Method and system for tropical cyclone precipitation forecast diagnostic evaluation

Abstract

The application provides a tropical cyclone precipitation forecast diagnosis evaluation method and system, the method comprises the following steps: sequentially eliminating the position, magnitude and structure error of the forecast precipitation field in steps, and calculating the evaluation index promotion quantity after each correction, and the error specifically comprises the spatial deviation related to the path of the tropical cyclone precipitation forecast, the magnitude deviation related to the model systematic deviation and the intensity of the tropical cyclone, the radial distribution deviation related to the radial structure of the model precipitation, and the azimuth deviation related to the precipitation asymmetry. The tropical cyclone precipitation forecast error is decomposed into the errors related to the tropical cyclone path, precipitation magnitude, precipitation radial distribution and precipitation asymmetric structure prediction, the source of the tropical cyclone precipitation forecast error is better explained, the error contribution after each step correction can better reflect the contribution degree of the error source related to the tropical cyclone itself attribute or the environment field to the total precipitation error, and the correction reference is provided for real-time tropical cyclone precipitation forecast.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of meteorological research, and in particular to a tropical cyclone precipitation forecast diagnosis and evaluation method and system. BACKGROUND

[0002] Tropical cyclone quantitative precipitation forecast is very important for protecting life and property safety and effectively utilizing water resources. At present, the tropical cyclone quantitative precipitation forecast mainly relies on numerical models. Objectively evaluating the forecast ability of the model and clearly diagnosing the error source of the model forecast is of great significance for using and improving the model forecast.

[0003] For the verification of tropical cyclone quantitative precipitation forecast, the business generally uses the TS (Threat Score) score based on point-to-point or the CRA (Contiguous Rain Area) or MODE (Mode-based Object Diagnostic Assessment) verification method based on objects. These methods do not consider the correlation between the tropical cyclone precipitation distribution and the properties such as the position and intensity of the tropical cyclone at the beginning of the design, and the physical meaning is not clear enough, which has certain limitations.

[0004] Therefore, it is necessary to design a tropical cyclone precipitation forecast diagnosis and evaluation method, which can decompose and diagnose the tropical cyclone precipitation forecast error, and associate the precipitation forecast error with the forecast error of the properties of the tropical cyclone itself, so as to better explain the source of the precipitation forecast error. SUMMARY

[0005] In view of the problems existing in the prior art, the present application provides a tropical cyclone precipitation forecast diagnosis and evaluation method and system, which can calculate the contribution of the tropical cyclone path forecast error, the systematic deviation of the model precipitation intensity, the radial precipitation structure error and the asymmetric precipitation structure error to the total precipitation forecast error. The method can provide more rich diagnostic information for tropical cyclone precipitation forecast evaluation, and is helpful for forecasters to interpret the forecast and for model developers to improve the model.

[0006] The present application provides a tropical cyclone precipitation forecast diagnosis and evaluation method, which comprises:

[0007] According to the observed tropical cyclone path, the observed precipitation field and the numerical model predicted precipitation field are cut respectively to obtain the observed precipitation field and the predicted precipitation field .

[0008] According to the offset between the numerical model predicted tropical cyclone center and the observed tropical cyclone center, the predicted precipitation field translation, the forecast precipitation field after translation is extracted with the tropical cyclone center as the center and R as the radius to obtain a forecast precipitation revised field after path deviation removal ;

[0009] The forecast precipitation revised field is adjusted based on the quantile mapping method to adjust the precipitation magnitude of the forecast precipitation revised field , so that the adjusted precipitation magnitude distribution of the forecast precipitation revised field is consistent with the precipitation magnitude distribution of the observed precipitation field , to obtain a forecast precipitation revised field after model magnitude deviation removal . ;

[0010] The observed precipitation field in the polar coordinate system and the forecast precipitation revised field are respectively subjected to azimuth angle averaging to obtain radial precipitation distribution profiles, and the forecast precipitation revised field is adjusted according to the ratio of the radial precipitation distribution profiles of the two, to obtain a forecast precipitation revised field . , which is interpolated back to the original latitude-longitude grid coordinates to obtain a radial distribution-adjusted forecast precipitation revised field .

[0011] The azimuth 1 wave asymmetric components of the observed precipitation field in the polar coordinate system and the forecast precipitation revised field are respectively calculated, and the forecast precipitation revised field is rotated according to the offset angle between the azimuth angles at which the maximum values of the azimuth 1 wave asymmetric components of the two are located, to obtain a forecast precipitation revised field . , which is interpolated back to the original latitude-longitude grid coordinates to obtain an azimuth angle asymmetric structure-adjusted forecast precipitation revised field .

[0012] The evaluation indexes of the forecast precipitation field and the forecast precipitation revised field after each revision are calculated , , , and , and the contribution value of the associated factor of each revision to the precipitation forecast error is obtained according to the change amount of the evaluation index after each revision.

