Convective heavy rainfall monitoring and early warning method and system based on satellite remote sensing

Abstract

The invention relates to the technical field of disaster monitoring, in particular to a convective heavy rainfall monitoring and early warning method and system based on satellite remote sensing, and the method comprises the following steps: calculating the cloud top brightness temperature change of a convective cloud cluster in image data based on the detection data of GPM, FY-3G and H8 satellites, analyzing the time sequence of the cloud top brightness temperature according to the cloud cluster dynamic tracking data, and calculating the time sequence of the cloud top brightness temperature; and obtaining a cloud top temperature feature set. According to the method, through the time sequence analysis of the cloud top brightness temperature, tiny meteorological changes can be rapidly captured, the early-stage potential rainfall risk identification capability is enhanced, the classification process of weather conditions is optimized through the introduction of principal component analysis, the risk assessment accuracy is improved, and through the application of the Kalman filter, the risk assessment accuracy is improved. The dynamic tracking of the cloud cluster is more accurate, the VAR model provides higher reliability in predicting the future meteorological threshold, the deep Q network is combined with the real-time satellite data, the early warning information is updated in real time, and an instant and effective disaster prevention countermeasure is provided.

Description

Technical Field

[0001] The present invention relates to the field of disaster monitoring technology, and in particular to a convective heavy rainfall monitoring and early warning method and system based on satellite remote sensing. Background Art

[0002] The field of disaster monitoring technology focuses on applying various methods to predict, detect, assess and respond to natural and man-made disasters, including the use of ground sensors, satellite remote sensing, aircraft reconnaissance and other telemetry technologies to monitor environmental and meteorological conditions in real time. Disaster monitoring technology can provide critical data to help predict the likelihood and severity of disasters such as floods, earthquakes, fires and storms. The data is essential for developing effective emergency response plans to reduce casualties and property losses.

[0003] Among them, the convective heavy precipitation monitoring and early warning method focuses on the detection and early warning of convective heavy precipitation events, such as heavy rain, thunderstorms, etc. It uses satellite images and data from ground meteorological stations to predict the occurrence of heavy precipitation by analyzing the morphology, temperature and movement trends of clouds. Its main purpose is to provide accurate weather forecasts and alarms in a timely manner so as to take appropriate preventive measures to avoid or reduce the impact of floods and other related disasters caused by sudden heavy precipitation.

[0004] The data update frequency and real-time data processing capabilities of existing technologies limit their effectiveness in the rapid development stage of emergency meteorological events. The lack of efficient data fusion and analysis methods makes the information extracted from ground sensors and basic satellite data insufficient to support accurate decision-making. This limitation results in the early warning system's inability to respond in a timely manner and unable to provide sufficient preparation time, increasing the risks and losses caused by slow response. Existing methods also have deficiencies in the accuracy of multi-dimensional data analysis and risk prediction, which is particularly prominent when responding to sudden and rapidly developing weather disasters, resulting in untimely implementation or low effectiveness of disaster prevention and mitigation measures. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a convective heavy rainfall monitoring and early warning method and system based on satellite remote sensing.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme: a convective heavy rainfall monitoring and early warning method based on satellite remote sensing, comprising the following steps:

[0007] S1: Based on the detection data of GPM, FY-3G and H8 satellites, the cloud top brightness temperature changes of convective clouds in the image data are calculated. According to the cloud cluster dynamic tracking data, the time series of cloud top brightness temperature is analyzed to obtain the cloud top temperature feature set;

[0008] S2: Based on the cloud top temperature feature set, the time series data of cloud parameters, including particle radius and cloud water path, are analyzed through ERA-5 reanalysis data, and compared with historical threshold data to obtain a time series parameter analysis set;

[0009] S3: Based on the time series parameter analysis set, principal component analysis is used to classify the current weather situation, identify categories with high precipitation potential, and output climate risk categories;

[0010] S4: Based on the climate risk categories, a Kalman filter is used to analyze the tracked clouds, calculate the changes in optical thickness and particle radius, evaluate the potential impact of heavy precipitation, and obtain cloud impact indicators;

[0011] S5: Based on the cloud impact indicators, the VAR model is used to calculate the key precipitation thresholds of cloud top brightness and optical thickness, and the future thresholds of key parameters are predicted to establish a precipitation threshold prediction model;

[0012] S6: Based on the precipitation threshold prediction model and recent satellite data, the deep Q network is used to analyze the potential and intensity of heavy precipitation events, and the warning level is adjusted according to the ground conditions to generate warning decision information.

[0013] As a further solution of the present invention, the cloud top temperature feature set includes cloud top temperature peak value, temperature change rate, and tracking coordinates; the climate risk category includes precipitation potential level, risk occurrence frequency, and climate trend analysis results; the cloud impact indicators include optical property distribution, dynamic change speed, and influence rating; the early warning decision information includes early warning signal type, response time window, and early warning effectiveness evaluation results.

