Terrain interpolation assisted rainstorm forecast optimization method and system

Abstract

The invention provides a terrain interpolation-assisted rainstorm forecast optimization method and system, and relates to the technical field of meteorological prediction.The terrain interpolation-assisted rainstorm forecast optimization method comprises the steps that remote sensing data and historical ground observation data of a target area are collected, and the historical ground observation data comprise rainfall data; performing interpolation processing on the remote sensing data by using an interpolation method to generate topographic data in a grid format; performing mask processing on the topographic data in the grid format by using the administrative division vector data of the target area, and extracting topographic information in the target area; dynamically updating the topographic data in the grid format through the real-time ground observation data, and screening topographic features from the updated topographic data in the grid format and the historical ground observation data; and carrying out training processing on the rainstorm forecasting model based on the topographic features, the rainfall data in the historical ground observation data and the topographic information in the target area, and carrying out rainstorm distribution forecasting by using the trained rainstorm forecasting model to realize accurate forecasting of rainstorm distribution.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological forecasting, and in particular to a terrain interpolation-assisted rainstorm forecasting optimization method and system thereof. Background Art

[0002] As an extreme weather phenomenon, rainstorm has a significant impact on human life, social and economic activities and the natural environment. Accurately predicting the distribution and intensity of rainstorms is of great significance for disaster prevention and mitigation, water resources management and ecological environmental protection. However, the occurrence and development of rainstorms are affected by many factors, among which terrain characteristics are an important factor that cannot be ignored. The height, slope, and aspect of the terrain can significantly affect the atmospheric circulation, water vapor transport and precipitation process, thereby affecting the formation and distribution of rainstorms.

[0003] Traditional rainstorm forecasting methods mainly rely on meteorological observation data, numerical weather forecast models and statistical forecasting methods. However, these methods often ignore the impact of terrain characteristics on rainstorm forecasting, resulting in certain errors in the forecast results. In order to improve the accuracy and reliability of rainstorm forecasting, it is necessary to develop a rainstorm forecasting optimization method that can comprehensively consider terrain characteristics and other meteorological and hydrological data.

[0004] Therefore, it is necessary to provide a terrain interpolation-assisted rainstorm forecast optimization method and system to solve the above technical problems. Summary of the invention

[0005] To solve the above technical problems, the present invention provides a terrain interpolation-assisted rainstorm forecast optimization method and system, which not only improves the accuracy and reliability of rainstorm forecast, but also can comprehensively consider multiple influencing factors, providing new technical means and ideas for rainstorm forecast.

[0006] The present invention provides a terrain interpolation-assisted rainstorm forecast optimization method, the optimization method comprising the following steps:

[0007] Collecting remote sensing data and historical ground observation data of a target area, wherein the historical ground observation data includes rainfall data;

[0008] Performing interpolation processing on the remote sensing data by using an interpolation method to generate terrain data in a raster format;

[0009] Using the administrative division vector data of the target area to perform mask processing on the terrain data in the raster format, and extracting the terrain information in the target area;

[0010] Dynamically updating the terrain data in raster format through real-time ground observation data, and filtering terrain features from the updated terrain data in raster format and historical ground observation data, wherein the terrain features represent features that affect rainfall;

[0011] The rainstorm forecast model is trained based on the terrain features, the rainfall data in the historical ground observation data and the terrain information in the target area, and the trained rainstorm forecast model is used to predict the distribution of rainstorm.

[0012] Preferably, the collecting of remote sensing data and historical ground observation data of the target area includes:

[0013] Collecting remote sensing data of the target area, wherein the remote sensing data includes satellite images, radar images and aerial images;

[0014] Collecting historical ground observation data of the target area, wherein the historical ground observation data includes meteorological elements and hydrological data of a specified time series, wherein the meteorological elements include rainfall data, temperature data, and humidity data, and the hydrological data includes surface runoff and soil moisture;

[0015] The collected remote sensing data and the historical ground observation data are imported into a geographic information system.

[0016] Preferably, the interpolation processing of the remote sensing data by using an interpolation method to generate terrain data in a raster format includes:

[0017] Importing the remote sensing data into a geographic information system and preprocessing the imported remote sensing data;

[0018] Performing interpolation processing on the pre-processed remote sensing data using an interpolation algorithm in a geographic information system;

[0019] The remote sensing data after interpolation processing is converted into a raster format to generate terrain data in a raster format.

