Joint analysis method and system for agricultural disasters based on meteorological data

Abstract

The application discloses a meteorological data-based agricultural disaster joint analysis method and system, and belongs to the technical field of data processing and risk assessment. The method comprises the following steps: obtaining multi-dimensional meteorological observation data of multiple observation points in a target region, generating a multi-dimensional meteorological element field, performing background trend fitting and residual separation, generating a pure meteorological anomaly field tensor, identifying a meteorological precursor disturbance bubble structure, performing dynamic coding, generating a meteorological disturbance collaborative evolution characteristic vector, simulating disaster inter-disaster based on abnormal energy transmission, generating a disaster causal excitation atlas, combining a crop distribution layer, and generating a target region multi-disaster agricultural collaborative risk level. The method adopts a physical consistency element field reconstruction and residual separation technology to extract a pure meteorological anomaly signal, combines dynamic coding to construct a disaster causal excitation atlas, and performs counterfactual crop state reasoning, so that the collaborative stress degree of a multi-disaster chain can be quantified, and a multi-disaster agricultural collaborative risk level can be output.

Description

Technical Field

[0001] This invention relates to the field of data processing and risk assessment technology, and in particular to a method and system for joint analysis of agricultural disasters based on meteorological data. Background Technology

[0002] Joint analysis of agricultural disasters based on meteorological data refers to the data processing process that uses computer systems to process multi-dimensional meteorological observation information and combine it with crop growth patterns to conduct risk assessments and trend predictions of potential agricultural disasters. This technology is a core component of the modern agricultural risk management and yield forecasting system. Its operating mechanism relies on in-depth mining of time-series data such as temperature, humidity, and wind speed, aiming to provide data support for agricultural insurance pricing, disaster prevention and mitigation scheduling, and macro-agricultural economic planning.

[0003] Existing agricultural disaster analysis schemes typically employ static threshold comparisons or statistical regression models based on single meteorological indicators to assess disaster risk. Conventional data processing mainly involves collecting historical data from weather stations in a specific region, calculating the average or extreme values ​​of various meteorological elements, and triggering an early warning signal for the corresponding single disaster type when real-time monitoring data exceeds a preset fixed safety range. Some schemes introduce simple crop growth cycle parameters and use empirical formulas to calculate the probability of meteorological disasters occurring at specific stages.

[0004] Existing technical solutions lack constraints on the physical laws of fluid dynamics when processing complex meteorological data. This leads to spatial structure distortions in the interpolated meteorological element fields and fails to isolate background climate trends, making it difficult to accurately pinpoint the true signals of hazardous disturbances. Conventional solutions often treat various agricultural disasters as isolated events for static assessment, ignoring the cascading excitation mechanisms of meteorological anomalies in the transmission and transformation between different disaster forms, and failing to construct causal evolution paths between multiple disasters. This fragmented analytical model cannot accurately reflect the synergistic stress effects of multiple disasters on crop physiological development, resulting in biased agricultural risk assessment results. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method and system for joint analysis of agricultural disasters based on meteorological data. It employs physical consistency element field reconstruction and residual separation techniques to extract pure meteorological anomaly signals, combines dynamic coding to construct a disaster causal excitation map, and performs counterfactual crop state reasoning. This method can quantify the degree of synergistic stress of multiple disaster chains and output the synergistic risk level of multi-hazard agriculture.

[0006] The above objectives can be achieved through the following approach: The agricultural disaster joint analysis method based on meteorological data includes acquiring multidimensional meteorological observation data from multiple observation points within the target area, reconstructing a physically consistent multidimensional element field, and generating a multidimensional meteorological element field. Background trend fitting and residual separation are performed on the multidimensional meteorological element field to generate a pure meteorological anomaly field tensor. In the pure meteorological anomaly field tensor, the bubble structure of meteorological precursor disturbances is identified, and the co-evolution trajectory of meteorological elements inside the bubble structure is dynamically encoded to generate a feature vector of meteorological disturbance co-evolution. The meteorological disturbance co-evolution feature vector is input into the disaster cascade excitation generator, and a disaster causal excitation spectrum is generated by simulating the causal excitation intensity based on abnormal energy transfer between disasters. Based on the disaster causal triggering map and crop distribution layer, counterfactual crop state reasoning is performed to compare the expected crop state under no disaster chain with the predicted crop state under stress, and generate a multi-hazard agricultural collaborative risk level for the target area.

[0007] Based on the same inventive concept, the present invention also provides a joint analysis system for agricultural disasters based on meteorological data, the system comprising: The physical field reconstruction module is used to acquire multidimensional meteorological observation data from multiple observation points within the target area, perform physical consistency multidimensional element field reconstruction, and generate a multidimensional meteorological element field. An anomaly separation module is used to perform background trend fitting and residual separation on the multidimensional meteorological element field to generate a pure meteorological anomaly field tensor. The precursor coding module is used to identify the weather precursor disturbance bubble structure in the pure weather anomaly field tensor, and to perform dynamic coding on the co-evolution trajectory of meteorological elements inside the bubble structure to generate a weather disturbance co-evolution feature vector. The cascaded inference module is used to input the meteorological disturbance co-evolution feature vector into the disaster cascade excitation generator, and generate a disaster causal excitation spectrum by simulating the causal excitation intensity based on abnormal energy transfer between disasters; The risk assessment module is used to perform counterfactual crop state reasoning based on the disaster causal triggering map and crop distribution layer, compare the expected crop state under no disaster chain with the predicted crop state under stress, and generate a multi-hazard agricultural collaborative risk level for the target area.

[0008] Compared with the prior art, the present invention has the following advantages: This invention reconstructs a multidimensional meteorological element field with a spatial structure conforming to physical laws by performing geographic weighted spatiotemporal interpolation on multi-source time-series observation data and applying fluid dynamic constraints. Based on this, background trend fitting and residual separation are performed to extract the pure meteorological anomaly field tensor. This data processing mechanism eliminates interference from the conventional climate background, locks in information on deviating from the norm and causing hazardous disturbances, and provides a high-fidelity data foundation for subsequent disaster prediction, overcoming the physical distortion defects inherent in the direct application of traditional meteorological data.

[0009] This invention identifies the bubble structure of meteorological precursor disturbances in the pure meteorological anomaly field tensor, extracts the cooperative rate of change of elements as they move with the disturbance body for dynamic encoding, and inputs the generated feature vectors into a disaster cascade excitation generator. By calculating the anomalous energy transfer flux and material transport efficiency, a disaster causal excitation map is constructed. This mechanism transforms isolated meteorological anomalies into cascaded transmission paths with spatiotemporal continuity and causal correlation, enabling dynamic extrapolation of the successive occurrence of multiple disasters and improving the predictive ability of complex disaster chain evolution trends.