[0014] As a further solution of the present invention, based on the detection data of GPM, FY-3G and H8 satellites, the cloud top brightness temperature changes of convective clouds in the image data are calculated, and according to the cloud cluster dynamic tracking data, the time series of cloud top brightness temperature is analyzed to obtain the cloud top temperature feature set. Specifically, the steps are as follows:

[0015] S101: Collect the detection data of GPM, FY-3G and H8 satellites, integrate the data, unify the data format and clean the data, and verify the data quality to obtain the cloud top brightness temperature data set;

[0016] S102: Based on the cloud top brightness temperature data set, extract cloud cluster dynamics, calculate the brightness temperature differences at consecutive time points to construct a time series, use a sliding window method to smooth the data, and identify key change points to obtain a temperature series set;

[0017] S103: Based on the temperature sequence set, analyze the cloud top temperature variation characteristics, identify the temperature extremes by using the temperature range segmentation, determine the cloud top key temperature index, and generate the cloud top temperature feature set.

[0018] As a further solution of the present invention, based on the cloud top temperature feature set, the time series data of cloud parameters, including particle radius and cloud water path, are analyzed through ERA-5 reanalysis data, and compared with historical threshold data to obtain the time series parameter analysis set. Specifically, the steps are as follows:

[0019] S201: Based on the cloud top temperature feature set and combined with ERA-5 reanalysis data, analyze the cloud parameter time series, monitor the change trend of cloud parameters, including particle size and cloud water path, and generate a cloud parameter time series set;

[0020] S202: Based on the cloud parameter time series set, compare the historical thresholds, analyze and identify data points that exceed the normal range by setting the thresholds, identify abnormal values, and obtain time series abnormality indicators;

[0021] S203: Based on the time series anomaly index, the multi-source data results are merged, the parameter analysis process is continuously optimized, and a time series parameter analysis set is generated.

[0022] As a further solution of the present invention, based on the time series parameter analysis set, principal component analysis is used to classify the current weather situation, identify categories with high precipitation potential, and output the climate risk category in the following steps:

[0023] S301: Based on the time series parameter analysis set, key variables are selected for data screening, principal component analysis is used to distinguish different weather types, and weather types are calibrated through frequent pattern mining to generate weather category distribution information;

[0024] S302: Based on the weather category distribution information, categories with high precipitation potential are screened, historical data comparison is performed on the categories, precipitation frequency and intensity are determined, and key weather categories with precipitation potential are identified to generate high precipitation potential information;

[0025] S303: Based on the high precipitation potential information, multiple categories of climate risk information are summarized, and climate risks are classified by analyzing key risk factors to obtain climate risk categories.

[0026] As a further solution of the present invention, the principal component analysis is performed according to the formula:

[0027] ZD=U(Σ+αE)V T

[0028] The principal component feature ZD is obtained, where U is the left singular vector, Σ is the original singular value diagonal matrix, α is the regularization coefficient, E is the unit matrix, and V T are right singular vectors.

[0029] As a further solution of the present invention, based on the climate risk category, a Kalman filter is used to analyze the tracked cloud clusters, calculate the changes in optical thickness and particle radius, and evaluate the potential impact of heavy precipitation. The steps of obtaining the cloud cluster impact index are specifically as follows:

[0030] S401: Based on the climate risk category, a Kalman filter is used to analyze cloud tracking data, and by continuously monitoring changes in cloud structure, optical thickness and particle radius are calculated to obtain precipitation potential impact data;

[0031] S402: Based on the precipitation potential impact data, analyzing the change trend of the optical thickness and the particle radius, evaluating the impact on the precipitation by analyzing the changes in the data details, and obtaining the cloud precipitation probability index;

[0032] S403: Based on the cloud cluster precipitation probability index, the relationship between cloud cluster characteristics and precipitation data is analyzed, the probability of heavy precipitation events is predicted, and the cloud cluster impact index is obtained.

[0033] As a further solution of the present invention, based on the cloud impact index, a VAR model is used to calculate the key precipitation thresholds of cloud top brightness and optical thickness, and to predict the future thresholds of the key parameters. The steps of establishing a precipitation threshold prediction model are specifically as follows:

[0034] S501: Based on the cloud impact index, the VAR model is used to analyze the historical data of cloud top brightness temperature and optical thickness, compare the historical thresholds with the current observation data, determine the key precipitation threshold, and generate precipitation threshold data;

[0035] S502: Based on the precipitation threshold data, adjust and optimize the parameters of the prediction model, verify the stability of the adjusted model through multi-period data, and establish a prediction optimization model;

[0036] S503: Based on the prediction optimization model, by simulating the threshold response under different climate conditions, the adaptability of the model in a changing environment is determined, and a precipitation threshold prediction model is established.

[0037] As a further solution of the present invention, based on the precipitation threshold prediction model and recent satellite data, a deep Q network is used to analyze the potential and intensity of heavy precipitation events, and the warning level is adjusted according to the ground conditions. The steps of generating warning decision information are specifically as follows:

[0038] S601: Based on the precipitation threshold prediction model and recent satellite data, using a deep Q network, analyzing the potential of heavy precipitation under current climate conditions, combining historical trends and real-time meteorological data for analysis, and obtaining a precipitation potential index;

[0039] S602: Based on the precipitation potential index, by comparing current and historical precipitation event data, quantitatively evaluate the potential intensity of future heavy precipitation events to obtain a precipitation intensity index;

[0040] S603: Based on the precipitation intensity index, combined with the current ground environment conditions and warning needs, the warning level is adjusted in real time to obtain warning decision information.

