Method, system, device and medium for short-time severe convective precipitation intensity prediction

Abstract

The present application relates to the field of weather prediction, and specifically relates to a short-time strong convective precipitation intensity prediction method, system, device and medium, comprising: obtaining a plurality of sets of historical meteorological data of a target area; calculating an atmospheric stability index based on atmospheric vertical profile data; constructing a training sample of the atmospheric stability index, ground observation data, radar echo data and precipitation intensity label corresponding to each set of historical meteorological data; inputting the training sample into a pre-constructed ensemble tree model to predict the predicted precipitation intensity; training the ensemble tree model based on the predicted precipitation intensity and the precipitation intensity label to obtain a precipitation intensity prediction model; inputting the meteorological data of the target area into the precipitation intensity prediction model to obtain the short-time strong convective precipitation intensity of the target area. The present application can adjust the precipitation intensity output according to the real stability state of the atmosphere, thereby significantly improving the estimation ability of the strong convective precipitation intensity.

Description

Technical Field

[0001] This invention relates to the field of meteorological forecasting, specifically to a method, system, equipment, and medium for predicting the intensity of short-duration severe convective precipitation. Background Technology

[0002] Short-duration severe convective precipitation refers to precipitation events that occur within tens of minutes to several hours, characterized by high intensity, strong localization, and often accompanied by severe weather phenomena such as thunderstorms, strong winds, and hail. Typical precipitation intensity can reach over 20 millimeters per hour, and even exceed 100 millimeters. These weather events are characterized by their suddenness, rapid evolution, and high destructiveness, easily triggering secondary disasters such as urban flooding, flash floods, landslides, and mudslides, posing a serious threat to people's lives and property, agricultural production, transportation, water conservancy projects, and urban operations. Therefore, accurately predicting the intensity and occurrence area of ​​short-duration severe convective precipitation has become one of the core needs in the field of meteorological disaster prevention and mitigation.

[0003] Currently, the prediction of short-duration severe convective precipitation intensity mainly relies on numerical weather prediction models, radar echo extrapolation techniques, and empirical forecasting methods based on statistics or machine learning. Numerical weather prediction models simulate future weather evolution by solving atmospheric dynamic equations, but their analytical capabilities for convective-scale processes are limited and computational resources are enormous. Radar echo extrapolation methods use radar images from consecutive time intervals to predict the movement and evolution of precipitation systems, but they struggle to reflect changes in atmospheric thermal stratification. Existing machine learning methods attempt to incorporate various meteorological data, but typically only use radar echo characteristics and ground observation parameters as input features. However, these features mainly reflect the distribution of condensate and near-surface conditions at current or historical moments, and the models cannot distinguish the differences in actual precipitation intensity that may occur under the same radar echo intensity, thus limiting the accuracy of severe convective precipitation intensity prediction. Summary of the Invention

[0004] To address the above problems, this invention provides a method, system, equipment, and medium for predicting the intensity of short-duration severe convective precipitation.

[0005] The first aspect of this invention discloses a method for predicting the intensity of short-duration severe convective precipitation, comprising: Acquire multiple sets of historical meteorological data for the target area; wherein each set of historical meteorological data includes atmospheric vertical profile data, ground observation data, radar echo data, and precipitation intensity labels; The atmospheric stability index is calculated based on the atmospheric vertical profile data. Training samples were constructed using the atmospheric stability index, ground observation data, radar echo data, and precipitation intensity labels corresponding to each set of historical meteorological data. The training samples are input into a pre-built ensemble tree model for prediction, and the predicted precipitation intensity corresponding to each training sample is obtained. Based on the predicted precipitation intensity and the precipitation intensity label, the ensemble tree model is trained to obtain a precipitation intensity prediction model; The meteorological data of the target area is input into the precipitation intensity prediction model to obtain the short-term severe convective precipitation intensity of the target area.

