CN113447931A - Short-time strong precipitation identification method based on Doppler radar data - Google Patents
Short-time strong precipitation identification method based on Doppler radar data Download PDFInfo
- Publication number
- CN113447931A CN113447931A CN202110648016.3A CN202110648016A CN113447931A CN 113447931 A CN113447931 A CN 113447931A CN 202110648016 A CN202110648016 A CN 202110648016A CN 113447931 A CN113447931 A CN 113447931A
- Authority
- CN
- China
- Prior art keywords
- reflectivity
- monomer
- short
- heavy precipitation
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001556 precipitation Methods 0.000 title claims abstract description 120
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000012549 training Methods 0.000 claims abstract description 20
- 238000012360 testing method Methods 0.000 claims abstract description 15
- 239000000178 monomer Substances 0.000 claims description 137
- 238000002310 reflectometry Methods 0.000 claims description 90
- 239000007788 liquid Substances 0.000 claims description 22
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 22
- 238000000528 statistical test Methods 0.000 claims description 4
- 230000001186 cumulative effect Effects 0.000 claims 2
- 238000002474 experimental method Methods 0.000 abstract description 3
- 238000012353 t test Methods 0.000 abstract description 2
- 238000000692 Student's t-test Methods 0.000 abstract 1
- 230000008569 process Effects 0.000 description 8
- 238000010801 machine learning Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/95—Radar or analogous systems specially adapted for specific applications for meteorological use
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/14—Rainfall or precipitation gauges
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Environmental & Geological Engineering (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Electromagnetism (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Physics & Mathematics (AREA)
- Hydrology & Water Resources (AREA)
- Atmospheric Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
本发明公开了一种基于多普勒雷达数据的短时强降水识别方法，包括：收集短时强降水实况信息以及与之相匹配的雷达数据，将雷达数据转换为三维格点数据；识别出所有时刻的对流单体，与相应的实况信息进行匹配并标记；将标记完成的强降水对流单体记为正样本，非强降水单体记为负样本，提取所有单体的特征，通过统计学t检验选择有效特征；将数据集分为训练集和测试集，根据训练集正负样本的有效特征，训练一个分类器模型，应用该分类器模型进行短时强降水对流单体的识别。本发明实现了短时强降水对流单体和非强降水单体的分类，能够识别出产生短时强降水的对流单体，并通过实验验证了本方法的有效性。
The invention discloses a method for identifying short-term heavy precipitation based on Doppler radar data. The convective cells at all times are matched and marked with the corresponding live information; the marked heavy precipitation convective cells are recorded as positive samples, and the non-heavy precipitation cells are recorded as negative samples, the characteristics of all cells are extracted, and the statistical Learn t-test to select effective features; divide the data set into training set and test set, train a classifier model according to the effective features of positive and negative samples in the training set, and apply the classifier model to identify convective cells with short-term heavy precipitation. The invention realizes the classification of short-term heavy precipitation convective cells and non-heavy precipitation cells, can identify convective cells that produce short-term heavy precipitation, and verifies the effectiveness of the method through experiments.
Description
Technical Field
The invention relates to the field of meteorology and machine learning, in particular to a method for identifying a convection monomer generating short-time strong precipitation by using Doppler weather radar data.
Background
The short-time strong rainfall means that the rainfall is large in 1h20mm or 3h rainfall is more than or equal to 50mm precipitation event[1]It has the characteristics of strong burst property, strong hour rain, strong disaster causing property and the like. The short-time strong rainfall can form flood bursts in a short time, cause urban waterlogging and farmland flooding, even possibly cause geological disasters such as torrential flood, debris flow and the like, and cause great economic loss and casualties to the local[2]Therefore, the method has important significance for strengthening the research on the short-time strong precipitation identification method and the proximity monitoring early warning.