[0010] This invention performs counterfactual crop state inference based on disaster causal excitation maps and crop distribution layers. It inputs the stress intensity of disaster types as a correction factor into the crop growth model, comparing the expected crop state under no disaster chain with the predicted crop state under stress. This assessment method directly maps meteorological dynamics evolution to the actual impact on crop physiological development. By quantifying yield and quality differences, it generates a multi-hazard agricultural collaborative risk level for the target area, providing quantitative decision support for disaster prevention and mitigation planning and risk management in agricultural production. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating the agricultural disaster joint analysis method based on meteorological data, according to an embodiment of the present invention.

[0013] Figure 2 This is a schematic diagram of the structure of the agricultural disaster joint analysis system based on meteorological data according to an embodiment of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Reference Figure 1 One embodiment of the present invention proposes a joint analysis method for agricultural disasters based on meteorological data. It uses physical consistency element field reconstruction and residual separation technology to extract pure meteorological anomaly signals, combines dynamic coding to construct a disaster causal excitation map, and performs counterfactual crop state reasoning. This method can quantify the degree of synergistic stress of multiple disaster chains and output the synergistic risk level of multi-hazard agriculture.

[0016] The method described in this embodiment specifically includes: S1. Acquire multidimensional meteorological observation data from multiple observation points within the target area, reconstruct a physically consistent multidimensional element field, and generate a multidimensional meteorological element field. In one embodiment of the present invention, step S1 includes the following steps: Acquire multi-source time-series observation data of temperature, humidity, wind speed, and precipitation from multiple observation points within the target area; Geographically weighted spatiotemporal interpolation is performed on the multi-source time-series observation data to obtain a regular gridded four-dimensional tensor. The dimensions of the four-dimensional tensor include time, longitude, latitude, and meteorological parameter types. Fluid dynamic constraints based on the Navier-Stokes equations and the continuity equation are applied to the four-dimensional tensor, and variational calibration is performed to generate a multi-dimensional meteorological element field with a spatial structure that conforms to physical laws.

[0017] Specifically, the system acquires multidimensional meteorological observation data from multiple observation points within the target area. This multidimensional meteorological observation data manifests as multi-source time-series observation data of temperature, humidity, wind speed, and precipitation. The target area refers to the specific geographic spatial range for which meteorological monitoring and analysis are required. Multiple observation points are distributed across different geographical locations within the target area to collect real-time meteorological information. The multi-source time-series observation data is a set of data containing temperature, humidity, wind speed, and precipitation values ​​recorded by these observation points over continuous time steps. After acquiring the multi-source time-series observation data, the system performs geographically weighted spatiotemporal interpolation on the data to obtain a regularly gridded four-dimensional tensor. Geographically weighted spatiotemporal interpolation refers to the process of weighting the data from discrete observation points by combining spatial distance and time intervals to estimate meteorological values ​​on a continuous spatiotemporal grid. The regularly gridded four-dimensional tensor is a multidimensional data matrix composed of equally spaced spatial grids and time steps. The dimensions of the four-dimensional tensor include time, longitude, latitude, and meteorological parameter types. Time represents the time series of data collection, longitude and latitude determine spatial location, and meteorological parameter types correspond to temperature, humidity, wind speed, and precipitation. The calculation relationship of geographic weighted spatiotemporal interpolation satisfies the following formula.

[0018] , in the formula The longitude index in a regular gridded four-dimensional tensor is: Latitude index is The time step is Meteorological parameter types are The value. This indicates the total number of observation points. Indicates the first Each observation point at the time step Meteorological parameter types Multi-source time-series observation data. Indicates the first The dimensionless spatiotemporal weight coefficients of each observation point are set based on the spatial distance and time difference between the observation point and the target grid point. The closer the distance and the smaller the time difference, the greater the weight. The sum of the spatiotemporal weight coefficients of all observation points is equal to 1.

[0019] To ensure the generated data conforms to the real physical environment, the system applies fluid dynamic constraints based on the Navier-Stokes equations and the continuity equation to the four-dimensional tensor, performing variational calibration to generate a multi-dimensional meteorological element field with a spatial structure conforming to physical laws. The Navier-Stokes equations describe the conservation of fluid momentum, and the continuity equation describes the conservation of fluid mass. Fluid dynamic constraints utilize these two physical laws to limit the spatiotemporal variations of meteorological elements. Variational calibration is an optimization process that adjusts meteorological values ​​by minimizing the deviation between interpolated data and physical laws. After variational calibration, the system completes the reconstruction of a physically consistent multi-dimensional element field, ultimately generating a multi-dimensional meteorological element field, which is a comprehensive meteorological data set that satisfies the laws of fluid dynamics in its spatiotemporal distribution. The optimization process of variational calibration is achieved by minimizing a cost function, the calculation relationship of which satisfies the following formula.

[0020] , in the formula This represents the dimensionless cost function for variational calibration. The longitude index in the multidimensional meteorological element field is represented as Latitude index is The time step is Meteorological parameter types are The value. Indicates meteorological parameter type The historical observation standard deviation is used to eliminate dimensional differences. Its setting is based on the statistical analysis of multi-source time-series observation data of the target area over the past many years. The dimensionless penalty coefficient representing the fluid dynamics constraint is set based on the model convergence test. and These represent the physical components of the wind speed parameter in the longitude and latitude directions, respectively, in the multidimensional meteorological field. and These represent spatial distances along the longitude and latitude directions, respectively. This represents the characteristic time constant, used to transform the divergence term into a dimensionless quantity; its setting is based on the average lifetime of the meteorological system within the target area. This is achieved by solving the cost function. The minimum value of the system outputs a multi-dimensional meteorological element field whose spatial structure conforms to physical laws.