[0006] Furthermore, the steps for constructing training samples from the atmospheric stability index, ground observation data, radar echo data, and precipitation intensity labels corresponding to each set of historical meteorological data include: Ground wind speed, ground temperature, and relative humidity are extracted from the ground observation data as ground features; Radar echo intensity, echo top height, and vertical cumulative liquid water content are extracted from the radar echo data as radar features. Training samples were constructed by combining the atmospheric stability index, ground features, radar features, and precipitation intensity labels corresponding to each set of historical meteorological data.

[0007] Furthermore, the steps for calculating the atmospheric stability index based on the atmospheric vertical profile data include: Calculate the atmospheric convective available potential energy of the target region based on the atmospheric vertical profile data. The atmospheric K index of the target area The atmospheric lifting index of the target area ; The atmospheric stability index is calculated using the following formula. : ; in, The preset standard atmospheric convective effective potential energy, The preset standard atmospheric K index, The preset standard atmospheric lifting index, The preset layer stability weighting coefficients, This is the preset weighting coefficient.

[0008] Furthermore, the steps of training the ensemble tree model to obtain a precipitation intensity prediction model include: All training samples are divided into training set and test set according to a preset ratio; The hyperparameters of the ensemble tree model are adjusted based on the training set, and the candidate model is obtained by training with the adjusted hyperparameters. The prediction accuracy index of the candidate model is calculated using the test set; The hyperparameters of the integrated tree model, and / or the stratification stability weight coefficient, and / or the lifting weight coefficient are iteratively adjusted until the prediction accuracy index meets the preset conditions, thus obtaining the precipitation intensity prediction model.

[0009] Furthermore, the step of adjusting the hyperparameters of the ensemble tree model based on the training set includes: The learning rate, tree depth, and number of leaf nodes of the ensemble tree model are adjusted by iteratively validating the training set by dividing it into multiple sub-training sets and validation sets.

[0010] Furthermore, the step of iteratively adjusting the layer stability weight coefficient and / or the lifting weight coefficient includes: By utilizing the deviation between the predicted precipitation intensity and the corresponding precipitation intensity label of the candidate model on the test set, the least squares method is used to update the stratification stability weight coefficient and / or the lifting weight coefficient. Based on the updated stratification stability weight coefficients and / or the updated lifting weight coefficients, the atmospheric stability index is calculated, and the training samples are updated. The candidate model is retrained using the updated training samples, and the prediction accuracy index is calculated again until the prediction accuracy index meets the preset conditions.

[0011] Furthermore, the prediction accuracy index Calculate using the following formula: ; in, This indicates the number of training samples in the test set. This indicates that the candidate model is the first in the test set. Predicted precipitation intensity for each training sample. Indicates the first test set Precipitation intensity labels for each training sample.

[0012] A second aspect of this invention discloses a short-duration severe convective precipitation intensity prediction system, comprising: The acquisition module is used to acquire multiple sets of historical meteorological data for the target area; wherein each set of historical meteorological data includes atmospheric vertical profile data, ground observation data, radar echo data, and precipitation intensity labels; The calculation module is used to calculate the atmospheric stability index based on the atmospheric vertical profile data; The module is used to construct training samples from the atmospheric stability index, ground observation data, radar echo data, and precipitation intensity labels corresponding to each set of historical meteorological data. The prediction module is used to input the training samples into a pre-built ensemble tree model for prediction, and obtain the predicted precipitation intensity corresponding to each training sample. The training module is used to train the ensemble tree model based on the predicted precipitation intensity and the precipitation intensity label to obtain a precipitation intensity prediction model. The input module is used to input meteorological data of the target area into the precipitation intensity prediction model to obtain the short-term severe convective precipitation intensity of the target area.

[0013] A third aspect of the present invention discloses an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the short-duration severe convective precipitation intensity prediction methods disclosed in the first aspect of the present invention.

[0014] The fourth aspect of the present invention discloses a storage medium storing a computer program that, when executed by a processor, implements the steps of any of the short-duration severe convective precipitation intensity prediction methods disclosed in the first aspect of the present invention.