[0020] Preferably, the interpolation method is one of an inverse distance interpolation method and a spline interpolation method.

[0021] Preferably, the step of using the administrative division vector data of the target area to perform mask processing on the terrain data in raster format and extracting terrain information in the target area includes:

[0022] Importing administrative division vector data of the target area into a geographic information system;

[0023] Using a mask tool in a geographic information system, the administrative division vector data is used as a mask to process the terrain data in a raster format, so as to remove the terrain data outside the target area;

[0024] The terrain information in the target area is extracted from the terrain data processed by the mask by using the extraction tool in the geographic information system, wherein the terrain information includes the terrain height, slope and slope direction.

[0025] Preferably, dynamically updating the terrain data in raster format by using real-time ground observation data, and selecting terrain features from the updated terrain data in raster format and historical ground observation data, comprises:

[0026] Collect and import real-time ground observation data of the target area in real time;

[0027] In the geographic information system, the real-time ground observation data is compared and integrated with the original raster-format terrain data, and the raster-format terrain data is dynamically updated through assimilation technology;

[0028] Combining the updated terrain data in raster format with historical ground observation data, the analysis tools in the geographic information system are used to screen out the terrain features that affect rainfall.

[0029] Preferably, the training process of the rainstorm forecast model based on the terrain features, the rainfall data in the historical ground observation data and the terrain information in the target area, and the prediction of rainstorm distribution using the trained rainstorm forecast model includes:

[0030] Integrate the terrain features, rainfall data from historical ground observation data, and terrain information in the target area in a geographic information system to form comprehensive data;

[0031] Using the integrated data as input features and building a machine learning model, the machine learning model is used to learn the relationship between terrain features and rainfall;

[0032] Dividing the input features into a training set and a validation set, and using the training set to train the machine learning model;

[0033] Using the validation set to perform performance evaluation on the trained machine learning model, and adjusting machine learning model parameters according to the performance evaluation results, wherein the performance evaluation includes accuracy, recall, and F1 score;

[0034] Use the trained and adjusted machine learning model to predict the distribution of heavy rain in the target area and output the prediction results.

[0035] The present invention also provides a terrain interpolation-assisted rainstorm forecast optimization system, which is used to execute the terrain interpolation-assisted rainstorm forecast optimization method. The optimization system includes:

[0036] A data collection module, used for collecting remote sensing data and historical ground observation data of a target area, wherein the historical ground observation data includes rainfall data;

[0037] An interpolation processing module is used to perform interpolation processing on the remote sensing data using an interpolation method to generate terrain data in a raster format;

[0038] A mask processing module, used to perform mask processing on the terrain data in the raster format using the administrative division vector data of the target area, and extract terrain information in the target area;

[0039] A terrain feature acquisition module, used to dynamically update the terrain data in raster format through real-time ground observation data, and filter out terrain features from the updated terrain data in raster format and historical ground observation data, wherein the terrain features represent features that affect rainfall;

[0040] The rainstorm prediction module is used to train the rainstorm forecast model based on the terrain features, the rainfall data in the historical ground observation data and the terrain information in the target area, and use the trained rainstorm forecast model to predict the distribution of rainstorm.

[0041] Compared with the related art, the terrain interpolation-assisted rainstorm forecast optimization method and system provided by the present invention have the following beneficial effects:

[0042] The present invention collects remote sensing data and historical ground observation data of the target area, generates terrain data in raster format using interpolation, and performs mask processing and terrain information extraction in combination with administrative division vector data. The terrain data is dynamically updated through real-time ground observation data, and the terrain features that affect rainfall are screened out. Finally, the rainstorm forecast model is trained and optimized based on the terrain features, historical rainfall data, and terrain information in the target area to achieve accurate prediction of rainstorm distribution. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flow chart of a terrain interpolation-assisted rainstorm forecast optimization method provided by the present invention;

[0044] Figure 2 A module structure diagram of a terrain interpolation-assisted rainstorm forecast optimization system provided by the present invention. DETAILED DESCRIPTION

[0045] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only the parts related to the present invention, rather than all structures, are shown in the accompanying drawings. In addition, the embodiments of the present invention and the features in the embodiments may be combined with each other without conflict.

