A kind of Doppler radar quantitative precipitation estimation method based on Precipitation Clouds identificationTechnical field
The present invention relates to weather radar field of signal processing, specifically a kind of Doppler day based on Precipitation Clouds identificationGas radar quantitative precipitation estimation method.
Background technique
Traditional Doppler radar mostly uses single Z-I method to estimate precipitation and its intensity, failsIn view of the phase information of precipitation particles, estimation result precision fluctuation is larger, and confidence level is lower.
In recent years, arranging net successively with Dual-Polarized Doppler Weather Radar, has been widely used in Quantitative PrecipitationIn estimation.Dual-Polarized Doppler Weather Radar passes through alternate emission or simultaneously emission level and vertically polarized wave, and partially to twoThe signal processing of the echo-signal in vibration direction in different ways, so that reflectivity Z can be obtainedH, Analysis of Differential Reflectivity Factor Measured ZDR, zeroLag correlation coefficient ρHV(0) and difference travel phase constant KDPEtc. multiple polarization parameters.These polarization parameters reflect bigThe features such as the size, shape, phase of precipitation particles and orientation in headroom.It is double inclined compared to traditional Doppler radarThe Doppler radar that shakes utilizes R (ZH)、R(KDP)、R(KDP, ZDR) and R (ZH, ZDR) etc. precipitation intensities equation to precipitationAnd its intensity carries out Combined estimator, has fully taken into account the physical characteristic of precipitation particles, substantially increases the accurate of estimation resultDegree and confidence level.But these methods fail the physical change process in view of precipitation particles, as zero_dynamics system phenomenon causesCertain polarization parameters change, to reduce the precision of final precipitation estimated result.
Summary of the invention
Aiming at the defects existing in the prior art, the technical problem to be solved by the present invention is in existing dual-polarizationOn the basis of Doppler radar quantitative precipitation estimation method, a kind of Doppler radar based on Precipitation Clouds identification is providedQuantitative precipitation estimation method effectively improves the precision of Precipitation estimation.
Present invention technical solution used for the above purpose is: a kind of Doppler weather based on Precipitation Clouds identificationRadar quantitative precipitation estimation method, comprising the following steps:
To the Dual-Polarized Doppler Weather Radar PPI body total number of input according to interpolation processing is carried out, to obtain multiple heightCAPPI data；
The zero_dynamics system in the CAPPI data of the multiple height is identified respectively；
Precipitus classification is carried out to the non-zero-degree layer bright band region in more height CAPPI data respectively, is classified as stratiformCloud and convective cloud two major classes；
Zero_dynamics system is set to stratiform clouds, and the Precipitation Clouds recognition result of more height is merged, obtains Precipitation CloudsDistribution map；
To in Precipitation Clouds distribution map stratiform clouds and convective cloud region, respectively carry out precipitation estimation, completely droppedIrrigation water distribution map.
The interpolation processing uses VHI method interpolation
The CAPPI data of a certain height include that the level of the height penetrates rate factor ZH, Analysis of Differential Reflectivity Factor Measured ZDR, difference passBroadcast phase-shift constant KDPAnd zero-lag correlation coefficient ρHV(0)。
4. a kind of Doppler radar quantitative precipitation estimation side based on Precipitation Clouds identification according to claim 1Method, which is characterized in that the zero_dynamics system in the CAPPI data to the multiple height carries out identification using fuzzy logicMethod.
The non-zero-degree layer bright band region in more height CAPPI data carries out precipitus classification and uses based on CNN'sPrecipitation Clouds method of identification.
It is described that precipitation estimation is carried out to convective cloud region, specifically:
In formula, ZHFor the horizontal reflection rate factor, ZDRFor Analysis of Differential Reflectivity Factor Measured, KDPIndicate difference travel phase-shift constant.
It is described that precipitation estimation is carried out to stratiform clouds region, specifically:
In formula, ZHFor the horizontal reflection rate factor, ZDRFor Analysis of Differential Reflectivity Factor Measured, R is precipitation intensity.