[0041] A convective heavy rainfall monitoring and early warning system based on satellite remote sensing, wherein the convective heavy rainfall monitoring and early warning system based on satellite remote sensing is used to execute the above-mentioned convective heavy rainfall monitoring and early warning method based on satellite remote sensing, and the system comprises:

[0042] The cloud top temperature tracking module calculates the cloud top brightness temperature changes in the image data based on the detection data of GPM, FY-3G and H8 satellites, and analyzes the time series of cloud cluster dynamic tracking data and cloud top brightness temperature to generate a cloud top temperature feature set;

[0043] The time series data analysis module analyzes the cloud parameter time series data in the ERA-5 reanalysis data, including particle radius and cloud water path, based on the cloud top temperature feature set, and compares the historical threshold data to generate a time series parameter analysis set;

[0044] The risk category determination module classifies the current weather situation based on the time series parameter analysis set using principal component analysis, identifies categories with high precipitation potential, and generates climate risk categories;

[0045] The cloud dynamic assessment module uses a Kalman filter to analyze the tracked clouds based on the climate risk categories, calculate the changes in optical thickness and particle radius, assess the potential impact of heavy precipitation, and generate cloud impact indicators;

[0046] The heavy precipitation warning decision module is based on the cloud impact indicators and uses the VAR model to calculate the key precipitation thresholds of cloud top brightness temperature and optical thickness, predict the future thresholds of key parameters, and use the deep Q network based on recent satellite data to analyze the potential and intensity of heavy precipitation events, adjust the warning level, and generate warning decision information.

[0047] Compared with the prior art, the advantages and positive effects of the present invention are:

[0048] In the present invention, by utilizing the time series analysis of cloud top brightness temperature, small meteorological changes can be quickly captured, and the ability to identify potential precipitation risks at an early stage is enhanced. The introduction of principal component analysis optimizes the classification process of weather conditions and improves the accuracy of risk assessment. Through the application of Kalman filter, the dynamic tracking of cloud clusters is more accurate, making the assessment of the impact of potential heavy precipitation more detailed. The VAR model provides higher reliability in predicting future meteorological thresholds, ensuring that the early warning system can effectively notify before a disaster occurs. The deep Q network combined with real-time satellite data enables the early warning information to be updated in real time, providing immediate and effective disaster prevention measures, and greatly improving the ability to respond to rapidly changing weather events. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0050] Figure 2 This is a detailed flow chart of S1 of the present invention;

[0051] Figure 3 This is a detailed flow chart of S2 of the present invention;

[0052] Figure 4 This is a detailed flow chart of S3 of the present invention;

[0053] Figure 5 This is a detailed flow chart of S4 of the present invention;

[0054] Figure 6 This is a detailed flow chart of S5 of the present invention;

[0055] Figure 7 This is a detailed flow chart of S6 of the present invention;

[0056] Figure 8 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0058] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0059] Embodiment 1

[0060] See also Figure 1 The present invention provides a technical solution: a convective heavy rainfall monitoring and early warning method based on satellite remote sensing, comprising the following steps:

[0061] S1: Based on the detection data of GPM, FY-3G and H8 satellites, the cloud top brightness temperature changes of convective clouds in the image data are calculated. According to the cloud cluster dynamic tracking data, the time series of cloud top brightness temperature is analyzed to obtain the cloud top temperature feature set;

[0062] S2: Based on the cloud top temperature feature set, the time series data of cloud parameters, including particle radius and cloud water path, are analyzed through ERA-5 reanalysis data, and compared with historical threshold data to obtain the time series parameter analysis set;

[0063] S3: Based on the time series parameter analysis set, principal component analysis is used to classify the current weather situation, identify categories with high precipitation potential, and output climate risk categories;

[0064] S4: Based on the climate risk category, the Kalman filter is used to analyze the tracked clouds, calculate the changes in optical thickness and particle radius, evaluate the potential impact of heavy precipitation, and obtain cloud impact indicators;

[0065] S5: Based on the cloud impact index, the VAR model is used to calculate the key precipitation thresholds of cloud top brightness temperature and optical thickness, and the future thresholds of key parameters are predicted to establish a precipitation threshold prediction model;

[0066] S6: Based on the precipitation threshold prediction model and recent satellite data, the deep Q network is used to analyze the potential and intensity of heavy precipitation events, adjust the warning level according to the ground conditions, and generate warning decision information.

[0067] The cloud top temperature feature set includes cloud top temperature peak, temperature change rate, and tracking coordinates. The climate risk categories include precipitation potential level, risk occurrence frequency, and climate trend analysis results. Cloud impact indicators include optical property distribution, dynamic change speed, and influence rating. Early warning decision information includes early warning signal type, response time window, and early warning effectiveness evaluation results.