[0021] For example, the system acquires multidimensional meteorological observation data from multiple observation points within a target area. Specifically, this multidimensional meteorological observation data represents multi-source time-series observation data of temperature, humidity, wind speed, and precipitation. The target area is set as an agricultural planting base, and the system collects temperature, humidity, wind speed, and precipitation values ​​over continuous time steps from multiple observation points distributed within that base. After acquiring the data, the system performs geographically weighted spatiotemporal interpolation on the multi-source time-series observation data to obtain a regularly gridded four-dimensional tensor. During the interpolation process, the total number of observation points... Set to 20, time step Set to hourly, meteorological parameter type It covers temperature, humidity, wind speed, and precipitation. The system calculates dimensionless spatiotemporal weighting coefficients based on spatial distance and time difference. Observations closer to the target grid point and with smaller time differences are assigned greater weights, with the sum of the spatiotemporal weight coefficients of all observations equal to 1. After calculation, the system generates a regularly gridded four-dimensional tensor, whose dimensions include time, longitude, latitude, and meteorological parameter types. To ensure the data conforms to the real physical environment, the system applies fluid dynamic constraints based on the Navier-Stokes equations and the continuity equation to the four-dimensional tensor, performing variational calibration to generate a multi-dimensional meteorological element field with a spatial structure conforming to physical laws. During the variational calibration process, the system constructs a dimensionless cost function. To eliminate dimensional differences, the system statistically obtains the historical observation standard deviation based on multi-source time-series observation data of the target area over the past 10 years. For example, the standard deviation of historical temperature observations is set to a value of 5 degrees Celsius. The dimensionless penalty coefficient for fluid dynamic constraints. The model convergence test setting is 0.8. Characteristic time constant. The average lifespan of the meteorological system within the target area is set at 3600 seconds. The system minimizes the dimensionless cost function. By adjusting meteorological values ​​to satisfy the conservation of fluid momentum and mass, a multidimensional element field with physical consistency is reconstructed, and finally, a multidimensional meteorological element field with spatial structure conforming to physical laws is output.

[0022] S2. Perform background trend fitting and residual separation on the multidimensional meteorological element field to generate a pure meteorological anomaly field tensor. In one embodiment of the present invention, step S2 includes the following steps: On the multidimensional meteorological element field, a sliding window filter is applied to the meteorological parameter sequence at each spatial grid point along the time axis to extract the long-period low-pass signal as the climate background trend field. The multidimensional meteorological element field and the climate background trend field are subtracted element by element on the same spatiotemporal coordinates to obtain the abnormal residual field that represents the deviation from the normal state. The abnormal residual field is subjected to a statistical significance test. The abnormal signals that pass the significance test are retained, and a pure meteorological anomaly field tensor containing only statistically significant catastrophic disturbance information is generated.

[0023] Specifically, the system applies a sliding window filter to the per-meteorological parameter sequence at each spatial grid point along the time axis in a multi-dimensional meteorological element field, extracting a long-period low-pass signal as the climate background trend field. The multi-dimensional meteorological element field is a comprehensive meteorological data set reconstructed with physical consistency. A per-meteorological parameter sequence refers to the set of values ​​for a single meteorological indicator that changes continuously over time at a specific spatial location. Sliding window filtering is a processing method that calculates the average value of a fixed-length data segment in the time series to smooth short-term fluctuations. The long-period low-pass signal is the slow trend retained after filtering out high-frequency short-term fluctuations. The climate background trend field reflects the normal meteorological state without the influence of sudden weather systems. The calculation relationship of the sliding window filter satisfies the following formula: , in the formula The longitude index in the climate background trend field is Latitude index is The time step is Meteorological parameter types are The value. This represents the total time step of the sliding window, which is set based on statistical analysis of historical meteorological cycles in the target area. It represents the numerical value of the corresponding spatiotemporal location and meteorological parameter type in a multidimensional meteorological element field. This represents the time step variable within the sliding window.

[0024] The system performs element-wise subtraction between the multidimensional meteorological element field and the climate background trend field on the same spatiotemporal coordinates to obtain an anomaly residual field representing deviations from the norm. The anomaly residual field represents the difference matrix between the actual meteorological state and the normal climate background. The calculation relationship for element-wise subtraction satisfies the following formula: , in the formula This represents the numerical value of the corresponding spatiotemporal location and meteorological parameter type in the abnormal residual field.

[0025] After acquiring the anomalous residual field, the system performs a statistical significance test on the residual field, retaining the anomalous signals that pass the test, and generating a pure meteorological anomaly field tensor containing only statistically significant catastrophic disturbance information. The statistical significance test is a mathematical verification process to determine whether the anomalous residuals exceed the range of random fluctuations. An anomalous signal refers to the residual value that passes this verification process. The pure meteorological anomaly field tensor is a data matrix containing catastrophic disturbance information after filtering out random noise. The calculation relationship for the statistical significance test satisfies the following formula: , in the formula This represents a dimensionless test statistic. This represents the local standard deviation of the anomalous residual field within a set time period, used to eliminate dimensional differences. Its setting is based on historical residual data for the corresponding spatial grid points and meteorological parameter types. The system sets a dimensionless significance threshold. The significance threshold is set based on the confidence interval of the standard normal distribution. When the absolute value of the test statistic is greater than the significance threshold, the pure meteorological anomaly field tensor... The value is equal to the abnormal residual field. The value, otherwise the pure meteorological anomaly field tensor The value is 0.

[0026] For example, the system applies a sliding window filter to the per-meteorological parameter sequence at each spatial grid point along the time axis in a multi-dimensional meteorological element field, extracting a long-period low-pass signal as the climate background trend field. The system extracts the temperature per-meteorological parameter sequence at spatial grid points of an agricultural planting base. Based on statistical analysis of historical meteorological cycles in the region, the system calculates the total number of time steps for the sliding window. The time window is set to 720 hours. The system calculates the average temperature within this time window, extracts long-period low-pass signals, and generates a climate background trend field. The system subtracts the multidimensional meteorological element field from the climate background trend field element by element on the same spatiotemporal coordinates to obtain anomaly residual fields representing deviations from the norm. For example, if the temperature at a certain moment in the multidimensional meteorological element field is 35 degrees Celsius, and the corresponding temperature in the climate background trend field is 28 degrees Celsius, the element-wise subtraction yields anomaly residual field value of 7 degrees Celsius. The system performs a statistical significance test on the anomaly residual field, retains the anomalous signals that pass the significance test, and generates a pure meteorological anomaly field tensor containing only statistically significant catastrophic disturbance information. The system calculates the local standard deviation based on historical residual data. The value is 2 in Celsius. The system calculates the dimensionless test statistic. The significance threshold is 3.5. The system uses the confidence interval of the standard normal distribution to set the dimensionless significance threshold. The value is set to 1.96. Since 3.5 is greater than 1.96, the value of the anomalous residual field is retained as an anomalous signal, and the value of the pure meteorological anomaly field tensor at this spatiotemporal coordinate is assigned the value of 7 degrees Celsius.