With the rapid development of the doppler weather radar technology and the continuous expansion of the business application, the weather radar has become one of the mainstream weather research tools at present, and the new generation of doppler weather radar in China is laid out and plays an important role in strong convection weather monitoring and early warning. Some scholars discuss the mechanism of the strong short-time precipitation by analyzing radar features of the short-time strong precipitation, such as radar echo intensity, medium and small-scale wind speed shear, cyclone convergence and the like, and provide a theoretical basis for identification of the short-time strong precipitation. The method is characterized in that a learner analyzes some characteristics of a hail convection monomer by using Doppler weather radar data, such as radar echo morphological structure, monomer nuclear height, reflectivity intensity and the like, in order to improve the recognition rate of the hail monomer and reduce the empty reporting rate, an effective method takes hail as a positive example and strong short-time rainfall as a counter example, and the characteristics are used for training a classifier to recognize the hail convection monomer and the strong rainfall convection monomer[3][4]。
In the process of implementing the invention, the inventor finds that at least the following disadvantages and shortcomings exist in the prior art:
at present often with the hail as the contrast in the identification method of short-time heavy precipitation, mainly utilize the radar characteristic of hail, discern the short-time heavy precipitation as the counterexample of hail, do not consider the characteristic of short-time heavy precipitation convection current monomer itself and the difference between short-time heavy precipitation and the non-strong convection weather, consequently, can be mixed with non-strong precipitation monomer type in the monomer of the convection current monomer through hail-short-time heavy precipitation identification model discernment for non-hail to the input, the event has the space of improvement based on the short-time heavy precipitation identification method of Doppler weather radar.
[ reference documents ]
[1] Shu Xiaoding, the idea and method of short-term strong precipitation approach forecast [ J ] storm disaster, 2013,32(03): 202-.
[2] Haoyingying, Yaoye green, Zheng Tquality, Roujun, short-time heavy precipitation multi-scale analysis and proximity early warning [ J ] Meteorological, 2012,38(08):903 + 912.
[3] Wangping, Pangzhou. hail identification model [ J ] based on significant features, physical bulletin, 2013,62(06): 515-.
[4] Shijunzhi, an intelligent approach forecast method for hail and rainstorm based on weather radar research [ D ]. Tianjin university, 2020.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a short-time strong precipitation identification method based on Doppler weather radar data, which solves the problem that the radar characteristics of hail are mainly utilized, the short-time strong precipitation is identified as an example of the hail, the characteristics of short-time strong precipitation convection monomers and the difference between the short-time strong precipitation and non-strong convection weather are not considered, and therefore, the type of the non-strong precipitation monomers can be mixed in the input convection monomers which are identified as the non-hail through a hail-short-time strong precipitation identification model. The method can detect the convection monomer generating strong short-time rainfall in the non-hail monomer by using Doppler weather radar data and a machine learning algorithm, and can perform early warning on disasters in time. The technical scheme of the invention is described in detail as follows:
the invention provides a short-time heavy precipitation identification method based on Doppler radar data, which comprises the following steps of:
step one, collecting short-time strong precipitation live information and Doppler weather radar data matched with the short-time strong precipitation live information, and converting the radar data into three-dimensional grid point data;
the short-time strong precipitation live information comprises precipitation starting time, precipitation ending time, an automatic station number, and 5-minute accumulated precipitation of the automatic station from the starting time to the precipitation ending time.
The step of converting the radar data into the three-dimensional lattice point data refers to the step of performing bilinear interpolation operation on the reflectivity data of 9 elevation angles of the radar to obtain 512 x 31 three-dimensional lattice point data, wherein the data resolution is 1km x 0.5 km.
Step two, identifying convection monomers at all moments from the radar data of each short-time strong precipitation event, matching the convection monomers with corresponding live information and marking the convection monomers;
the steps of matching and marking the convection current monomer with the live information are as follows:
calculating the total rainfall accumulated by each mobile station for one hour from the current moment according to the original live information, and using the total rainfall as a type label of a relevant monomer of each mobile station at the current moment;
the types of monomers include heavy precipitation convection monomers and non-heavy precipitation convection monomers.
For the convection monomer at each moment, recording the positions and the hourly rainfall of all automatic stations corresponding to the same or the nearest moment in the live information, and marking the strong precipitation convection monomer and the non-strong precipitation convection monomer according to the following rules:
rule 1-1: recording the position of the automatic station in the range of the monomer area, and marking the monomer as a strong precipitation monomer if the hourly rainfall of the automatic station is greater than or equal to 20 mm;
rule 1-2: the upper limit of the non-precipitation monomer is set to 18mm/h, considering that the automation station may have large or missing errors in the recorded data due to temporary instrument failure, and that the public is not sensitive to the difference between the 20mm/h threshold and the 19mm/h threshold of the short-term precipitation. Therefore, the single cells which are within the single cell area and have the hour rainfall less than 18mm are marked as non-strong rainfall single cells;
rules 1-3: the rainfall amount of the sample is the maximum value in all automatic station hour rainfall amount records which meet the conditions.