[0046] It should also be noted that, for ease of description, only the part relevant to the present invention but not all content is shown in the accompanying drawings. It should be mentioned before discussing exemplary embodiments in more detail that some exemplary embodiments are described as processing or methods depicted as flow charts. Although the flow chart describes each operation (or step) as sequential processing, many operations therein can be implemented in parallel, concurrently or simultaneously. In addition, the order of each operation can be rearranged. When its operation is completed, the processing can be terminated, but it can also have additional steps not included in the accompanying drawings. The processing can correspond to methods, functions, procedures, subroutines, subprograms, etc.

[0047] Embodiment 1

[0048] The present invention provides a terrain interpolation-assisted rainstorm forecast optimization method, referring to Figure 1 As shown, the optimization method includes the following steps:

[0049] S1: Collect remote sensing data and historical ground observation data of the target area, where the historical ground observation data includes rainfall data.

[0050] Specifically, step S1 includes the following steps:

[0051] S11: Collect remote sensing data of the target area, where the remote sensing data includes satellite images, radar images, and aerial images.

[0052] In this embodiment, remote sensing data is surface information obtained from high altitude or outer space, which can be used to monitor and study various phenomena on the earth's surface. For rainstorm forecasting, remote sensing data provides a large range of continuous spatial coverage, which helps to understand the surface changes before and after precipitation. Remote sensing data mainly includes satellite images, radar images, and aerial images.

[0053] When collecting data, satellites or other remote sensing platforms with appropriate spatial resolution, temporal resolution, and spectral resolution should be selected based on the characteristics of the target area and the meteorological conditions that need to be monitored.

[0054] Regularly obtain the latest remote sensing image data from relevant agencies (such as the National Meteorological Administration, International Space Station, etc.) to ensure the timeliness and accuracy of the data.

[0055] Establish a special data warehouse or database to store the collected remote sensing data to facilitate subsequent data processing and analysis.

[0056] S12: Collect historical ground observation data of the target area. The historical ground observation data include meteorological elements and hydrological data of a specified time series. The meteorological elements include rainfall data, temperature data, and humidity data. The hydrological data include surface runoff and soil moisture.

[0057] In this embodiment, historical ground observation data refers to meteorological and hydrological data recorded at a specific location over a period of time in the past. These data are crucial for understanding regional climate characteristics and analyzing the historical laws of extreme weather events.

[0058] During the collection process, meteorological stations are rationally arranged according to the geographical distribution of the target area to ensure that representative meteorological element data can be collected, including but not limited to meteorological elements such as rainfall, temperature, humidity, wind speed, and hydrological data such as surface runoff and soil moisture.

[0059] The collected data are cleaned, verified and standardized to eliminate the impact of outliers and missing values ​​and ensure data quality. At the same time, historical data provides important training samples for the rainstorm forecast model, helping the model to better learn the regular characteristics of rainstorm occurrence.

[0060] S13: Import the collected remote sensing data and historical ground observation data into the geographic information system.

[0061] Specifically, Geographic Information System (GIS) is a spatial data analysis tool that can effectively manage and analyze information with spatial attributes. Integrating remote sensing data and ground observation data into GIS can achieve spatial visualization of data and facilitate spatial analysis.

[0062] During implementation, the interface provided by the GIS software is used to import the processed remote sensing data and ground observation data into the system. In the GIS environment, data from different sources are spatially aligned and superimposed to facilitate spatial analysis. The visualization function of GIS is used to generate various maps and charts to intuitively display the spatial distribution of the data.

[0063] Geographic information systems can promote the effective integration of multi-source data and provide a unified basic platform for subsequent analysis and processing. In addition, spatial visualization capabilities enhance the interpretability of data and help researchers quickly discover spatial patterns in the data.

[0064] S2: Use interpolation method to interpolate remote sensing data and generate terrain data in raster format.

[0065] Specifically, step S2 includes the following steps:

[0066] S21: Import remote sensing data into the geographic information system and pre-process the imported remote sensing data.

