MQVP data set of the short-time stormwater weather background, an MQVP probability model for representing the short-time stormwater weather background is constructed; and constructing an MQVP early warning data set, calculating an MQVP comprehensive similarity probability index CP (TH), and judging the short-time stormwater weather background according to an MQVP probability model. According to the judging method provided by the invention, the MQVP probability modeling is performed by fully utilizing the polarization quantity characteristics of the dual-polarization radar, the automatic identification of the weather background of the short-time strong precipitation of the rainfall division level is realized according to the MQVP probability modeling, and the advantages of the polarization radar in the fields of monitoring and early warning of the short-time strong precipitation can be better exerted.

Description

Automatic discriminating method, system, equipment and terminal for short-time heavy rain weather background

Technical Field

The invention belongs to the technical field of short-time storm recognition, and particularly relates to a method, a system, equipment and a terminal for automatically distinguishing a short-time storm weather background.

Background

At present, the precipitation classification (divided into convection and layer cloud) method based on the vertical structure of reflectivity and precipitation rate can be used for identifying warm cloud precipitation, improving the quantitative radar precipitation estimation precision to a certain extent, filling a radar beam shielding area and further improving the radar data quality. However, the conventional method still has a certain limitation because of the lack of identification capability of short-time heavy rain micro physical characteristics. With the maturation of the dual-polarization radar technology, the method provides possibility for providing extremely short-time heavy rain refinement cloud physical parameters. The quasi-vertical profile (quasi-vertical profiles, QVP) method of the dual-polarization radar can fully exert the characteristic of high space-time resolution of the dual-polarization radar, and the micro-physical characteristics of short-time storm are monitored from a multi-dimensional angle, so that the method provides possibility for early identification of the characteristics of short-time storm.

Because shallow precipitation and clouds and convective precipitation have significant differences in vertical structure and microphysical characteristics, ground precipitation, particularly extreme precipitation microphysical characteristics, have significant space-time variations and diversity. The micro-physical characteristics of the extreme convection system in the plum rain period are studied to find that the existence of multiple types of precipitation can be observed in plain areas in a certain period of time. And the combination of raindrop spectrum data discovers that the intensity of precipitation in the plum rain period and the depth of convection development are not necessarily related, and extreme precipitation is mainly caused by medium-height convection. Dual polarization radars are widely used In the study of precipitation microphysics, the horizontal polarization reflectance factor (Z _H ) Differential reflectance factor (Z _DR ) Phase shift of difference propagation (phi) _DP ) Differential propagation phase shift (K) _DP ) Parameters, and the like, which are closely related to factors such as precipitation particle type, shape, spatial orientation and distribution, and falling motion. Z is Z _DR Is an important index for judging the phase states of different types of cloud precipitation particles; the breaking and collision of raindrops tend to balance during strong precipitation, and the increase of the rain intensity depends on the increase of the concentration of the raindrops, K _DP Can be used as an index for judging whether the rain intensity is increased. Z is Z _DR Column and K _DP The evolution of the column is predictive of the change in ground rain intensity, particularly during sustained precipitation, Z _DR (K _DP ) The re-development of the column predicts a re-enhancement of the precipitation system. In recent years, a weather student tries to study the weather field by using a deep learning method. The research shows that the radar quantitative precipitation estimation deep learning network architecture is constructed by using different combinations of the two polarization parameters of 3 polarization as input factors, and the multi-parameter network architecture has a good estimation effect on quantitative precipitation estimation through inspection. However, this method does not construct a multiparameter vertical structural analysis and has certain limitations.

At present, the application of the vertical profile of the polarization amount in China mainly focuses on the analysis and comparison of micro-physical characteristics in the period of the strong convection occurrence process by using a QVP method, and an automatic service method for identifying QVP short-time heavy rain weather background characteristics is lacking. The traditional QVP construction method is based on an average strategy on vertical different equal-altitude surfaces (namely, calculating average polarization amount information on the equal-altitude surfaces by using the equal-elevation surfaces), so that the average state of the micro-physical characteristics of precipitation particles is obtained. However, short-term storms often have local characteristics, and thus, the conventional QVP construction method has certain disadvantages for directly identifying the characteristics of short-term storms. As the layer heights of the layers of 0 ℃ and minus 10 ℃ and minus 20 ℃ in different seasons are different, the QVP characteristics in different seasons have larger difference, QVP analysis under the strong convection weather background is carried out according to the existing method, the seasons are needed to be separated, the classification research is carried out by combining with sounding information such as temperature layer junction, and the research process analysis process is more complex. Although quantitative precipitation estimation methods based on low-layer polarization (such as KDP) can relatively accurately estimate the time-by-time rainfall, the method cannot predict and early warn extremely-short-time rainfall in advance.

Aiming at weather background characteristics of extremely short-time heavy rain, how to fully utilize polarization amount information and design a reasonable QVP statistical construction method. In addition, the conventional QVP model characterizes more water vapor, but lacks information about characterizing water vapor dynamics (such as water vapor flux divergence), and thus is one of the difficulties to be solved by the present invention. A QVP model matching method suitable for the background characteristics of short-time heavy rain weather is established, so that rapid and accurate automatic discrimination of extremely short-time heavy rain is realized. In addition, according to the service requirement, the magnitude of local maximum storm in the area is pre-judged in advance, which is also a difficult problem to be solved by the invention. Aiming at the two problems, how to combine the meteorological background knowledge and the computer image recognition technology to design an efficient and accurate algorithm model is another important problem to be solved by the invention.

Through the above analysis, the problems and defects existing in the prior art are as follows:

(1) Conventional methods of precipitation classification (divided into convection and clouding) based on vertical structure of reflectivity and precipitation rate remain limited due to lack of identification of short-term stormwater microphysical characteristics.

(2) The method for constructing the radar quantitative precipitation estimation deep learning network architecture by using different combinations of the two polarization parameters as input factors does not construct multi-parameter vertical structure analysis and has limitations.

(3) At present, the application of the vertical profile of the polarization amount in China lacks an automatic business method for identifying the background characteristics of QVP short-time heavy rain weather.

(4) The traditional QVP construction method has the defect of directly identifying the short-time storm features, lacks the consideration of the characteristic water vapor dynamic information, and cannot predict and early warn the extreme short-time storm in advance.

Disclosure of Invention

Aiming at the problems existing in the prior art, the invention provides a method, a system, equipment and a terminal for automatically distinguishing the short-time heavy rain weather background, in particular to a method, a system, a medium, equipment and a terminal for automatically distinguishing the short-time heavy rain weather background based on polarization profile feature recognition.

The invention is realized in such a way that the automatic distinguishing method of the short-time heavy rain weather background comprises the following steps: utilizing historical process example data of short-time storm to construct MQVP (standardized large-value strategy QVP) corresponding to each double-polarization radar time and time, and further constructing an MQVP data set of short-time storm weather background; according to the MQVP data set of the short-time stormwater weather background, an MQVP probability model for representing the short-time stormwater weather background is constructed; and constructing an MQVP early warning data set, calculating an MQVP comprehensive similarity probability index CP (TH), and judging the short-time stormwater weather background according to an MQVP probability model.

Further, the automatic distinguishing method for the short-time heavy rain weather background comprises the following steps:

firstly, processing MQVP of each dual-polarization radar data, and constructing an MQVP probability model for representing short-time stormwater weather background;

and secondly, calculating an MQVP comprehensive similarity probability index CP (TH), and judging the short-time stormwater weather background according to the MQVP probability model.

In the first step, the MQVP corresponding to each double-polarization radar time and time is built by utilizing historical process example data of short-time storm, an MQVP data set of short-time storm weather background is built, and an MQVP probability model is built according to the data set.

Further, the MQVP process of each dual-polarized radar material in step one includes:

(1) Pretreatment of

Performing bilinear interpolation on the single radar data to obtain equal-altitude lattice point polarization data; generating particle phase identification data by adopting a phase inversion method based on fuzzy logic; generating vertical wind profile data representing atmospheric environment wind field information by adopting a speed and direction display method; wherein, the constant-altitude data are uniformly processed from 1km to 17km with a vertical interval of 500 m.

(2) Convection kernel region lattice marking

Judging convection kernel areas with different heights according to the equal-altitude surface reflectivity and the particle phase state identification result, and marking lattice points. The original particle phase states identified in the product include ground objects, organisms, dry snow, wet snow, ice crystals, aragonite, heavy droplets, rain and heavy rain; the criteria for determining the lattice points of the kernel area are: for layer heights below 0deg.C, if Z _H >42dBZ, and meanwhile, the particle phase state is identified as hail or rain or heavy rain, and is marked as a grid point of a core area; for layer heights above 0deg.C, if Z _H >42dBZ, while satisfying the identification of the particle phase as hail or aragonite, is marked as the kernel region lattice point.