Rules 1-4: samples of radar-based data corruption are deleted.
Recording the strong precipitation convection monomer as a positive sample, recording the non-strong precipitation monomer as a negative sample, and extracting the characteristics of all monomers;
the characteristics of the monomer include a reflectivity density characteristic, a reflectivity intensity characteristic, a reflectivity gradient characteristic, a distance characteristic and a liquid water content characteristic.
The reflectivity density type characteristics are obtained from radar three-dimensional lattice point data and comprise single 30dBZ reflectivity density, single 40dBZ reflectivity density and space single reflectivity 40dBZ ratio;
the reflectivity intensity class characteristics are obtained from radar combined reflectivity data and comprise a monomer combined reflectivity mean value, a 90% quantile of the monomer maximum reflectivity intensity, an 85% quantile of the monomer maximum reflectivity intensity, an 80% quantile of the monomer maximum reflectivity intensity, a proportion of the monomer reflectivity being more than 40dBZ and a proportion of the monomer reflectivity being more than 45 dBZ;
the reflectivity gradient characteristics are obtained from radar combined reflectivity data and comprise reflectivity gradient _ th1, reflectivity gradient _ th2 and reflectivity gradient _ th 3;
the distance type characteristics are obtained from radar combined reflectivity data and comprise a monomer core point monomer 30dBZ contour line average distance and a monomer core point monomer 40dBZ contour line average distance;
the liquid water content characteristic is obtained from radar three-dimensional lattice point data and comprises area vertical accumulated liquid water content, liquid water content density _1 and liquid water content density _ 2.
Step four, taking each feature of all positive and negative sample feature sets as a group of input, respectively carrying out statistical test, taking the feature as an original hypothesis that the features have no significant difference on the respective population, taking the significant difference as a candidate hypothesis, and defining the statistical quantity obeying t distribution as:
in the formula, x1、x2Respectively the mean of the sample features from the two populations,
is the corresponding variance, n1And n2Tests of confidence level (1- α) were developed for both types of samples.The significance level alpha is taken to be 0.01, and the table look-up can obtain ta/2(n1+n2-2) if the value of t of a feature is greater than ta/2(n1+n2-2), the original hypothesis is overridden at a confidence level of "0.99" and the alternative hypothesis is considered to be true. To make the statistical difference of the features on the two sample sets more significant, the value of t is selected to be larger than ta/2(n1+n2-2) as a valid feature of the positive and negative sample sets.
And step five, dividing the data set into a training set and a testing set, training a classifier model according to the effective characteristics of positive and negative samples of the training set, and identifying the short-time heavy precipitation by using the classifier model.
Compared with the prior art, the technical scheme provided by the invention has the beneficial effects that: the method changes the conventional method of generally constructing characteristics aiming at hail convection monomers and identifying the short-time strong precipitation convection monomers as counter examples, constructs the characteristics aiming at the short-time strong precipitation convection monomers, distinguishes the short-time strong precipitation convection monomers and non-strong precipitation convection monomers by utilizing the characteristics, trains a classifier model by combining a machine learning method, realizes the identification of the short-time strong precipitation convection monomers, and verifies the effectiveness of the method through experiments.
Drawings
FIG. 1 is a result of identifying a convective cell on a radar reflectance image;
FIG. 2 is a sample of radar data corruption;
FIG. 3a shows feature y1To y9Schematic representation of the significance analysis of (1);
FIG. 3b shows feature y10To y17Schematic representation of the significance analysis of (1);
FIG. 4 is an example of a partial precipitation process test result;
FIG. 5 is a combined reflectance graph and three-dimensional monolithic example of a missing intense precipitation monolithic.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
The invention provides a short-time heavy precipitation identification method based on Doppler radar data, which comprises the following steps of:
step one, collecting short-time strong precipitation live information and Doppler weather radar data matched with the short-time strong precipitation live information, and converting the radar data into three-dimensional grid point data;
the short-time strong precipitation live information comprises precipitation starting time, precipitation ending time, an automatic station number, and 5-minute accumulated precipitation of the automatic station from the starting time to the precipitation ending time.