[0015] This invention provides a method for predicting the intensity of short-duration severe convective precipitation. The method acquires multi-source historical meteorological data, including atmospheric vertical profile data, and calculates an atmospheric stability index based on the atmospheric vertical profile data to quantify the degree of atmospheric convective instability. This index, along with ground observation data and radar echo data, forms a training sample used to train an ensemble tree model. Since the atmospheric stability index directly reflects the instability of atmospheric stratification—a higher index indicates greater atmospheric instability and is more conducive to the development of severe convection—incorporating this physically meaningful characteristic into the prediction model is equivalent to adding prior knowledge of atmospheric vertical dynamics and thermodynamic conditions to the model. Compared with existing technologies, the model of this invention can adjust the precipitation intensity output according to the actual atmospheric stability state. That is, even with similar radar echo characteristics, the model can provide differentiated precipitation intensity predictions based on the stability index, thereby significantly improving the ability to predict the intensity of severe convective precipitation. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating a method for predicting the intensity of short-duration severe convective precipitation disclosed in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a short-term severe convective precipitation intensity prediction system disclosed in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device disclosed in the embodiments of the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, or product comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, apparatus, or products.

[0020] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0021] Please see Figure 1 As shown, Figure 1 This is a flowchart illustrating a method for predicting the intensity of short-duration severe convective precipitation disclosed in an embodiment of the present invention. Figure 1 As shown, this method for predicting the intensity of short-duration severe convective precipitation may include the following operations: S101. Acquire multiple sets of historical meteorological data for the target area; wherein each set of historical meteorological data includes atmospheric vertical profile data, ground observation data, radar echo data, and precipitation intensity labels; In an optional embodiment, the step of constructing training samples from the atmospheric stability index, ground observation data, radar echo data, and precipitation intensity labels corresponding to each set of historical meteorological data includes: Ground wind speed, ground temperature, and relative humidity are extracted from the ground observation data as ground features; Radar echo intensity, echo top height, and vertical cumulative liquid water content are extracted from the radar echo data as radar features. Training samples were constructed by combining the atmospheric stability index, ground features, radar features, and precipitation intensity labels corresponding to each set of historical meteorological data.

[0022] This optional embodiment further defines the construction of training samples by extracting surface wind speed, surface temperature, and relative humidity as surface features from surface observation data, and extracting radar echo intensity, echo top height, and vertically accumulated liquid water content as radar features from radar echo data. The atmospheric stability index is then used in conjunction with these features to construct the training samples. Surface wind speed reflects the convergence and shear intensity of the low-level wind field, while surface temperature and relative humidity jointly determine the lower-level water vapor content and potential unstable energy. Radar echo intensity directly characterizes the reflectivity of precipitation particles, echo top height reflects the vertical extension of convection development, and vertically accumulated liquid water content quantitatively estimates the total mass of liquid water in clouds. The atmospheric stability index characterizes stratification instability in the vertical direction, while the aforementioned surface and radar features provide complementary information in the horizontal direction and instantaneous state. The precipitation intensity label is the actual precipitation intensity value observed through surface rain gauges. This optional embodiment organically combines these multi-source features, enabling the training samples to simultaneously contain information on atmospheric thermal stratification, near-surface water vapor dynamics, and the distribution of condensates within clouds. This provides the subsequent ensemble tree model with information-rich and physically meaningful input. Compared to using only partial features, the feature combination defined in this optional embodiment can more comprehensively characterize the precursor signals of short-term severe convection, helping the ensemble tree model learn the nonlinear mapping relationship between different meteorological conditions and precipitation intensity, thereby improving the model's generalization ability and prediction accuracy.

[0023] In an optional embodiment, after acquiring multiple sets of historical meteorological data for the target area, to ensure the validity and consistency of the data, preprocessing is required for the atmospheric vertical profile data, ground observation data, and radar echo data in each set of historical meteorological data. Specifically, firstly, outlier values ​​in parameters such as radar echo intensity and ground wind speed are identified and removed using outlier removal criteria based on statistical distribution. For missing values ​​resulting from the removal, as well as other missing values ​​in the original data, linear interpolation based on adjacent data points is used to fill in the missing values, forming a continuous and complete data sequence. Then, extreme value normalization is used to transform the values ​​of all meteorological parameters to a unified range to eliminate the influence of different physical dimensions on subsequent model training. Finally, spatiotemporal matching processing is performed on the radar echo data, resampling its temporal resolution to an hourly scale consistent with the ground observation data, and resampling its spatial resolution to a unified spatial grid scale, so that the atmospheric vertical profile data, ground observation data, and radar echo data have the same resolution in both time and space. After the above preprocessing operations, the historical meteorological data of each group are comparable and can be used as standardized input for subsequent calculation of atmospheric stability index and construction of training samples.