(3) Statistics of maximum quasi-vertical profiles for various variables of MQVP within a convection kernel region

Carrying out numerical feature statistics on polarized quantity of the flow core points marked on different contour planes to generate corresponding MQVP _ZH 、MQVP _PS 、MQVP _KDP 、MQVP _ZDR 、MQVP _RHV MQVP _KV The method comprises the steps of carrying out a first treatment on the surface of the Wherein the quasi-vertical profile sampling strategy is to employ a maximum value.

(4) The MQVP is standardized according to the reference temperature layer junction

And taking the 0 ℃ layer and the-20 ℃ height in the temperature layer junction curve as references, and carrying out standardization processing on the original quasi-vertical profile of each variable through a nearby interpolation strategy. The normalized quasi-vertical profile is 3 parts: a) Ground to 0 ℃ layer height; b) Layer height of 0 ℃ to layer height of-20 ℃; c) -20 ℃ layer height to 17km height. Each vertical profile data segment is divided equally by n, which is preferably set to 10.

Further, the quasi-vertical profile sampling strategy in step (3) includes:

1)MQVP _ZH is a sampling strategy of (a)

Z for marked convection nuclei on different contours _H Numerical feature statistics, statistics of maximum Z on different contour heights _H Numerical values. To count to 0 degree C layerConvective microphysical structural features near altitude, effective statistical horizon of MQVP [ D _min ，D _max ]The height of the layer is determined according to the actual specific temperature:

D _min ＝(arccos((R _m +H _R )*cos(θ _19.5 )/(H ₀ +R _m ))-θ _19.5 )*R _m ；

D _max ＝(arccos((R _m +H _R )*cos(θ _0.5 )/(H ₀ +R _m ))-θ _0.5 )*R _m ；

wherein D is _min Representing a minimum radius centered on the radar in a statistical level range, in km; d (D) _max Representing the maximum radius centered on the radar in the statistical level range, in km; arccosis represents an inverse cosine function; cos represents a cosine function; r is R _m Representing equivalent earth radius, unit km; h _R Representing the height of the radar station, and the unit km; h ₀ The height of the sounding observation layer at 0 ℃ corresponding to the statistical data time period is expressed, and the unit is km; θ _19.5 The radian value corresponding to the angle 19.5 degrees is represented; θ _0.5 The radian value corresponding to the angle 0.5 deg. is indicated.

2)MQVP _PS Is a sampling strategy of (a)

Utilizing MQVP _ZH Statistical level range algorithm of (1) to determine MQVP _PS And (5) horizontal statistics range information. And carrying out remarkable particle phase state characteristic statistics on the convection core points marked on different contour surfaces, and counting the most remarkable particle phase state values on different contour surface heights. Wherein the order of decreasing significance of the different particle phases in the flow kernel is: hail, aragonite, heavy rain, heavy drops, ice crystals, wet snow, dry snow.

3)MQVP _KDP Is a sampling strategy of (a)

Utilizing MQVP _ZH Statistical level range algorithm of (1) to determine MQVP _KDP And (5) horizontal statistics range information. K for marked convection kernel points on different contour surfaces _DP Numerical feature statistics, statistics of maximum K on different contour heights _DP Numerical values. Wherein ρ is increased _hv Numerical monitoring to realize polarization quantityIs a simple quality control constraint; if ρ _hv The values are in the range of [0.6,1.0]The current polarization is considered to be involved in MQVP statistics.

4)MQVP _ZDR Is a sampling strategy of (a)

Utilizing MQVP _ZH Statistical level range algorithm of (1) to determine MQVP _ZDR And (5) horizontal statistics range information. Z for marked convection nuclei on different contours _DR Numerical feature statistics, statistics of maximum Z on different contour heights _DR Numerical values. Wherein ρ is increased _hv Numerical monitoring is carried out to realize simple substance control constraint of the polarization quantity; if ρ _hv The values are in the range of [0.6,1.0]The current polarization is considered to be involved in MQVP statistics.

5)MQVP _RHV Is a sampling strategy of (a)

Utilizing MQVP _ZH Statistical level range algorithm of (1) to determine MQVP _RHV And (5) horizontal statistics range information. ρ for flow kernel points marked on different contours _hv Numerical feature statistics, statistics of maximum rho on different contour heights _hv A numerical value; wherein ρ is _hv Is in the reasonable numerical range of [0.6,1.0 ] ]。

6)MQVP _KV Is a sampling strategy of (a)

Will K _DP Multiplying the ambient wind speed value VADL to obtain a polarized water condensate flux KV that approximates the characteristics of the water vapor flux distribution. According to K _DP The water content characteristics of the condensate in the atmosphere are detected, so that the KV large value indicates the existence of the region combination characteristic.

KV(h)＝MQVP _KDP (h)*VADL(h),h∈[0,17km]；

Where h is the height of the quasi-vertical profile in km.

By MQVP corresponding to height h _KDP Multiplying the ambient wind speed VADL obtained by calculation of the speed and direction display method, and further calculating to obtain polarized condensate flux KV corresponding to the height h.

Further, the MQVP probability model in the first step is MQVP probability model PMD (TH) of the rainfall grade, and statistical construction is carried out according to a threshold TH of the rainfall grade of the storm; TH was selected to be 20mm/h, 30mm/h, 40mm/h and 50mm/h in this order. And constructing a corresponding data set by taking the automatic rainfall station monitoring rainfall on a time-by-time basis, and if the maximum hour rainfall in the radar monitoring range reaches the TH standard, classifying the MQVP of all radar data within 1 hour before the moment of the storm to be an MQVP probability model statistical data set under a threshold TH.

Further, the constructing of the MQVP probability model in the step one includes:

(1)PMD _PS construction of (TH)

Probability statistics of particle phases are used to analyze constraints. The probability model consists of the following three types of statistical properties:

1) Variable D _G The method is used for representing the phase probability distribution of the aragonite particles, and the calculation formula is as follows:

D _G ＝N _G /N；

wherein N is _G The number of aragonite phases for each sublayer in each MQVP in the statistical dataset is represented, and N represents the number of MQVP samples in the statistical dataset.

2) Variable D _W The method is used for representing the phase probability distribution of the pure liquid water particles, and the calculation formula is as follows:

D _W ＝N _W /N；

wherein N is _W The number of the phase states of each sublayer of pure liquid water particles, the number of the phase states of heavy rain particles, the number of the phase states of rain particles and the number of the phase states of heavy drops in each MQVP in the statistical data set are represented; n represents the MQVP sample number in the statistical dataset.

3) Variable D _S The method is used for representing the phase probability distribution of the remarkable solid water particles, and the calculation formula is as follows:

D _S ＝N _S /N；

wherein N is _S Representing the number of significant solid water particle phases for each sublayer in each MQVP in the statistical dataset, the number of hail particle phases + the number of ice crystal particle phases; n represents the MQVP sample number in the statistical dataset.

(2) Construction of asymmetric Gaussian distribution probability model

PMD generation using asymmetric Gaussian distribution probability model construction method _ZH 、PMD _KDP 、PMD _ZDR 、PMD _RHV 、PMD _KV . In the method for constructing the asymmetric Gaussian distribution probability model, probability distribution of the same layer is centered on the statistical average value of the MQVP, and Gaussian distribution is between the minimum value on the left side and the average value and between the maximum value on the right side and the average value.

The calculation steps of the asymmetric Gaussian distribution probability model are as follows: the normalized MQVP in the data set is counted, and the minimum value V of each layer is calculated _MIN Maximum value V _MAX Average value V _AVE The method comprises the steps of carrying out a first treatment on the surface of the At a minimum value V _MIN And average value V _AVE Calculating left side distribution probability P of the layer _L A curve; with average value V _AVE And a maximum value V _MAX Calculating the right side distribution probability P of the layer _R And finally, generating a corresponding MQVP probability model.

