Method and system for identifying sensitive area of strong precipitation based on convective scale ensemble prediction and ensemble transformation sensitivity

Abstract

The application provides a method and system for identifying a sensitive area of heavy rain based on convection scale ensemble prediction and sensitivity of ensemble transformation, which comprises the following steps: obtaining convection-resolvable ensemble prediction data based on a convection scale ensemble prediction system; determining a verification area, which is a heavy rain occurrence area; determining an analysis time and a verification time, and calculating an ensemble perturbation matrix at the analysis time and the verification time, wherein the verification time is a heavy rain occurrence time; selecting a metric standard, a perturbation variable and a vertical level suitable for heavy rain, initializing a projection matrix, and giving an initial analysis error variance matrix; calculating a gradient value of a reduction amount of a prediction error variance of all state variable positions relative to a reduction amount of the analysis error variance, and selecting an area with a gradient value exceeding a preset threshold as a heavy rain sensitive area. The application improves the calculation efficiency of heavy rain sensitive area identification, reduces the calculation cost, and improves the calculation accuracy.

Description

Technical Field

[0001] This invention relates to the field of weather data processing technology, and in particular to a method and system for identifying areas sensitive to heavy precipitation based on convective-scale ensemble forecasting and ensemble transformation sensitivity. Background Technology

[0002] Heavy rainfall is characterized by its suddenness and localized nature, easily causing serious damage to people's lives and property. Therefore, effective heavy rainfall forecasting is of great significance. Numerical forecasting is an important method for heavy rainfall forecasting. To improve the accuracy of numerical forecasts, it is usually necessary to identify sensitive areas of weather systems in a specified region. If the distribution of these sensitive areas can be identified in advance and meteorological observations are increased within these areas, it will help improve initial conditions and thus reduce numerical forecast errors.

[0003] Currently, there are two main types of sensitive area identification methods: adjoint-based methods and set-based methods. Adjoint-based methods require obtaining the adjoint of the pattern and then solving the price function to determine the sensitive area, which consumes significant computational resources in solving the price function. Set-based methods, on the other hand, require obtaining a set of ensemble forecasts beforehand and then using these forecasts to identify potential sensitive areas. Their computational efficiency is higher than that of adjoint-based methods; examples include ensemble divergence methods, ensemble transformation methods, and ensemble transformation Kalman filtering methods.

[0004] In the existing technology, the above methods are difficult to apply to the identification of areas sensitive to heavy precipitation. This is because the identification of areas sensitive to heavy precipitation requires high resolution. Even the above methods based on ensembles are mainly applied to typhoons or large-scale weather systems. When facing specific areas or high-impact weather systems such as heavy precipitation that require high resolution, there are still computational efficiency issues. For example, the ensemble transformation method and the ensemble transformation Kalman filter method need to traverse all possible observation schemes to obtain the final sensitive area. Therefore, as the resolution increases, the number of vertical layers increases, and the number of ensemble members increases, the computational cost will increase exponentially. Summary of the Invention

[0005] This invention provides a method and system for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity, in order to solve the technical problems of low computational efficiency, high computational cost, and poor computational accuracy in the identification of heavy precipitation sensitive areas in the prior art.

[0006] The present invention solves the above-mentioned technical problems through the following technical solution: The first aspect of this invention provides a method for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity, comprising the following steps: Based on the convective-scale ensemble forecast system, obtain convectively resolvable ensemble forecast data; The verification area is defined as the area where heavy rainfall occurs. Determine the analysis time and the verification time, and calculate the set perturbation matrix of the analysis time and the verification time, wherein the verification time is the time when heavy precipitation occurs; Select a measurement standard, perturbation variable, and vertical layer suitable for heavy precipitation, initialize the projection matrix, and provide an initial analysis error variance matrix; Calculate the gradient value of the reduction in the forecast error variance of all state variable locations relative to the reduction in the analysis error variance, and select areas where the gradient value exceeds a preset threshold as areas sensitive to heavy precipitation.

[0007] Preferably, the horizontal resolution of the convective-scale ensemble forecasting system is 2-4 km.

[0008] Preferably, the area where heavy precipitation occurs is the region where the probability of precipitation occurrence obtained by the convective-scale ensemble forecasting system is greater than a predetermined threshold.

[0009] Preferably, the metric is the total energy of dry air, and the disturbance variables include meridional wind, zonal wind, and temperature.