The present invention has the following advantages and beneficial effects:
1, the present invention has fully considered the feature of Precipitation Clouds, to stratiform clouds, convection current compared with current precipitation amount estimation methodCloud is respectively adopted different precipitation amount estimation methods and carries out precipitation estimation, can effectively improve the precision of precipitation estimation.
2, the present invention in advance identifies zero_dynamics system, avoids and missed in the Classification and Identification of Precipitation CloudsIt is judged to convective cloud；Meanwhile using the method for deep learning, the feature ginseng of precipitation cloud-type can be reflected by not needing artificial searchingNumber, such as maximum reflectivity, reflectivity gradient and echo high characteristic parameter, substantially increase the efficiency of judgement.
Detailed description of the invention
Fig. 1 is the flow chart of the method for the present invention；
Fig. 2 is the VHI method interpolation schematic diagram in the method for the present invention；
Fig. 3 is that the zero_dynamics system based on fuzzy logic in the method for the present invention differentiates schematic diagram；
Fig. 4 is the Precipitation Clouds type identification algorithm flow chart based on CNN in the method for the present invention；
Fig. 5 is more height Precipitation Clouds distribution maps in the method for the present invention；
Fig. 6 is the Precipitation Clouds Map of Distributions of Types in the method for the present invention；
Fig. 7 is the precipitation intensity distribution map in the method for the present invention.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and embodiments.
As shown in Figure 1, the method for the present invention is input with Dual-Polarized Doppler Weather Radar PPI body total number evidence, finally with dropIrrigation water distribution map is output, and specific implementation flow is as follows:
Step 1: to the input PPI body total number according to VHI method interpolation is used, Interpolation Principle is as shown in fig.2, Z is reflection in figureRate, R are the space length apart from radar station, and θ is azimuth corresponding to current radial data, and φ is right for current radial dataThe elevation angle answered.To the data point Z (R, θ, φ) of a certain interpolation, by being disposed below elevation angle φiOn data Z (R, θ, φi)、Z(R2, θ, φi) and be positioned above elevation angle φi+1On data Z (R, θ, φi+1)、Z(R1, θ, φi+1) linear interpolation obtains,Interpolation expression is as shown in Equation 1.
In formula 1,WithRespectively Z (R, θ, φi) and Z (R, θ, φi+1) interpolation weights coefficient, expression formula such as 2Shown in formula, andWithRespectively Z (R1, θ, φi+1) and Z (R2, θ, φi) interpolation weights coefficient, expression formula such as 3 formula institutesShow.
By above-mentioned Interpolation Process, the reflectivity CAPPI data of a certain height can be obtained, similarly, can also interpolation obtainThe Analysis of Differential Reflectivity Factor Measured Z of arbitrary heightDR, difference travel phase shift KDPAnd zero-lag correlation coefficient ρHV(0) CAPPI data.In this embodiment, it is highly selected as 2KM, 3KM, 4KM, 5KM, 6KM and 7KM, horizontal resolution is 100 meters.
Step 2: zero_dynamics system is a kind of characteristic feature of Stratiform Cloud Precipitation, sufficiently reflects precipitation particles from solid-stateIt is converted into the physical process of liquid.In this process, solid precipitation particle appearance is wrapped in water membrane, to generate a kind of voidFalse echo information, and there is stronger similitudes for the echo character of the echo character and Convective Cloud Precipitation.After in order to preventZero_dynamics system is mistaken for convective cloud in continuous Precipitation Clouds identification process, precipitation particles exist in article combination zero_dynamics systemZH、ZDR、ρHV(0) and the regularity of distribution on height H, zero_dynamics system is identified using fuzzy logic algorithm.It is fuzzyLogical algorithm schematic diagram is as shown in Figure 3.In this embodiment, it selects subordinating degree function for Beta type subordinating degree function, expressesFormula is as shown in Equation 4, and each to input subordinating degree function parameter setting corresponding to parameter refering to shown in table 1, wherein H is CAPPI data instituteCorresponding vertical height.