[0027] S3. Identify the weather precursor disturbance bubble structure in the pure weather anomaly field tensor, and perform dynamic encoding on the co-evolution trajectory of weather elements inside the bubble structure to generate a weather disturbance co-evolution feature vector. In one embodiment of the present invention, step S3 includes the following steps: Extract the anomalous energy concentration degree from the pure meteorological anomaly field tensor, locate the spatiotemporal connected domains where the anomalous energy concentration degree exceeds a preset identification threshold, and mark them as meteorological precursor disturbance bubble structures. The meteorological precursor disturbance bubble structure consists of an envelope surface describing the spatial morphology and a time duration interval. The multidimensional meteorological element sequence within the structure of the meteorological precursor disturbance bubble is extracted, and the body derivative analysis is performed from the perspective of Lagrange. The cooperative change rate of the elements as they move with the disturbance body is extracted. The cooperative change rate characterizes how the change of one meteorological element changes with the change of another meteorological element. Within the time duration interval of the meteorological precursor disturbance bubble structure, the cooperative rate of change of all spatial grid points at each time step is extracted to form a set of cooperative rates of change. Principal component analysis was performed on the covariance set to extract feature vector sequences that can characterize the main patterns of the corresponding evolution process; Feature extraction is performed on the feature vector sequence to form a fixed-length vector that can characterize the potential triggering ability of multiple disasters within the bubble structure, which serves as the feature vector for the co-evolution of meteorological disturbances.

[0028] Specifically, the system extracts the anomalous energy concentration from the pure meteorological anomaly field tensor. The pure meteorological anomaly field tensor is a data matrix generated in step S2 that contains only statistically significant catastrophic disturbance information. The anomalous energy concentration is a dimensionless index that measures the comprehensive intensity of multidimensional meteorological anomaly signals at a specific spatiotemporal location. The system calculates the sum of squares of the anomaly values ​​for each meteorological parameter type to obtain the anomalous energy concentration, and the calculation relationship satisfies the following formula: , in the formula Indicates longitude index as Latitude index is The time step is The abnormal energy concentration. This represents the numerical value of the corresponding spatiotemporal location and meteorological parameter type in the pure meteorological anomaly field tensor. Indicates meteorological parameter type The historical extreme values ​​are used to eliminate dimensional differences, and their setting is based on the statistical analysis of multi-source time-series observation data of the target area over the past many years.

[0029] After acquiring the abnormal energy concentration, the system locates spatiotemporal connected regions where the abnormal energy concentration exceeds a preset identification threshold and marks them as meteorological precursor disturbance bubble structures. The identification threshold is a critical value used to determine whether abnormal energy is sufficient to trigger a disaster; its setting is based on the statistical distribution of energy before historical disaster events. A spatiotemporal connected region refers to a set of spatially adjacent and temporally continuous grid points that meet the threshold conditions. A meteorological precursor disturbance bubble structure is a three-dimensional, spatially and temporally combined dynamic entity representing a potential disaster source. The meteorological precursor disturbance bubble structure consists of an envelope surface describing its spatial morphology and a temporal duration interval. The envelope surface is the outermost boundary enclosing all spatial grid points that meet the threshold conditions, and the temporal duration interval is the time span from the appearance to the dissipation of the connected region.

[0030] For the identified dynamic entities, the system extracts multidimensional meteorological element sequences within the bubble structure of meteorological precursor disturbances, performs volume derivative analysis from a Lagrange perspective, and extracts the cooperative rate of change of elements as they move with the disturbance. The multidimensional meteorological element sequence is continuous data containing parameters such as temperature and humidity within the bubble structure. Volume derivative analysis from a Lagrange perspective refers to a method of calculating the comprehensive rate of change of meteorological elements with time and spatial location, following the movement of meteorological fluid particles. The cooperative rate of change characterizes how the change of one meteorological element changes with the change of another, reflecting the coupling relationship between different meteorological parameters in the evolution process. The calculation relationship between the volume derivative and the cooperative rate of change satisfies the following formula: , , in the formula Indicates meteorological parameter type The volume derivative of . Represents the types of meteorological parameters in a multidimensional meteorological element field. The numerical difference. Indicates the time step interval. and These represent the physical components of the wind speed parameter in the longitude and latitude directions, respectively, in the multidimensional meteorological field. and These represent spatial distances along the longitude and latitude directions, respectively. Indicates meteorological parameter type With meteorological parameter types The dimensionless co-variable rate of change between them. and These represent the types of meteorological parameters. and The historical maximum absolute value of the volume derivative is used to eliminate dimensional differences. Its setting is based on statistical analysis of the historical meteorological evolution process of the target area.

[0031] Within the temporal duration of the meteorological precursor disturbance bubble structure, the system extracts the coordinated rate of change of all spatial grid points at each time step, forming a covariance rate set. This covariance rate set is a data matrix that gathers parametric coupling change information at all spatiotemporal locations within the bubble structure. The system performs principal component analysis on the covariance rate set to extract eigenvector sequences that characterize the main patterns of the corresponding evolution process. Principal component analysis is a mathematical method that uses orthogonal transformation to convert correlated variables into linearly uncorrelated variables, thus achieving data dimensionality reduction. The main patterns of the corresponding evolution process refer to the most representative trends with the largest variance contribution in complex meteorological changes. The eigenvector sequences are numerical sequences reflecting the continuous temporal evolution characteristics of these main patterns.

[0032] The system extracts features from the feature vector sequence to form a fixed-length vector that characterizes the potential triggering capacity of multiple hazards within the bubble structure, serving as the feature vector for the co-evolution of meteorological disturbances. Feature extraction involves statistically aggregating the sequence data along the time dimension to unify the data dimensions. The potential triggering capacity of multiple hazards refers to the likelihood that the meteorological disturbance will trigger various subsequent agricultural disasters. The feature vector for the co-evolution of meteorological disturbances is the standardized data input ultimately used for subsequent disaster prediction. The calculation relationship of feature extraction satisfies the following formula: , in the formula The first characteristic vector representing the co-evolution of meteorological disturbances Each element. This represents the total number of time steps within the duration interval of the weather precursor disturbance bubble structure. Represents the sequence of feature vectors at time step The The values ​​of each principal component.

[0033] For example, the system extracts the anomalous energy concentration from the pure meteorological anomaly field tensor. The temperature anomaly value at a certain spatiotemporal location in the pure meteorological anomaly field tensor is 7 degrees Celsius. The system obtains historical extreme values ​​of temperature based on multi-source time-series observation data of the target area over the past ten years. The temperature value is 14 degrees Celsius. The system calculates the square of the temperature parameter at this location to be 0.25, and combines this with other parameters to obtain the dimensionless anomalous energy concentration. The threshold is set to 0.85. Based on the statistical distribution of energy before historical disaster events, the system sets the dimensionless identification threshold to 0.6. Since 0.85 is greater than 0.6, the system locates spatiotemporal connected regions where the abnormal energy concentration exceeds the preset identification threshold, marking them as meteorological precursor disturbance bubble structures. These structures consist of an envelope describing the spatial morphology and a time duration interval. The envelope covers a specific grid area above the agricultural planting base, and the time duration interval is 48 hours. The system extracts multidimensional meteorological element sequences within the meteorological precursor disturbance bubble structure, performs volume derivative analysis from a Lagrange perspective, and extracts the cooperative change rate of elements as they move with the disturbance body. The system calculates the volume derivatives of temperature and humidity, and based on statistical analysis of historical meteorological evolution, identifies the historical maximum absolute value of the volume derivative of temperature. The humidity was set to a historical maximum absolute value based on a temperature of 2 degrees Celsius per hour. The value is set to 5 per hourly percentage. The system calculates the dimensionless co-variance rate of change between temperature and humidity. The value is 0.4. Within the time duration of the meteorological precursor disturbance bubble structure, the system extracts the coordinated rate of change of all spatial grid points at each time step, forming a set of coordinated rate of change. The system performs principal component analysis on the set of coordinated rate of change to extract feature vector sequences that characterize the main patterns of the corresponding evolution process. The system then performs feature extraction on the feature vector sequences, determining the total number of time steps within the time duration interval. The system calculates the average value of the feature vector sequence over 48 time steps, forming a fixed-length vector that characterizes the potential triggering capability of multiple hazards within the bubble structure, serving as the feature vector for the co-evolution of meteorological disturbances. The first digit of this vector... element The calculation result is 0.15.