[0013] According to the tropical cyclone precipitation forecast diagnosis and evaluation method provided by the application, the forecast precipitation revised field is adjusted according to the ratio of the radial precipitation distribution profiles of the two, to obtain a forecast precipitation revised field . , which comprises:

[0014] The radial precipitation distribution profile of the observed precipitation field in the polar coordinate system and the forecast precipitation revised field are calculated .​​​​ a ratio between the radial precipitation distribution profiles of the observed precipitation field and the forecast precipitation field;

[0015] the forecast precipitation correction field in the polar coordinate system is multiplied by the ratio to obtain a forecast precipitation correction field . .

[0016] According to the tropical cyclone precipitation forecast diagnostic evaluation method provided by the application, before the observed precipitation field in the polar coordinate system and the forecast precipitation correction field are respectively azimuthally averaged to obtain the radial precipitation distribution profiles, the method further comprises the following steps:

[0017] The observed precipitation field and the forecast precipitation correction field are interpolated into the polar coordinate system with the observed tropical cyclone center as the origin and R as the radius, and R is 500 km or is set according to the observed precipitation distribution range. According to the tropical cyclone precipitation forecast diagnostic evaluation method provided by the application, the radial resolution of the polar coordinate system is the same as the grid spacing of the precipitation field at the origin and gradually increases along the radial direction to the periphery.

[0018] According to the tropical cyclone precipitation forecast diagnostic evaluation method provided by the application, the evaluation indexes of the forecast precipitation field and the forecast precipitation correction field after each correction are calculated, and the contribution value of the associated factor of each correction to the precipitation forecast error is obtained according to the change amount of the evaluation index after each correction, comprising:

[0019] According to the tropical cyclone precipitation forecast diagnostic evaluation method provided by the application, the evaluation indexes of the forecast precipitation field and the forecast precipitation correction field after each correction are calculated, and the contribution value of the associated factor of each correction to the precipitation forecast error is obtained according to the change amount of the evaluation index after each correction, comprising:

[0020] The evaluation index SSIM score between the forecast precipitation field and the observed precipitation field is calculated.

[0021] The evaluation index SSIM score between the forecast precipitation correction field and the observed precipitation field is calculated.

[0022] The evaluation index SSIM score between the forecast precipitation correction field and the observed precipitation field is calculated.

[0023] The evaluation index SSIM score between the forecast precipitation correction field and the observed precipitation field is calculated. ​​​​​​​​​​​​​​​​​​​;

[0024] computing the evaluation index SSIM score between the predicted precipitation field and the observed precipitation field ; ;

[0025] taking the difference between the predicted precipitation field and the observed precipitation field as the overall error ; taking the proportion of the difference between the predicted precipitation field and the observed precipitation field in the overall error as the contribution value corresponding to the i-th correction, i=1, 2, 3 and 4.

[0026] The tropical cyclone precipitation prediction diagnosis evaluation method provided by the application further comprises:

[0027] computing the difference between the predicted precipitation field and the observed precipitation field as the overall error ; taking the proportion of the difference between the predicted precipitation field and the observed precipitation field in the overall error as the cumulative contribution value corresponding to the i-th correction.

[0028] The tropical cyclone precipitation prediction diagnosis evaluation method provided by the application further comprises:

[0029] taking the difference between 1 and the predicted precipitation field and the observed precipitation field as the unexplained residual error.

[0030] The application further provides a tropical cyclone precipitation prediction diagnosis evaluation system, comprising:

[0031] an initial scoring module configured to crop the observed precipitation field and the precipitation field predicted by the numerical model according to the observed tropical cyclone path with R as the radius, to obtain the observed precipitation field and the predicted precipitation field ; ; ;

[0032] a first correction module configured to perform translation on the predicted precipitation field according to the offset between the numerical model predicted tropical cyclone center and the observed tropical cyclone center, and extract precipitation from the translated predicted precipitation field with the observed tropical cyclone center as the center and R as the radius, to obtain the predicted precipitation correction field after removing the path deviation ; ;

[0033] a second correction module configured to adjust the precipitation magnitude of the predicted precipitation correction field based on the quantile mapping method, so that the adjusted precipitation magnitude distribution of the predicted precipitation correction field is consistent with the precipitation magnitude distribution of the observed precipitation field, to obtain the predicted precipitation correction field after removing the model magnitude deviation ; ; ;

[0034] ​​​​​The third correction module is used to correct the observed precipitation field in polar coordinates. and Forecasted Precipitation Correction Field Radial precipitation profiles are obtained by azimuth averaging. The ratio of the two radial precipitation profiles is used to correct the forecast precipitation field. Adjustments were made to obtain the revised precipitation forecast field. Interpolate it back to the original latitude and longitude grid coordinates to obtain the radially distributed adjusted forecast correction field. ;

[0035] The fourth correction module is used to calculate the observed precipitation field in polar coordinates. and Forecasted Precipitation Correction Field The azimuth 1-wave asymmetric component, and the offset angle between the azimuth angles where the maximum values ​​of the azimuth 1-wave asymmetric components of the two components are located in each radial direction, are used to correct the forecast precipitation field. Rotate to obtain the forecast precipitation correction field. Interpolate it back to the original latitude and longitude grid coordinates to obtain the forecast precipitation correction field with azimuth asymmetric structure adjustment. ;

[0036] The evaluation module is used to calculate the forecast precipitation field. And the revised precipitation forecast field after each revision , , and The evaluation indicators are used to determine the contribution of factors associated with each correction to the precipitation forecast error based on the changes in the evaluation indicators after each correction.