[0024] S102. Calculate the atmospheric stability index based on the atmospheric vertical profile data; In an optional embodiment, the step of calculating the atmospheric stability index based on the atmospheric vertical profile data includes: Calculate the atmospheric convective available potential energy of the target region based on the atmospheric vertical profile data. The atmospheric K index of the target area The atmospheric lifting index of the target area ; The atmospheric stability index is calculated using the following formula. : ; in, The preset standard atmospheric convective effective potential energy, The preset standard atmospheric K index, The preset standard atmospheric lifting index, The preset layer stability weighting coefficients, This is the preset weighting coefficient.

[0025] In this optional embodiment, the standard atmospheric convective potential energy Standard atmospheric K index Standard atmospheric lifting index The value can be set based on the statistical average or climatological standard value of long-term historical meteorological data. For example, the average convective effective potential energy under conditions of no precipitation or weak precipitation in the target area can be used as the standard value. ; Stratification stability weighting coefficient and the weighting coefficient The stability index can be obtained by conducting sensitivity tests on historical severe convective events or by fitting the data using the least squares method, so that the calculated stability index has the highest correlation with the actual precipitation intensity.

[0026] In this optional embodiment, atmospheric convective potential energy This represents the total work done by buoyancy during the free ascent of the air parcel. It can be calculated by integrating the vertical work done by buoyancy as the air parcel rises from the free convection height along the wet adiabatic line to the equilibrium height, based on the temperature and humidity distribution in the atmospheric vertical profile data. The larger the value, the more abundant the energy in the atmosphere that can be converted into kinetic energy.

[0027] Atmospheric K Index This data comprehensively reflects the temperature decrease, humidity, and dew point differences at key layers such as 850hPa, 700hPa, and 500hPa. It is highly sensitive to thunderstorm potential and can be calculated using the temperatures and dew point temperatures of the three isobaric surfaces (850hPa, 700hPa, and 500hPa) in the atmospheric vertical profile data, according to the following empirical formula: ; in, This represents the atmospheric temperature on the 850 hPa isobaric surface. This represents the atmospheric temperature on the 500 hPa isobaric surface. This represents the dew point temperature on an 850 hPa isobaric surface. This represents the atmospheric temperature on the 700 hPa isobaric surface. This indicates the dew point temperature on the 700 hPa isobaric surface.

[0028] Atmospheric Lifting Index When the value is negative, it indicates that the air parcel is warmer than the ambient air at 500 hPa. The larger the absolute value, the more unstable it is. It can be obtained by raising a specified air parcel from the free convection height along the wet adiabatic line to 500 hPa and calculating the difference between the temperature of the air parcel and the ambient temperature of the isobaric surface.

[0029] This optional embodiment introduces a preset standard value. , , and weighting coefficients , amplified by the exponential term The nonlinear effects deviating from the standard value are coupled together through a product term, linking the contributions of the three indices. This ensures that the final atmospheric stability index retains the physical meaning of each individual index while comprehensively reflecting the synergistic enhancement effect of all three on atmospheric instability. This calculation method, compared to simple linear summation, better reflects the complex nonlinear coupling relationships between energy, stratification, and lifting conditions in severe convective weather, thus providing more sensitive and discriminative feature inputs for subsequent prediction models.