[0068] See also Figure 2 Based on the detection data of GPM, FY-3G and H8 satellites, the cloud top brightness temperature changes of convective clouds in the image data are calculated. According to the cloud cluster dynamic tracking data, the time series of cloud top brightness temperature is analyzed. The steps to obtain the cloud top temperature feature set are as follows:

[0069] S101: Collect the detection data of GPM, FY-3G and H8 satellites, integrate the data, unify the data format and clean the data, and verify the data quality to obtain the cloud top brightness temperature data set;

[0070] The detection data of GPM, FY-3G and H8 satellites were collected and the data format was unified. All data were unified into CSV format, including satellite name, detection time, geographic coordinates and cloud top brightness temperature. Preliminary data screening was carried out to remove invalid data caused by satellite occlusion or sensor failure, such as abnormal brightness temperature values ​​that exceeded the normal cloud top temperature range (minus ninety degrees Celsius to thirty degrees Celsius). The quality of the screened data was further verified by comparing it with historical data to check the frequency of abnormal values, ensure the accuracy and consistency of the data set, and obtain the cloud top brightness temperature data set.

[0071] S102: Based on the cloud top brightness temperature dataset, extract cloud cluster dynamics, calculate the brightness temperature differences at consecutive time points to construct a time series, use the sliding window method to smooth the data, and identify key change points to obtain a temperature series set;

[0072] Based on the cloud top brightness temperature dataset, the observation data every thirty minutes is taken as a unit, the brightness temperature data at consecutive time points are extracted, the difference between the brightness temperature at each time point and the brightness temperature at the previous time point is calculated, and the brightness temperature change sequence is constructed. The sliding window method with a period of three hours is applied, and the average and standard deviation of the brightness temperature change in the window are calculated every half hour. This method is used to smooth the data and reduce the impact of instantaneous fluctuations. By setting the threshold, which is set to twice the standard deviation of the average change value, the key change points in the sequence, such as the formation or dissipation of clouds, are identified, and the temperature series set is obtained.

[0073] S103: Based on the temperature sequence set, analyze the cloud top temperature variation characteristics, identify the temperature extremes by using the temperature range segmentation, determine the cloud top key temperature index, and generate the cloud top temperature feature set;

[0074] The quartile method is used to locate the extreme points of cloud top temperature, and the temperature range is divided into four intervals. The highest and lowest values ​​in each interval are identified respectively, and the key temperature extreme points of the cloud top are determined. Combined with the physical characteristics of the cloud top, the temperature extreme points are set as key temperature indicators for subsequent climate models and weather forecasts. Through the analysis of the extreme points, a feature set containing key temperature indicators of the cloud top is constructed.

[0075] See also Figure 3 Based on the cloud top temperature feature set, the time series data of cloud parameters, including particle radius and cloud water path, are analyzed through ERA-5 reanalysis data, and compared with historical threshold data. The specific steps to obtain the time series parameter analysis set are as follows:

[0076] S201: Based on the cloud top temperature feature set and combined with ERA-5 reanalysis data, analyze the cloud parameter time series, monitor the changing trends of cloud parameters, including particle size and cloud water path, and generate a cloud parameter time series set;

[0077] The cloud top temperature feature set that has been constructed is integrated with the ERA-5 reanalysis data, and a time series analysis is performed on the integrated data set, including extracting cloud top particle size and cloud water path parameters from daily data. Parameters are considered to be key factors affecting weather patterns and climate change. The cloud parameter time series is constructed by calculating the daily average and rate of change of cloud top particle size and cloud water path. The daily rate of change is calculated by comparing the cloud top parameters of the day with the parameters of the previous day. This is used to monitor the changing trend of cloud parameters and generate a cloud parameter time series set with time tags.

[0078] S202: Based on the cloud parameter time series set, compare the historical thresholds, analyze and identify data points that exceed the normal range by setting thresholds, identify abnormal values, and obtain time series abnormality indicators;

[0079] Based on the cloud parameter time series set, the normal variation range of each parameter is set by comparing with the historical data of the same period. For example, based on the data of the same period in the past five years, the average value and standard deviation of the particle size and cloud water path are calculated to set the threshold. The threshold setting is intended to capture data points that are beyond the normal variation range. By comparing the daily recorded parameters with the set thresholds, outliers are identified, recorded and marked, and a set of time series anomaly indicators is formed.

[0080] S203: Based on the time series anomaly index, the multi-source data results are merged, the parameter analysis process is continuously optimized, and a time series parameter analysis set is generated;

[0081] Based on the time series anomaly indicators, multi-source data are further merged, including ground observation station data, satellite remote sensing data, and reanalysis data. The parameter analysis process is continuously optimized through cross-validation and comparative analysis of the parameters provided by each data source. By introducing algorithms to perform weighted processing on each parameter, the data integration and analysis process is optimized, the accuracy and efficiency of data processing are improved, and a time series parameter analysis set is generated.