The step of converting the radar data into the three-dimensional lattice point data refers to the step of performing bilinear interpolation operation on the reflectivity data of 9 elevation angles of the radar to obtain 512 x 31 three-dimensional lattice point data, wherein the data resolution is 1km x 0.5 km.
Step two, identifying convection monomers at all moments from the radar data of each short-time strong precipitation event, matching the convection monomers with corresponding live information and marking the convection monomers;
the steps of matching and marking the convection current monomer with the live information are as follows:
calculating the total rainfall accumulated by each mobile station for one hour from the current moment according to the original live information, and using the total rainfall as a type label of a relevant monomer of each mobile station at the current moment;
the types of monomers include heavy precipitation convection monomers and non-heavy precipitation convection monomers.
For the convection monomer at each moment, recording the positions and the hourly rainfall of all automatic stations corresponding to the same or the nearest moment in the live information, and marking the strong precipitation convection monomer and the non-strong precipitation convection monomer according to the following rules:
rule 1-1: recording the position of the automatic station in the range of the monomer area, and marking the monomer as a strong precipitation monomer if the hourly rainfall of the automatic station is greater than or equal to 20 mm;
rule 1-2: the upper limit of the non-precipitation monomer is set to 18mm/h, considering that the automation station may have large or missing errors in the recorded data due to temporary instrument failure, and that the public is not sensitive to the difference between the 20mm/h threshold and the 19mm/h threshold of the short-term precipitation. Therefore, the single cells which are within the single cell area and have the hour rainfall less than 18mm are marked as non-strong rainfall single cells;
rules 1-3: the rainfall amount of the sample is the maximum value in all automatic station hour rainfall amount records which meet the conditions.
Rules 1-4: deleting samples of radar-based data corruption, as shown in FIG. 2;
recording the strong precipitation convection monomer as a positive sample, recording the non-strong precipitation monomer as a negative sample, and extracting the characteristics of all monomers;
the characteristics of the monomer include a reflectivity density characteristic, a reflectivity intensity characteristic, a reflectivity gradient characteristic, a distance characteristic and a liquid water content characteristic.
The reflectivity density type characteristics are obtained from radar three-dimensional lattice point data and comprise single 30dBZ reflectivity density, single 40dBZ reflectivity density and space single reflectivity 40dBZ ratio;
the reflectivity intensity class characteristics are obtained from radar combined reflectivity data and comprise a monomer combined reflectivity mean value, a 90% quantile of the monomer maximum reflectivity intensity, an 85% quantile of the monomer maximum reflectivity intensity, an 80% quantile of the monomer maximum reflectivity intensity, a monomer reflectivity 40dBZ ratio and a monomer reflectivity 45dBZ ratio;
the reflectivity gradient characteristics are obtained from radar combined reflectivity data and comprise reflectivity gradient _ th1, reflectivity gradient _ th2 and reflectivity gradient _ th 3;
the distance type characteristics are obtained from radar combined reflectivity data and comprise the average distance between a monomer core point and a monomer 30dBZ contour line and the average distance between the monomer core point and the monomer 40dBZ contour line;
the liquid water content characteristic is obtained from radar three-dimensional lattice point data and comprises a vertically accumulated liquid water content, a liquid water content density _1 and a liquid water content density _ 2.
All the characteristics are shown in table 1.
TABLE 1
Step four, taking each feature of all positive and negative sample feature sets as a group of input, respectively carrying out statistical test, taking the feature as an original hypothesis that the features have no significant difference on the respective population, taking the significant difference as a candidate hypothesis, and defining the statistical quantity obeying t distribution as:
in the formula, x1、x2Respectively the mean of the sample features from the two populations,
is the corresponding variance, n1And n2Tests of confidence level (1- α) were developed for both types of samples. The significance level alpha is taken to be 0.01, and the table look-up can obtain ta/2(n1+n2-2) if the value of t of a feature is greater than ta/2(n1+n2-2), the original hypothesis is overridden at a confidence level of "0.99" and the alternative hypothesis is considered to be true. To make the statistical difference of the features on the two sample sets more significant, the value of t is selected to be larger than ta/2(n1+n2-2) as a valid feature of the positive and negative sample sets.