[0030] S103. Construct training samples from the atmospheric stability index, ground observation data, radar echo data, and precipitation intensity labels corresponding to each set of historical meteorological data. In this optional embodiment, for each set of historical meteorological data, firstly, surface wind speed, surface temperature, and relative humidity are extracted from ground observation data as features reflecting lower-level water vapor and dynamic conditions. Secondly, radar echo intensity, echo top height, and vertically accumulated liquid water content—three features that directly describe the distribution and vertical development of condensates within clouds—are extracted from radar echo data. Then, the previously calculated atmospheric stability index is used as another independent feature, combined with the above six features to form a multidimensional prediction feature vector. Finally, this feature vector is paired with the precipitation intensity label corresponding to that set of historical meteorological data to form a complete training sample.

[0031] S104. Input the training samples into the pre-built ensemble tree model for prediction to obtain the predicted precipitation intensity corresponding to each training sample; In this optional embodiment, the ensemble tree model is an ensemble learning model based on decision trees. Its basic idea is to improve overall prediction accuracy and generalization ability by combining the prediction results of multiple decision trees. The ensemble tree model is sequentially fed with the specific values ​​of seven features: atmospheric stability index, surface wind speed, surface temperature, relative humidity, radar echo intensity, echo top height, and vertically accumulated liquid water content. After receiving these inputs, the ensemble tree model performs nonlinear mapping calculations layer by layer according to the existing tree structure within the current model: each decision tree splits downwards from the root node according to the input feature values ​​until it reaches the leaf node; the value stored in the leaf node is the tree's prediction value for that sample. The prediction values ​​of all trees are then integrated according to certain rules to obtain the predicted precipitation intensity corresponding to that sample. In this way, a predicted precipitation intensity value can be output for each training sample. This value is compared with the actual precipitation intensity label stored in the sample to calculate the prediction error, thereby driving the subsequent model parameter update process.

[0032] S105. Based on the predicted precipitation intensity and the precipitation intensity label, the integrated tree model is trained to obtain a precipitation intensity prediction model; In an optional embodiment, the step of training the ensemble tree model to obtain a precipitation intensity prediction model includes: All training samples are divided into training set and test set according to a preset ratio; The hyperparameters of the ensemble tree model are adjusted based on the training set, and the candidate model is obtained by training with the adjusted hyperparameters. The prediction accuracy index of the candidate model is calculated using the test set; The hyperparameters of the integrated tree model, and / or the stratification stability weight coefficient, and / or the lifting weight coefficient are iteratively adjusted until the prediction accuracy index meets the preset conditions, thus obtaining the precipitation intensity prediction model.

[0033] This optional embodiment introduces a closed-loop feedback mechanism that simultaneously optimizes the model's internal hyperparameters and weight coefficients in feature engineering during training. Since the prediction accuracy index quantitatively reflects the candidate model's performance on a test set not used in training, when the accuracy does not meet preset conditions, two types of parameters can be adjusted: the hyperparameters of the ensemble tree model are used to control the model's complexity and learning speed, while the stratification stability weight coefficient and the lifting weight coefficient directly affect the magnitude of the atmospheric stability index, thereby changing the numerical distribution of this feature in the training samples. By alternately or jointly iteratively adjusting these two types of parameters, this optional embodiment enables the model structure and feature representation to synchronously adapt to the statistical laws of actual data, avoiding systematic biases caused by unreasonable feature calculations, and ultimately obtaining a precipitation intensity prediction model that is stable and physically self-consistent on the test set.

[0034] In an optional embodiment, the step of adjusting the hyperparameters of the ensemble tree model based on the training set includes: The learning rate, tree depth, and number of leaf nodes of the ensemble tree model are adjusted by iteratively validating the training set by dividing it into multiple sub-training sets and validation sets.

[0035] As can be seen, this optional embodiment, through a systematic iterative verification strategy, can reliably search for the optimal combination of these three hyperparameters while ensuring computational efficiency. This enables the ensemble tree model to have sufficient capacity to learn the complex nonlinear relationship between multi-source meteorological features and precipitation intensity, while not overly relying on noise in the training data, thereby improving the model's predictive stability for unseen samples.