[0082] See also Figure 4 Based on the time series parameter analysis set, principal component analysis is used to classify the current weather situation and identify the categories with high precipitation potential. The specific steps for outputting climate risk categories are as follows:

[0083] S301: Based on the time series parameter analysis set, key variables are selected for data screening, principal component analysis is used to distinguish different weather types, and weather types are calibrated through frequent pattern mining to generate weather category distribution information;

[0084] Representative key variables, such as cloud top temperature, cloud water path and particle size, are selected from the time series parameter analysis set for data screening. The principal component analysis method is used to extract the main components from multiple variables to distinguish different weather types. Based on the extracted principal components, frequent pattern mining technology is applied to calibrate weather types by analyzing the co-occurrence frequency between variables and identify frequently co-occurring weather patterns. Through this method, weather category distribution information is generated, which reveals the distribution of different weather types in the data set.

[0085] Principal component analysis, according to the formula:

[0086] ZD=U(Σ+αE)V T

[0087] The principal component feature ZD is obtained, where U is the left singular vector, Σ is the original singular value diagonal matrix, α is the regularization coefficient, E is the unit matrix, and V T are right singular vectors.

[0088] Execution process:

[0089] The three components of the data matrix X are obtained by singular value decomposition: U, Σ, and V T Then, add the regularization coefficient α and the unit matrix E to the singular value Σ to increase the stability of the model. T Combined, an improved principal component score ZD is obtained, where the value of α is determined through cross-validation to ensure that the improved equation can effectively deal with the existing overfitting problem, thereby more accurately extracting and distinguishing different weather types and obtaining more accurate distribution information of weather categories.

[0090] S302: Based on the weather category distribution information, categories with high precipitation potential are screened, historical data comparison is performed on the categories, precipitation frequency and intensity are determined, and key weather categories with precipitation potential are identified to generate high precipitation potential information;

[0091] Based on the generated weather category distribution information, weather categories with high precipitation potential are screened out. By comparing with historical data, including comparing the precipitation frequency and intensity of the categories in history, the categories with high precipitation potential are determined. By analyzing the weather records of the same period in history, the precipitation performance of each category is evaluated, thereby identifying the key weather categories related to high precipitation. After calibrating the categories, a high precipitation potential information set is formed.

[0092] S303: Based on the high precipitation potential information, summarize the climate risk information of multiple categories, classify the climate risks by analyzing the key risk factors, and obtain the climate risk categories;

[0093] Based on the high precipitation potential information, multiple categories of climate risk information are further summarized and analyzed. Climate risks are classified by detailed analysis of key risk factors in each category, such as high temperature, strong winds and extreme precipitation. The classification process is not only based on a single risk factor, but also considers the interactive influence of multiple factors, thus obtaining a comprehensive set of climate risk categories. The categories reflect the combination of different risk factors and their impact on climate conditions.

[0094] See also Figure 5 Based on the climate risk category, the Kalman filter is used to analyze the tracked clouds, calculate the changes in optical thickness and particle radius, and evaluate the potential impact of heavy precipitation. The specific steps to obtain the cloud impact index are as follows:

[0095] S401: Based on the climate risk category, the Kalman filter is used to analyze the cloud tracking data, and the optical thickness and particle radius are calculated by continuously monitoring the changes in cloud structure to obtain the potential impact data on precipitation;

[0096] The Kalman filter is used to track and analyze the continuous changes in cloud structure. It is possible to estimate the state of clouds from a series of dynamic observation data, especially when the clouds are changing rapidly and the data is interfered by noise. By continuously monitoring the changes in cloud structure, the optical thickness and particle radius at each time point are calculated. The reflectivity data and optical scattering characteristics of the clouds at each moment are calculated. By extracting the optical characteristic data of clouds, data points with potential impact on precipitation are generated.

[0097] S402: Based on the precipitation potential impact data, analyze the change trend of optical thickness and particle radius, evaluate the impact on precipitation by analyzing the changes in data details, and obtain the cloud precipitation probability index;

[0098] Based on the obtained data on potential impact on precipitation, the changing trends of optical thickness and particle radius are analyzed. By carefully examining the change details of each data point, such as the sudden increase or decrease in optical thickness and particle radius, the potential impact of the changes on precipitation is evaluated. The analysis takes into account how the physical and chemical properties of clouds affect the precipitation process. For example, an increase in particle radius indicates that the condensation of water droplets in clouds is enhanced, thereby increasing precipitation. Through this method, a cloud precipitation probability index is obtained, which reflects the specific impact of changes in different cloud characteristics on precipitation probability.

[0099] S403: Based on the cloud cluster precipitation probability index, the relationship between cloud cluster characteristics and precipitation data is analyzed, the probability of heavy precipitation events is predicted, and the cloud cluster impact index is obtained.