And step five, dividing the data set into a training set and a testing set, training a classifier model according to the effective characteristics of positive and negative samples of the training set, and identifying the short-time heavy precipitation by using the classifier model.
Experimental example: the implementation of the method is described in detail below with reference to specific experimental data, and the steps are as follows:
1) collecting short-time strong precipitation live information and Doppler weather radar data matched with the short-time strong precipitation live information, and converting the radar data into three-dimensional lattice point data;
the information collected about historical sample data for short-term heavy and non-heavy precipitation is: the rainfall actual condition information of 2018, 2019 and 2020, 2 months to 10 months corresponds to the radar data of the time period, and the reflectivity is converted into three-dimensional lattice point data.
2) Calculating the total rainfall accumulated by each mobile station for one hour from the current moment according to the original live information;
3) identifying the convection current monomers at all the moments, matching the convection current monomers with corresponding live information and marking a sample;
for all data collected in step 1), radar convective cells were identified, and a total of 4792 convective cells were detected. The result of identifying a single convection cell on the radar reflectivity image is shown in fig. 1, where a single convection cell is identified within a rectangular frame.
For each detected convection monomer, recording the positions and the hourly rainfall of all automatic stations corresponding to the same or the latest moment in the live information, and marking the convection monomer with strong rainfall and the monomer with non-strong rainfall according to the following rules:
rule 1-1: recording the position of the automatic station in the range of the monomer area, and marking the monomer as a strong precipitation monomer if the hourly rainfall of the automatic station is greater than or equal to 20 mm;
rule 1-2: the upper limit of the non-precipitation monomer is set to 18mm/h, considering that the automation station may have large or missing errors in the recorded data due to temporary instrument failure, and that the public is not sensitive to the difference between the 20mm/h threshold and the 19mm/h threshold of the short-term precipitation. Therefore, the single cells which are within the single cell area and have the hour rainfall less than 18mm are marked as non-strong rainfall single cells;
rules 1-3: the rainfall amount of the sample is the maximum value in the hourly rainfall records of all the automatic stations meeting the conditions.
Rules 1-4: deleting samples of radar-based data corruption, as shown in FIG. 2;
881 short-time strong precipitation convection monomers and 1228 non-strong precipitation monomers are marked by the rules.
4) Recording short-time strong precipitation convection monomers as positive samples, recording non-strong precipitation monomers as negative samples, and extracting the characteristics of all monomers;
the convective monomer extraction of fig. 1 is characterized by: the monomer 30dBZ reflectivity density is 80.36, the monomer 40dBZ reflectivity density is 91.78, the space monomer reflectivity 40dBZ fraction is 0.13, the monomer combination reflectivity mean is 43.22, the 90% quantile of the maximum reflectivity intensity is 51.76, the 85% quantile of the maximum reflectivity intensity is 50.15, the 80% quantile of the maximum reflectivity intensity is 49.05, the monomer reflectivity 40dBZ fraction is 0.67, the monomer reflectivity 45dBZ fraction is 0.42, the reflectivity gradient _ th1 is 0.62, the reflectivity gradient _ th2 is 0.48, the reflectivity gradient _ th3 is 0.37, the average distance of the monomer core point from the monomer 30dBZ contour line is 0.15, the average distance of the monomer core point from the monomer 40dBZ contour line is 0.12, the vertically accumulated liquid water content is 509.85, the liquid water content density _1 is 6.34, and the liquid water content density _2 is 5.56. As shown in table 2.