[0034] S4. Input the meteorological disturbance co-evolution feature vector into the disaster cascade excitation generator, and generate a disaster causal excitation spectrum by simulating the causal excitation intensity based on abnormal energy transfer between disasters. In one embodiment of the present invention, step S4 includes the following steps: The meteorological disturbance co-evolution feature vector is used as an initial condition and input into a preset disaster cascade excitation generator. The disaster cascade excitation generator includes nodes for characterizing different disaster states and directed edges connecting the nodes. The solver extracts the co-evolutionary feature vectors of meteorological disturbances at the time of the preceding disaster and the co-evolutionary feature vectors of meteorological disturbances at the time of the subsequent disaster, forming the analysis sample. Based on the analysis samples, the covariance matrices of anomalous vorticity, divergence, and vertical motion in the two spatiotemporal windows are calculated to construct the anomalous energy transfer flux. Calculate the advection transport of water vapor or heat between two spatiotemporal windows as the mass transport efficiency; By combining and normalizing the abnormal energy transfer flux and the material transport efficiency, the causal excitation intensity between the two disaster state nodes is obtained. Using the identified disaster state as a node and the causal excitation intensity exceeding a preset intensity threshold as the weight of the directed edges connecting the nodes, a disaster causal excitation graph is constructed. The disaster causal excitation graph depicts the cascaded paths of the transmission and transformation of disturbance energy between different disaster forms.

[0035] Optionally, after generating the disaster causal excitation map, the method further includes:

[0036] Network topology analysis is performed on the disaster causal triggering graph to identify cascaded source nodes with preset out-degree values ​​and cascaded convergence nodes with preset in-degree values; Extract the structured features of the cascaded source nodes and cascaded convergence nodes; The parameters of the solver in the disaster cascade excitation generator are adjusted based on the structured features.

[0037] Specifically, the system uses the meteorological disturbance co-evolution feature vector as initial conditions and inputs it into a pre-defined disaster cascade triggering generator. The meteorological disturbance co-evolution feature vector is standardized data generated in step S3, representing the potential triggering capacity of multiple disasters. The disaster cascade triggering generator is a graph-structured computational model used to simulate the successive occurrence of multiple disasters. It contains nodes representing different disaster states and directed edges connecting these nodes. The nodes representing different disaster states represent specific agricultural disaster types such as drought, flood, and pests. The directed edges connecting the nodes represent the transmission direction of secondary disasters triggered by preceding disasters.

[0038] The system uses a solver to extract the co-evolutionary feature vectors of meteorological disturbances at the time of the preceding disaster and at the time of the subsequent disaster, forming the analysis sample. The solver is an algorithm module that performs time series matching and feature extraction. The time of the preceding disaster refers to the time when the initial disaster manifests. The time of the subsequent disaster refers to the time when the subsequent disaster triggered by the initial disaster manifests. The analysis sample is a data set composed of the pairing of feature vectors at these two specific times, used to quantify the evolutionary relationship between disasters.

[0039] Based on the analyzed samples, the system calculates the covariance matrices of anomalous vorticity, divergence, and vertical motion within two consecutive spatiotemporal windows, constructing the anomalous energy transfer flux. Anomalous vorticity characterizes the degree of anomalous airflow rotation intensity. Divergence characterizes the degree of anomalous outward divergence or inward convergence of airflow. Vertical motion characterizes the degree of anomalous vertical ascent and descent of airflow. The two spatiotemporal windows correspond to the time periods and spatial extents of the preceding and subsequent disasters, respectively. The covariance matrix is ​​a mathematical tool for measuring the degree of coordinated change of the above meteorological dynamic variables within the two spatiotemporal windows. The anomalous energy transfer flux is a dimensionless index quantifying the scale of energy transfer from the preceding to the subsequent state of meteorological disturbance. The calculation relationship of the anomalous energy transfer flux satisfies the following formula: , in the formula It represents the dimensionless anomalous energy transfer flux. This represents the total number of data points extracted within the spatiotemporal window. Indicating the time of occurrence of the preceding disaster in the analysis sample, the first... The combined dynamic component values ​​of the data points integrate the characteristics of anomalous vorticity, divergence, and vertical motion. Indicating the time of occurrence of secondary disasters in the analysis sample. The combined dynamic component values ​​of each data point. and These represent the average values ​​of the integrated dynamic components within the two spatiotemporal windows, respectively. and These represent the standard deviations of the integrated dynamic components within the two spatiotemporal windows, respectively, used to eliminate dimensional differences.

[0040] The system calculates the advection transport of water vapor or heat between two spatiotemporal windows, which is used as the mass transport efficiency. Advection transport refers to the total physical amount of water vapor or heat moving horizontally with the airflow. Mass transport efficiency is a dimensionless index that measures the smoothness of the transfer of disaster-causing substances between disaster nodes. The calculation relationship for mass transport efficiency satisfies the following formula: , in the formula This represents the dimensionless efficiency of material transport. and These represent the first and second spatiotemporal windows, respectively. The advection transport value of water vapor or heat at each data point. This represents the historical maximum advection transport volume of the target area, used to eliminate dimensional differences. Its setting is based on statistical data obtained from meteorological observations of the target area over the past many years.

[0041] The system combines and normalizes anomalous energy transfer flux and material transport efficiency to obtain the causal excitation intensity between two disaster state nodes. Causal excitation intensity is a numerical indicator that quantifies the inevitability of a preceding disaster triggering a secondary disaster. The calculation relationship of causal excitation intensity satisfies the following formula: , in the formula It represents the dimensionless intensity of causal excitation. The dimensionless weighting coefficient representing the flux of abnormal energy transfer. A dimensionless weighting coefficient representing the efficiency of material transport. and The setting is based on the statistical analysis of the relative contributions of energy and material transport in historical disaster cascade events, and the sum of the two is equal to 1.