[0100] Based on the cloud precipitation probability index, the relationship between cloud characteristics and actual precipitation data is further analyzed. Through statistical analysis and model simulation, the probability of heavy precipitation events is predicted, taking into account the combined effects of cloud optical thickness, particle radius and other meteorological conditions. The prediction is not only based on a single indicator, but also through a comprehensive analysis of the interaction of multiple related indicators, so as to more accurately predict heavy precipitation events and generate a comprehensive cloud impact index. The index provides a quantitative tool for evaluating precipitation events caused by specific clouds in the future.

[0101] See also Figure 6 Based on the cloud impact index, the VAR model is used to calculate the key precipitation thresholds of cloud top brightness temperature and optical thickness, and predict the future thresholds of key parameters. The specific steps to establish the precipitation threshold prediction model are as follows:

[0102] S501: Based on the cloud impact index, the VAR model is used to analyze the historical data of cloud top brightness temperature and optical thickness, compare the historical thresholds with the current observation data, determine the key precipitation threshold, and generate precipitation threshold data;

[0103] A vector autoregression (VAR) model is used to analyze historical data of cloud top brightness temperature and optical thickness to reveal the dynamic relationship between the two variables over time. The key precipitation threshold is determined through in-depth analysis of historical data. By comparing past thresholds with current observation data, the changing trends and periodic fluctuations of the thresholds are analyzed to capture the complexity of the interaction between key variables and how they jointly affect the occurrence of precipitation events. Based on the analysis results, precipitation threshold data is generated.

[0104] S502: Based on the precipitation threshold data, the parameters of the prediction model are adjusted and optimized, the stability of the adjusted model is verified through multi-period data, and a prediction optimization model is established;

[0105] Based on the generated precipitation threshold data, the parameters of the prediction model are adjusted and optimized. The response and stability of the model are verified by analyzing data sets of different periods, including the use of long-term and short-term historical data to evaluate the adaptability of the model parameters and ensure the effectiveness of the model at different time scales. Through this multi-period data verification, a prediction optimization model is established and adapted to various climate change situations.

[0106] S503: Based on the prediction optimization model, by simulating the threshold response under different climate conditions, the adaptability of the model in a variable environment is determined, and a precipitation threshold prediction model is established;

[0107] Based on the prediction optimization model, the adaptability of the model is further determined by simulating the threshold responses under different climatic conditions, including testing the model's prediction ability under extreme climate events and mild climate conditions to evaluate the model's performance in a changing environment. Through simulation, the effectiveness and reliability of the model under different climate change scenarios are established, and a precipitation threshold prediction model is established. The model can not only predict the precipitation threshold under normal weather conditions, but also predict precipitation activities under extreme or uncommon climate conditions.

[0108] See also Figure 7 Based on the precipitation threshold prediction model and recent satellite data, the deep Q network is used to analyze the potential and intensity of heavy precipitation events, and the warning level is adjusted according to the ground conditions. The specific steps for generating warning decision information are as follows:

[0109] S601: Based on the precipitation threshold prediction model and recent satellite data, the deep Q network is used to analyze the potential of heavy precipitation under current climate conditions, combined with historical trends and real-time meteorological data to obtain precipitation potential indicators;

[0110] A deep Q-network (DQN) is used to analyze recent satellite data and existing precipitation threshold prediction models to determine the potential for heavy precipitation under current climate conditions, allowing the model to update its predictions while continuously receiving new data, so that it can capture subtle climate changes and respond quickly. Combining historical trends and real-time meteorological data, such as key parameters such as cloud top brightness temperature, optical thickness and wind speed, the deep Q-network analyzes the changing trends and interactions of the data to generate precipitation potential indicators.

[0111] S602: Based on the precipitation potential index, by comparing the current and historical precipitation event data, the potential intensity of future heavy precipitation events is quantitatively evaluated to obtain a precipitation intensity index;

[0112] Based on the precipitation potential index, by comparing the current meteorological conditions with data of similar precipitation events in history, a quantitative assessment of the potential intensity of future heavy precipitation events is made, including analyzing the frequency, duration and intensity of precipitation events, and comparing the data with historical events. Through comparison, the climate patterns that lead to high-intensity precipitation can be identified, and a precipitation intensity index can be generated accordingly. The index provides an important quantitative basis for predicting the severity of upcoming precipitation events.

[0113] S603: Based on the precipitation intensity index, combined with the current ground environment conditions and warning needs, the warning level is adjusted in real time to obtain warning decision information;

[0114] Based on the precipitation intensity index, combined with the current ground environmental conditions such as soil moisture, urban drainage capacity and warning needs, the warning level is adjusted in real time, including assessing the carrying capacity of the ground environment, predicting flood risks, and adjusting the warning level according to the assessment results. In this way, early warning decision-making information is generated, which can guide government departments and the public to take appropriate preventive measures, such as activating flood control systems, issuing public safety information and preparing emergency rescue resources. This process ensures the timeliness and accuracy of warning information and improves the efficiency and effectiveness of disaster response.