TABLE 2
| Feature name | Feature numbering | Characteristic value |
| Monomer 30dBZ reflectance density | y1 | 80.36 |
| Monomer 40dBZ reflectance density | y2 | 91.78 |
| Spatial monomer reflectivity 40dBZ fraction | y3 | 0.13 |
| Mean value of reflectance of monomer combination | y4 | 43.22 |
| 90% quantile of maximum reflectance intensity | y5 | 51.76 |
| 85% quantile of maximum reflectance intensity | y6 | 50.15 |
| 80% quantile of maximum reflectance intensity | y7 | 49.05 |
| Monomer reflectivity 40dBZ fraction | y8 | 0.67 |
| Monomer reflectivity 45dBZ ratio | y9 | 0.42 |
| Reflectivity gradient th1 | y10 | 0.62 |
| Reflectivity gradient th2 | y11 | 0.48 |
| Reflectivity gradient th3 | y12 | 0.37 |
| Average distance between core point of monomer and 30dBZ contour line of monomer | y13 | 0.15 |
| Average distance between core point of monomer and contour line of 40dBZ of monomer | y14 | 0.12 |
| Vertical accumulation of liquid water content | y15 | 509.85 |
| Liquid Water content Density _1 | y16 | 6.34 |
| Liquid Water content Density _2 | y17 | 5.56 |
5) Statistical tests were performed using statistics obeying the t-distribution, defined as:
in the formula, x1、x2Respectively the mean of the sample features from the two populations,
is the corresponding variance, n1And n2Tests of confidence level (1- α) were developed for both types of samples. The significance level alpha is taken to be 0.01, and the table look-up can obtain ta/2(n1+n2-2)＝t0.005(2107)＜t0.005When the t value of a certain feature is greater than 2.576, (∞) — 2.576, the original hypothesis is inverted at a confidence level of "0.99", and the alternative hypothesis is considered to be true. To make the statistical difference of the features on the two sample sets more significant, the value of t is selected to be larger than ta/2(n1+n2-2) features of 5 times or 12.88 as effective features for the positive and negative sample sets.
An example of the significance of each feature in the positive and negative sample feature sets is analyzed. Fig. 3a and 3b show the distribution of all features over the positive and negative examples. The curve with dots in the figure is the positive sample feature distribution and the curve without dots is the negative sample feature distribution. As can be seen from FIG. 3b, feature y13(average distance of monomer core point from monomer 30dBZ contour line) and feature y14(average distance between monomer core point and monomer 40dBZ contour line) has poor discrimination effect on two types of samples, and the result is calculated through statistical t test, and the characteristic y is13Has a t value of 6.05 and a characteristic y14The t value of (a) is 8.06, all other features are clearly distinguished on the two types of samples, and the t value is greater than 12.88, which shows that the statistical difference of the features on the short-time strong precipitation sample set and the non-strong precipitation sample set is obvious enough, so that the feature y is deleted13、y14And other features are retained.
After passing the screen, 15 features were finally retained.
8) The data set is divided into a training set and a testing set, a classifier model is trained according to the effective characteristics of positive and negative samples of the training set, and the classifier model is applied to identify short-time heavy precipitation.
The classifier model adopted in the invention is a support vector machine model with a Gaussian kernel, and the purpose of training is to find out two optimal model parameters. The specific method comprises the following steps: the positive and negative sample sets are divided by 4: 1, dividing the ratio into a training set and a test set, and extracting effective characteristics of each sample on the training set as an input vector of a classifier; then, performing ten-fold cross validation on the training set for training, searching for a C gamma value and a gamma value which enable the classification accuracy of the classifier on the training set to be highest, and respectively obtaining two optimal model parameters: c is 1.48, and gamma is 0.47.
In order to verify the feasibility of the short-time heavy precipitation identification method based on Doppler radar data in meteorology for identifying short-time heavy precipitation, the following test experiments are carried out:
the test set contained 560 samples from 5, 7, 9, 4, 8, 2019 and 3,6 of 2020 with 240 positive samples and 320 negative samples that did not participate in training.
The evaluation indexes of the classifier include a hit rate (POD), a false positive rate (FAR), and a Critical Success Index (CSI) which are calculated by the formula of POD ═ TP/(TP + FN), FAR ═ FP/(TP + FP), and CSI ═ TP/(TP + FN + FP), where TP is the number of positive samples classified into positive samples, FP is the number of negative samples classified into positive samples, and FN is the number of negative samples classified into positive samples. The variation range of the hit rate POD, the empty report rate FAR and the critical success index CSI is 0-1, and the higher the hit rate and the critical success index are, the lower the empty report rate is, and the best performance of the classifier is.