[0067] In the rainstorm forecast optimization method, remote sensing data (such as satellite images, radar images, etc.) provide a large geographical coverage, which can capture the slight changes in the surface and the evolution of weather systems. However, the original remote sensing data usually contains noise or has certain uncertainties due to the influence of sensor characteristics and atmospheric conditions, so it needs to be preprocessed to improve the accuracy of subsequent analysis.

[0068] Use GIS software, including but not limited to ArcGIS or QGIS, to import various types of remote sensing data collected. This data may come from different sources such as weather satellites, radar stations, drones, etc.

[0069] Then the imported data is geometrically corrected to eliminate image deformation caused by factors such as the earth's curvature and sensor tilt. At the same time, radiation correction is performed to correct radiation errors caused by atmospheric scattering and absorption.

[0070] If there are data from multiple sources, data fusion is required to ensure that the data from different sources are comparable under the same spatial reference system.

[0071] Apply filtering techniques to remove noise from images, including but not limited to median filtering, mean filtering, etc.

[0072] By adjusting the image contrast and brightness, certain features can be enhanced to make subsequent analysis more effective.

[0073] The preprocessed remote sensing data removes unnecessary interference factors, making the image features more obvious and helping to improve the accuracy of subsequent terrain interpolation.

[0074] S22: Use the interpolation algorithm in the geographic information system to interpolate the pre-processed remote sensing data.

[0075] Interpolation is a method of estimating data at unknown points based on data from known points. For remote sensing data, interpolation can be used to fill in the gaps between sample points to generate a continuous terrain surface.

[0076] When implementing, choose a suitable interpolation algorithm based on the data characteristics and research needs. For example, you can use the inverse distance weighted (IDW) interpolation method, which is suitable for situations where the data distribution is relatively uniform; or use spline interpolation, which is suitable for situations where a smooth surface is required.

[0077] During the interpolation process, the selected interpolation algorithm is applied in the GIS software to generate a continuous surface based on the preprocessed remote sensing data points. The parameters of the interpolation algorithm, including the search radius, weight factor, etc., are adjusted to achieve the best interpolation effect. Finally, the quality of the interpolation result is checked to ensure that there are no outliers or unreasonable changes.

[0078] The raster data obtained after interpolation processing can more accurately reflect the actual terrain characteristics and provide a basis for subsequent terrain analysis and heavy rain forecasting.

[0079] S23: Convert the interpolated remote sensing data into a raster format to generate terrain data in a raster format.

[0080] Raster data is a data structure that divides geographic space into regular grids, making it easier for computers to process and analyze.

[0081] During conversion, determine the resolution, origin coordinates, number of rows and columns and other properties of the raster. In the GIS software, convert the interpolated vector data into a raster format. For each raster cell, assign the corresponding attribute value (such as elevation) and store the generated raster data as a file for subsequent use.

[0082] After conversion into raster format, terrain data is more suitable for spatial analysis, such as terrain feature extraction, slope calculation, etc., thus providing more accurate support for heavy rain forecasting.

[0083] S3: Use the administrative division vector data of the target area to mask the terrain data in raster format and extract the terrain information in the target area.

[0084] Specifically, step S3 includes the following steps:

[0085] S31: Import administrative division vector data of the target area into the geographic information system.

[0086] In the process of optimizing rainstorm forecast, it is crucial to determine the forecast range. Due to the complexity and diversity of the natural environment, there may be significant topographic differences in different administrative regions, which will affect the rainfall pattern and its distribution. Therefore, before making a rainstorm forecast, it is necessary to first clarify the specific boundaries of the study area so that subsequent data processing and analysis are limited to this specific range. Administrative division vector data provides precise geographic boundary definitions, which is helpful for subsequent terrain analysis and rainfall forecasting.

[0087] This step involves importing the administrative division information of the target area into the Geographic Information System (GIS) software in the form of vector data. Vector data is a way of storing geographic information that can accurately represent the location, shape, and attributes of terrain features such as points, lines, and surfaces. In actual operation, researchers will obtain the latest administrative division electronic map files of the area (such as .shp files) and load these files into the system through the data import function provided by the GIS platform. After loading, the vector data will be displayed in a visual form on the GIS platform, clearly marking the boundaries of the study area.