[0036] In an optional embodiment, the step of iteratively adjusting the layer stability weight coefficient and / or the lifting weight coefficient includes: By utilizing the deviation between the predicted precipitation intensity and the corresponding precipitation intensity label of the candidate model on the test set, the least squares method is used to update the stratification stability weight coefficient and / or the lifting weight coefficient. Based on the updated stratification stability weight coefficients and / or the updated lifting weight coefficients, the atmospheric stability index is calculated, and the training samples are updated. The candidate model is retrained using the updated training samples, and the prediction accuracy index is recalculated until the prediction accuracy index meets the preset conditions.

[0037] Stratification stability weighting coefficient and the weighting coefficient These values ​​are preset based on prior experience. However, convective development conditions vary across different regions and seasons, and fixed coefficients may not accurately reflect the relative contributions of various indices to instability in the actual atmosphere. This optional embodiment introduces a closed-loop calibration mechanism based on prediction bias feedback: the prediction bias of the test set contains information about the insufficient fit between the current atmospheric stability index and precipitation intensity. The least squares method, by minimizing the sum of squares of the differences between the predicted and actual values, can inversely solve for the value that minimizes the bias. and Value selection. Since the atmospheric stability index is directly involved in model prediction as an input feature, its numerical changes will systematically affect the output of the entire model. This optional embodiment uses calibration... and Making the stability index more accurately match the actual precipitation intensity is equivalent to performing adaptive optimization under physical constraints at the feature level. This is more interpretable than simply adjusting the internal parameters of the model, and it can ensure that even if the hyperparameters remain unchanged, the model can still achieve improved accuracy through more reasonable feature representation.

[0038] In an optional embodiment, the prediction accuracy index Calculate using the following formula: ; in, This indicates the number of training samples in the test set. This indicates that the candidate model is the first in the test set. Predicted precipitation intensity for each training sample. Indicates the first test set Precipitation intensity labels for each training sample.

[0039] In this optional embodiment, the mean absolute error reflects the average deviation of the predicted value in absolute terms, with units consistent with precipitation intensity, intuitively representing the baseline deviation of the model's prediction; the root mean square error, by summing and then taking the square root of the squared terms, amplifies the weight of large error samples, enabling sensitive detection of a small number of severely inaccurate predictions; the mean relative error eliminates the influence of dimensions, assessing the percentage deviation of the predicted value from the actual value, particularly suitable for scenarios with large variations in precipitation intensity. This optional embodiment combines these three indicators into a comprehensive index in the form of an equal-weighted sum, taking into account the error characteristics of different dimensions while avoiding the evaluation bias that may result from using a single indicator. Through this comprehensive index, the accuracy of the candidate model can be quantified simultaneously from three perspectives: absolute error, sensitivity to large errors, and relative error. When the prediction accuracy index... When the preset conditions are met, it means that the model has reached an acceptable level in all three indicators, thus ensuring that the final output precipitation intensity prediction model has comprehensive and balanced prediction performance.

[0040] S106. Input the meteorological data of the target area into the precipitation intensity prediction model to obtain the short-term severe convective precipitation intensity of the target area.

[0041] Please see Figure 2 As shown, Figure 2 This is a schematic diagram of a short-duration severe convective precipitation intensity prediction system disclosed in an embodiment of the present invention, comprising: The acquisition module 201 is used to acquire multiple sets of historical meteorological data for the target area; wherein each set of historical meteorological data includes atmospheric vertical profile data, ground observation data, radar echo data, and precipitation intensity labels; Calculation module 202 is used to calculate the atmospheric stability index based on the atmospheric vertical profile data; Module 203 is used to construct training samples from the atmospheric stability index, ground observation data, radar echo data and precipitation intensity labels corresponding to each set of historical meteorological data. Prediction module 204 is used to input the training samples into a pre-built ensemble tree model for prediction, and obtain the predicted precipitation intensity corresponding to each training sample; Training module 205 is used to train the ensemble tree model based on the predicted precipitation intensity and the precipitation intensity label to obtain a precipitation intensity prediction model; Input module 206 is used to input meteorological data of the target area into the precipitation intensity prediction model to obtain the short-term severe convective precipitation intensity of the target area. Specific limitations regarding the short-duration severe convective precipitation intensity prediction system can be found in the limitations of the short-duration severe convective precipitation intensity prediction method described above, and will not be repeated here. Each module in the aforementioned short-duration severe convective precipitation intensity prediction system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the electronic device in hardware format or independent of it, or stored in the memory of the electronic device in software format, so that the processor can call the corresponding operations of each module.