The classifier prediction results are shown in table 3, where TP is 213, FP is 49, and FN is 27.
TABLE 3
From table 3, the performance indexes of the classifiers in the test set are respectively: the POD is 88.75%, the FAR is 18.70% and the CSI is 73.70%, which shows that the classifier can effectively distinguish the short-time strong precipitation convection monomer from the non-strong precipitation monomer, although the air report rate is slightly high, in real life, the short-time strong precipitation has the disaster causing property, and the cost of missing report is FAR higher than that of the air report, so the classifier still has use value in service forecast.
To further illustrate the effect of the classifier, a part of the short-time strong precipitation monomers used for testing are analyzed according to the process, and the arrangement of the processes of part of the strong precipitation is shown in fig. 4, so that the classifier can correctly identify most of the processes, and a small part of the strong precipitation monomers can identify errors at the end of the processes, wherein the 10# strong precipitation process is identified as non-strong precipitation, and the reason for identifying errors can be found mainly by analyzing the reflectivity image and the corresponding three-dimensional monomer structure as shown in fig. 5 and combining the values of all the characteristics is that the monomer space structure in the process is loose, so that the density characteristic value of the reflectivity of the monomer 40dBZ is small, the content of liquid water is small, and the ratio of the reflectivity exceeding 40dBZ in the monomer area is small, so that the classifier identifies the monomer as the non-strong precipitation.
Those skilled in the art will appreciate that the drawings are only schematic illustrations of preferred embodiments, and the above-described embodiments of the present invention are merely provided for description and do not represent the merits of the embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (7)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110648016.3A CN113447931A (en) | 2021-06-10 | 2021-06-10 | Short-time strong precipitation identification method based on Doppler radar data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110648016.3A CN113447931A (en) | 2021-06-10 | 2021-06-10 | Short-time strong precipitation identification method based on Doppler radar data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113447931A true CN113447931A (en) | 2021-09-28 |
Family
ID=77811230
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110648016.3A Pending CN113447931A (en) | 2021-06-10 | 2021-06-10 | Short-time strong precipitation identification method based on Doppler radar data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113447931A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114217319A (en) * | 2022-02-22 | 2022-03-22 | 中国气象局公共气象服务中心(国家预警信息发布中心) | Method and device for correcting weather radar minute rainfall forecast value |
| CN116049726A (en) * | 2023-04-03 | 2023-05-02 | 中国科学技术大学 | Classification method, device, equipment and storage medium of precipitation types on the Qinghai-Tibet Plateau in summer |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS63101790A (en) * | 1986-10-20 | 1988-05-06 | Suga Shikenki Kk | Rainfall display device |
| JP2002070029A (en) * | 2000-08-31 | 2002-03-08 | Dai Nippon Construction | Risk decision of slope and measuring system and safety evaluation method of measuring works |
| CN108665486A (en) * | 2018-03-31 | 2018-10-16 | 天津大学 | A kind of labeling method of strong convection monomer sample |
| CN109164450A (en) * | 2018-09-12 | 2019-01-08 | 天津大学 | A kind of downburst prediction technique based on Doppler Radar Data |
| CN110687618A (en) * | 2019-09-25 | 2020-01-14 | 天津大学 | An automatic nowcasting method for short-term heavy precipitation events in multi-cell convective systems |
| CN112651463A (en) * | 2021-01-05 | 2021-04-13 | 天津大学 | Construction method of double-forecast model of hail weather in plateau area |
-
2021
- 2021-06-10 CN CN202110648016.3A patent/CN113447931A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS63101790A (en) * | 1986-10-20 | 1988-05-06 | Suga Shikenki Kk | Rainfall display device |
| JP2002070029A (en) * | 2000-08-31 | 2002-03-08 | Dai Nippon Construction | Risk decision of slope and measuring system and safety evaluation method of measuring works |
