A Method for Constructing a Composite Agricultural Meteorological Disaster Monitoring Index System

Abstract

This invention relates to the field of agricultural meteorological disaster monitoring technology, and particularly to a method for constructing a composite agricultural meteorological disaster monitoring indicator system. The method includes the following steps: acquiring regional agricultural component data and performing spatial distribution analysis to identify agricultural distribution boundaries, thereby delineating overlapping agricultural areas; then collecting historical meteorological disaster data and generating spatiotemporal sequences of meteorological disasters; identifying the correlation patterns between meteorological and agricultural disasters; predicting composite agricultural boundary development data based on overlapping agricultural areas; identifying usable resources and conducting heterogeneous component interaction simulations; constructing an agricultural interaction effect field; inferring the meteorological disaster absorption capacity; and finally, predicting composite agricultural disasters based on historical correlation patterns and absorption capacity, constructing a monitoring indicator system, and realizing the systematic monitoring and management of agricultural meteorological disasters. This invention achieves systematic monitoring and assessment of agricultural meteorological disasters, promoting the intelligent and scientific management of agriculture.

Description

Technical Field

[0001] This invention relates to the field of agricultural meteorological disaster monitoring technology, and in particular to a method for constructing a composite agricultural meteorological disaster monitoring indicator system. Background Technology

[0002] Traditional technologies have many shortcomings in monitoring and forecasting meteorological disasters. Traditional methods often rely on a single data source and lack comprehensive analysis of the complex relationship between agricultural components and meteorological factors, resulting in insufficient early warning capabilities for meteorological disasters. In particular, when faced with the interactive influence of multiple meteorological factors, existing monitoring methods cannot identify potential agricultural risks in a timely and accurate manner, which makes agricultural production face high uncertainty and risk of loss. In practical application scenarios, the meteorological and ecological environment of agricultural areas is complex and changeable. Existing technologies rely heavily on static models for spatial distribution analysis of regional agricultural components, which cannot reflect dynamic changes in a timely manner. They also lack effective identification of resources in overlapping agricultural areas, resulting in inadequate management and allocation of usable resources. The utilization of historical meteorological disaster data is not systematic enough, and there is a lack of effective spatiotemporal sequence analysis methods, which prevents in-depth exploration of the correlation patterns between meteorological and agricultural disasters, thus affecting the scientificity and effectiveness of disaster response strategies. Summary of the Invention

[0003] Therefore, it is necessary to provide a method for constructing a composite agricultural meteorological disaster monitoring index system to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a method for constructing a composite agricultural meteorological disaster monitoring index system includes the following steps:

[0005] Step S1: Obtain regional agricultural component data; perform spatial distribution analysis on the regional agricultural component data and identify agricultural distribution boundaries; delineate overlapping agricultural areas based on agricultural distribution boundaries and regional agricultural component data;

[0006] Step S2: Collect a collection of historical meteorological disasters; discretize the historical meteorological disaster collection into a spatiotemporal sequence to generate a spatiotemporal sequence of meteorological disasters; identify the correlation pattern between historical meteorological and agricultural disasters based on the spatiotemporal sequence of meteorological disasters;

[0007] Step S3: Predict the development data of the complex agricultural boundary based on the agricultural overlapping area; identify the agricultural resources available in the agricultural overlapping area through the complex agricultural boundary development data, and conduct heterogeneous component interaction simulation based on the agricultural resources available in the area to generate heterogeneous agricultural component interaction data;

[0008] Step S4: Construct an agricultural interaction effect field using data from interactions between different agricultural components; infer meteorological disaster mitigation capacity based on the agricultural interaction effect field and regional agricultural component data;

[0009] Step S5: Based on the historical meteorological-agricultural disaster correlation model and the meteorological disaster absorption capacity, conduct composite agricultural disaster prediction and generate predicted agricultural meteorological disasters; construct a composite agricultural meteorological disaster monitoring index system based on the predicted agricultural meteorological disasters.