[0037] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the tropical cyclone precipitation forecasting, diagnosis and evaluation method as described above.

[0038] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the tropical cyclone precipitation forecasting, diagnosis and evaluation method as described above.

[0039] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the tropical cyclone precipitation forecasting, diagnosis and evaluation method as described above.

[0040] The application provides a tropical cyclone precipitation forecast diagnosis evaluation method and system, which decomposes tropical cyclone precipitation forecast error into errors associated with tropical cyclone path, precipitation order, precipitation radial distribution and precipitation asymmetric structure prediction, better explains the source of the tropical cyclone precipitation forecast error, and has a more explicit physical meaning compared with previous methods. Through actual implementation, it can be found that the precipitation field corrected step by step has high similarity with the observation field, the error contribution calculated after each correction process can better reflect the contribution degree of the error source related to the properties of the tropical cyclone or the environment field to the total precipitation error, and can provide correction reference for real-time tropical cyclone precipitation prediction. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can also be obtained by those skilled in the art without any creative effort.

[0042] Figure 1 It is a flowchart of the tropical cyclone precipitation forecast diagnosis evaluation method provided by the application;

[0043] Figure 2 It is a complete flowchart of the tropical cyclone precipitation forecast diagnosis evaluation method provided by the application;

[0044] Figure 3 It is a schematic diagram of the observation and forecast precipitation field in each step of the tropical cyclone precipitation forecast diagnosis evaluation method provided by the application, wherein part a is the original observation precipitation field , part b is the original forecast precipitation field , part c is the forecast precipitation correction field after eliminating the path error without clipping, part d is the clipped observation precipitation field , part e is the clipped forecast precipitation field , part f is the forecast precipitation correction field after eliminating the path error , part g is the precipitation correction field after eliminating the precipitation order deviation , part h is the precipitation correction field after eliminating the radial distribution error , part i is the precipitation correction field after eliminating the precipitation azimuth error , and the plus sign indicates the position of the tropical cyclone center.

[0045] Figure 4 It is a schematic diagram of the quantile mapping model established by the observation precipitation field and the forecast precipitation correction field provided by the application, which is applied to the forecast precipitation quantity. the revised curve after the correction, wherein the dotted line is the predicted revised precipitation and the mapping relationship of the predicted precipitation , the solid line is the standard unbiased line;

[0046] Figure 5 is the radial distribution profile of the observed and predicted axisymmetric precipitation obtained in the tropical cyclone precipitation prediction diagnosis and evaluation method provided by the application (solid line) and (dotted line), and the radial distribution profile ratio of the observed and predicted azimuthally averaged fields (dot-dash line) schematic diagram;

[0047] Figure 6 is the schematic diagram of the corrected asymmetric precipitation error under the polar coordinates in the tropical cyclone precipitation prediction diagnosis and evaluation method provided by the application, wherein part a is the observed azimuthal 1-wave asymmetric component field , part b is the predicted azimuthal 1-wave asymmetric component field , part c is the predicted azimuthal 1-wave asymmetric component field corrected for the asymmetric precipitation error, part d is the observed precipitation field , part e is the predicted precipitation corrected field , and part f is the precipitation prediction field corrected for the asymmetric precipitation error ;

[0048] Figure 7 is the structural schematic diagram of the tropical cyclone precipitation prediction diagnosis and evaluation system provided by the application. DETAILED DESCRIPTION

[0049] To make the objectives, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely below with reference to the drawings in the present application. Obviously, the described embodiments are some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the protection scope of the present application.

[0050] A tropical cyclone precipitation prediction diagnosis and evaluation method provided by the present application will be described below with reference to Figure 1 and Figure 2 , which comprises:

[0051] Step 101, according to the observed tropical cyclone path, the observed precipitation field and the precipitation field predicted by the numerical model are respectively cut with R as the radius, to obtain the observed precipitation field and the predicted precipitation field ;

[0052] This embodiment evaluates and diagnoses errors in tropical cyclone precipitation forecasts for a specific time period. A cumulative precipitation duration of 1 hour or 3 hours is preferred. If the user requires results for a longer time period, they can first process the 3-hour data and then overlay it onto the desired timeframe.

[0053] First, tropical cyclone data for a specific time period is acquired, including numerical model forecasts, cumulative precipitation observations, and the tropical cyclone positions at the start and end of that period. The observed cumulative precipitation data is interpolated to the same grid as the forecasts. The GFDL vortex tracking algorithm is used to obtain the model-predicted tropical cyclone paths, and the average observed and predicted tropical cyclone positions for that time period are calculated.

[0054] Observed and forecasted precipitation data were cropped separately based on the observed tropical cyclone path. The observed precipitation field... and forecast precipitation field Precipitation was extracted using the observed tropical cyclone location as the center and R as the radius, resulting in the observed precipitation field. and forecast precipitation field .

[0055] The specific method for extracting precipitation fields is as follows: if the distance between a grid point and the center of a tropical cyclone does not exceed the radius R, then the precipitation at that grid point is retained; otherwise, it is set as a missing value.