[0010] Preferably, the metric is the total energy of moist air, and the disturbance variables include meridional wind, zonal wind, temperature, and humidity.

[0011] Preferably, the vertical layer is the key vertical atmospheric stratification corresponding to the heavy precipitation process.

[0012] Preferably, the ensemble perturbation matrix is ​​calculated by the difference between the ensemble forecast data and the ensemble mean, wherein the ensemble mean is the mean of each ensemble member in the ensemble forecast data; The projection matrix is ​​a diagonal matrix, and the diagonal elements of the projection matrix are set according to whether the state variable is located in the verification area and the variable type; The initial analysis error variance matrix is ​​the analysis field error variance matrix after adding a new observation.

[0013] Preferably, the method further includes: An observation system simulation experiment was conducted, and numerical experiments were carried out on the heavy precipitation sensitive area to analyze the nonlinear characteristics of the heavy precipitation sensitive area.

[0014] A second aspect of this invention provides a system for identifying sensitive areas of heavy precipitation based on convective-scale ensemble forecasting and ensemble transformation sensitivity, used to implement the aforementioned method for identifying sensitive areas of heavy precipitation based on convective-scale ensemble forecasting and ensemble transformation sensitivity, the system comprising: The data acquisition module is used to acquire convection-resolvable ensemble forecast data based on the convection-scale ensemble forecast system. The verification area determination module is used to determine the verification area, which is the area where heavy precipitation occurs; The perturbation matrix calculation module is used to determine the analysis time and the verification time, and to calculate the set perturbation matrix of the analysis time and the verification time, wherein the verification time is the time when heavy precipitation occurs; The measurement standard selection module is used to select a measurement standard, perturbation variable, and vertical level suitable for heavy precipitation, initialize the projection matrix, and provide an initial analysis error variance matrix. The gradient value calculation module is used to calculate the gradient value of the reduction in the forecast error variance of all state variable locations relative to the reduction in the analysis error variance, and selects areas where the gradient value exceeds a preset threshold as areas sensitive to heavy precipitation.

[0015] Preferably, the system further includes: The verification module is used to conduct observation system simulation experiments, carry out numerical experiments on the heavy precipitation sensitive area, and analyze the nonlinear characteristics of the heavy precipitation sensitive area.

[0016] The positive and progressive effects of this invention are as follows: This invention is based on a convective-scale ensemble forecast system to acquire convectively distinguishable ensemble forecast data and uses ensemble transformation sensitivity to identify areas sensitive to heavy precipitation. This overcomes the shortcomings of traditional sensitive area identification methods in the prior art, which are not applicable to the identification of areas sensitive to heavy precipitation. It improves the computational efficiency of identifying areas sensitive to heavy precipitation, reduces computational costs, and improves computational accuracy. Attached Figure Description

[0017] Figure 1 The flowchart shows the method for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity provided in Embodiment 1 of the present invention.

[0018] Figure 2 This is a schematic diagram of the structure of the heavy precipitation sensitive area identification system based on convective-scale ensemble forecasting and ensemble transformation sensitivity provided in Embodiment 2 of the present invention.

[0019] Figure 3 This is a diagram of the simulation region of the CMA-BJ-set v1.0 on which embodiments 1 and 2 of the present invention are based in specific applications.

[0020] Figure 4 The above are the actual precipitation maps of Beijing used in specific applications of Embodiments 1 and 2 of the present invention.

[0021] Figure 5 This is a horizontal distribution map of the heavy precipitation sensitive area obtained in specific applications of Embodiments 1 and 2 of the present invention.

[0022] Figure 6This is a random radiosonde distribution diagram of Embodiments 1 and 2 of the present invention in a specific application.

[0023] Figure 7 The graph shows the verification results of Embodiments 1 and 2 of the present invention in a specific application (based on TS score).

[0024] Figure 8 The graph shows the verification results of Embodiments 1 and 2 of the present invention in a specific application (based on Bias score).

[0025] Explanation of reference numerals in the attached figures: A heavy precipitation sensitive area identification system based on convective-scale ensemble forecasting and ensemble transformation sensitivity 100 Data acquisition module 101 Verification Area Determination Module 102 Perturbation matrix calculation module 103 Measurement Standard Selection Module 104 Gradient value calculation module 105 Verification module 106 Detailed Implementation Before introducing specific embodiments, the inventive concept and core principles of this invention will be explained.