In sigma _L Get (V) _AVE -V _MIN )/3，σ _R Get (V) _MAX -V _AVE )/3。

Further, the judging of the short-time heavy rain weather background according to the MQVP probability model in the second step comprises the following steps:

calculating polarized radar data based on the current analysis moment, and calculating a standardized MQVP; according to PMD (TH), calculating MQVP comprehensive similarity probability indexes CP (TH) under different rainfall levels TH; and then judging: when the CP (TH) exceeds a judging threshold R (TH) under the rainfall level TH, the event of the rainfall level TH is considered to occur in a radar monitoring area, and the event is currently in a weather background of short-time heavy rain; and (3) screening MQVP data before and after 1 hour before the moment of short-time heavy rain to form an MQVP early warning data set.

When the determination threshold condition is satisfied under the occurrence of a plurality of rain levels TH, it is determined that a relatively large rain level TH event will occur in the radar monitoring area.

(1) Particle phase similarity probability index CP _PS (TH)

PMD-based _PS (TH) probability model, counting each radar data in MQVP early warning data set, calculating probability value of each layer, accumulating layer by layer, and generating particle phase similarity probability index CP _PS (TH). CP corresponding to each radar data of MQVP early warning data set _PS Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the particle phase similarity probability index _PS (TH). Wherein, the calculation formula of each layer probability value is:

wherein H represents the total number of layers of the standardized quasi-vertical profile; h represents MQVP _PS The phase of the particles of the H layer; in the case of the aragonite phase, CP _G D equal to layer H _G Probability value, and CP _W 、CP _S All are 0 probability values; if it is in the form of heavy rain particle phase or heavy drop particle phase, CP _W D equal to layer H _W Probability value, and CP _G 、CP _S All are 0 probability values; in the case of hail particle phase or ice crystal particle phase, CP _S D equal to layer H _S Probability value, and CP _G 、CP _W All are 0 probability values.

(2) Horizontal reflectivity factor similarity probability index CP _ZH (TH)

PMD-based _ZH The (TH) probability model is used for counting each radar data in the MQVP early warning data set, calculating the probability value of each layer, accumulating layer by layer and generating a horizontal reflectivity factor similarity probability index CP _ZH (TH). CP corresponding to each radar data of MQVP early warning data set _ZH Counting frequency and selecting the value of (TH)A value with a pass rate of 90% or more is used as a determination threshold R of the similarity probability index of the horizontal reflectivity factor _ZH (TH). Wherein, the calculation formula of each layer probability value is:

(3) Propagation phase shift rate similarity probability index CP _KDP (TH)

PMD-based _KDP The (TH) probability model is used for counting each radar data in the MQVP early warning data set, calculating the probability value of each layer, accumulating layer by layer and generating a propagation phase shift rate similarity probability index CP _KDP (TH). CP corresponding to each radar data of MQVP early warning data set _KDP Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the propagation phase shift rate similarity probability index _KDP (TH). Wherein, the calculation formula of each layer probability value is:

(4) Propagation phase shift rate similarity probability index CP _ZDR (TH)

PMD-based _ZDR (TH) probability model, counting each radar data in MQVP early warning data set, calculating probability value of each layer, accumulating layer by layer, and generating differential reflectivity similarity probability index CP _ZDR (TH). CP corresponding to each radar data of MQVP early warning data set _ZDR Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the differential reflectivity similarity probability index _ZDR (TH). Wherein, the calculation formula of each layer probability value is:

(5) Correlation coefficientSimilarity probability index CP _RHV (TH)

PMD-based _RHV (TH) probability model, counting each radar data in MQVP early warning data set, calculating probability value of each layer, accumulating layer by layer, generating correlation coefficient similarity probability index CP _RHV (TH). CP corresponding to each radar data of MQVP early warning data set _RHV Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the similarity probability index of the correlation coefficient _RHV (TH). Wherein, the calculation formula of each layer probability value is:

(6) Polarized condensate flux similarity probability index CP _KV (TH)

PMD-based _KV (TH) probability model, counting each radar data in MQVP early warning data set, calculating probability value of each layer, accumulating layer by layer, and generating polarized condensate flux similarity probability index CP _KV (TH). CP corresponding to each radar data of MQVP early warning data set _KV Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judgment threshold value R of the polarized condensate flux similarity probability index _KV (TH). Wherein, the calculation formula of each layer probability value is:

another object of the present invention is to provide an automatic short-time heavy rain weather background discrimination system applying the automatic short-time heavy rain weather background discrimination method, the automatic short-time heavy rain weather background discrimination system comprising:

the data set construction module is used for constructing MQVP corresponding to each double-polarization radar time by utilizing historical process case data of short-time storm, so as to construct an MQVP data set of short-time storm weather background;

the MQVP probability model building module is used for building an MQVP probability model for representing the short-time stormwater weather background according to the MQVP data set of the short-time stormwater weather background;

and the short-time stormwater weather background judging module is used for constructing an MQVP early warning data set, calculating an MQVP comprehensive similarity probability index CP (TH), and judging the short-time stormwater weather background according to the MQVP probability model.

Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, causes the processor to execute the steps of the automatic short-time stormwater weather background discriminating method.

Another object of the present invention is to provide a computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of the method for automatically discriminating a short-time heavy rain weather background.

The invention further aims at providing an information data processing terminal which is used for realizing the automatic distinguishing system for the short-time heavy rain weather background.

In combination with the technical scheme and the technical problems to be solved, the technical scheme to be protected has the following advantages and positive effects:

first, aiming at the technical problems in the prior art and the difficulty of solving the problems, the technical problems solved by the technical proposal of the invention are analyzed in detail and deeply by tightly combining the technical proposal to be protected, the results and data in the research and development process, and the like, and some technical effects brought after the problems are solved have creative technical effects. The specific description is as follows:

(1) The traditional QVP method is based on an average statistical strategy, and the MQVP (normalized large value strategy QVP) method proposed in the invention is selected based on a maximum value strategy on different layers of equal-height surfaces. The benefit of the maximum strategy is that the likelihood of local short-term stormwater characteristics can be monitored in advance more timely.

(2) The temperature layer knots (0 ℃ layer height, -10 ℃ layer height, -20 ℃ layer height) in different seasons are different, so that QVP of short-time heavy rain weather background generated in different seasons has certain difference, and the construction of a unified discrimination model is not facilitated. The MQVP method provided by the invention takes the layer height at 0 ℃ and the layer height at-20 ℃ as references, and performs standardized treatment on the vertical profile, so that the method is more beneficial to constructing a unified feature recognition model of short-time heavy rain weather background.

(3) The method can provide the identification of the short-time heavy rain weather background in the radar monitoring area and the pre-judging information of the maximum rainfall level in the area, further provides technical support for the early prediction and early warning of the short-time heavy rain disasters, and has positive significance for improving the disaster prevention and reduction of the weather strong weather disasters.

Secondly, the technical scheme is regarded as a whole or from the perspective of products, and the technical scheme to be protected has the following technical effects and advantages:

according to the judging method provided by the invention, the MQVP probability modeling is performed by fully utilizing the polarization quantity characteristics of the dual-polarization radar, and the automatic identification of the short-time strong rainfall weather background of the rainfall grade is realized according to the MQVP probability modeling. Compared with the traditional QVP method, the novel method can better exert the advantages of the polarized radar in the fields of monitoring and early warning of short-time strong precipitation, and expand the application neighborhood of QVP in the early warning of strong convection. The probability model is designed to be an intelligent recognition model based on radar weather service logic, and has a good demonstration effect on how to reasonably and scientifically integrate and apply computer recognition technology and the like in weather service in the future.

Thirdly, as inventive supplementary evidence of the claims of the present invention, the following important aspects are also presented:

(1) The expected benefits and commercial values after the technical scheme of the invention is converted are as follows:

the method fully utilizes the advantages of the double-polarization radar, constructs a short-time heavy rain weather background recognition model, can better play the advantages of the polarization radar in the fields of monitoring and early warning of short-time strong rainfall, and expands the application neighborhood of QVP in the early warning of strong convection. In addition, the model construction method can be further upgraded in the future, and a rainfall-based recognition model can be constructed, so that the model can have a certain recognition capability on weather background recognition of extremely short-time strong rainfall (> 50 mm/h) in theory, and has positive significance on short-time early warning of the short-time strong rainfall.

The method is characterized in that QVP is firstly applied to early warning of short-time strong rainfall in China, and the application neighborhood of QVP in early warning of strong convection is further expanded.

(2) Whether the technical scheme of the invention solves the technical problems that people want to solve all the time but fail to obtain success all the time is solved:

a large number of monitoring devices can be used for monitoring short-time strong precipitation live disasters in the current business, but the corresponding method is limited for short-time strong precipitation early warning, so that the invention is designed according to the business requirement of short-time heavy precipitation early warning.