[0056] Preferably, the location of the tropical cyclone can be obtained from the best track data of tropical cyclones IBTRASC (International BestTrack Archive for Climate Stewardship), and the gridded precipitation data can be obtained from the Global Precipitation Measurement (GPM) program, with R being 500 km or set according to the observed precipitation distribution range.

[0057] Step 102: Based on the offset between the tropical cyclone center predicted by the numerical model and the observed tropical cyclone center, adjust the predicted precipitation field. The forecast precipitation field is shifted, and precipitation is extracted from the shifted forecast precipitation field with the observed tropical cyclone center as the center and R as the radius, to obtain the path bias-removed forecast precipitation correction field. ;

[0058] The specific method for shifting the forecast precipitation field is as follows: calculate the offset of the forecast tropical cyclone center relative to the observed tropical cyclone center, spatially shift the latitude and longitude of the forecast precipitation field grid according to the offset, and interpolate the shifted precipitation field to the original grid to ensure the consistency of the verification grid.

[0059] Step 103: Adjust the forecast precipitation correction field based on the quantile mapping method the precipitation magnitude of the forecast precipitation adjustment field is consistent with the observed precipitation field, and the forecast precipitation adjustment field after removing the bias of the magnitude of the model is obtained ;

[0060] Based on the quantile mapping method, the precipitation magnitude of the forecast precipitation adjustment field is adjusted to the magnitude distribution of the observation, and the forecast precipitation adjustment field after removing the bias of the magnitude of the model is obtained . The use of the quantile mapping method eliminates the bias of the magnitude of the tropical cyclone precipitation and calculates the evaluation index improvement amount after adjusting the magnitude bias.

[0061] Step 104, the observed precipitation field in the polar coordinate system and the forecast precipitation adjustment field are respectively azimuthally averaged to obtain radial precipitation distribution profiles, and the forecast precipitation adjustment field is adjusted according to the ratio of the radial precipitation distribution profiles of the two, to obtain the forecast precipitation adjustment field , which is interpolated back to the original latitude and longitude grid coordinates to obtain the radial distribution adjusted forecast precipitation adjustment field ;

[0062] The forecast precipitation adjustment field and the observation field are established in a polar coordinate system with the center of the observed tropical cyclone as the origin. The radial resolution of the polar coordinate system is set to be the same as the grid resolution of the forecast field at the origin, and gradually transitions to a coarse resolution along the radial direction to the periphery, to obtain the forecast precipitation adjustment field and the observation field in the polar coordinate system. Preferably, the azimuthal resolution of the polar coordinate system is set to 2.5 degrees.

[0063] Step 105, the azimuthal 1-wave asymmetric components of the observed precipitation field and the forecast precipitation adjustment field in the polar coordinate system are respectively calculated, and the forecast precipitation adjustment field is rotated according to the offset angle between the azimuth angles at which the maximum values of the azimuthal 1-wave asymmetric components of the two in each radial direction are located, to obtain the forecast precipitation adjustment field , which is interpolated back to the original latitude and longitude grid coordinates to obtain the azimuthally asymmetrically structured adjusted forecast precipitation adjustment field ;

[0064] The observed precipitation field and the forecast precipitation adjustment field in the polar coordinate system are converted to frequency space using Fourier transform, the 1-wave component is retained, and then converted back to the time domain through inverse transform to obtain the azimuthal 1-wave asymmetric component fields of the observed and forecast fields in the polar coordinate system and .

[0065] Finding the azimuthal position of the maximum of the 1-wave asymmetric component in each radial and Calculating the error of the azimuthal position of the maximum of the observed field and the azimuthal position of the maximum of the forecast field in each radial .

[0066] Rotating the forecast field according to the azimuthal error in each radial to obtain the azimuthally distributed corrected precipitation forecast field , and interpolating it back to the original latitude-longitude grid to obtain the precipitation field corrected for the error of the asymmetric distribution of precipitation .

[0067] Step 106, calculating the evaluation index of the forecast precipitation field and the corrected forecast precipitation field after each correction , , and According to the change in the evaluation index after each correction, the contribution value of the associated factor of each correction to the precipitation forecast error is obtained.

[0068] This embodiment eliminates the position, magnitude, and structure errors step by step and calculates the evaluation index improvement value, which specifically includes the spatial deviation related to the path of the tropical cyclone precipitation forecast, the magnitude deviation related to the systematic deviation of the model, the radial distribution deviation related to the radial structure of the model precipitation, and the azimuthal deviation related to the asymmetric precipitation.

[0069] This embodiment eliminates the position, magnitude, and structure errors of the forecast precipitation field step by step and calculates the evaluation index improvement value before and after each correction, thereby obtaining the contribution of the tropical cyclone path forecast error, the systematic deviation of the model precipitation intensity, the radial precipitation structure forecast error, and the asymmetric precipitation structure forecast error to the total precipitation forecast error. Compared with the commonly used CRA or MODE verification method, the tropical cyclone precipitation forecast verification evaluation is more targeted and interpretable, the physical meaning is more explicit, it is helpful for the forecasters to interpret the forecast and the model developers to improve the model, and it can provide strong support for real-time error analysis and correction of tropical cyclone precipitation forecast.