In formula (4), X is the input of the algorithm, respectively corresponds ZH、ZDR、ρHV(0) and H data, for different types ofData, using different a, b, m parameter carries out Fuzzy processing to it, and parameter setting is as shown in table 1.At blurringReason, can convert degree of membership for actual detection data, the judgement for zero_dynamics system.Article is the more of 2-7KM to heightThe CAPPI data of a elevation plane have carried out the identification of zero_dynamics system.
1 zero_dynamics system identification parameter table of table
Step 3: precipitus classification being carried out to the non-zero-degree layer bright band region in more height CAPPI data respectively, by its pointFor stratiform clouds and convective cloud two major classes, classification method is the Precipitation Clouds recognition methods based on CNN.Article uses Python+Pytorch development platform has built the Precipitation Clouds identifying system based on CNN, and functional block diagram is refering to shown in Fig. 4.The precipitationCloud identifying system is mainly by two convolutional layers, two residual block layers, two full articulamentums and a Sofatmax classifier structureAt.By a convolution algorithm module Conv2d, (there is important parameter each convolutional layer in convolution algorithm: convolution kernel sizeKernel, step-length Stride and zero padding position Pad) and an activation computing module ReLU composition.Each residual block layer is by oneConvolution algorithm module Conv2d, a standardization computing module InstanceNorm and an activation computing module ReLU groupAt.Full articulamentum mainly converts a feature vector for multiple special systems of battle formations, for the classification of Softmax classifier, finallyRealize the differentiation to precipitation cloud-type.
In this embodiment, using reflectivity ZHDistinguishing rule of the CAPPI data as precipitation cloud-type, system is defeatedEnter the Z for 32 × 32HMatrix, input matrix pass sequentially through the fortune of convolutional layer, two residual block layers and another convolutional layerIt calculates, finally obtains 16 3 × 3 characteristic patterns.Then, system is converted 16 3 × 3 characteristic patterns to using two full articulamentums1 × 64 feature vector finally obtains the type of Precipitation Clouds for the classification of Softmax classifier.To 2-7KM height in textNon-zero-degree layer bright band region in interior multiple CAPPI data has carried out the identification of precipitation cloud-type, and recognition result is refering to Fig. 5 instituteShow.
Step 4: zero_dynamics system being set to stratiform clouds, and the Precipitation Clouds recognition result of more height is merged, is obtainedMore complete Precipitation Clouds distribution map, final fusion results are refering to shown in Fig. 6.
Step 5: in Precipitation Clouds distribution map stratiform clouds and convective cloud region, different precipitation estimation sides is respectively adoptedMethod carries out precipitation estimation, finally obtains complete Rainfall distribution figure.In this embodiment, 2 kilometers of height are selectedCAPPI data estimate precipitation, wherein Convective Cloud Precipitation amount estimation method is Stratiform Cloud Precipitation amount shown in formula (5)Estimation method is shown in formula (6), and final precipitation estimation result is refering to shown in Fig. 7.
Z in articleH、KDP、ZDR、ρHV(0) etc. parameters are all the data that dual-polarization weather radar detects, wherein ZHFor waterThe flat fire rate factor；ZDRFor Analysis of Differential Reflectivity Factor Measured；KDPFor difference travel phase-shift constant；ρHVIt (0) is zero-lag correlation coefficient.TableUp in formula 3, R is precipitation intensity, can not directly obtain, need by ZH、ZDRAnd KDPEqual polarization parameters data estimate itMeter, here according to ZH、ZDR、KDPValue range it is different, R is estimated using different equatioies.It is largely counted and shows benefitWhen being estimated respectively precipitation intensity R corresponding to different type Precipitation Clouds with 3 formulas and 4 formulas, estimation result is more accurate.
In this way by being carried out to stratiform clouds and the first Classification and Identification of convective cloud progress, then using different precipitation amount estimation methodsPrecipitation estimation, can effectively improve the precision of precipitation estimation.