[0042] The system constructs a disaster causal excitation graph by using identified disaster states as nodes and the causal excitation intensity exceeding a preset intensity threshold as the weight of the directed edges connecting these nodes. Identified disaster states refer to specific disaster types confirmed to be possible within the target area. The intensity threshold is a critical value used to determine whether a causal relationship exists between disasters; its setting is based on the statistical distribution of historical disaster chain occurrence probabilities. The disaster causal excitation graph is a network topology depicting the cascading paths of disturbance energy transmission and transformation between different disaster forms.

[0043] After generating the disaster causal activation graph, the system performs network topology analysis on the graph to identify cascaded source nodes with preset out-degree values ​​and cascaded sink nodes with preset in-degree values. Network topology analysis is a mathematical method for studying the node connectivity and network structure in a graph. The preset out-degree value refers to a standard value for the number of directed edges a node can point to other nodes. A cascaded source node is a starting disaster node that outputs disaster-causing energy and has an out-degree that reaches the preset out-degree value. The preset in-degree value refers to a standard value for the number of directed edges a node can receive from other nodes. A cascaded sink node is a terminal disaster node that receives disaster-causing energy from multiple sources and has an in-degree that reaches the preset in-degree value. The preset out-degree and preset in-degree values ​​are determined based on connectivity statistics of historical disaster networks in the target area.

[0044] The system extracts structured features from cascade source nodes and cascade convergence nodes. Structured features refer to quantitative indicators reflecting the importance and connectivity of nodes in the network, such as node centrality values. Based on these structured features, the system adjusts the parameters of the solver in the disaster cascade trigger generator. The parameter adjustment process involves using the structured features to optimize the solver's time window length when extracting and analyzing samples, thereby improving the accuracy of subsequent disaster matching.

[0045] For example, the system uses the meteorological disturbance co-evolution feature vector as initial conditions and inputs it into a pre-defined disaster cascade excitation generator. The disaster cascade excitation generator contains nodes representing different disaster states and directed edges connecting these nodes. The system extracts the meteorological disturbance co-evolution feature vectors at the time of the preceding disaster and the time of the subsequent disaster through a solver, forming analysis samples. The preceding disaster is set as extreme high temperature, and the subsequent disaster is set as agricultural drought. Based on the analysis samples, the system calculates the covariance matrices of anomalous vorticity, divergence, and vertical motion in the preceding and following spatiotemporal windows, constructing the anomalous energy transfer flux. The total number of data points extracted within the spatiotemporal window is... The value is 24. The system calculates the average value of the integrated dynamic components within the spatiotemporal window of the preceding and subsequent disasters. and and using standard deviation and Dimensionless anomalous energy transfer flux was calculated by eliminating dimensions. The value is 0.75. The system calculates the advection transport of water vapor or heat between two spatiotemporal windows as the mass transport efficiency. The system obtains the historical maximum advection transport value based on meteorological observation data of the target area over the past ten years. The dimensionless mass transport efficiency was calculated. The value is 0.60. The system combines and normalizes the abnormal energy transfer flux and material transport efficiency to obtain the causal excitation intensity between two disaster state nodes. The system uses dimensionless weighting coefficients based on historical disaster cascade event statistics. Set the dimensionless weighting coefficient to 0.6. The value was set to 0.4. The system calculated the dimensionless causal excitation intensity. The intensity threshold is set to 0.69. Based on the historical probability of disaster chains, the system sets the intensity threshold to 0.5. Since 0.69 is greater than 0.5, the system uses the identified disaster state as a node and the causal excitation intensity exceeding the preset intensity threshold as the weight of the directed edges connecting the nodes to construct a disaster causal excitation graph. This graph depicts the cascaded paths of disturbance energy transmission and transformation between different disaster forms. After generating the disaster causal excitation graph, the system performs network topology analysis to identify cascaded source nodes with preset out-degree values ​​and cascaded convergence nodes with preset in-degree values. The system sets the preset out-degree value to 3 and the preset in-degree value to 4. The extreme high-temperature node, with an out-degree of 4, is identified as a cascaded source node. The system extracts the structured features of the cascaded source nodes and cascaded convergence nodes, obtaining a centrality value of 0.82 for the extreme high-temperature node. Based on these structured features, the system adjusts the parameters of the solver in the disaster cascade excitation generator, increasing the time window length for extracting analysis samples from 48 hours to 72 hours.

[0046] S5. Based on the disaster causal triggering map and crop distribution layer, perform counterfactual crop state reasoning, compare the expected crop state under no disaster chain with the predicted crop state under stress, and generate the multi-hazard agricultural collaborative risk level of the target area.

[0047] In one embodiment of the present invention, step S5 includes the following steps: Obtain crop planting type map and phenological period information of the target area as a crop distribution layer; By calling a pre-trained crop growth model and inputting a meteorological field containing only normal climate trends, the crop growth trajectory under the influence of no disaster chain is deduced as the expected state of the crop. The stress intensity, occurrence time, and duration of each disaster type involved in the disaster causal triggering map are used as correction factors and input into the crop growth model to deduce the predicted state of the crop under stress. By comparing the yield and quality differences between the predicted and expected crop states, and based on these differences, a multi-hazard agricultural collaborative risk level for the target area is generated, which quantifies the degree of collaborative stress from the disaster chain.

[0048] Specifically, the system acquires crop planting type maps and phenological information for the target area as a crop distribution layer. The crop planting type map refers to the geographic information data showing the spatial distribution of different crops within the target area. The phenological information refers to the time-specific data of crops at different growth stages. The crop distribution layer is a comprehensive data layer that integrates spatial planting distribution and temporal growth stages.

[0049] After obtaining the crop distribution layer, the system calls a pre-trained crop growth model, inputting only a meteorological field containing normal climate trends, to deduce the crop growth trajectory under conditions free of disaster chains, representing the expected crop state. The pre-trained crop growth model is a computational program that optimizes parameters using historical agricultural data to simulate the physiological development of crops. The meteorological field representing normal climate trends refers to conventional meteorological data that excludes extreme weather disturbances. The crop growth trajectory refers to the dynamic curve of crop biomass changes over time. The expected crop state refers to the yield and quality indicators that the crop can ultimately achieve under ideal climate conditions. The calculation relationship for deduce the expected crop state satisfies the following formula: , in the formula This indicates the expected yield of the crop under the expected conditions. This indicates the total number of days in the growth cycle as defined by phenological information. The meteorological field representing normal climate trends is in the first... The value of a day. Represents the crop distribution layer. Physiological parameters of the phenological period of the day. This represents the daily biomass accumulation function in the pre-trained crop growth model.