[0026] In existing technologies, whether based on adjoint methods or ensemble transformation methods and ensemble transformation Kalman filtering methods for sensitive area identification, all suffer from low computational efficiency when dealing with high-resolution specific regions or high-impact weather systems such as heavy precipitation. This invention aims to introduce the concept of ensemble transformation sensitivity based on ensemble methods. Using multi-member forecast data generated by a convective-scale ensemble forecast system, it analyzes the relationship between forecast errors and initial field disturbances, identifying the regions that contribute the most to forecast errors and designating these regions as sensitive areas for heavy precipitation. The core of this invention lies in utilizing the diversity among ensemble forecast members to estimate the sensitivity of forecast errors, thereby providing guidance for observations, particularly aiming to improve the accuracy of specific forecast events. It estimates sensitive areas for heavy precipitation by using the sensitivity (gradient) of the forecast error variance. This process only requires minimizing the valence function, i.e., the partial derivative of the cost function. Therefore, this invention does not require traversing the locations of all state variables, effectively improving computational efficiency.

[0027] The present invention will be described more clearly and completely below by way of embodiments and in conjunction with the accompanying drawings, but the present invention is not limited to the scope of the embodiments described herein.

[0028] Example 1 like Figure 1 As shown, Embodiment 1 of the present invention provides a method for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity. The method includes the following steps: Step S1: Obtain convection-resolvable ensemble forecast data based on the convection-scale ensemble forecast system; Step S2: Determine the verification area, which is the area where heavy rainfall occurs; Step S3: Determine the analysis time and the verification time, and calculate the set perturbation matrix of the analysis time and the verification time. The verification time is the time when the heavy precipitation occurs. Step S4: Select a measurement standard, perturbation variable, and vertical layer suitable for heavy precipitation, initialize the projection matrix, and provide an initial analysis error variance matrix; Step S5: Calculate the gradient value of the reduction in the forecast error variance of all state variable locations relative to the reduction in the analysis error variance, and select areas where the gradient value exceeds a preset threshold as areas sensitive to heavy precipitation. Step S6: Conduct a simulation experiment of the observation system, carry out numerical experiments on the heavy precipitation sensitive area, and analyze the nonlinear characteristics of the heavy precipitation sensitive area.

[0029] Specifically, in step S1, the convective-scale ensemble forecasting system is a multi-member probabilistic forecasting system with a horizontal resolution of 2–4 km, capable of explicitly simulating convective processes. It is used to quantify the forecast uncertainty of severe convective weather (such as short-duration heavy rainfall, thunderstorms, and hail). The selection of the convective-scale ensemble forecasting system should be based on operational objectives, computing resources, observational foundation, and regional weather characteristics, comprehensively considering horizontal resolution, initial value / model perturbation schemes, assimilation capabilities, lead time coverage, and local validation performance. Furthermore, the identification of sensitive areas for heavy precipitation requires a high-resolution convective-scale ensemble forecasting system, making traditional sensitive area identification methods difficult to apply. This is because traditional sensitive area identification methods suffer from low computational efficiency and high computational costs when dealing with high-resolution systems.

[0030] Specifically, in step S1, a set of ensemble forecasts is obtained from the acquired convection-resolved ensemble forecast data. ,in, For forecast timeliness. For one OK Column matrix This represents the value of all state variables at each grid point. Indicates the number of members in the set.

[0031] Specifically, in step S2, the area where heavy precipitation occurs is defined as the region where the probability of precipitation occurrence obtained from the convective-scale ensemble forecasting system is greater than a predetermined threshold. Based on the obtained ensemble forecast data, the probability of precipitation occurring at each grid point can be calculated. The region where the probability of precipitation occurrence is greater than the predetermined threshold is defined as the verification region. For example, if the predetermined threshold is set to 0.5, then it is considered that heavy precipitation will occur at each grid point where the probability of precipitation occurrence is greater than 0.5. Corresponding to the verification region, the moment when heavy precipitation occurs in the aforementioned heavy precipitation area is the verification moment.

[0032] Specifically, in step S3, the ensemble forecast data includes data obtained from adaptive observations. Adaptive observations refer to supplementary and enhanced observations conducted on a specific region / time period based on existing observation systems in order to specifically improve the quality of numerical forecasts for high-impact weather. The adaptive observation time is the same as the analysis time.