[0010] This invention comprehensively understands the composition and structure of regional agriculture by acquiring regional agricultural component data, providing a data foundation for subsequent analysis. Spatial distribution analysis and agricultural distribution boundary identification clarify the geographical characteristics of agricultural resources, delineate overlapping agricultural areas, and enhance the targeted utilization of agricultural resources. Collecting historical meteorological disaster data and discretizing its spatiotemporal sequence to generate a meteorological disaster spatiotemporal sequence improves the understanding of meteorological disaster occurrence patterns. Identifying historical meteorological-agricultural disaster correlation patterns provides historical evidence for risk assessment. Predicting the development of composite agricultural boundaries based on overlapping agricultural areas effectively identifies usable resources and promotes optimal resource allocation. Through heterogeneous component interaction simulation, the generated heterogeneous agricultural component interaction data provides insights into the interactions between different agricultural components. Based on scientific evidence, an agricultural interaction field was constructed, enhancing the understanding of the mutual influence of various factors within the agricultural ecosystem. Inferring meteorological disaster absorption capacity based on the agricultural interaction field and regional agricultural component data provides crucial support for disaster prevention and mitigation. Composite agricultural disaster prediction based on historical meteorological-agricultural disaster correlation patterns and meteorological disaster absorption capacity significantly improves the accuracy of agricultural meteorological disaster early warning. A composite agricultural meteorological disaster monitoring indicator system was constructed, enabling systematic monitoring and assessment of agricultural meteorological disasters. This promotes the intelligent and scientific management of agriculture, forming a complete monitoring framework that provides effective protection for agricultural production, enhances the resilience of agricultural production, promotes sustainable development, and ultimately provides strong support for agricultural ecological security and resource management. Attached Figure Description

[0011] Figure 1 A schematic diagram illustrating the steps involved in constructing a composite agricultural meteorological disaster monitoring index system;

[0012] Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S2.

[0013] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0014] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0015] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0016] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0017] To achieve the above objectives, please refer to Figures 1 to 2 A method for constructing a composite agricultural meteorological disaster monitoring index system includes the following steps:

[0018] Step S1: Obtain regional agricultural component data; perform spatial distribution analysis on the regional agricultural component data and identify agricultural distribution boundaries; delineate overlapping agricultural areas based on agricultural distribution boundaries and regional agricultural component data;

[0019] Step S2: Collect a collection of historical meteorological disasters; discretize the historical meteorological disaster collection into a spatiotemporal sequence to generate a spatiotemporal sequence of meteorological disasters; identify the correlation pattern between historical meteorological and agricultural disasters based on the spatiotemporal sequence of meteorological disasters;

[0020] Step S3: Predict the development data of the complex agricultural boundary based on the agricultural overlapping area; identify the agricultural resources available in the agricultural overlapping area through the complex agricultural boundary development data, and conduct heterogeneous component interaction simulation based on the agricultural resources available in the area to generate heterogeneous agricultural component interaction data;

[0021] Step S4: Construct an agricultural interaction effect field using data from interactions between different agricultural components; infer meteorological disaster mitigation capacity based on the agricultural interaction effect field and regional agricultural component data;

[0022] Step S5: Based on the historical meteorological-agricultural disaster correlation model and the meteorological disaster absorption capacity, conduct composite agricultural disaster prediction and generate predicted agricultural meteorological disasters; construct a composite agricultural meteorological disaster monitoring index system based on the predicted agricultural meteorological disasters.

[0023] This invention comprehensively understands the composition and structure of regional agriculture by acquiring regional agricultural component data, providing a data foundation for subsequent analysis. Spatial distribution analysis and agricultural distribution boundary identification clarify the geographical characteristics of agricultural resources, delineate overlapping agricultural areas, and enhance the targeted utilization of agricultural resources. Collecting historical meteorological disaster data and discretizing its spatiotemporal sequence to generate a meteorological disaster spatiotemporal sequence improves the understanding of meteorological disaster occurrence patterns. Identifying historical meteorological-agricultural disaster correlation patterns provides historical evidence for risk assessment. Predicting the development of composite agricultural boundaries based on overlapping agricultural areas effectively identifies usable resources and promotes optimal resource allocation. Through heterogeneous component interaction simulation, the generated heterogeneous agricultural component interaction data provides insights into the interactions between different agricultural components. Based on scientific evidence, an agricultural interaction field was constructed, enhancing the understanding of the mutual influence of various factors within the agricultural ecosystem. Inferring meteorological disaster absorption capacity based on the agricultural interaction field and regional agricultural component data provides crucial support for disaster prevention and mitigation. Composite agricultural disaster prediction based on historical meteorological-agricultural disaster correlation patterns and meteorological disaster absorption capacity significantly improves the accuracy of agricultural meteorological disaster early warning. A composite agricultural meteorological disaster monitoring indicator system was constructed, enabling systematic monitoring and assessment of agricultural meteorological disasters. This promotes the intelligent and scientific management of agriculture, forming a complete monitoring framework that provides effective protection for agricultural production, enhances the resilience of agricultural production, promotes sustainable development, and ultimately provides strong support for agricultural ecological security and resource management.