[0042] It should be noted that, in order to highlight the innovative aspects of this invention, this embodiment does not include modules that are not closely related to solving the technical problems proposed by this invention, but this does not mean that there are no other modules in this embodiment.

[0043] like Figure 3 As shown, the electronic device 1 provided by the present invention may include a memory 12, a processor 13 and a bus, and may also include a computer program stored in the memory 12 and executable on the processor 13, such as a short-term severe convective precipitation intensity prediction program.

[0044] The memory 12 includes at least one type of readable storage medium, such as flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 1, such as a portable hard drive. In other embodiments, the memory 12 can be an external storage device of the electronic device 1, such as a plug-in portable hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 1. Furthermore, the memory 12 can include both internal and external storage units of the electronic device 1. The memory 12 can be used not only to store application software and various types of data installed on the electronic device 1, such as code for predicting short-term severe convective precipitation intensity, but also to temporarily store data that has been output or will be output.

[0045] In some embodiments, the processor 13 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 13 is the control unit of the electronic device 1, connecting various components of the electronic device 1 through various interfaces and lines. It executes programs or modules stored in the memory 12 (e.g., short-term severe convective precipitation intensity prediction programs) and calls data stored in the memory 12 to perform various functions and process data of the electronic device 1.

[0046] The processor 13 executes the operating system of the electronic device 1 and various installed applications. The processor 13 executes the applications to implement the steps in the above-described method for predicting the intensity of short-duration severe convective precipitation.

[0047] For example, the computer program may be divided into one or more modules, which are stored in the memory 12 and executed by the processor 13 to complete this application. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the electronic device 1. For example, the computer program may be divided into an acquisition module 201, a calculation module 202, a construction module 203, a prediction module 204, a training module 205, and an input module 206.

[0048] The integrated unit implemented as a software functional module can be stored in a computer-readable storage medium, which can be non-volatile or volatile. The software functional module, stored in the storage medium, includes several instructions to cause a computer device (which may be a personal computer, computer equipment, or network device, etc.) or processor to execute some functions of the short-term severe convective precipitation intensity prediction method described in the various embodiments of this application.

[0049] In summary, the present invention discloses a method, system, device, and medium for predicting short-duration severe convective precipitation intensity. The model of this invention can adjust the precipitation intensity output according to the actual atmospheric stability state. That is, even with similar radar echo characteristics, the model can provide differentiated precipitation intensity predictions based on the stability index, thereby significantly improving the ability to predict severe convective precipitation intensity. Therefore, this invention effectively overcomes the various shortcomings of existing technologies and has high industrial application value.

[0050] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A method for predicting the intensity of short-duration severe convective precipitation, characterized in that, The method includes: Acquire multiple sets of historical meteorological data for the target area; wherein each set of historical meteorological data includes atmospheric vertical profile data, ground observation data, radar echo data, and precipitation intensity labels; The atmospheric stability index is calculated based on the atmospheric vertical profile data. Training samples were constructed using the atmospheric stability index, ground observation data, radar echo data, and precipitation intensity labels corresponding to each set of historical meteorological data. The training samples are input into a pre-built ensemble tree model for prediction, and the predicted precipitation intensity corresponding to each training sample is obtained. Based on the predicted precipitation intensity and the precipitation intensity label, the ensemble tree model is trained to obtain a precipitation intensity prediction model; The meteorological data of the target area is input into the precipitation intensity prediction model to obtain the short-term severe convective precipitation intensity of the target area.