[0033] Specifically, in step S3, the ensemble perturbation matrix is ​​calculated by the difference between the ensemble forecast data and the ensemble mean, where the ensemble mean is the average of each ensemble member in the ensemble forecast data. (ensemble perturbation matrix) Also one OK A column matrix, where each column represents the deviation of a set member from the average state; the entire matrix is ​​used to characterize the prediction system's performance. Spatial distribution of uncertainty at any given time. Ensemble perturbation matrix. It can be represented as: (1) in, It is a set average, which is also a OK A column matrix, where each column is the average vector of the same set.

[0034] Therefore, the adaptive observation time can be calculated. The ensemble perturbation matrix at (i.e., the analysis time) and verification time set perturbation matrix .

[0035] At this point, the error covariance matrix can be calculated.

[0036] In the ensemble space, the analysis error covariance matrix is ​​based on ensemble forecast data. It can be represented as: (2) Assume that model forecast errors are negligible and that the growth of ensemble forecast errors is quasi-linear. Define the model operator. The model starts from adaptive observation time Points until verification time The prediction error covariance matrix based on the set It can be represented as: (3) Define a new set perturbation matrix This matrix is ​​the ensemble perturbation matrix generated after assimilating adaptive observation data. The purpose of the ensemble transformation method is to obtain the transformation matrix C, which characterizes the ensemble perturbation. To the new set of perturbations The transformation between sets. Assuming the set has a sufficiently large number of members, and the perturbations of each member are orthogonal, then this transformation matrix is ​​unique. Assuming the transformation matrix exists, we have: (4) Where C is The matrix. Similar to formula (2), after assimilating the new observation data, the analysis error covariance matrix can be expressed as: (5) In practical applications of adaptive observation, the analytical error covariance matrix of the real atmosphere is crucial. It is unknown. At the adaptive observation moment. The error covariance of set-based methods is mathematically similar to that of the actual analysis. Therefore, in data assimilation applications, the analysis error covariance matrix can be estimated based on the initial guess. The initial analysis error covariance matrix is ​​thus given: if an adaptive observation is added at a certain position of a state variable, the analysis error variance at that position will decrease by 50%.

[0037] According to formula (3), the forecast error covariance matrix after adding observations is: (6) Because the new observation data was assimilated, the forecast error was reduced by [amount missing]. for: (7) Assume the set perturbation matrix If the matrix is ​​a full-rank matrix, then the matrix... Invertible. According to formula (5), the dot product of the transformation matrix and its transpose is: (8) In the application of the traditional set transformation method, formula (8) can only calculate the signal variance corresponding to a certain adaptive observation scheme. For a new adaptive observation scheme, it is necessary to construct a new analysis error covariance matrix, then calculate the transformation matrix, and finally obtain the signal variance matrix. This also illustrates the problem of low computational efficiency in the traditional set transformation method.

[0038] Specifically, in step S4, in order to obtain the optimal adaptive observation scheme, it is necessary to select a metric to measure the sensitivity of the adaptive observation scheme. The metric is usually in the form of total perturbation energy. In this embodiment, the metric is selected as total dry air energy, and the perturbation variables include meridional wind, zonal wind, and temperature. The vertical layer is the key vertical atmospheric stratification corresponding to the heavy precipitation process. For example, the key vertical atmospheric stratification is selected as the atmospheric stratification between 200-925 hPa.

[0039] The total energy function of dry air is as follows: (9) in, These represent the meridional wind, zonal wind, and temperature, respectively. This is the specific heat of dry air at constant pressure (taken as 1005.7 in this embodiment). K), For reference temperature; Indicates the horizontal direction of the calculation area. This represents the vertical direction of the calculation area. In this embodiment, the variable used for sensitive area identification is wind speed, and the reference temperature used is 270 K. Specifically, in other alternative embodiments, air pressure can also be considered as a disturbance factor when determining the total energy function of dry air.

[0040] Define vector This represents the rate at which the analytical error covariance is reduced by adaptive observation data. For example: This indicates that the analysis error has not decreased. This indicates that the analysis error has been reduced by 70%. Therefore... This indicates a reduction rate of The analysis error covariance matrix is ​​defined as follows: its off-diagonal elements represent the covariance between different state variables (different altitudes, different locations, and different variables); its magnitude also characterizes the relationship between different physical quantities. The variance of the state variables themselves has the most significant impact on improving model forecast errors, while the covariance matrix makes calculations very complex. Therefore, covariance information can be ignored when conducting adaptive observations. Furthermore, in traditional heavy precipitation forecasting, the analysis error covariance matrix used in the ensemble transformation method is a diagonal matrix (i.e., the analysis error variance matrix), where the diagonal elements are the variance of the state variable perturbation, and the off-diagonal elements are 0. It is worth noting that the traditional ensemble transformation Kalman filtering method also simplifies the analysis error covariance matrix to a diagonal matrix when determining the sensitive area. In practical applications, for ease of calculation and discussion, the analysis error covariance matrix used is a diagonal matrix, i.e., the analysis error variance matrix. The initial analysis error variance matrix is ​​the analysis field error variance matrix after adding a new observation.