[0115] See also Figure 8 , a convective heavy rainfall monitoring and early warning system based on satellite remote sensing, the system includes:

[0116] The cloud top temperature tracking module calculates the cloud top brightness temperature changes in the image data based on the detection data of GPM, FY-3G and H8 satellites, and analyzes the time series of cloud cluster dynamic tracking data and cloud top brightness temperature to generate a cloud top temperature feature set;

[0117] The time series data analysis module analyzes the cloud parameter time series data in the ERA-5 reanalysis data, including particle radius and cloud water path, based on the cloud top temperature feature set, and compares it with the historical threshold data to generate a time series parameter analysis set;

[0118] The risk category determination module uses principal component analysis to classify the current weather situation based on the time series parameter analysis set, identifies categories with high precipitation potential, and generates climate risk categories;

[0119] The cloud dynamic assessment module uses the Kalman filter to analyze the tracked clouds based on the climate risk category, calculate the changes in optical thickness and particle radius, evaluate the potential impact of heavy precipitation, and generate cloud impact indicators;

[0120] The heavy precipitation warning decision module is based on cloud impact indicators and uses the VAR model to calculate the key precipitation thresholds of cloud top brightness temperature and optical thickness, predict the future thresholds of key parameters, and use the deep Q network based on recent satellite data to analyze the potential and intensity of heavy precipitation events, adjust the warning level, and generate warning decision information.

[0121] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A convective heavy rainfall monitoring and early warning method based on satellite remote sensing, characterized in that: The following steps are involved: Based on the detection data of GPM, FY-3G and H8 satellites, the cloud top brightness temperature changes of convective clouds in the image data are calculated. According to the cloud cluster dynamic tracking data, the time series of cloud top brightness temperature is analyzed to obtain the cloud top temperature feature set. Based on the cloud top temperature feature set, the time series data of cloud parameters, including particle radius and cloud water path, are analyzed through ERA-5 reanalysis data, and compared with historical threshold data to obtain a time series parameter analysis set; Based on the time series parameter analysis set, principal component analysis is used to classify the current weather situation, identify categories with high precipitation potential, and output climate risk categories; Based on the climate risk categories, a Kalman filter is used to analyze the tracked clouds, calculate the changes in optical thickness and particle radius, evaluate the potential impact of heavy precipitation, and obtain cloud impact indicators; Based on the cloud impact indicators, the VAR model is used to calculate the key precipitation thresholds of cloud top brightness and optical thickness, and the future thresholds of key parameters are predicted to establish a precipitation threshold prediction model; Based on the precipitation threshold prediction model and recent satellite data, a deep Q network is used to analyze the potential and intensity of heavy precipitation events, adjust the warning level according to ground conditions, and generate warning decision information.

2. The satellite remote sensing-based convective heavy rainfall monitoring and early warning method according to claim 1 is characterized in that: The cloud top temperature feature set includes cloud top temperature peak, temperature change rate, and tracking coordinates; the climate risk category includes precipitation potential level, risk occurrence frequency, and climate trend analysis results; the cloud impact indicators include optical property distribution, dynamic change speed, and influence rating; the early warning decision information includes early warning signal type, response time window, and early warning effectiveness evaluation results.

3. The satellite remote sensing-based convective heavy rainfall monitoring and early warning method according to claim 1 is characterized in that: Based on the detection data of GPM, FY-3G and H8 satellites, the cloud top brightness temperature changes of convective clouds in the image data are calculated. According to the cloud cluster dynamic tracking data, the time series of cloud top brightness temperature is analyzed. The steps to obtain the cloud top temperature feature set are as follows: Collect detection data from GPM, FY-3G and H8 satellites, integrate the data, standardize the data format, clean the data, and verify the data quality to obtain the cloud top brightness temperature data set; Based on the cloud top brightness temperature dataset, cloud cluster dynamics are extracted, brightness temperature differences at consecutive time points are calculated to construct a time series, the data are smoothed using a sliding window method, and key change points are identified to obtain a temperature series set; Based on the temperature sequence set, the cloud top temperature variation characteristics are analyzed, the temperature extremes are identified by using temperature range segmentation, the cloud top key temperature indicators are determined, and the cloud top temperature feature set is generated.

4. The satellite remote sensing-based convective heavy rainfall monitoring and early warning method according to claim 1 is characterized in that: Based on the cloud top temperature feature set, the time series data of cloud parameters, including particle radius and cloud water path, are analyzed through ERA-5 reanalysis data, and compared with historical threshold data. The specific steps to obtain the time series parameter analysis set are as follows: Based on the cloud top temperature feature set and combined with ERA-5 reanalysis data, the cloud parameter time series are analyzed to monitor the changing trends of cloud parameters, including particle size and cloud water path, and generate a cloud parameter time series set; Based on the cloud parameter time series set, compare historical thresholds, analyze and identify data points beyond the normal range by setting thresholds, identify outliers, and obtain time series anomaly indicators; Based on the time series anomaly indicators, multi-source data results are merged, the parameter analysis process is continuously optimized, and a time series parameter analysis set is generated.