| CN108665486A (en) * | 2018-03-31 | 2018-10-16 | 天津大学 | A kind of labeling method of strong convection monomer sample |
| CN109164450A (en) * | 2018-09-12 | 2019-01-08 | 天津大学 | A kind of downburst prediction technique based on Doppler Radar Data |
| CN110687618A (en) * | 2019-09-25 | 2020-01-14 | 天津大学 | An automatic nowcasting method for short-term heavy precipitation events in multi-cell convective systems |
| CN112651463A (en) * | 2021-01-05 | 2021-04-13 | 天津大学 | Construction method of double-forecast model of hail weather in plateau area |
Non-Patent Citations (7)
| Title |
|---|
| 徐琼芳等: "潜江市中小尺度强降水短临预警指标技术研究", 《湖北农业科学》 * |
| 王萍: "一种描述强对流三维流场的非刚性曲面匹配方法", 《天津大学学报(自然科学与工程技术版)》 * |
| 王萍: "基于多普勒天气雷达数据的中层径向辐合自动识别及其与强对流天气的相关性研究", 《物理学报》 * |
| 王萍等: "50km以内雷暴系统的分类识别方法研究", 《气象》 * |
| 王萍等: "基于显著性特征的大冰雹识别模型", 《物理学报》 * |
| 路志英: "基于深度学习的短时强降水天气识别", 《天津大学学报(自然科学与工程技术版)》 * |
| 陈永仁等: "基于SWAN产品的短时强降水雷达特征及预警分析", 《高原山地气象研究》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114217319A (en) * | 2022-02-22 | 2022-03-22 | 中国气象局公共气象服务中心(国家预警信息发布中心) | Method and device for correcting weather radar minute rainfall forecast value |
| CN116049726A (en) * | 2023-04-03 | 2023-05-02 | 中国科学技术大学 | Classification method, device, equipment and storage medium of precipitation types on the Qinghai-Tibet Plateau in summer |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2022184088A1 (en) | Flood sensitivity risk assessment method based on ensemble learning | |
| CN109765559B (en) | A kind of mountain area disastrous rainstorm recognition methods based on Doppler radar and deep learning | |
| CN103163278B (en) | Atmospheric heavy pollution forecast method based on combination of numerical model and statistic analysis | |
| CN108831115B (en) | Adaboost-based power transmission line rainstorm disaster risk early warning method | |
| Lavaysse et al. | Statistical downscaling of the French Mediterranean climate: assessment for present and projection in an anthropogenic scenario | |
| CN107526083B (en) | A kind of strong convection wind scale prediction technique based on weather radar data | |
| CN113447931A (en) | Short-time strong precipitation identification method based on Doppler radar data | |
| CN115689293B (en) | An assessment method for urban waterlogging resilience based on the stress-state-response framework | |
| CN109164450B (en) | A Downburst Prediction Method Based on Doppler Radar Data | |
| CN109856702A (en) | A kind of division of precipitation Day change type and spatial distribution extracting method based on cluster | |
| CN111814407B (en) | Flood forecasting method based on big data and deep learning | |
| CN105046259A (en) | Coronal mass ejection (CME) detection method based on multi-feature fusion | |
| CN116205136A (en) | Large-scale river basin deep learning flood forecasting method based on runoff lag information | |
| CN115423272A (en) | Flooding risk assessment method and system integrating historical flooding intensity | |
| CN112233381B (en) | A Debris Flow Early Warning Method and System Based on Mechanism and Machine Learning Coupling | |
| CN115358507B (en) | Production construction project disturbance map spot soil erosion risk identification and assessment method | |
| CN112686192B (en) | A Landslide Stability Classification Method Based on Fine Terrain Features | |
| CN115601544A (en) | A Method of Landslide Detection and Segmentation in High Resolution Image | |
| CN113158899B (en) | A method for measuring the development status of villages and towns based on remote sensing night light and dark target enhancement technology | |
| CN114897395A (en) | A Storm Surge Risk Assessment and Disaster Mitigation Method | |
| CN118982234A (en) | A method, device and medium for evaluating and predicting tree obstacle hazards of overhead power transmission lines | |
| Li et al. | Improving multi-model ensemble probabilistic prediction of Yangtze River valley summer rainfall | |
| Wu et al. | A new and efficient method for tropical cyclone detection and tracking in gridded datasets | |
| CN116061953A (en) | A Discrimination and Evaluation Method for Dangerous Driving Behavior of Trucks Based on Trajectory Data | |
| CN113268867A (en) | Glacier debris flow area early warning method based on hydrothermal combination |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20210928 |