[0041] if ( ) represents the analysis error covariance matrix The diagonal elements. Then .in, , For a given adaptive observation scheme, Values ​​of n at different positions of the state variable There will also be differences. For example, This indicates that there was no reduction in the variance of the analysis error at the position of the nth state variable. This indicates that due to the assimilation of adaptive observation data, the variance of the positional analysis error for the nth state variable has decreased, and its value is [value missing]. Similarly, with The corresponding transformation matrix is The value of the matrix can be obtained using formula (8). Then, the prediction error variance matrix and the signal variance can be obtained.

[0042] Next, the projection matrix needs to be initialized. The projection matrix is ​​a diagonal matrix. The diagonal elements of the projection matrix are set according to whether the state variable is located in the verification area and the variable type, and are represented as follows: .in, Represents the first in the projection matrix Each element.

[0043] The purpose of the projection matrix is ​​to "focus" on the forecast area and physical quantities of interest. The elements on the diagonal... The validation range is determined by whether the corresponding state variable lies within a predefined validation region and the type of the variable. For example, if a variable at a certain grid point is within the validation region, its corresponding diagonal element is non-zero (e.g., weighted according to the energy norm); otherwise, it is zero. In this way, matrix multiplication can restrict the computational sensitivity to the validation region.

[0044] The steps for estimating the sensitive region using the metric formula (9) are as follows: First, define the projection matrix. : (10) If the state variable is within the validation region, then The value is defined as follows: (11) In other cases, .

[0045] After that, set , It is a matrix The The sequence is then used to calculate the trace of the forecast error variance matrix. : (12) Next, the reduction in error is calculated. According to formula (12), the gradient of the trace of the prediction error variance matrix can be obtained: (13) The corresponding reduction in forecast error is: (14) From formula (14), it can be seen that in order to obtain , needs to be calculated That is, it is necessary to calculate the transformation matrix and its transpose relative to the transformation matrix. The derivative of .

[0046] To simplify the derivation, a new matrix is ​​defined. : (15) The dot product of the transformation matrix and its transpose can then be expressed as: For different analytical error variance reduction coefficients Using the derivative formula of the inverse of a matrix, we know that: (16) in, (17) In summary, by combining formulas (14) to (17), the formula for calculating sensitivity can be derived: (18) Therefore, the above formula can be used to obtain the signal variance of all state variable locations, that is, the gradient value of the reduction in prediction error variance relative to the reduction in analysis error variance, and finally used to determine the sensitive area.

[0047] The above calculation process reveals that this invention aims to obtain the signal variance of all state variable locations through a series of matrix operations. The core of this process is calculating the gradient of the reduction in prediction error variance relative to the reduction in initial analysis error variance, i.e. Here This represents the total energy of the prediction error defined by a specific metric within the validation region, while Is it the first time from the initial moment? l The analysis error variance at the position of the state variable decreases the correlation coefficient. The physical meaning of this gradient is: at the initial time, the first... l Investing one unit of observational resources at each location (i.e., reducing a certain amount of analytical error) will result in a reduction of forecast error energy within the verification area at the verification time. Regions with larger gradient values ​​have a greater impact on the forecast results at the verification time; these are the "sensitive areas for heavy precipitation" we are looking for. By calculating this gradient value at all state variable locations, a spatial distribution map can be drawn. Regions in the map whose gradient values ​​exceed a preset threshold are identified as sensitive areas for heavy precipitation at the verification time.

[0048] In addition, in other alternative implementations, the metric can also be the total energy of moist air, as shown in the following formula: (19) in, u, v, T and q These represent meridional wind, zonal wind, temperature, and humidity, respectively. The specific heat at constant pressure of dry air; Indicates the horizontal direction of the calculation area. Indicates the vertical direction of the calculation region; The reference temperature is 300 K; L is the latent heat of phase change. For reference specific humidity. Specifically, in other alternative embodiments, the disturbance factor of air pressure may also be taken into account when determining the total energy function of moist air.