[0050] To assess the actual impact of disasters, the system uses the stress intensity, occurrence time, and duration of each disaster type involved in the disaster causal activation map as correction factors, inputting them into the crop growth model to deduce the predicted crop state under stress. The disaster causal activation map is a network diagram depicting the cascading paths of disasters generated in step S4. Stress intensity refers to the degree of damage the disaster inflicts on the physiological functions of crops. Occurrence time refers to the specific date of the disaster. Duration refers to the time span of the disaster's impact. The correction factor is a dimensionless attenuation coefficient used to adjust the normal growth rate. The predicted crop state under stress refers to the final yield and quality indicators of crops under the superimposed influence of multiple disasters. The calculation relationship for deduce the predicted crop state under stress satisfies the following formula: , in the formula This indicates the predicted yield of crops under stress. Indicates the first A set of disaster types that occur under the influence of the elements. Indicates the type of disaster The dimensionless stress intensity is determined based on the transformation of causal excitation intensity in the disaster causal excitation spectrum. Indicates the type of disaster The duration in days. This indicates a multiplication operation, used to synthesize the synergistic attenuation effect of multiple disasters.

[0051] After obtaining data from both states, the system compares the yield and quality differences between the predicted and expected crop states. Based on these differences, it generates a multi-hazard agricultural synergy risk level for the target area, quantifying the degree of synergistic stress from the disaster chain. The yield and quality difference refers to the difference in yield values ​​and quality scores between the expected and predicted states. The multi-hazard agricultural synergy risk level for the target area is a quantitative grading indicator that comprehensively assesses the degree of damage caused by the disaster chain to agricultural production. The calculation relationship for generating the risk level satisfies the following formula: , in the formula This represents a dimensionless comprehensive risk index. This represents the expected quality score in the expected state of the crop. This represents the predicted quality score in the predicted state of crops under stress. The dimensionless weighting coefficient representing output. A dimensionless weighting coefficient representing quality. and The parameters are set based on historical statistics of agricultural economic indicators in the target area, and the sum of the two is equal to 1. The system outputs the final multi-hazard agricultural collaborative risk level of the target area based on the numerical range mapping of the comprehensive risk index.

[0052] For example, the system acquires crop planting type maps and phenological information for the target area as a crop distribution layer. The target area is set as a corn-growing region, and the system extracts the spatial distribution of corn in that area and its corresponding time points, fusing them into a crop distribution layer. The system calls a pre-trained crop growth model, inputting a meteorological field containing only normal climate trends, to extrapolate the crop growth trajectory under conditions without disaster chains, representing the expected crop state. (Total number of days in the growth cycle) The timeframe is set to 120 days. The system calculates the expected yield under the expected crop conditions. Expected quality score of 10,000 kg per hectare The score is 90. The system uses the stress intensity, occurrence time, and duration of each disaster type involved in the disaster causal triggering map as correction factors, inputs them into the crop growth model, and deduce the predicted crop state under stress. Disaster types include drought and pests / diseases. The dimensionless stress intensity of drought is also included. Set to 0.4, duration in days The timeframe is set to 20 days. The system integrates the synergistic attenuation effects of multiple hazards to calculate the predicted yield under stress conditions. Predicted quality score: 7500 kg per hectare The score is 72. The system compares the yield and quality differences between the predicted and expected crop states, and based on these differences, generates a multi-hazard agricultural synergistic risk level for the target area, quantifying the degree of synergistic stress in the disaster chain. The system also uses a dimensionless weighting coefficient for yield based on historical statistics of agricultural economic indicators in the target area. Setting it to 0.7 sets the dimensionless weighting coefficient for quality. Set to 0.3. The system calculates the dimensionless comprehensive risk index. The value is 0.235. Based on the preset risk classification standard, the system maps 0.235 to medium risk, and finally outputs that the multi-hazard agricultural collaborative risk level of the target area is medium risk.

[0053] Based on the same inventive concept, such as Figure 2 As shown, the present invention also provides a joint analysis system for agricultural disasters based on meteorological data, the system comprising: The physical field reconstruction module is used to acquire multidimensional meteorological observation data from multiple observation points within the target area, perform physical consistency multidimensional element field reconstruction, and generate a multidimensional meteorological element field. An anomaly separation module is used to perform background trend fitting and residual separation on the multidimensional meteorological element field to generate a pure meteorological anomaly field tensor. The precursor coding module is used to identify the weather precursor disturbance bubble structure in the pure weather anomaly field tensor, and to perform dynamic coding on the co-evolution trajectory of meteorological elements inside the bubble structure to generate a weather disturbance co-evolution feature vector. The cascaded inference module is used to input the meteorological disturbance co-evolution feature vector into the disaster cascade excitation generator, and generate a disaster causal excitation spectrum by simulating the causal excitation intensity based on abnormal energy transfer between disasters; The risk assessment module is used to perform counterfactual crop state reasoning based on the disaster causal triggering map and crop distribution layer, compare the expected crop state under no disaster chain with the predicted crop state under stress, and generate a multi-hazard agricultural collaborative risk level for the target area.

[0054] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for joint analysis of agricultural disasters based on meteorological data, characterized in that, The method includes: Acquire multidimensional meteorological observation data from multiple observation points within the target area, reconstruct a physically consistent multidimensional element field, and generate a multidimensional meteorological element field. Background trend fitting and residual separation are performed on the multidimensional meteorological element field to generate a pure meteorological anomaly field tensor. In the pure meteorological anomaly field tensor, the bubble structure of meteorological precursor disturbances is identified, and the co-evolution trajectory of meteorological elements inside the bubble structure is dynamically encoded to generate a feature vector of meteorological disturbance co-evolution. The meteorological disturbance co-evolution feature vector is input into the disaster cascade excitation generator, and a disaster causal excitation spectrum is generated by simulating the causal excitation intensity based on abnormal energy transfer between disasters. Based on the disaster causal triggering map and crop distribution layer, counterfactual crop state reasoning is performed to compare the expected crop state under no disaster chain with the predicted crop state under stress, and generate a multi-hazard agricultural collaborative risk level for the target area.

2. The method for joint analysis of agricultural disasters based on meteorological data according to claim 1, characterized in that, The generated multidimensional meteorological element field includes: Acquire multi-source time-series observation data of temperature, humidity, wind speed, and precipitation from multiple observation points within the target area; Geographically weighted spatiotemporal interpolation is performed on the multi-source time-series observation data to obtain a regular gridded four-dimensional tensor. The dimensions of the four-dimensional tensor include time, longitude, latitude, and meteorological parameter types. Fluid dynamic constraints based on the Navier-Stokes equations and the continuity equation are applied to the four-dimensional tensor, and variational calibration is performed to generate a multi-dimensional meteorological element field with a spatial structure that conforms to physical laws.