5. The satellite remote sensing-based convective heavy rainfall monitoring and early warning method according to claim 1 is characterized in that: Based on the time series parameter analysis set, principal component analysis is used to classify the current weather situation, identify categories with high precipitation potential, and output the climate risk category in the following steps: Based on the time series parameter analysis set, key variables are selected for data screening, principal component analysis is used to distinguish different weather types, and weather types are calibrated through frequent pattern mining to generate weather category distribution information; Based on the weather category distribution information, categories with high precipitation potential are screened, historical data comparison is performed on the categories, precipitation frequency and intensity are determined, and weather categories with key precipitation potential are identified to generate high precipitation potential information; Based on the high precipitation potential information, multiple categories of climate risk information are summarized, and by analyzing key risk factors, climate risks are classified to obtain climate risk categories.

6. The satellite remote sensing-based convective heavy rainfall monitoring and early warning method according to claim 5 is characterized in that: The principal component analysis is based on the formula: ZD=U(Σ+αE)V T The principal component feature ZD is obtained, where U is the left singular vector, Σ is the original singular value diagonal matrix, α is the regularization coefficient, E is the unit matrix, and V T are right singular vectors.

7. The satellite remote sensing-based convective heavy rainfall monitoring and early warning method according to claim 1, characterized in that: Based on the climate risk categories, the Kalman filter is used to analyze the tracked clouds, calculate the changes in optical thickness and particle radius, and evaluate the potential impact of heavy precipitation. The specific steps to obtain the cloud impact index are as follows: Based on the climate risk categories, Kalman filter is used to analyze cloud tracking data, and by continuously monitoring the changes in cloud structure, optical thickness and particle radius are calculated to obtain precipitation potential impact data; Based on the potential impact data of precipitation, the changing trends of optical thickness and particle radius are analyzed, and the impact on precipitation is evaluated by analyzing the changes in data details to obtain the cloud cluster precipitation probability index; Based on the cloud cluster precipitation probability index, the relationship between cloud cluster characteristics and precipitation data is analyzed, the probability of heavy precipitation events is predicted, and the cloud cluster impact index is obtained.

8. The satellite remote sensing-based convective heavy rainfall monitoring and early warning method according to claim 1, characterized in that: Based on the cloud impact index, the VAR model is used to calculate the key precipitation thresholds of cloud top brightness temperature and optical thickness, and predict the future thresholds of key parameters. The specific steps of establishing the precipitation threshold prediction model are as follows: Based on the cloud impact indicators, the VAR model is used to analyze the historical data of cloud top brightness temperature and optical thickness, compare the historical thresholds with the current observation data, determine the key precipitation threshold, and generate precipitation threshold data; Based on the precipitation threshold data, the parameters of the prediction model are adjusted and optimized, the stability of the adjusted model is verified through multi-period data, and a prediction optimization model is established; Based on the prediction optimization model, by simulating the threshold response under different climate conditions, the adaptability of the model in a changing environment is determined, and a precipitation threshold prediction model is established.

9. The satellite remote sensing-based convective heavy rainfall monitoring and early warning method according to claim 1, characterized in that: Based on the precipitation threshold prediction model and recent satellite data, the deep Q network is used to analyze the potential and intensity of heavy precipitation events, and the warning level is adjusted according to the ground conditions. The specific steps of generating warning decision information are as follows: Based on the precipitation threshold prediction model and recent satellite data, the deep Q network is used to analyze the potential for heavy precipitation under current climate conditions, and the precipitation potential index is obtained by combining historical trends and real-time meteorological data for analysis; Based on the precipitation potential index, by comparing current and historical precipitation event data, the potential intensity of future heavy precipitation events is quantitatively evaluated to obtain a precipitation intensity index; Based on the precipitation intensity index, combined with the current ground environmental conditions and warning needs, the warning level is adjusted in real time to obtain warning decision information.

10. The convective heavy rainfall monitoring and early warning system based on satellite remote sensing is characterized by: According to any one of claims 1 to 9, the satellite remote sensing-based convective heavy rainfall monitoring and early warning method comprises: The cloud top temperature tracking module calculates the cloud top brightness temperature changes in the image data based on the detection data of GPM, FY-3G and H8 satellites, and analyzes the time series of cloud cluster dynamic tracking data and cloud top brightness temperature to generate a cloud top temperature feature set; The time series data analysis module analyzes the cloud parameter time series data in the ERA-5 reanalysis data, including particle radius and cloud water path, based on the cloud top temperature feature set, and compares the historical threshold data to generate a time series parameter analysis set; The risk category determination module classifies the current weather situation based on the time series parameter analysis set using principal component analysis, identifies categories with high precipitation potential, and generates climate risk categories; The cloud dynamic assessment module uses a Kalman filter to analyze the tracked clouds based on the climate risk categories, calculate the changes in optical thickness and particle radius, assess the potential impact of heavy precipitation, and generate cloud impact indicators; The heavy precipitation warning decision module is based on the cloud impact indicators and uses the VAR model to calculate the key precipitation thresholds of cloud top brightness temperature and optical thickness, predict the future thresholds of key parameters, and use the deep Q network based on recent satellite data to analyze the potential and intensity of heavy precipitation events, adjust the warning level, and generate warning decision information.

Images