[0049] Specifically, in step S6, the Observing System Simulation Experiment (OSSE) involves interpolating reanalysis data as "real atmosphere" to the observation station to obtain simulated radiosonde observation data. The forecast effects of the two sets of experiments—one with assimilated simulated radiosonde observation data and the other without—are compared and analyzed to verify the correctness of the sensitive area. If the effect of assimilated simulated radiosonde observation data is better than that of unassimilated simulated radiosonde observation data, then the accuracy of the sensitive area identification is proven.

[0050] Example 2 like Figure 2 As shown, this embodiment provides a heavy precipitation sensitive area identification system 100 based on convective-scale ensemble forecasting and ensemble transformation sensitivity. This system is used to implement the aforementioned heavy precipitation sensitive area identification method based on convective-scale ensemble forecasting and ensemble transformation sensitivity. The system includes: The data acquisition module 101 is used to acquire convection-resolvable ensemble forecast data based on the convection-scale ensemble forecast system. The verification area determination module 102 is used to determine the verification area, which is the area where heavy precipitation occurs. The perturbation matrix calculation module 103 is used to determine the analysis time and the verification time, and to calculate the set perturbation matrix of the analysis time and the verification time, wherein the verification time is the time when the heavy precipitation occurs. The metric selection module 104 is used to select a metric suitable for heavy precipitation, perturbation variables and vertical layers, initialize the projection matrix, and provide an initial analysis error variance matrix. The gradient value calculation module 105 is used to calculate the gradient value of the reduction in the forecast error variance of all state variable locations relative to the reduction in the analysis error variance, and selects areas where the gradient value exceeds a preset threshold as areas sensitive to heavy precipitation; and The verification module 106 is used to conduct observation system simulation experiments, carry out numerical experiments on areas sensitive to heavy precipitation, and analyze the nonlinear characteristics of areas sensitive to heavy precipitation.

[0051] This system is based on a convective-scale ensemble forecast system, which acquires convectively resolvable ensemble forecast data and uses ensemble transformation sensitivity to identify areas sensitive to heavy precipitation. This overcomes the shortcomings of traditional sensitive area identification methods in existing technologies, which are not applicable to the identification of areas sensitive to heavy precipitation. It improves the computational efficiency of identifying areas sensitive to heavy precipitation, reduces computational costs, and improves computational accuracy.

[0052] To make it easier to understand, the present invention will be further illustrated below by performing a specific application of the method provided in Example 1 on the system provided in Example 2.

[0053] The following explains the selection criteria for convective-scale ensemble forecasting systems.

[0054] The convective-scale ensemble forecasting system used in this application is primarily based on the North China Convective-Scale Ensemble Forecasting System (CMA-BJ-ensemble v1.0), which is operationally managed by the Beijing Urban Meteorological Research Institute. The CMA-BJ-ensemble v1.0 system has a horizontal resolution of 3 km and 59 vertical layers, covering most of North China (e.g., [missing information]). Figure 3 (As shown).

[0055] This system primarily uses data from the National Center for Atmospheric Prediction (NCEP) Global Ensemble Forecast System (GEFS) as the driving field. It is built upon the WRF mesoscale model and employs core technologies such as global ensemble forecast data preprocessing, generation of background field and lateral boundary data, initial perturbation obtained by combining conventional observational perturbations with ensemble data assimilation, and perturbation of stochastic physical processes tending the model. This results in ensemble forecast data with 1-hour output frequencies covering North China, providing rich probabilistic forecast products for the region.

[0056] The technical specifications of the CMA-BJ-Collection v1.0 system are shown in Table 1.

[0057]

[0058] The following is an explanation of the localization deployment of the heavy precipitation sensitive area identification system based on convective-scale ensemble forecasting and ensemble transformation sensitivity.

[0059] Based on the ensemble transformation sensitivity model (represented in Example 1 as a matrix calculation method for gradient values) and the North China convective-scale ensemble forecasting system, a high-resolution heavy precipitation sensitive area identification system (i.e., the system described in Example 2) was constructed. Currently, the ensemble transformation sensitivity model is mostly used for sensitive area identification research of typhoon systems. In this application, the ensemble transformation sensitivity model was ported to a high-performance computer, the code was compiled, and a data interface between the ensemble transformation sensitivity model and the convective-scale ensemble forecasting system was developed, realizing offline coupling between the ensemble transformation sensitivity model and the convective-scale ensemble forecasting system.