3. The method for joint analysis of agricultural disasters based on meteorological data according to claim 2, characterized in that, The generated pure meteorological anomaly field tensor includes: On the multidimensional meteorological element field, a sliding window filter is applied to the meteorological parameter sequence at each spatial grid point along the time axis to extract the long-period low-pass signal as the climate background trend field. The multidimensional meteorological element field and the climate background trend field are subtracted element by element on the same spatiotemporal coordinates to obtain the abnormal residual field that represents the deviation from the normal state. The abnormal residual field is subjected to a statistical significance test. The abnormal signals that pass the significance test are retained, and a pure meteorological anomaly field tensor containing only statistically significant catastrophic disturbance information is generated.

4. The method for joint analysis of agricultural disasters based on meteorological data according to claim 3, characterized in that, The generated meteorological disturbance co-evolution feature vector includes: Extract the anomalous energy concentration degree from the pure meteorological anomaly field tensor, locate the spatiotemporal connected domains where the anomalous energy concentration degree exceeds a preset identification threshold, and mark them as meteorological precursor disturbance bubble structures. The meteorological precursor disturbance bubble structure consists of an envelope surface describing the spatial morphology and a time duration interval. The multidimensional meteorological element sequence within the structure of the meteorological precursor disturbance bubble is extracted, and the body derivative analysis is performed from the perspective of Lagrange. The cooperative change rate of the elements as they move with the disturbance body is extracted. The cooperative change rate characterizes how the change of one meteorological element changes with the change of another meteorological element. The coordinated rate of change within the envelope is extracted temporally to form a fixed-length vector that can characterize the potential triggering ability of multiple hazards within the bubble structure, which serves as the feature vector for the coordinated evolution of meteorological disturbances.

5. The method for joint analysis of agricultural disasters based on meteorological data according to claim 4, characterized in that, The step of extracting features from the coordinated rate of change within the envelope surface over time to form a fixed-length vector that characterizes the potential triggering capability of multiple hazards within the bubble structure, serving as a feature vector for the coordinated evolution of meteorological disturbances, includes: Within the time duration interval of the meteorological precursor disturbance bubble structure, the cooperative rate of change of all spatial grid points at each time step is extracted to form a set of cooperative rates of change. Principal component analysis was performed on the covariance set to extract feature vector sequences that can characterize the main patterns of the corresponding evolution process; Feature extraction is performed on the feature vector sequence to form a fixed-length vector that can characterize the potential triggering ability of multiple disasters within the bubble structure, which serves as the feature vector for the co-evolution of meteorological disturbances.

6. The method for joint analysis of agricultural disasters based on meteorological data according to claim 5, characterized in that, The generated disaster causal triggering map includes: The meteorological disturbance co-evolution feature vector is used as an initial condition and input into a preset disaster cascade excitation generator. The disaster cascade excitation generator includes nodes for characterizing different disaster states and directed edges connecting the nodes. The solver calculates the anomalous energy transfer flux and material transport efficiency between any two disaster state nodes that occur sequentially in time, and calculates the causal excitation intensity of the evolution from one disaster state to another. Using the identified disaster state as a node and the causal excitation intensity exceeding a preset intensity threshold as the weight of the directed edges connecting the nodes, a disaster causal excitation graph is constructed. The disaster causal excitation graph depicts the cascaded paths of the transmission and transformation of disturbance energy between different disaster forms.

7. The method for joint analysis of agricultural disasters based on meteorological data according to claim 6, characterized in that, The calculation of the causal triggering intensity of the evolution from one disaster state to another includes: The co-evolutionary feature vectors of meteorological disturbances at the time of the preceding disaster and the co-evolutionary feature vectors of meteorological disturbances at the time of the subsequent disaster are extracted to form the analysis sample; Based on the analysis samples, the covariance matrices of anomalous vorticity, divergence, and vertical motion in the two spatiotemporal windows are calculated to construct the anomalous energy transfer flux. Calculate the advection transport of water vapor or heat between two spatiotemporal windows as the mass transport efficiency; By combining and normalizing the abnormal energy transfer flux and the material transport efficiency, the causal excitation intensity between the two disaster state nodes is obtained.

8. The method for joint analysis of agricultural disasters based on meteorological data according to claim 6, characterized in that, After generating the disaster causal activation map, the method further includes: performing network topology analysis on the disaster causal activation map to identify cascaded source nodes with preset out-degree values ​​and cascaded convergence nodes with preset in-degree values; Extract the structured features of the cascaded source nodes and cascaded convergence nodes; The parameters of the solver in the disaster cascade excitation generator are adjusted based on the structured features.

9. The method for joint analysis of agricultural disasters based on meteorological data according to claim 6, characterized in that, The generated target area's multi-hazard agricultural collaborative risk level includes: Obtain crop planting type map and phenological period information of the target area as a crop distribution layer; By calling a pre-trained crop growth model and inputting a meteorological field containing only normal climate trends, the crop growth trajectory under the influence of no disaster chain is deduced as the expected state of the crop. The stress intensity, occurrence time, and duration of each disaster type involved in the disaster causal triggering map are used as correction factors and input into the crop growth model to deduce the predicted state of the crop under stress. By comparing the yield and quality differences between the predicted and expected crop states, and based on these differences, a multi-hazard agricultural collaborative risk level for the target area is generated, which quantifies the degree of collaborative stress from the disaster chain.

10. A joint analysis system for agricultural disasters based on meteorological data, characterized in that: The system includes: The physical field reconstruction module is used to acquire multidimensional meteorological observation data from multiple observation points within the target area, perform physical consistency multidimensional element field reconstruction, and generate a multidimensional meteorological element field. An anomaly separation module is used to perform background trend fitting and residual separation on the multidimensional meteorological element field to generate a pure meteorological anomaly field tensor. The precursor coding module is used to identify the weather precursor disturbance bubble structure in the pure weather anomaly field tensor, and to perform dynamic coding on the co-evolution trajectory of meteorological elements inside the bubble structure to generate a weather disturbance co-evolution feature vector. The cascaded inference module is used to input the meteorological disturbance co-evolution feature vector into the disaster cascade excitation generator, and generate a disaster causal excitation spectrum by simulating the causal excitation intensity based on abnormal energy transfer between disasters; The risk assessment module is used to perform counterfactual crop state reasoning based on the disaster causal triggering map and crop distribution layer, compare the expected crop state under no disaster chain with the predicted crop state under stress, and generate a multi-hazard agricultural collaborative risk level for the target area.

Images