(3) The technical scheme of the invention overcomes the technical bias:

whether it is weather or in the computer neighborhood, a great deal of researchers focus on improvement of machine learning method models and learning method parameters, but in the weather field, the method for constructing a weather identification model is more important because of the problems of large data volume, uneven data quality, large data feature quantity and the like. The probability model is designed to be an intelligent recognition model based on radar weather service logic, and has a good demonstration effect on how to reasonably and scientifically integrate and apply computer recognition technology and the like in weather service in the future.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments of the present invention will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a flow chart of a method for automatically discriminating the background of short-time heavy rain weather provided by the embodiment of the invention;

FIG. 2 is a standardized quasi-vertical profile schematic provided by an embodiment of the present invention; FIG. (a) is a normalized horizontal reflectance factor quasi-vertical profile and FIG. (b) is a horizontal reflectance factor quasi-vertical profile;

FIG. 3 is a schematic diagram of the statistical level range of MQVP provided by an embodiment of the present invention;

FIG. 4 is a schematic diagram of an MQVP particle phase probability model provided by an embodiment of the invention;

FIG. 5 is a schematic diagram of an asymmetric Gaussian probability distribution curve provided by an embodiment of the invention;

FIG. 6 is a schematic diagram of the MQVP probability model (1%, 20%, 40%, 60%, 80% and 100% probability characteristic curves) provided by an embodiment of the present invention;

FIG. 7 is a graph of a quasi-vertical profile probability model (1%, 20%, 40%, 60%, 80%, and 100% probability signature curves) for maximum correlation coefficients provided by an embodiment of the present invention;

FIG. 8 is a graph of a quasi-vertical profile probability model (1%, 20%, 40%, 60%, 80%, and 100% probability signature curves) of the maximum propagation phase shift rate provided by an embodiment of the present invention;

FIG. 9 is a graph of quasi-vertical profile probability models (1%, 20%, 40%, 60%, 80%, and 100% probability signature curves) for maximum differential reflectivity provided by embodiments of the present invention;

FIG. 10 is a graph of a quasi-vertical profile probability model (1%, 20%, 40%, 60%, 80%, and 100% probability characteristics) of the maximum horizontal reflectance factor provided by an embodiment of the invention;

FIG. 11 is a schematic representation of a quasi-vertical profile probability model (1%, 20%, 40%, 60%, 80% and 100% probability characteristics) of maximum polarized water condensate flux provided by an embodiment of the present invention;

FIG. 12 is a schematic diagram of a quasi-vertical profile probability model of the most significant particle phase (aragonite, pure liquid particle phase, and significant solid particle phase probability characteristics) provided by an embodiment of the present invention.

Fig. 13 is a comparison of a short-time automatic rainstorm weather background discrimination result and an hour-by-hour automatic station rainstorm monitoring result of a 2022, 6, 1 day weather process provided by an embodiment of the invention.

Fig. 14 is a comparison of a short-time automatic rainstorm weather background discrimination result and an hour-by-hour automatic station rainstorm monitoring result of a 2022, 8, 21-day weather process provided by an embodiment of the invention.

Detailed Description

The present invention will be described in further detail with reference to the following examples in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

Aiming at the problems existing in the prior art, the invention provides a method, a system, equipment and a terminal for automatically distinguishing the background of short-time heavy rain weather, and the invention is described in detail below with reference to the accompanying drawings.

In order to fully understand how the invention may be embodied by those skilled in the art, this section is an illustrative embodiment in which the claims are presented for purposes of illustration.

As shown in fig. 1, the method for automatically distinguishing the short-time heavy rain weather background provided by the embodiment of the invention comprises the following steps:

s101, constructing MQVP corresponding to time and time of the double-polarization radar by utilizing historical process example data of short-time heavy rain, and further constructing an MQVP data set of short-time heavy rain weather background;

s102, constructing an MQVP probability model for representing the short-time stormwater weather background according to an MQVP data set of the short-time stormwater weather background;

s103, an MQVP early warning data set is constructed, an MQVP comprehensive similarity probability index CP (TH) is calculated, and short-time storm weather background discrimination is carried out according to an MQVP probability model.

As a preferred embodiment, the method for automatically distinguishing the short-time heavy rain weather background provided by the embodiment of the invention specifically comprises the following steps:

1. MQVP probability model construction for representing short-time storm weather background

According to the method, the historical process example data of short-time storm is utilized to construct MQVP corresponding to each double-polarization radar time and time, so that an MQVP data set of short-time storm weather background is constructed, and then an MQVP probability model is constructed according to the data set. The automatic judging method for the short-time heavy rain weather background provided by the embodiment of the invention selects the variables listed in the table 1 to form the MQVP for representing the short-time heavy rain weather background.

TABLE 1 variable List of MQVP

1.1MQVP Process

The MQVP process for each dual-polarization radar profile is divided into 4 steps:

1) Pretreatment of

Firstly, carrying out bilinear interpolation on single radar data to obtain equal-altitude lattice point polarization data; then generating particle phase state identification data by adopting a phase state inversion method based on fuzzy logic; vertical wind profile data representing atmospheric environmental wind field information is generated using a velocity-azimuth display (VAD) method. The contour data is uniformly processed from 1km to 17km at a vertical interval of 500 m. The combined raindrop spectrum data analysis shows that the intensity of precipitation in the plum rainy period is not necessarily related to the depth of convection development, and extreme precipitation is mainly caused by medium-height convection. Therefore, the contour data can be processed to 17km height to meet the analysis requirement.

2) Convection kernel region lattice marking

Firstly, judging convection kernel areas with different heights according to the equal-altitude surface reflectivity and the particle phase state identification result, and marking lattice points. The particle phase states identified in the original particle phase state identification product are mainly: ground, living things, dry snow, wet snow, ice crystals, aragonite, heavy drops, rain, heavy rain. The determination is here as a kernel region latticeThe criteria for the dots are: for layer heights below 0deg.C, if Z _H >42dBZ, and meanwhile, the particle phase state is identified as hail or rain or heavy rain, and is marked as a grid point of a core area; for layer heights above 0deg.C, if Z _H >42dBZ, while satisfying the identification of the particle phase as hail or aragonite, is marked as the kernel region lattice point.

3) Statistics of maximum quasi-vertical profiles for various variables of MQVP within a convection kernel region

Carrying out numerical feature statistics on polarized quantity of the flow core points marked on different contour planes to generate corresponding MQVP _ZH 、MQVP _PS 、MQVP _KDP 、MQVP _ZDR 、MQVP _RHV MQVP _KV . The quasi-vertical profile sampling strategy here takes the maximum, see in particular steps a) to f).

4) The MQVP is standardized according to the reference temperature layer junction

As shown in fig. 2, the quasi-vertical profile of each original variable is normalized by a nearest interpolation strategy with reference to the 0 ℃ layer and-20 ℃ height in the temperature layer junction curve. The normalized quasi-vertical profile is 3 parts: a) Ground to 0 ℃ layer height; b) Layer height of 0 ℃ to layer height of-20 ℃; c) -20 ℃ layer height to 17km height. Each vertical profile data segment is n-aliquoted (10 by default) to ensure the equalization of data characteristic information for different data segments.

a)MQVP _ZH Is a sampling strategy of (a)

Z for marked convection nuclei on different contours _H Numerical feature statistics, statistics of maximum Z on different contour heights _H Numerical values. To maximize the statistics of convective microphysical structural features near the more complete 0 ℃ layer height, the effective statistical level range of MQVP [ D _min ，D _max ]And according to the height calculation of the actual specific temperature layer, determining:

D _min ＝(arccos((R _m +H _R )*cos(θ _19.5 )/(H ₀ +R _m ))-θ _19.5 )*R _m ；

D _max ＝(arccos((R _m +H _R )*cos(θ _0.5 )/(H ₀ +R _m ))-θ _0.5 )*R _m ；

wherein D is _min Representing a radar-centric minimum radius (in km) in the statistical level range; d (D) _max Representing a radar-centric maximum radius (in km) in the statistical level range; arccosis represents an inverse cosine function; cos represents a cosine function; r is R _m Representing the equivalent earth radius (in km); h _R Representing radar station altitude (in km); h ₀ A layer height (in km) of 0 ℃ representing a sounding observation corresponding to the statistical data period; θ _19.5 The radian value corresponding to the angle 19.5 degrees is represented; θ _0.5 The radian value corresponding to the angle 0.5 deg. is indicated.

b)MQVP _PS Is a sampling strategy of (a)