[0060] Tests show that the system can stably operate the heavy precipitation sensitive area identification method of Example 1, and the identification results are basically reasonable. It can be used for research on heavy precipitation sensitive area identification algorithms and sensitive area characteristics. The constructed heavy precipitation sensitive area identification system can flexibly select ensemble forecast isobaric surface data and their resolutions for calculating heavy precipitation sensitive areas. This lays the foundation for exploring the sensitivity of heavy precipitation sensitive area identification results to different resolutions and different isobaric surface ensemble forecast data, thus providing a theoretical basis for subsequent sensitive area calculation strategies for individual cases of heavy precipitation with different weather types. Simultaneously, the model can calculate the total sensitive area distribution based on the selection of multiple isobaric surface ensemble forecast data, or it can calculate the sensitive area distribution at each layer.

[0061] The precipitation conditions selected for this application and the specific test results are as follows.

[0062] The study selected an extreme rainstorm that occurred in the northern mountainous area of ​​Beijing from 15:00 on July 26 to 02:00 on July 27, 2025. Figure 4 This shows the cumulative precipitation in Beijing from 15:00 on July 26, 2025 to 02:00 on July 27, 2025. Most areas in northern Beijing received over 50 mm of cumulative precipitation, with some areas exceeding 250 mm, mainly distributed in Miyun, Huairou, and Yanqing districts. The citywide average precipitation was 17.7 mm. The highest cumulative precipitation was recorded at Huangtuliang in Miyun District (315.3 mm). This rainfall was extremely heavy, prolonged, and highly localized. Based on this case, a heavy precipitation sensitive area identification experiment was conducted using the method described in Example 1. The specific experimental settings are shown in Table 2. The calculation area is the region corresponding to the obtained ensemble forecast data, and the verification area is the region considered to be prone to heavy precipitation after calculating the probability of heavy precipitation. The analysis time is the adaptive observation time; multiple analysis times were selected for this application, such as... Figure 5 As shown, the dynamic changes in the sensitive areas for heavy precipitation can be displayed. The figure numbering "ets-12-27" is used as an example, signifying that: using the method described in Example 1, namely the method of introducing ensemble transformation sensitivity, based on the convective-scale ensemble forecast system, forecasts were initiated at 20:00 on July 25, 2025 (i.e., the initiation time 2025072520BJT, where BJT represents Beijing time). A 12-hour forecast (i.e., 2025072608BJT) was used as the analysis time, and 27 represents 27 hours from the initiation time of the ensemble forecast (i.e., 2025072623BJT), serving as the verification time. The distribution of sensitive areas at the analysis time (2025072608BJT) when heavy precipitation occurred was calculated using the 12-hour and 27-hour ensemble forecast data.

[0063]

[0064] Based on the identification results of areas sensitive to heavy precipitation, a verification experiment was conducted to confirm the correctness of these areas. Specifically, the observation system simulation experiment described in Example 1 was used. The specific experimental setup is shown in Table 3. The random radiosonde distribution is as follows: Figure 6 As shown, the high-signal area is the heavy precipitation sensitive area identified according to the method of Example 1. Experiment 1 added 10 random radiosondes in the low-signal area (non-sensitive area), Experiment 2 added 10 random radiosondes in the high-signal area (heavy precipitation sensitive area), while the control experiment did not assimilate any simulated observation data. The three sets of experiments were compared to verify the heavy precipitation sensitive area identification method of the present invention.

[0065]

[0066] The verification results are as follows Figure 7 and Figure 8 As shown. Among them, Figure 7 To use the TS rating chart, Figure 8 This is a Bias score chart. It should be noted that the TS score and Bias score are two commonly used methods in this field to evaluate the accuracy of precipitation forecasts. The higher the TS score (ranging from 0 to 1), the higher the forecast accuracy, and the closer the Bias score is to 1, the smaller the forecast deviation.

[0067] Depend on Figure 7 It can be seen that when the precipitation threshold is 25mm and 50mm (generally, a precipitation threshold above 25mm is considered heavy precipitation), the TS score of Experiment 2 (assimilating simulated observation data in a heavy precipitation sensitive area) is higher than that of Experiment 1 (assimilating simulated observation data in a non-sensitive area) and the TS score of the control experiment (not assimilating any observation data). This indicates that Experiment 2 has a higher forecast accuracy and further demonstrates the accuracy of the heavy precipitation sensitive area identification method of the present invention.