2. The method for predicting the intensity of short-duration severe convective precipitation according to claim 1, characterized in that, The steps for constructing training samples from each set of historical meteorological data, including atmospheric stability index, ground observation data, radar echo data, and precipitation intensity labels, are as follows: Ground wind speed, ground temperature, and relative humidity are extracted from the ground observation data as ground features; Radar echo intensity, echo top height, and vertical cumulative liquid water content are extracted from the radar echo data as radar features. Training samples were constructed by combining the atmospheric stability index, ground features, radar features, and precipitation intensity labels corresponding to each set of historical meteorological data.

3. The method for predicting the intensity of short-duration severe convective precipitation according to claim 1, characterized in that, The steps for calculating the atmospheric stability index based on the atmospheric vertical profile data include: Calculate the atmospheric convective available potential energy of the target region based on the atmospheric vertical profile data. The atmospheric K index of the target area The atmospheric lifting index of the target area ; The atmospheric stability index is calculated using the following formula. : ； in, The preset standard atmospheric convective effective potential energy, The preset standard atmospheric K index, The preset standard atmospheric lifting index, The preset layer stability weighting coefficients, This is the preset weighting coefficient.

4. The method for predicting the intensity of short-duration severe convective precipitation according to claim 3, characterized in that, The steps for training the ensemble tree model to obtain a precipitation intensity prediction model include: All training samples are divided into training set and test set according to a preset ratio; The hyperparameters of the ensemble tree model are adjusted based on the training set, and candidate models are obtained by training with the adjusted hyperparameters. The prediction accuracy index of the candidate model is calculated using the test set; The hyperparameters of the integrated tree model, and / or the stratification stability weight coefficient, and / or the lifting weight coefficient are iteratively adjusted until the prediction accuracy index meets the preset conditions, thus obtaining the precipitation intensity prediction model.

5. The method for predicting the intensity of short-duration severe convective precipitation according to claim 4, characterized in that, The steps for adjusting the hyperparameters of the ensemble tree model based on the training set include: The learning rate, tree depth, and number of leaf nodes of the ensemble tree model are adjusted by iteratively validating the training set by dividing it into multiple sub-training sets and validation sets.

6. The method for predicting the intensity of short-duration severe convective precipitation according to claim 4, characterized in that, The steps of iteratively adjusting the layer stability weight coefficient and / or the lifting weight coefficient include: By utilizing the deviation between the predicted precipitation intensity and the corresponding precipitation intensity label of the candidate model on the test set, the least squares method is used to update the stratification stability weight coefficient and / or the lifting weight coefficient. Based on the updated stratification stability weight coefficients and / or the updated lifting weight coefficients, the atmospheric stability index is calculated, and the training samples are updated. The candidate model is retrained using the updated training samples, and the prediction accuracy index is recalculated until the prediction accuracy index meets the preset conditions.

7. The method for predicting the intensity of short-duration severe convective precipitation according to claim 4, characterized in that, The prediction accuracy index Calculate using the following formula: ； in, This indicates the number of training samples in the test set. This indicates that the candidate model is the first in the test set. Predicted precipitation intensity for each training sample. Indicates the first test set Precipitation intensity labels for each training sample.

8. A short-duration severe convective precipitation intensity prediction system, characterized in that, include: The acquisition module is used to acquire multiple sets of historical meteorological data for the target area; wherein each set of historical meteorological data includes atmospheric vertical profile data, ground observation data, radar echo data, and precipitation intensity labels; The calculation module is used to calculate the atmospheric stability index based on the atmospheric vertical profile data; The module is used to construct training samples from the atmospheric stability index, ground observation data, radar echo data, and precipitation intensity labels corresponding to each set of historical meteorological data. The prediction module is used to input the training samples into a pre-built ensemble tree model for prediction, and obtain the predicted precipitation intensity corresponding to each training sample. The training module is used to train the ensemble tree model based on the predicted precipitation intensity and the precipitation intensity label to obtain a precipitation intensity prediction model. The input module is used to input meteorological data of the target area into the precipitation intensity prediction model to obtain the short-term severe convective precipitation intensity of the target area.

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 computer program, it implements the steps of the short-duration severe convective precipitation intensity prediction method as described in any one of claims 1 to 7.

10. A storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the short-duration severe convective precipitation intensity prediction method as described in any one of claims 1 to 7.

Images