Reference and MQVP _ZH The same statistical level range algorithm (see equations 1-2) determines MQVP _PS And (5) horizontal statistics range information. And carrying out remarkable particle phase state characteristic statistics on the convection core points marked on different contour surfaces, and counting the most remarkable particle phase state values on different contour surface heights. Here is given a decreasing order of significance ordering of the different particle phases in the flow kernel: hail, aragonite, heavy rain, heavy drops, ice crystals, wet snow, dry snow.

c)MQVP _KDP Is a sampling strategy of (a)

Reference and MQVP _ZH The same statistical level range algorithm (see equations 1-2) determines MQVP _KDP And (5) horizontal statistics range information. K for marked convection kernel points on different contour surfaces _DP Numerical feature statistics, statistics of maximum K on different contour heights _DP Numerical values. Here increase ρ _hv Numerical monitoring is carried out to realize simple substance control constraint of the polarization quantity; if ρ _hv The values are in the range of [0.6,1.0]It is believed that the current polarization amount may participate in MQVP statistics.

d)MQVP _ZDR Is a sampling strategy of (a)

Reference and MQVP _ZH The same statistical level range algorithm (see equations 1-2) determines MQVP _ZDR And (5) horizontal statistics range information. Z for marked convection nuclei on different contours _DR Numerical feature statistics, statistics of maximum Z on different contour heights _DR Numerical values. Here increase ρ _hv Numerical monitoring is carried out to realize simple substance control constraint of the polarization quantity; if ρ _hv The values are in the range of [0.6,1.0]It is believed that the current polarization amount may participate in MQVP statistics.

e)MQVP _RHV Is a sampling strategy of (a)

Reference and MQVP _ZH The same statistical level range algorithm (see equations 1-2) determines MQVP _RHV And (5) horizontal statistics range information. ρ for flow kernel points marked on different contours _hv Numerical feature statistics, statistics of maximum rho on different contour heights _hv Numerical values. ρ _hv Is in the reasonable numerical range of [0.6,1.0 ]]。

f)MQVP _KV Is a sampling strategy of (a)

Will K _DP Multiplying it by the ambient wind speed value VADL, a polarized condensate flux (KV, equation 3) that approximates the characteristics of the water vapor flux distribution can be obtained. In addition, according to K _DP Is the detection of the moisture content characteristics of the condensate in the atmosphere, so that a large KV value is also indicative of the presence of this region irradiance feature.

KV(h)＝MQVP _KDP (h)*VADL(h),h∈[0,17km](3)

H in equation 3 is the height (km) of the quasi-vertical profile; by MQVP corresponding to height h _KDP Multiplying the ambient wind speed VADL calculated by the velocity-azimuth display (VAD) method can calculate and obtain polarized condensate flux (KV) corresponding to the height h.

1.2 construction of MQVP probability model

The MQVP probability model in the invention is MQVP probability model PMD (TH) of the rainfall grade, and is statistically constructed according to a threshold TH of the rainfall grade of the storm. The TH is selected to be 20mm/h, 30mm/h, 40mm/h and 50mm/h in sequence in the method. And constructing a corresponding data set by taking the automatic rainfall station monitoring rainfall on a time-by-time basis, and if the maximum hour rainfall in the radar monitoring range reaches the TH standard, classifying the MQVP of all radar data within 1 hour before the moment of the storm to be an MQVP probability model statistical data set under a threshold TH. Because the MQVP is a standard processing quasi-vertical profile, the problem of junction difference of seasonal temperature layers is not needed to be considered in MQVP probability model processing, and the MQVP probability model can be directly and uniformly output as a unique MQVP probability model. The probability model is composed mainly of the variable probability models listed in table 2.

TABLE 2 variable list of PMD

Variable name	Chinese name
		PMD ZH	Maximum Z H Quasi-vertical profile probability model (of horizontal reflectivity factor)
PMD PS	Quasi-vertical profile probability model for most significant PS (particle phase)
		PMD KDP	Maximum K DP Quasi-vertical profile probability model (propagation phase shift rate)
PMD ZDR	Maximum Z DR Quasi-vertical profile probability model (differential reflectivity)
		PMD RHV	Maximum ρ hv Quasi-vertical profile probability model (correlation coefficient)
PMD KV	Quasi-vertical profile probability model for maximum KV (polarized hydrogel flux)

1)PMD _PS (TH) construction method

Probability statistics of particle phases are mainly used for analysis constraints. The probability model consists of three types of statistical properties (see table 3), and the constructed schematic diagram is shown in fig. 4.

TABLE 3PMD _PS Statistical attribute table

2) Asymmetric Gaussian distribution probability model construction method

PMD generation using asymmetric Gaussian distribution probability model construction method _ZH 、PMD _KDP 、PMD _ZDR 、PMD _RHV 、PMD _KV . The asymmetric gaussian distribution probability model construction method assumes that the probability distribution of the same layer is asymmetric with the statistical average value of MQVP as the center, the gaussian distribution between the minimum value to the average value on the left side and the gaussian distribution between the maximum value to the average value on the right side (see fig. 5).

The calculation steps of the asymmetric Gaussian distribution probability model are as follows: firstly, carrying out statistics on MQVP after standardization in a data set, and calculating the minimum value V of each layer _MIN Maximum value V _MAX Average value V _AVE . Then, at the minimum value V _MIN And average value V _AVE Calculating left side distribution probability P of the layer _L Curves (see formulas 4 to 5, fig. 5); with average value V _AVE And a maximum value V _MAX Calculating the right side distribution probability P of the layer _R Curve (see public)

Formulas 4 to 5, fig. 5). Finally, a corresponding MQVP probability model is generated (see fig. 6).

Sigma in formula (4) _L Default can take (V) _AVE -V _MIN ) 3; sigma in formula (5) _R Default can take (V) _MAX -V _AVE )/3。

2. Short-time heavy rain weather background discrimination method based on MQVP probability model

In practical business application, firstly, calculating polarized radar data based on the current analysis moment, and calculating standardized MQVP; then, according to PMD (TH), calculating MQVP comprehensive similarity probability index CP (TH) under different rainfall levels TH; finally, judging: when CP (TH) exceeds a decision threshold R (TH) at the rain level TH, then it is considered that an event at the rain level TH will occur within the radar surveillance area (i.e., currently in the weather background of short-term heavy rain).

In view of the early warning of the weather early warning business, it is necessary to construct an MQVP early warning data set so as to monitor and judge the weather background of short-time strong rainfall earlier. Because the MQVP probability model statistical data set contains all MQVP data within 1 hour before all short-time storm occurrence moments, MQVP data before and after 1 hour before the short-time storm occurrence moments are filtered to form an MQVP early warning data set.

2.1MQVP comprehensive similarity probability index CP (TH) calculation method

The MQVP comprehensive similarity probability index CP (TH) consists of the following 6 indices: particle phase similarity probability index CP _PS (TH), horizontal reflectivity factor similarity probability index CP _ZH (TH) propagation phase Shift Rate similarity probability index CP _KDP (TH), differential reflectivity similarity probability index CP _ZDR (TH) correlation coefficient similarity probability index CP _RHV (TH) and polarized-water-condensate flux similarity probability index CP _KV (TH). The specific judgment standard is as follows:

if CP _ZH (TH)、CP _PS (TH)、CP _RHV (TH) and CP _KV (TH) are all greater than the corresponding threshold while satisfying CP _KDP (TH) or CP _ZDR If a similarity probability index in (TH) is greater than the corresponding threshold, it may be determined that an event of the current rainfall level TH will occur in the radar surveillance area (i.e., currently in the weather background of short-term heavy rain).

When the determination threshold condition is satisfied at all of the plurality of rain levels TH, it is determined that a relatively large rain level TH event will occur in the radar monitoring area.

(1) Particle phase similarity probability index CP _PS (TH)

PMD-based _PS (TH) probability model, counting each radar data in the MQVP early warning data set, calculating probability value of each layer according to formula (6), accumulating layer by layer, and generating particle phase similarity probability index CP _PS (TH). CP corresponding to each radar data of MQVP early warning data set _PS Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the particle phase similarity probability index _PS (TH)。

In equation (6), H represents the total number of layers of the normalized quasi-vertical profile; h represents MQVP _PS The phase of the particles of the H layer; if it is in the aragonite phase, CP _G D equal to layer H _G Probability values, at the same time CP _W 、CP _S Are all 0 probability values; if it is in a heavy rain particle phase or a heavy drop particle phase, CP _W D equal to layer H _W Probability values, at the same time CP _G 、CP _S Are all 0 probability values; if it is hail particle phase or ice crystal particle phase, then CP _S D equal to layer H _S Probability values, at the same time CP _G 、CP _W Are all 0 probability values.