[0070] On the basis of the above embodiment, the forecast precipitation correction field is adjusted according to the ratio of the radial precipitation distribution profiles of the two in this embodiment, to obtain the forecast precipitation correction field , which includes:

[0071] Calculating the observed precipitation field radial precipitation distribution profiles of the observed field and the forecast corrected field a ratio between the radial precipitation distribution profiles of the observed field and the forecast corrected field

[0072] multiplying the forecast corrected field in polar coordinates by the ratio to obtain a forecast corrected field of the radial deviation of azimuthally averaged radial precipitation distribution .

[0073] azimuthally averaging the observed field and the forecast corrected field to obtain radial precipitation distribution profiles of the observed and forecast axisymmetric precipitation respectively and calculating a ratio of the radial distribution profiles of the azimuthally averaged observed and forecast fields .

[0074] multiplying the forecast corrected field in polar coordinates by the ratio of the distribution profiles to obtain a forecast corrected field of the radial deviation of azimuthally averaged radial precipitation distribution, i.e. interpolating the forecast corrected field back to the original latitude-longitude grid coordinates to obtain a forecast corrected field .

[0075] On the basis of the above-mentioned embodiments, before azimuthally averaging the observed precipitation field in polar coordinates and the forecast corrected precipitation field to obtain radial precipitation distribution profiles, the present embodiment further comprises:

[0076] interpolating the observed precipitation field and the forecast corrected precipitation field to a polar coordinate system with the observed tropical cyclone center as the origin and R as the radius.

[0077] On the basis of the above-mentioned embodiments, the radial resolution of the polar coordinate system in the present embodiment is the same as the grid spacing at the origin and gradually coarsens along the radial direction to the periphery.

[0078] When constructing the polar coordinates of precipitation, a gradually coarsening radial resolution is set to make the construction of the 1-wave asymmetric component and the azimuthal rotation process more robust.

[0079] On the basis of the above-mentioned embodiments, the present embodiment calculates an evaluation index of the forecast precipitation field and the forecast corrected precipitation field after each correction, and according to the change amount of the evaluation index after each correction, obtains the contribution value of the associated factor of each correction to the precipitation forecast error, comprising:

[0080] calculating the evaluation index SSIM score between the forecast precipitation field and the observed precipitation field . ;

[0081] The evaluation index SSIM score between the forecast precipitation revised field and the observation precipitation field ; The evaluation index SSIM score between the forecast precipitation revised field and the observation precipitation field ;

[0082] The evaluation index SSIM score between the forecast precipitation revised field and the observation precipitation field ;

[0083] The evaluation index SSIM score between the forecast precipitation revised field and the observation precipitation field ; The evaluation index SSIM score between the forecast precipitation revised field and the observation precipitation field ;

[0084] The evaluation index SSIM score between the forecast precipitation revised field and the observation precipitation field ;

[0085] The difference between and is taken as the overall explained error, and the proportion of the difference between and in the overall explained error is taken as the contribution value corresponding to the i-th revision, i=1, 2, 3 and 4.

[0086] The evaluation index SSIM is calculated between the numerical model precipitation forecast field and the observation field , , and , wherein i takes values of i=1, 2, 3 and 4.

[0087] is calculated as the overall error.

[0088] The contribution of the SSIM score promotion amount of each step revision to the overall error is calculated as the contribution of the forecast bias or error expressed by the step revision in the precipitation forecast error, wherein i takes values of i=1, 2, 3 and 4.

[0089] On the basis of the above-mentioned embodiments, the present embodiment further comprises:

[0090] The proportion of the difference between and in the overall error is calculated as the cumulative contribution value corresponding to the previous i revisions.​​​​​​

[0091] The cumulative contribution value of each step correction relative to the numerical model precipitation forecast is calculated: , where i takes values of i = 1, 2, 3 and 4.

[0092] On the basis of each of the above embodiments, the present embodiment further comprises:

[0093] The difference between 1 and is taken as the unexplained residual error.

[0094] The unexplained residual error is calculated.

[0095] For example, taking the 3 h cumulative precipitation of a super typhoon from 05 to 08 (UTC) on July 28, 2023 as an example. Numerical model prediction data can be obtained from a certain weather forecast center, and tropical cyclone position data can be obtained from the tropical cyclone best path data IBTrACS, with a time resolution of 3 h. Tropical cyclone cumulative precipitation data can be obtained from the Global Precipitation Measurement (GPM), with a time resolution of 30 min, and the corresponding period of precipitation is downloaded to obtain 3 h cumulative precipitation.

[0096] Using the 500 hPa potential height, 700 hPa and 850 hPa wind field, sea level pressure and 10 m wind field data predicted by the numerical model, the GFDL vortex tracking algorithm is used to calculate the predicted tropical cyclone center position and intensity information at 05 and 08 on the 28th, as shown in Table 1.

[0097] Table 1

[0098]

[0099] The average of the observed tropical cyclone positions at 05 and 08 is taken as the center of the observed tropical cyclone cumulative precipitation for the period, and the average of the predicted tropical cyclone positions at 05 and 08 is taken as the center of the predicted tropical cyclone cumulative precipitation for the period. The observed cumulative precipitation is interpolated to the predicted grid using nearest neighbor interpolation for evaluation, as shown in part a of Figure 3 .