[0088] By importing administrative division vector data, we can ensure that all subsequent processing processes are strictly limited to the target area, avoiding misanalysis of irrelevant peripheral areas. This not only improves the accuracy of rainstorm forecasts, but also simplifies the data processing process and reduces the consumption of computing resources.

[0089] S32: Using the mask tool in the geographic information system, the administrative division vector data is used as a mask to process the terrain data in raster format to remove the terrain data outside the target area.

[0090] After importing the vector data in the previous step, the next task is to clip the terrain data in raster format to ensure that only the data in the target area is retained for further analysis. This is because if the data is not clipped, noise or irrelevant data may be introduced, affecting the accuracy of the final forecast.

[0091] At this stage, the appropriate masking tool will be selected in the GIS platform, using the previously imported vector data as a clipping template to perform a clipping operation on the raster format terrain data. This means that all terrain data that falls outside the study area will be removed, leaving only the parts that meet the boundary conditions. This process usually involves spatial analysis functions, and the GIS platform will automatically identify which parts need to be retained and which parts should be deleted.

[0092] Masking effectively removes redundant terrain data, ensuring the pertinence and effectiveness of subsequent analysis. In addition, this method also improves data consistency because all analyses will be based on the same research area.

[0093] S33: Using extraction tools in the geographic information system, extracting terrain information in the target area from the masked terrain data, wherein the terrain information includes terrain height, slope and slope direction.

[0094] After importing vector data and clipping raster data, the next step is to extract key terrain information. This information is crucial to understanding how terrain affects precipitation patterns, especially in the application scenario of heavy rain forecasting.

[0095] In this step, the extraction tools on the GIS platform will be used to extract specific terrain attributes, such as terrain height, slope, and aspect, from the already clipped raster data. These attributes can be obtained through different GIS tools, such as the terrain analysis module. By setting the corresponding parameters, a dataset that accurately reflects the terrain characteristics of the study area can be obtained.

[0096] The extracted terrain information provides important input variables for subsequent rainstorm forecast modeling, helping the model to better understand and simulate the rainfall process. In addition, this data can also be used to calibrate the model and improve the accuracy of the forecast.

[0097] S4: Dynamically update the terrain data in raster format through real-time ground observation data, and screen out terrain features from the updated terrain data in raster format and historical ground observation data, wherein the terrain features represent features that affect rainfall.

[0098] Specifically, step S4 includes the following steps:

[0099] S41: Collect and import real-time ground observation data of the target area in real time.

[0100] In rainstorm forecasting, the importance of real-time data is self-evident. Due to the rapid changes in weather systems, especially the development of heavy precipitation systems, real-time data can provide the latest ground conditions, which is crucial to improving forecast accuracy. Therefore, real-time collection and import of ground observation data in the target area has become an indispensable part of the rainstorm forecasting system.

[0101] In this step, a network of multiple automatic weather stations is established. These stations are distributed in the target area and can monitor meteorological elements such as rainfall, temperature, and wind speed in real time.

[0102] Wireless transmission technology (including but not limited to GPRS and LoRa) is used to transmit the data collected by each site to the central server in real time. The central server receives the data from each site and imports it into the geographic information system (GIS) platform for subsequent processing.

[0103] Real-time monitoring is achieved to ensure the freshness of data and improve the temporal resolution of the forecast model.

[0104] After the data is imported into GIS, it is easy to integrate and analyze with other spatial data (such as terrain data), which helps to better understand the interaction between weather systems and terrain.

[0105] S42: In the geographic information system, the real-time ground observation data is compared and integrated with the original terrain data in raster format, and the terrain data in raster format is dynamically updated through assimilation technology.

[0106] In order to make the forecast model more accurate, it is necessary to combine the latest ground observation data with the existing terrain data to obtain a more accurate description of the ground state. This process usually involves the application of data assimilation technology.

[0107] In this step, spatial analysis capabilities in GIS are used to compare real-time observation data with existing terrain data and identify differences between the two.

[0108] Apply data assimilation algorithms (including but not limited to Kalman filtering and particle filtering) to integrate new observation data into existing terrain data to achieve dynamic updating. The quality and reliability of the data need to be considered during the updating process to ensure that the newly added data does not reduce the accuracy of the overall data set.