(2) Horizontal reflectivity factor similarity probability index CP _ZH (TH)

PMD-based _ZH (TH) probability model, counting each radar data in the MQVP early warning data set, calculating probability value of each layer according to formula (7), accumulating layer by layer, and generating horizontal reflectivity factor similarity probability index CP _ZH (TH). CP corresponding to each radar data of MQVP early warning data set _ZH Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the similarity probability index of the horizontal reflectivity factors _ZH (TH)。

(3) Propagation phase shift rate similarity probability index CP _KDP (TH)

PMD-based _KDP (TH) probability model, counting each radar data in the MQVP early warning data set, calculating probability value of each layer according to formula (8), accumulating layer by layer, and generating propagation phase shift rate similarity probability index CP _KDP (TH). CP corresponding to each radar data of MQVP early warning data set _KDP Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the propagation phase shift rate similarity probability index _KDP (TH)。

/>

(4) Propagation phase shift rate similarity probability index CP _ZDR (TH)

PMD-based _ZDR (TH) probability model, counting each radar data in the MQVP early warning data set, calculating probability value of each layer according to formula (9), accumulating layer by layer, and generating differential reflectivity similarity probability index CP _ZDR (TH)。CP corresponding to each radar data of MQVP early warning data set _ZDR Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the differential reflectivity similarity probability index _ZDR (TH)。

(5) Correlation coefficient similarity probability index CP _RHV (TH)

PMD-based _RHV (TH) probability model, counting each radar data in the MQVP early warning data set, calculating probability value of each layer according to formula (10), accumulating layer by layer, and generating correlation coefficient similarity probability index CP _RHV (TH). CP corresponding to each radar data of MQVP early warning data set _RHV Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the similarity probability index of the correlation coefficient _RHV (TH)。

(6) Polarized condensate flux similarity probability index CP _KV (TH)

PMD-based _KV (TH) probability model, counting each radar data in the MQVP early warning data set, calculating probability value of each layer according to formula (11), accumulating layer by layer to generate polarized water condensate flux similarity probability index CP _KV (TH). CP corresponding to each radar data of MQVP early warning data set _KV Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judgment threshold value R of the polarized condensate flux similarity probability index _KV (TH)。

The automatic distinguishing system for the short-time heavy rain weather background provided by the embodiment of the invention comprises the following steps:

the data set construction module is used for constructing MQVP corresponding to each double-polarization radar time by utilizing historical process case data of short-time storm, so as to construct an MQVP data set of short-time storm weather background;

the MQVP probability model building module is used for building an MQVP probability model for representing the short-time stormwater weather background according to the MQVP data set of the short-time stormwater weather background;

And the short-time stormwater weather background judging module is used for constructing an MQVP early warning data set, calculating an MQVP comprehensive similarity probability index CP (TH), and judging the short-time stormwater weather background according to the MQVP probability model.

In order to prove the inventive and technical value of the technical solution of the present invention, this section is an application example on specific products or related technologies of the claim technical solution.

Based on the method provided by the invention, the construction of the short-time storm weather background recognition model is carried out on the Taizhou area (radar ID Z9576) based on the process personal data of the short-time storm. The constructed model shows that the short-time strong precipitation polarization in the Taizhou area has the strong precipitation structural characteristics of low mass center.

After a short-time heavy rain weather background identification model of the Taizhou area is obtained, the method is applicable to the example of the two short-time heavy precipitation processes in the period of the present year, and the advance of the short-time heavy rain weather background is identified by the model. Two times process example period (Beijing time) 2022, 6, 1 and 2022, 8, 21. During month 1 of 2022, the strongly rainfall weather system moves from north to south, affecting the south-Zhejiang area approximately 14 to 20 hours of the day. Automatic rainfall station monitoring shows that at 14-20 hours, the southern China to the Wenzhou area of the Taizhou shows short-time strong precipitation of more than 20mm/h, so that the weather background type in the range monitored by the radar in the Taizhou is short-time strong precipitation. In the course of day 21, 8 of 2022, the strong precipitation system mainly affects the southeast region of Zhejiang. Automatic rainfall station monitoring shows that short-time strong precipitation of more than 20mm/h occurs in the state area from 13 hours to 19 hours, so the weather background type in the state radar monitoring range is short-time strong precipitation.

The short-time heavy rain weather background identification model of the Taizhou area is applied to body sweep data (scanned once in 6 minutes) of 2022, 6, 1, 8 and 23 days to judge the short-time heavy rain weather background. Comparison of the automatic discrimination result of the short-time heavy rain weather background and the automatic station heavy rain monitoring result (figure 13) from hour to hour shows that in the process, the method provided by the invention can identify the characteristics of the short-time heavy rain weather background in 13 hours and 15 minutes. By the method, short-time storm disasters can be judged to occur in the radar monitoring range of the area in advance by approximately 45 minutes. In addition, it is determined that the short-term storm occurrence period substantially coincides with the live monitoring.

The short-time heavy rain weather background identification model of the Taizhou area is applied to body sweep data (scanned once in 6 minutes) at the time of 2022, 8 and 21 days, 8 and 23 to judge the short-time heavy rain weather background. Comparison of the automatic discrimination result of the short-time heavy rain weather background and the automatic station heavy rain monitoring result (figure 14) from hour to hour shows that the method provided by the invention can identify the characteristics of the short-time heavy rain weather background in 12 hours and 20 minutes in the process. By the method, short-time storm disasters can be judged to occur in the radar monitoring range of the area in advance by approximately 40 minutes. In addition, it is determined that the short-term storm occurrence period substantially coincides with the live monitoring.

The embodiment of the invention has a great advantage in the research and development or use process, and has the following description in combination with data, charts and the like of the test process.

The existing objective short-time storm monitoring and forecasting method is mainly based on two types of automatic station rainfall monitoring and radar extrapolation forecasting methods. With the automatic station rainfall monitoring method, the accumulated rainfall for the past 1 hour is generally used as a standard for distinguishing the short-time heavy rainfall, so when the monitoring reaches the standard, a short-time heavy rainfall disaster (the early warning for the short-time heavy rainfall has a limited advance) has actually occurred. In the radar extrapolation prediction method, the guiding airflow field is extrapolated as an extrapolation field, and the convective cloud of the short-time storm actually has backward propagation characteristics, so that a certain deviation exists between the extrapolation field of the radar and the actual convection propagation direction and speed. Within 30 minutes, the radar extrapolated strong precipitation landing area is relatively close to the live condition, and after 30 minutes, the short-time storm prediction information also deviates more due to the fact that the extrapolation deviation is larger. The average statistics shows that the early warning advance of the objective method for short-time heavy rain is about 30 minutes.

Compared with the existing objective business method, the method provided by the invention has the advantages that the method has better short-time storm early warning advance (the advance time is increased to 40-45 minutes), and can provide more accurate short-time storm early warning information reference for a forecaster in a near forecast period. In addition, the invention can further subdivide and upgrade the statistical model in the future, and construct an extremely short-time strong precipitation recognition model (such as a 50mm/h rainfall recognition model), so that the invention has better expansion application prospect suitable for business.

It should be noted that the embodiments of the present invention can be realized in hardware, software, or a combination of software and hardware. The hardware portion may be implemented using dedicated logic; the software portions may be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or special purpose design hardware. Those of ordinary skill in the art will appreciate that the apparatus and methods described above may be implemented using computer executable instructions and/or embodied in processor control code, such as provided on a carrier medium such as a magnetic disk, CD or DVD-ROM, a programmable memory such as read only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The device of the present invention and its modules may be implemented by hardware circuitry, such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., as well as software executed by various types of processors, or by a combination of the above hardware circuitry and software, such as firmware.

The foregoing is merely illustrative of specific embodiments of the present invention, and the scope of the invention is not limited thereto, but any modifications, equivalents, improvements and alternatives falling within the spirit and principles of the present invention will be apparent to those skilled in the art within the scope of the present invention.