[0100] The error of the predicted tropical cyclone center relative to the observed center is calculated, and the longitude and latitude of the predicted precipitation grid field (part b of Figure 3 ) are subtracted by the error to obtain the moved longitude and latitude, and then interpolated back to the original grid longitude and latitude, completing the spatial translation of the predicted precipitation field, thereby correcting the precipitation prediction error caused by the tropical cyclone path error, as shown in part c of Figure 3 .

[0101] The observed, forecast, and track error-corrected forecast fields are interpolated into a polar coordinate system centered on the observed tropical cyclone with a radius of 5 latitudes (the radial resolution of the polar coordinate system is set to the model grid resolution at the origin, gradually becoming coarser outwards, and can be set to 4 times the origin at the edge; the azimuth resolution can be set to 2.5°).

[0102] The observed, forecast, and longitude error-corrected forecast fields in the polar coordinate system are re-interpolated back to the model's latitude and longitude grid, such as... Figure 3 The d and f parts in the data are used to obtain the observed precipitation field for verification. Forecast precipitation field And the predicted precipitation correction field after path error correction The SSIM scores for the numerical model forecast and the observed field after longitude path error correction are calculated separately in the latitude and longitude grid, and denoted as follows: and See rating Figure 3 The numerical values ​​indicated in the headings of sections e and f.

[0103] The quantile mapping method was used to correct the previously revised forecast precipitation field. Towards Figure 3 Observed precipitation in part d Adjustments are made to obtain the quantile-mapped precipitation correction field for each grid cell. ,like Figure 3 The g part. The predicted precipitation field obtained by the quantile mapping method. and correcting the precipitation field The mapping relationship is as follows Figure 4 Similarly, calculate the adjusted forecast field and the observed SSIM score, denoted as... The scoring results are shown below. Figure 3 The numerical value indicated in the title of section g.

[0104] Calculate observed precipitation in polar coordinates ( Figure 3 (part d in the previous step) and the revised forecast precipitation field after the previous step. ( Figure 3 The azimuth average of the g-part in the image is used to obtain the radial distribution profile of axisymmetric precipitation in the observed and corrected field. and The ratio of the observed profile to the corrected field profile is calculated as the radial distribution correction coefficient. ,like Figure 5 .

[0105] The revised precipitation forecast field after path correction and quantile mapping ( Figure 3The g part in the equation is multiplied by the corresponding radial distribution correction factor in each radial direction. The forecast correction field was obtained, which corrected the deviation of the azimuth mean radial precipitation distribution. The forecast will be revised. Interpolating back to the original latitude and longitude grid coordinates yields the revised forecast precipitation field after adjustment for axisymmetric radial distribution. ,like Figure 3 For the h part, the adjusted forecast field and the observed SSIM score are also calculated, denoted as The scoring results are shown below. Figure 3 The value indicated in the heading of the h section.

[0106] The observed precipitation and the previously corrected forecast precipitation fields in polar coordinates are transformed to the frequency domain using a real Fourier transform. All wavenumber components except the 1-wave component are removed. Then, an inverse real Fourier transform is performed to transform them back to the time domain in polar coordinates, resulting in the 1-wave asymmetric component fields of the observed and numerical model correction fields. and ,like Figure 6 Parts a and b are shown in the figure.

[0107] Calculate the azimuth maximum index in each radial direction for the observed and numerical model correction field's one-wave asymmetric component field in polar coordinates. Subtract the azimuth maximum index of the numerical model correction field from the observed azimuth maximum index to obtain the azimuth error characterized by grid offset. .

[0108] The numerical model correction field and its one-wave component in polar coordinates are adjusted in each radial direction according to the corresponding azimuth error. Rotation yields the revised precipitation field after removing the first-wave asymmetric structure error and its first-wave component, as follows: Figure 6 The f and c parts are shown in the diagram. To verify that the center of the asymmetric component of wave 1 has been adjusted to match the observation, the SSIM scores of the adjusted forecast and observation fields are calculated and denoted as... The scoring results are shown below. Figure 3 The numerical values ​​indicated in the heading of section i. Figure 6 The d and e parts in the figure are the observation fields. And the revised forecast .

[0109] After the above multi-step correction, the numerical model precipitation forecast deviation can be obviously eliminated, and the SSIM score of each step of correction can be obtained, including the numerical model original precipitation score SSIM0, the tropical cyclone path forecast error removal score SSIM1, the superimposed removal of the model precipitation intensity systematic deviation score SSIM2, the superimposed removal of the axisymmetric radial precipitation structure forecast error score SSIM3, and the superimposed removal of the asymmetric precipitation structure forecast error SSIM4.

[0110] The overall precipitation forecast error and the relative contribution of the four types of errors and biases to the overall precipitation forecast error are characterized using the following formula:

[0111] a) Overall precipitation forecast error:

[0112] b) Tropical cyclone path forecast error:

[0113] c) Model precipitation intensity systematic deviation:

[0114] d) Axisymmetric radial precipitation structure forecast error:

[0115] e) 1-wave asymmetric precipitation structure forecast error:

[0116] f) Error residual: .