[0109] Through assimilation technology, dynamic updating of data is achieved, which enhances the forecast model's ability to reflect current ground conditions. At the same time, it improves the consistency between terrain data and actual ground conditions, providing a more reliable basis for subsequent forecasts.

[0110] S43: Combine the updated terrain data in raster format with historical ground observation data and use the analysis tools in the geographic information system to screen out terrain features that affect rainfall.

[0111] In rainstorm forecasting, understanding which terrain features affect rainfall patterns is crucial to improving forecast accuracy. Analyzing the relationship between terrain features and rainfall can help us better understand the precipitation process.

[0112] By integrating updated terrain data and historical observation data on the GIS platform, using statistical analysis tools to explore the correlation between the two, and adopting data mining techniques (including but not limited to decision trees and random forests) to identify terrain features that significantly affect rainfall distribution, and visually displaying the selected features, we can intuitively understand the changing patterns of rainfall patterns under different terrain conditions.

[0113] S5: The heavy rain forecast model is trained based on the terrain features, the rainfall data in the historical ground observation data and the terrain information in the target area, and the heavy rain forecast model is used to predict the heavy rain distribution.

[0114] Specifically, step S5 includes the following steps:

[0115] S51: Integrate terrain features, rainfall data from historical ground observation data, and terrain information in the target area into the geographic information system to form comprehensive data.

[0116] The construction of a rainstorm forecast model requires a large amount of historical data as training samples, including topographic features, past rainfall records, and other relevant meteorological and hydrological information. These data together form the learning basis of the forecast model, helping the model understand the possibility and intensity of rainstorms under different conditions.

[0117] First, all terrain features extracted in the previous steps are integrated on the GIS platform, including terrain height, slope, aspect and other information.

[0118] Integrate rainfall data obtained from historical ground observations to ensure that these data are consistent in time and space with topographic features.

[0119] The topographic information in the target area is combined with the above datasets to form a comprehensive dataset, which contains the temporal and spatial distribution of topographic features and the corresponding rainfall records.

[0120] Preprocess the integrated data, including data cleaning, standardization, and feature engineering, to ensure data quality and availability.

[0121] S52: Use the synthetic data as input features and build a machine learning model, which is used to learn the relationship between terrain features and rainfall.

[0122] In the field of rainstorm forecasting, traditional statistical methods may have difficulty capturing the complex nonlinear relationship between terrain features and rainfall. As a powerful nonlinear model, neural networks can reveal this complex relationship through multi-level abstraction and feature learning. Therefore, neural networks are chosen as the basic model to build a rainstorm forecasting model.

[0123] Specific implementation content:

[0124] Data preparation: Ensure that all features in the comprehensive dataset are of numerical type. For categorical features (such as terrain type), one-hot encoding conversion is required.

[0125] The data is normalized to ensure that features of different dimensions are numerically comparable, which can also accelerate the convergence of the neural network.

[0126] Model design:

[0127] Input layer: The number of input nodes is equal to the number of features in the synthetic dataset.

[0128] Hidden layers: Multiple hidden layers are designed to capture deep feature relationships. Each layer can contain different numbers of nodes. ReLU activation function is used between hidden layers to increase the nonlinear expression ability of the model.

[0129] Output layer: For the rainfall prediction task, the output layer is a single node, which uses a linear activation function (i.e., the identity function) to directly predict rainfall.

[0130] Loss function: The mean squared error (MSE) is chosen as the loss function because it is suitable for regression tasks and can effectively measure the gap between the predicted value and the true value.

[0131] Optimizer: Choose the Adam optimizer or other adaptive learning rate method to speed up convergence and avoid local minima.

[0132] Model implementation:

[0133] Use a deep learning framework such as TensorFlow or PyTorch to implement the neural network model.

[0134] Write a training script and set hyperparameters such as the number of training epochs and batch size.

[0135] Implement an early stopping mechanism to terminate training early when the performance on the validation set no longer improves to prevent overfitting.

[0136] Model training:

[0137] Use the training set data to train the neural network model and adjust the weights and bias terms through the back-propagation algorithm.