Claims (8)

1. The automatic distinguishing method for the short-time heavy rain weather background is characterized by comprising the following steps of:

constructing a standardized large-value strategy QVP corresponding to each double-polarization radar time by utilizing historical process example data of short-time storm, namely an MQVP, and further constructing an MQVP data set of a short-time storm weather background; according to the MQVP data set of the short-time stormwater weather background, an MQVP probability model for representing the short-time stormwater weather background is constructed; constructing an MQVP early warning data set, calculating an MQVP comprehensive similarity probability index CP (TH), and judging the short-time stormwater weather background according to an MQVP probability model;

wherein the MQVP process for each dual-polarization radar profile comprises:

(1) Pretreatment of

Performing bilinear interpolation on the single radar data to obtain equal-altitude lattice point polarization data; generating particle phase identification data by adopting a phase inversion method based on fuzzy logic; generating vertical wind profile data representing atmospheric environment wind field information by adopting a speed and direction display method; wherein, the constant-altitude data are uniformly processed from 1km to 17km at a vertical interval of 500 m;

(2) Convection kernel region lattice marking

Judging convection kernel areas with different heights according to the equal-altitude surface reflectivity and the particle phase state identification result, and marking lattice points; the original particle phase states identified in the product include ground objects, organisms, dry snow, wet snow, ice crystals, aragonite, heavy droplets, rain and heavy rain; the criteria for determining the lattice points of the kernel area are: for layer heights below 0deg.C, if Z _H >42dBZ, and meanwhile, the particle phase state is identified as hail or rain or heavy rain, and is marked as a grid point of a core area; for layer heights above 0deg.C, if Z _H >42dBZ, while satisfying the recognition of the particle phase as hail or aragonite, is marked asKernel region lattice points; z is Z _H Is a symbol representing the weather radar level reflectivity factor, the unit is Dbz, and reflects the comprehensive effect of the number and the size of water drops or ice crystals in a volume element, and is used for estimating the rainfall intensity;

(3) Statistics of maximum quasi-vertical profiles for various variables of MQVP within a convection kernel region

Carrying out numerical feature statistics on polarized quantity of the flow core points marked on different contour planes to generate corresponding MQVP _ZH 、MQVP _PS 、MQVP _KDP 、MQVP _ZDR 、MQVP _RHV MQVP _KV The method comprises the steps of carrying out a first treatment on the surface of the The quasi-vertical profile sampling strategy adopts a maximum value; MQVP _ZH Representing the normalized maximum horizontal reflectance factor Z _H Is defined by a quasi-vertical profile of (a); characterization of the variation in vertical direction of maximum precipitation intensity in the atmosphere representing a strong weather feature within the monitoring range, MQVP _PS The representation is: the quasi-vertical profile of the standardized most significant particle phase PS characterizes the distribution in the vertical direction of the most significant precipitation particle phase in the atmosphere within the monitoring range, which can represent strong weather features; MQVP _KDP The representation is: normalized maximum propagation phase shift rate K _DP Representing the distribution of the maximum rainfall intensity representing the strong weather feature in the atmosphere in the monitoring range in the vertical direction; another measure reflecting the intensity of precipitation and monitoring the vertical structure of strong precipitation; MQVP _ZDR The representation is: normalized maximum differential reflectance Z _DR Representing the distribution of most obvious precipitation particle morphology capable of representing strong weather features in the atmosphere in the monitoring range in the vertical direction; MQVP _RHV The representation is: normalized maximum correlation coefficient ρ _hv Representing the vertical distribution of liquid precipitation particles representing strong weather features in the atmosphere within the monitoring range; MQVP _KV A quasi-vertical profile representing a normalized maximum polarized hydrogel flux KV, characterizing a vertical distribution of the maximum hydrogel flux representing strong weather features in the atmosphere within the monitoring range;

(4) The MQVP is standardized according to the reference temperature layer junction

Taking a layer at 0 ℃ and a height at-20 ℃ in a temperature layer junction curve as references, and carrying out standardization treatment on the original quasi-vertical profile of each variable through a nearby interpolation strategy; the normalized quasi-vertical profile is 3 parts: a) Ground to 0 ℃ layer height; b) Layer height of 0 ℃ to layer height of-20 ℃; c) -20 ℃ layer height to 17km height; n is divided equally for each vertical profile data segment, n is set to be 10 preferentially;

the short-time storm weather background discrimination according to the MQVP probability model comprises the following steps:

calculating polarized radar data based on the current analysis moment, and calculating a standardized MQVP; according to PMD (TH), calculating MQVP comprehensive similarity probability indexes CP (TH) under different rainfall levels TH; and then judging: when the CP (TH) exceeds a judging threshold R (TH) under the rainfall level TH, the event of the rainfall level TH is considered to occur in a radar monitoring area, and the event is currently in a weather background of short-time heavy rain; MQVP data before and after 1 hour before the moment of short-time heavy rain is screened to form an MQVP early warning data set;

when the determination threshold condition is satisfied under the occurrence of a plurality of rain levels TH, it is determined that a relatively large rain level TH event will occur in the radar monitoring area.

2. The method for automatically discriminating a short-time heavy rain weather background according to claim 1 wherein the quasi-vertical profile sampling strategy in step (3) comprises:

1)MQVP _ZH is a sampling strategy of (a)

Z for marked convection nuclei on different contours _H Numerical feature statistics, statistics of maximum Z on different contour heights _H A numerical value; to count convective microphysical structural features near 0 ℃ layer height, the effective statistical level range of MQVP [ D _min ，D _max ]The height of the layer is determined according to the actual specific temperature:

D _min ＝(arccos((R _m +H _R )*cos(θ _19.5 )/(H ₀ +R _m ))-θ _19.5 )*R _m ；

D _max ＝(arccos((R _m +H _R )*cos(θ _0.5 )/(H ₀ +R _m ))-θ _0.5 )*R _m ；

wherein D is _min Representing a minimum radius centered on the radar in a statistical level range, in km; d (D) _max Representing the maximum radius centered on the radar in the statistical level range, in km; arccosis represents an inverse cosine function; cos represents a cosine function; r is R _m Representing equivalent earth radius, unit km; h _R Representing the height of the radar station, and the unit km; h ₀ The height of the sounding observation layer at 0 ℃ corresponding to the statistical data time period is expressed, and the unit is km; θ _19.5 The radian value corresponding to the angle 19.5 degrees is represented; θ _0.5 The radian value corresponding to the angle of 0.5 degrees is represented;

2)MQVP _PS is a sampling strategy of (a)

Utilizing MQVP _ZH Statistical level range algorithm of (1) to determine MQVP _PS Level statistics range information; carrying out remarkable particle phase state feature statistics on the convection core points marked on different equal-altitude surfaces, and counting the most remarkable particle phase state values on different equal-altitude surface heights; wherein the order of decreasing significance of the different particle phases in the flow kernel is: hail, aragonite, heavy rain, heavy drops, ice crystals, wet snow, dry snow;

3)MQVP _KDP Is a sampling strategy of (a)

Utilizing MQVP _ZH Statistical level range algorithm of (1) to determine MQVP _KDP Level statistics range information; k for marked convection kernel points on different contour surfaces _DP Numerical feature statistics, statistics of maximum K on different contour heights _DP A numerical value; wherein ρ is increased _hv Numerical monitoring is carried out to realize simple substance control constraint of the polarization quantity; if ρ _hv The values are in the range of [0.6,1.0]The current polarization amount is considered to participate in MQVP statistics; k (K) _DP : the propagation phase shift rate is an inversion parameter of the polarization quantity of the double-polarization polarized radar, and the change rate of the differential phase along with the distance is represented; the correlation coefficient is one of the main polarization amounts observed by the dual-polarization radar, which is a parameter measuring the relation between the horizontal and vertical polarized echo power, ρ _hv The value range of the correlation coefficient is 0 to 1;

4)MQVP _ZDR is a sampling strategy of (a)

Utilizing MQVP _ZH Statistical level range algorithm of (1) to determine MQVP _ZDR Level statistics range information; z for marked convection nuclei on different contours _DR Numerical feature statistics, statistics of maximum Z on different contour heights _DR A numerical value; wherein ρ is increased _hv Numerical monitoring is carried out to realize simple substance control constraint of the polarization quantity; if ρ _hv The values are in the range of [0.6,1.0]The current polarization amount is considered to participate in MQVP statistics; z is Z _DR The differential reflectivity factor is one of the main polarization amounts observed by dual-polarization radars;

5)MQVP _RHV is a sampling strategy of (a)

Utilizing MQVP _ZH Statistical level range algorithm of (1) to determine MQVP _RHV Level statistics range information; ρ for flow kernel points marked on different contours _hv Numerical feature statistics, statistics of maximum rho on different contour heights _hv A numerical value; wherein ρ is _hv Is in the reasonable numerical range of [0.6,1.0 ]]；

6)MQVP _KV Is a sampling strategy of (a)

Will K _DP Multiplying the ambient wind speed value VADL to obtain polarized condensate flux KV which approximately represents the water vapor flux distribution characteristics; according to K _DP The water content characteristics of the condensate in the atmosphere are detected, so that the KV large value indicates the existence of the region combination characteristic;

KV(h)＝MQVP _KDP (h)*VADL(h),h∈[0,17km]；

wherein h is the height of the quasi-vertical profile, and the unit is km;

by MQVP corresponding to height h _KDP Multiplying the ambient wind speed VADL obtained by calculation of the speed and direction display method, and further calculating to obtain polarized condensate flux KV corresponding to the height h.