[0117] The tropical cyclone precipitation forecast diagnostic evaluation system provided by the present application is described below, and the tropical cyclone precipitation forecast diagnostic evaluation system described below can be mutually corresponding with the tropical cyclone precipitation forecast diagnostic evaluation method described above.

[0118] As Figure 7 described, the system includes an initial scoring module 701, a first correction module 702, a second correction module 703, a third correction module 704, a fourth correction module 705, and an evaluation module 706, wherein:

[0119] The initial scoring module 701 is used to cut the observed precipitation field and the numerical model predicted precipitation field according to the observed tropical cyclone path with R as the radius, respectively, to obtain the observed precipitation field and the predicted precipitation field .

[0120] The first correction module 702 is used to correct the predicted precipitation field according to the offset between the numerical model predicted tropical cyclone center and the observed tropical cyclone center.The forecast precipitation field is shifted, and precipitation is extracted from the shifted forecast precipitation field with the observed tropical cyclone center as the center and R as the radius, to obtain the path bias-removed forecast precipitation correction field. ;

[0121] The second correction module 703 is used to adjust the forecast precipitation correction field based on the quantile mapping method. The rainfall magnitude caused the forecast precipitation correction field Adjusted precipitation magnitude distribution and observed precipitation field The precipitation magnitude distribution is consistent, resulting in a revised precipitation field after removing model magnitude biases. ;

[0122] The third correction module 704 is used to correct the observed precipitation field in polar coordinates. and Forecasted Precipitation Correction Field Radial precipitation profiles are obtained by azimuth averaging. The ratio of the two radial precipitation profiles is used to correct the forecast precipitation field. Adjustments were made to obtain the revised precipitation forecast field. Interpolate it back to the original latitude and longitude grid coordinates to obtain the radially distributed adjusted forecast correction field. ;

[0123] The fourth correction module 705 is used to calculate the observed precipitation field in polar coordinates. and Forecasted Precipitation Correction Field The azimuth 1-wave asymmetric component, and the offset angle between the azimuth angles where the maximum values ​​of the azimuth 1-wave asymmetric components of the two components are located in each radial direction, are used to correct the forecast precipitation field. Rotate to obtain the forecast precipitation correction field. Interpolate it back to the original latitude and longitude grid coordinates to obtain the forecast precipitation correction field with azimuth asymmetric structure adjustment. ;

[0124] Evaluation module 706 is used to calculate the forecast precipitation field. And the revised precipitation forecast field after each revision , , and The evaluation indicators are used to determine the contribution of factors associated with each correction to the precipitation forecast error based on the changes in the evaluation indicators after each correction.

[0125] The embodiment eliminates the position, magnitude and structure errors of the predicted precipitation field in sequence by steps, and calculates the evaluation index improvement amount before and after each correction, thereby obtaining the contribution of the tropical cyclone path prediction error, the systematic deviation of the model precipitation intensity, the radial precipitation structure prediction error and the asymmetric precipitation structure prediction error to the total precipitation prediction error. Compared with the commonly used CRA or MODE test method, the embodiment is more targeted and interpretable in the tropical cyclone precipitation prediction test evaluation, has a more explicit physical meaning, is helpful for the prediction interpretation of the predictor and the improvement of the model for the model developer, can provide strong support for the real-time error analysis and correction of the tropical cyclone precipitation prediction.

[0126] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for tropical cyclone precipitation forecast diagnostic evaluation, characterized in that, include: According to the observed tropical cyclone path, the observed precipitation field and the precipitation field predicted by the numerical model are respectively cropped to obtain the observed precipitation field and the predicted precipitation field ; Based on the offset between the tropical cyclone center predicted by the numerical model and the observed tropical cyclone center, the predicted precipitation field is analyzed. The forecast precipitation field is shifted, and precipitation is extracted from the shifted forecast precipitation field with the observed tropical cyclone center as the center and R as the radius, to obtain the path bias-removed forecast precipitation correction field. ; Adjusting the Precipitation Forecast Field Based on Quantile Mapping Method The rainfall magnitude caused the forecast precipitation correction field Adjusted precipitation magnitude distribution and observed precipitation field The precipitation magnitude distribution is consistent, resulting in a revised precipitation field after removing model magnitude biases. ; Observed precipitation field in polar coordinate system and Forecasted Precipitation Correction Field Radial precipitation profiles are obtained by azimuth averaging. The ratio of the two radial precipitation profiles is used to correct the forecast precipitation field. Adjustments were made to obtain the revised precipitation forecast field. Interpolate it back to the original latitude and longitude grid coordinates to obtain the radially distributed adjusted forecast correction field. ; Calculate the observed precipitation field in polar coordinates respectively and Forecasted Precipitation Correction Field The azimuth 1-wave asymmetric component, and the offset angle between the azimuth angles where the maximum values ​​of the azimuth 1-wave asymmetric components of the two components are located in each radial direction, are used to correct the forecast precipitation field. Rotate to obtain the forecast precipitation correction field. Interpolate it back to the original latitude and longitude grid coordinates to obtain the forecast precipitation correction field with azimuth asymmetric structure adjustment. ; Calculate and predict precipitation fields And the revised precipitation forecast field after each revision , , and The evaluation indicators are used to determine the contribution of factors associated with each correction to the precipitation forecast error based on the changes in the evaluation indicators after each correction.