[0138] Regularly use the validation set to evaluate model performance and monitor the loss trend during training.

[0139] S53: Divide the input features into a training set and a validation set, and use the training set to train the machine learning model.

[0140] In order to ensure the generalization ability of the machine learning model, the dataset needs to be divided into a training set and a validation set, which are used for model training and performance evaluation respectively.

[0141] Specific implementation content:

[0142] A random sampling method is used to divide the comprehensive dataset into two parts according to a certain ratio (such as 70% training set and 30% validation set).

[0143] The model is trained using the training set data, and the model parameters are gradually adjusted to minimize the loss function.

[0144] During the training process, validation set data is regularly used to evaluate model performance to ensure that the model can perform well on unseen data.

[0145] S54: Use the validation set to perform performance evaluation on the trained machine learning model and adjust the machine learning model parameters based on the results of the performance evaluation, where the performance evaluation includes accuracy, recall, and F1 score.

[0146] After the model training is completed, the performance of the model needs to be fully evaluated to ensure that it can work reliably in practical applications.

[0147] Specific implementation content:

[0148] Use the validation set data divided previously to test the trained model and calculate the difference between the model prediction results and the actual observed values.

[0149] Evaluation indicators include but are not limited to accuracy, recall, F1 score, etc. These indicators can comprehensively measure the predictive performance of the model.

[0150] According to the evaluation results, adjust the model parameters, such as changing the learning rate, increasing or decreasing the number of hidden layer nodes, etc., in order to improve the model performance.

[0151] S55: Use the trained and adjusted machine learning model to predict the distribution of heavy rain in the target area and output the prediction results.

[0152] After all the previous steps, we now have a fully trained and optimized machine learning model that can be applied to the task of heavy rain forecasting in the target area.

[0153] Specific implementation content:

[0154] Collect the current terrain feature data of the target area and input it into the trained model for prediction.

[0155] Output the model's predictions, including a map of the expected rainfall distribution.

[0156] Analyze the prediction results, compare them with the actual observation data, and evaluate the actual forecasting effect of the model.

[0157] Embodiment 2

[0158] The present invention also provides a terrain interpolation-assisted rainstorm forecast optimization system for executing a terrain interpolation-assisted rainstorm forecast optimization method, referring to Figure 2 As shown, the optimization system includes:

[0159] The data collection module 100 is used to collect remote sensing data and historical ground observation data of the target area, wherein the historical ground observation data includes rainfall data.

[0160] The interpolation processing module 200 is used to perform interpolation processing on the remote sensing data using an interpolation method to generate terrain data in a raster format.

[0161] The mask processing module 300 is used to perform mask processing on the terrain data in raster format using the administrative division vector data of the target area, and extract the terrain information in the target area.

[0162] The terrain feature acquisition module 400 is used to dynamically update the terrain data in raster format through real-time ground observation data, and filter out terrain features from the updated terrain data in raster format and historical ground observation data, wherein the terrain features represent the features that affect rainfall.

[0163] The rainstorm prediction module 500 is used to train the rainstorm forecast model based on the terrain characteristics, rainfall data in the historical ground observation data and the terrain information in the target area, and use the trained rainstorm forecast model to predict the distribution of rainstorm.

[0164] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0165] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0166] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. A terrain interpolation-assisted rainstorm forecast optimization method, characterized in that: The optimization method comprises the following steps: Collecting remote sensing data and historical ground observation data of a target area, wherein the historical ground observation data includes rainfall data; Performing interpolation processing on the remote sensing data by using an interpolation method to generate terrain data in a raster format; Using the administrative division vector data of the target area to perform mask processing on the terrain data in the raster format, and extracting the terrain information in the target area; Dynamically updating the terrain data in raster format through real-time ground observation data, and filtering terrain features from the updated terrain data in raster format and historical ground observation data, wherein the terrain features represent features that affect rainfall; The rainstorm forecast model is trained based on the terrain features, the rainfall data in the historical ground observation data and the terrain information in the target area, and the trained rainstorm forecast model is used to predict the distribution of rainstorm.