3. The method for automatically distinguishing the short-time stormwater weather background according to claim 1, wherein the MQVP probability model in the first step is MQVP probability model PMD (TH) of the rainfall grade, and is statistically constructed according to a stormwater grade threshold TH; the TH is selected to be 20mm/h, 30mm/h, 40mm/h and 50mm/h in sequence; constructing a corresponding data set by taking the automatic rainfall station monitoring rainfall on a time-by-time basis, and if the maximum rainfall in the radar monitoring range reaches the TH standard, classifying the MQVP of all radar data within 1 hour before the moment of the storm as an MQVP probability model statistical data set under a threshold TH;

The construction of the MQVP probability model comprises the following steps:

(1)PMD _PS construction of (TH)

The probability statistics of the particle phases are used for analysis constraint, and the probability model consists of the following three types of statistical properties:

1) Variable D _G The method is used for representing the phase probability distribution of the aragonite particles, and the calculation formula is as follows:

D _G ＝N _G /N；

wherein N is _G The number of the aragonite phases of each sublayer in each MQVP in the statistical data set is represented, and N represents the number of MQVP samples in the statistical data set;

2) Variable D _W The method is used for representing the phase probability distribution of the pure liquid water particles, and the calculation formula is as follows:

D _W ＝N _W /N；

wherein N is _W The number of pure liquid water particle phase states of each sublayer in each MQVP in the statistical data set is represented: the number of phases of heavy rain particles, the number of phases of rain particles and the number of phases of heavy drops; n represents the number of MQVP samples in the statistical data set;

3) Variable D _S The method is used for representing the phase probability distribution of the remarkable solid water particles, and the calculation formula is as follows:

D _S ＝N _S /N；

wherein N is _S Representing the number of significant solid water particle phases per sublayer in each MQVP in the statistical dataset: the number of hail particle phases + the number of ice crystal particle phases; n represents the number of MQVP samples in the statistical data set;

(2) Construction of asymmetric Gaussian distribution probability model

Method for constructing asymmetric Gaussian distribution probability model PMD formation _ZH 、PMD _KDP 、PMD _ZDR 、PMD _RHV 、PMD _KV The method comprises the steps of carrying out a first treatment on the surface of the In the method for constructing the asymmetric Gaussian distribution probability model, probability distribution of the same layer is centered on the statistical average value of the MQVP, gaussian distribution is between the minimum value on the left side and the average value, and Gaussian distribution is between the maximum value on the right side and the average value;

the calculation steps of the asymmetric Gaussian distribution probability model are as follows: the normalized MQVP in the data set is counted, and the minimum value V of each layer is calculated _MIN Maximum value V _MAX Average value V _AVE The method comprises the steps of carrying out a first treatment on the surface of the At a minimum value V _MIN And average value V _AVE Calculating left side distribution probability P of the layer _L A curve; with average value V _AVE And a maximum value V _MAX Calculating the right side distribution probability P of the layer _R The curve is finally generated into a corresponding MQVP probability model;

in sigma _L Get (V) _AVE -V _MIN )/3，σ _R Get (V) _MAX -V _AVE )/3。

4. The method for automatically discriminating a background of a short-time heavy rain weather according to claim 1 wherein,

(1) Particle phase similarity probability index CP _PS (TH)

PMD-based _PS (TH) probability model, counting each radar data in MQVP early warning data set, calculating probability value of each layer, accumulating layer by layer, and generating particle phase similarity probability index CP _PS (TH); CP corresponding to each radar data of MQVP early warning data set _PS The magnitude of (TH) is used for frequency statistics,selecting a value with a passing rate of more than 90% as a judging threshold value R of the particle phase similarity probability index _PS (TH); wherein, the calculation formula of each layer probability value is:

wherein H represents the total number of layers of the standardized quasi-vertical profile; MQVP _PS (H) Represents MQVP _PS The phase of the particles of the H layer; in the case of the aragonite phase, CP _G D equal to layer H _G Probability value, and CP _W 、CP _S All are 0 probability values; if it is in the form of heavy rain particle phase or heavy drop particle phase, CP _W D equal to layer H _W Probability value, and CP _G 、CP _S All are 0 probability values; in the case of hail particle phase or ice crystal particle phase, CP _S D equal to layer H _S Probability value, and CP _G 、CP _W All are 0 probability values;

(2) Horizontal reflectivity factor similarity probability index CP _ZH (TH)

PMD-based _ZH The (TH) probability model is used for counting each radar data in the MQVP early warning data set, calculating the probability value of each layer, accumulating layer by layer and generating a horizontal reflectivity factor similarity probability index CP _ZH (TH); CP corresponding to each radar data of MQVP early warning data set _ZH Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the similarity probability index of the horizontal reflectivity factor _ZH (TH); wherein, the calculation formula of each layer probability value is:

(3) Propagation phase shift rate similarity probability index CP _KDP (TH)

PMD-based _KDP (TH) probability model, providing MQVPEach radar data in the front early warning data set is counted, probability values of each layer are calculated, accumulation is carried out layer by layer, and a propagation phase shift rate similarity probability index CP is generated _KDP (TH); CP corresponding to each radar data of MQVP early warning data set _KDP Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the propagation phase shift rate similarity probability index _KDP (TH); wherein, the calculation formula of each layer probability value is:

(4) Differential reflectivity similarity probability index CP _ZDR (TH)

PMD-based _ZDR (TH) probability model, counting each radar data in MQVP early warning data set, calculating probability value of each layer, accumulating layer by layer, and generating differential reflectivity similarity probability index CP _ZDR (TH); CP corresponding to each radar data of MQVP early warning data set _ZDR Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the differential reflectivity similarity probability index _ZDR (TH); wherein, the calculation formula of each layer probability value is:

(5) Correlation coefficient similarity probability index CP _RHV (TH)

PMD-based _RHV (TH) probability model, counting each radar data in MQVP early warning data set, calculating probability value of each layer, accumulating layer by layer, generating correlation coefficient similarity probability index CP _RHV (TH); CP corresponding to each radar data of MQVP early warning data set _RHV Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judging threshold value R of the similarity probability index of the correlation coefficient _RHV (TH); wherein, the calculation formula of each layer probability value is:

(6) Polarized condensate flux similarity probability index CP _KV (TH)

PMD-based _KV (TH) probability model, counting each radar data in MQVP early warning data set, calculating probability value of each layer, accumulating layer by layer, and generating polarized condensate flux similarity probability index CP _KV (TH); CP corresponding to each radar data of MQVP early warning data set _KV Counting the frequency of the value of (TH), and selecting the value with the passing rate of more than 90% as the judgment threshold value R of the polarized condensate flux similarity probability index _KV (TH); wherein, the calculation formula of each layer probability value is:

5. a short-time heavy rain weather background automatic discriminating system applying the short-time heavy rain weather background automatic discriminating method according to any one of claims 1 to 4, characterized in that the short-time heavy rain weather background automatic discriminating system comprises:

The data set construction module is used for constructing MQVP corresponding to each double-polarization radar time by utilizing historical process case data of short-time storm, so as to construct an MQVP data set of short-time storm weather background;

the MQVP probability model building module is used for building an MQVP probability model for representing the short-time stormwater weather background according to the MQVP data set of the short-time stormwater weather background;

and the short-time stormwater weather background judging module is used for constructing an MQVP early warning data set, calculating an MQVP comprehensive similarity probability index CP (TH), and judging the short-time stormwater weather background according to the MQVP probability model.

6. A computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the short-term stormwater weather background automatic discriminating method as claimed in any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of the short-time stormwater weather background automatic discriminating method as claimed in any one of claims 1 to 4.

8. An information data processing terminal, characterized in that the information data processing terminal is used for realizing the automatic discriminating system for short-time heavy rain weather background according to claim 5.