2. The method for forecasting, diagnosing, and evaluating tropical cyclone precipitation according to claim 1, characterized in that, The predicted precipitation field is corrected based on the ratio of the radial precipitation distribution profiles of the two. Adjustments were made to obtain the revised precipitation forecast field. ,include: Calculate the observed precipitation field Radial precipitation distribution profile and forecast precipitation correction field The ratio between radial precipitation profiles; Correction field for predicted precipitation in polar coordinates Multiplying by the ratio yields the forecast precipitation correction field. .

3. The method for forecasting, diagnosing, and evaluating tropical cyclone precipitation according to claim 1, characterized in that, Observation of precipitation field in polar coordinate system and Forecasted Precipitation Correction Field Before obtaining the radial precipitation distribution profile by azimuth averaging, the following steps are also included: Observing precipitation fields and Forecasted Precipitation Correction Field Interpolate to a polar coordinate system with the observed tropical cyclone center as the origin.

4. The method for forecasting, diagnosing, and evaluating tropical cyclone precipitation according to claim 1, characterized in that, The radial resolution of the polar coordinate system is the same as the precipitation field grid at the origin, and gradually becomes thicker radially outward.

5. The method for forecasting, diagnosing, and evaluating tropical cyclone precipitation according to any one of claims 1-4, characterized in that, Calculate and predict precipitation fields And the revised precipitation forecast field after each revision , , and The evaluation indicators are used to determine the contribution of factors associated with each correction to the precipitation forecast error, based on the changes in the evaluation indicators after each correction. These factors include: Calculate and predict precipitation fields With the observed precipitation field The evaluation metric between them is the SSIM score. ; Calculate the precipitation forecast correction field With the observed precipitation field The evaluation metric between them is the SSIM score. ; Calculate the precipitation forecast correction field With the observed precipitation field The evaluation metric between them is the SSIM score. ; Calculate the precipitation forecast correction field With the observed precipitation field The evaluation metric between them is the SSIM score. ; Calculate the precipitation forecast correction field With the observed precipitation field The evaluation metric between them is the SSIM score. ; Will As a total error, and The proportion of the difference between the two values ​​in the total error is taken as the contribution value corresponding to the i-th correction, i=1, 2, 3 and 4.

6. The method for forecasting, diagnosing, and evaluating tropical cyclone precipitation according to claim 5, characterized in that, Also includes: calculate and The proportion of the difference between the two errors in the total error is taken as the cumulative contribution value corresponding to the first i corrections.

7. The method for forecasting, diagnosing, and evaluating tropical cyclone precipitation according to claim 5, characterized in that, Also includes: Subtract 1 The difference is treated as an unexplained residual.

8. A tropical cyclone precipitation forecasting, diagnostic, and assessment system, characterized in that, include: The initial scoring module is used to score the observed precipitation field based on the observed tropical cyclone path with a radius of R. Precipitation field predicted by numerical model After cropping, the observed precipitation field is obtained. and forecast precipitation field ; The first correction module is used to adjust the forecast precipitation field based on the offset between the tropical cyclone center predicted by the numerical model and the observed tropical cyclone center. The forecast precipitation field is shifted, and precipitation is extracted from the shifted forecast precipitation field with the observed tropical cyclone center as the center and R as the radius, to obtain the path bias-removed forecast precipitation correction field. ; The second correction module is used to adjust the forecast precipitation correction field based on the quantile mapping method. The rainfall magnitude caused the forecast precipitation correction field Adjusted precipitation magnitude distribution and observed precipitation field The precipitation magnitude distribution is consistent, resulting in a revised precipitation field after removing model magnitude biases. ; The third correction module is used to correct the observed precipitation field in polar coordinates. and Forecasted Precipitation Correction Field Radial precipitation profiles are obtained by azimuth averaging. The ratio of the two radial precipitation profiles is used to correct the forecast precipitation field. Adjustments were made to obtain the revised precipitation forecast field. Interpolate it back to the original latitude and longitude grid coordinates to obtain the radially distributed adjusted forecast correction field. ; The fourth correction module is used to calculate the observed precipitation field in polar coordinates. and Forecasted Precipitation Correction Field The azimuth 1-wave asymmetric component, and the offset angle between the azimuth angles where the maximum values ​​of the azimuth 1-wave asymmetric components of the two components are located in each radial direction, are used to correct the forecast precipitation field. Rotate to obtain the forecast precipitation correction field. Interpolate it back to the original latitude and longitude grid coordinates to obtain the forecast precipitation correction field with azimuth asymmetric structure adjustment. ; The evaluation module is used to calculate the forecast precipitation field. And the revised precipitation forecast field after each revision , , and The evaluation indicators are used to determine the contribution of factors associated with each correction to the precipitation forecast error based on the changes in the evaluation indicators after each correction.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the tropical cyclone precipitation forecasting diagnostic assessment method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the tropical cyclone precipitation forecasting diagnostic assessment method as described in any one of claims 1 to 7.

Images