2. The terrain interpolation-assisted rainstorm forecast optimization method according to claim 1, characterized in that: The remote sensing data and historical ground observation data collected in the target area include: Collecting remote sensing data of the target area, wherein the remote sensing data includes satellite images, radar images and aerial images; Collecting historical ground observation data of the target area, wherein the historical ground observation data includes meteorological elements and hydrological data of a specified time series, wherein the meteorological elements include rainfall data, temperature data, and humidity data, and the hydrological data includes surface runoff and soil moisture; The collected remote sensing data and the historical ground observation data are imported into a geographic information system.

3. The terrain interpolation-assisted rainstorm forecast optimization method according to claim 2, characterized in that: The method of interpolating the remote sensing data to generate terrain data in a raster format includes: Importing the remote sensing data into a geographic information system and preprocessing the imported remote sensing data; Performing interpolation processing on the pre-processed remote sensing data using an interpolation algorithm in a geographic information system; The remote sensing data after interpolation processing is converted into a raster format to generate terrain data in a raster format.

4. The terrain interpolation-assisted rainstorm forecast optimization method according to claim 3 is characterized in that: The interpolation method is selected from one of the inverse distance interpolation method and the spline interpolation method.

5. The terrain interpolation-assisted rainstorm forecast optimization method according to claim 1, characterized in that: The step of using the administrative division vector data of the target area to perform mask processing on the terrain data in raster format and extracting terrain information in the target area includes: Importing administrative division vector data of the target area into a geographic information system; Using a mask tool in a geographic information system, the administrative division vector data is used as a mask to process the terrain data in a raster format, so as to remove the terrain data outside the target area; The terrain information in the target area is extracted from the terrain data processed by the mask by using the extraction tool in the geographic information system, wherein the terrain information includes the terrain height, slope and slope direction.

6. A terrain interpolation-assisted rainstorm forecast optimization method according to claim 5, characterized in that: The method of dynamically updating the terrain data in raster format by using real-time ground observation data, and selecting terrain features from the updated terrain data in raster format and historical ground observation data, includes: Collect and import real-time ground observation data of the target area in real time; In the geographic information system, the real-time ground observation data is compared and integrated with the original raster-format terrain data, and the raster-format terrain data is dynamically updated through assimilation technology; Combining the updated terrain data in raster format with historical ground observation data, the analysis tools in the geographic information system are used to screen out the terrain features that affect rainfall.

7. A terrain interpolation-assisted rainstorm forecast optimization method according to claim 6, characterized in that: The training process of the rainstorm forecast model based on the terrain features, the rainfall data in the historical ground observation data and the terrain information in the target area, and the prediction of rainstorm distribution using the trained rainstorm forecast model includes: Integrate the terrain features, rainfall data from historical ground observation data, and terrain information in the target area in a geographic information system to form comprehensive data; Using the integrated data as input features and building a machine learning model, the machine learning model is used to learn the relationship between terrain features and rainfall; Dividing the input features into a training set and a validation set, and using the training set to train the machine learning model; Using the validation set to perform performance evaluation on the trained machine learning model, and adjusting machine learning model parameters according to the performance evaluation results, wherein the performance evaluation includes accuracy, recall, and F1 score; Use the trained and adjusted machine learning model to predict the distribution of heavy rain in the target area and output the prediction results.

8. A terrain interpolation-assisted rainstorm forecast optimization system, used to execute a terrain interpolation-assisted rainstorm forecast optimization method as claimed in any one of claims 1 to 7, characterized in that: The optimization system comprises: A data collection module, used for collecting remote sensing data and historical ground observation data of a target area, wherein the historical ground observation data includes rainfall data; An interpolation processing module is used to perform interpolation processing on the remote sensing data using an interpolation method to generate terrain data in a raster format; A mask processing module, used to perform mask processing on the terrain data in the raster format using the administrative division vector data of the target area, and extract terrain information in the target area; A terrain feature acquisition module, used to dynamically update the terrain data in raster format through real-time ground observation data, and filter out terrain features from the updated terrain data in raster format and historical ground observation data, wherein the terrain features represent features that affect rainfall; The rainstorm prediction module is used to train the rainstorm forecast model based on the terrain features, the rainfall data in the historical ground observation data and the terrain information in the target area, and use the trained rainstorm forecast model to predict the distribution of rainstorm.

Images