[0068] Depend on Figure 8 It can be seen that when the precipitation thresholds are 25mm and 50mm, the Bias score of Experiment 2 is higher than that of Experiment 1 and the control experiment, and is closer to 1. This indicates that the forecast deviation of Experiment 2 is smaller and the underreporting phenomenon is alleviated. It also further demonstrates the accuracy of the heavy precipitation sensitive area identification method of the present invention.

[0069] Based on the evaluation results, it can be found that the heavy precipitation sensitive area identification method and system based on convective-scale ensemble forecasting and ensemble transformation sensitivity of the present invention has high accuracy in heavy precipitation forecasting.

[0070] While specific embodiments of the present invention have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of the present invention is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but all such changes and modifications fall within the scope of protection of the present invention.

Claims

1. A method for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity, characterized in that, Includes the following steps: Based on the convective-scale ensemble forecast system, obtain convectively resolvable ensemble forecast data; The verification area is defined as the area where heavy rainfall occurs. Determine the analysis time and the verification time, and calculate the set perturbation matrix of the analysis time and the verification time, wherein the verification time is the time when heavy precipitation occurs; Select a measurement standard, perturbation variable, and vertical layer suitable for heavy precipitation, initialize the projection matrix, and provide an initial analysis error variance matrix; Calculate the gradient value of the reduction in the forecast error variance of all state variable locations relative to the reduction in the analysis error variance, and select areas where the gradient value exceeds a preset threshold as areas sensitive to heavy precipitation.

2. The method for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity according to claim 1, characterized in that, The horizontal resolution of the convective-scale ensemble forecasting system is 2-4 km.

3. The method for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity according to claim 1, characterized in that, The area where heavy precipitation occurs is the region where the probability of precipitation occurrence obtained by the convective-scale ensemble forecasting system is greater than a predetermined threshold.

4. The method for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity according to claim 1, characterized in that, The metric used is the total energy of dry air, and the disturbance variables include meridional wind, zonal wind, and temperature.

5. The method for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity according to claim 1, characterized in that, The metric used is the total energy of moist air, and the disturbance variables include meridional wind, zonal wind, temperature, and humidity.

6. The method for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity according to claim 1, characterized in that, The vertical layer refers to the key vertical atmospheric stratification corresponding to the heavy precipitation process.

7. The method for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity according to claim 1, characterized in that, The ensemble perturbation matrix is ​​calculated by the difference between the ensemble forecast data and the ensemble mean, where the ensemble mean is the mean of each ensemble member in the ensemble forecast data. The projection matrix is ​​a diagonal matrix, and the diagonal elements of the projection matrix are set according to whether the state variable is located in the verification area and the variable type; The initial analysis error variance matrix is ​​the analysis field error variance matrix after adding a new observation.

8. The method for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity according to claim 1, characterized in that, The method further includes: An observation system simulation experiment was conducted, and numerical experiments were carried out on the heavy precipitation sensitive area to analyze the nonlinear characteristics of the heavy precipitation sensitive area.

9. A system for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity, characterized in that, The system is used to implement the method for identifying heavy precipitation sensitive areas based on convective-scale ensemble forecasting and ensemble transformation sensitivity as described in any one of claims 1 to 8, the system comprising: The data acquisition module is used to acquire convection-resolvable ensemble forecast data based on the convection-scale ensemble forecast system. The verification area determination module is used to determine the verification area, which is the area where heavy precipitation occurs; The perturbation matrix calculation module is used to determine the analysis time and the verification time, and to calculate the set perturbation matrix of the analysis time and the verification time, wherein the verification time is the time when heavy precipitation occurs; The measurement standard selection module is used to select a measurement standard, perturbation variable, and vertical level suitable for heavy precipitation, initialize the projection matrix, and provide an initial analysis error variance matrix. The gradient value calculation module is used to calculate the gradient value of the reduction in the forecast error variance of all state variable locations relative to the reduction in the analysis error variance, and selects areas where the gradient value exceeds a preset threshold as areas sensitive to heavy precipitation.

10. The heavy precipitation sensitive area identification system based on convective-scale ensemble forecasting and ensemble transformation sensitivity according to claim 9, characterized in that, Also includes: The verification module is used to conduct observation system simulation experiments, carry out numerical experiments on the heavy precipitation sensitive area, and analyze the nonlinear characteristics of the heavy precipitation sensitive area.

Images