[0024] In this embodiment of the invention, the method for constructing a composite agricultural meteorological disaster monitoring index system includes the following steps:

[0025] Step S1: Obtain regional agricultural component data; perform spatial distribution analysis on the regional agricultural component data and identify agricultural distribution boundaries; delineate overlapping agricultural areas based on agricultural distribution boundaries and regional agricultural component data;

[0026] In this embodiment, during the acquisition of regional agricultural component data, the SpectralView-S500 multispectral imaging device equipped on the ground remote sensing platform was used to collect data on the vegetation index NDVI (Normalized Difference Vegetation Index), soil moisture content, crop height, and leaf area index (LAI) of major crops in the region. The collection frequency was set to once a week, and the collection accuracy was set to a spatial resolution of 1 meter. The collected data was spatially distributed using the ArcGIS Pro 3.0 platform. The Kriging interpolation method was used to generate a continuous spatial distribution map of agricultural components. The distribution boundaries of major crops were extracted based on the threshold segmentation method. Areas with NDVI greater than 0.6 and leaf area index greater than 3 were set as the boundaries of high-yield crop areas. The Convex Hull algorithm was used to vectorize the boundaries of each crop type. The overlapping areas of agriculture were delineated based on the vector boundary overlap analysis. Specifically, the Intersect overlay analysis tool was used to extract the overlapping parts of the boundaries of each crop type, generating a vector dataset of agricultural overlapping areas.

[0027] Step S2: Collect a collection of historical meteorological disasters; discretize the historical meteorological disaster collection into a spatiotemporal sequence to generate a spatiotemporal sequence of meteorological disasters; identify the correlation pattern between historical meteorological and agricultural disasters based on the spatiotemporal sequence of meteorological disasters;

[0028] In this embodiment, during the collection of historical meteorological disaster data, disaster events such as drought, rainstorm, hail, strong wind, and frost damage that occurred in the region over a period of time are collected. The collected data includes five items: occurrence time, duration, affected area, type of affected crops, and degree of damage. The Python programming language combined with the Pandas library is used for data cleaning and standardization. Each meteorological disaster event is discretized according to four time scales: year, month, day, and hour, and with five decimal places of latitude and longitude precision. During the discretization process, a small grid method based on a fixed radius (50 kilometers) is used for spatial discretization. Disaster events within each grid cell are assigned weight values ​​to generate spatiotemporal sequence data of meteorological disasters. The DBSCAN density clustering algorithm is used to identify disaster clustering patterns and extract the correlation patterns between historical meteorological disasters and damage to specific agricultural components. The clustering parameters are set to a minimum sample size of 5 and a radius parameter of 0.1 degrees latitude and longitude.

[0029] Step S3: Predict the development data of the complex agricultural boundary based on the agricultural overlapping area; identify the agricultural resources available in the agricultural overlapping area through the complex agricultural boundary development data, and conduct heterogeneous component interaction simulation based on the agricultural resources available in the area to generate heterogeneous agricultural component interaction data;

[0030] In this embodiment, during the prediction of the development data of the composite agricultural boundary based on the agricultural overlapping areas, a Support Vector Regression (SVR) model is used to predict the development trend of the agricultural boundary areas. The input features are the area change rate, crop type change frequency, and land use index (LUI) of each agricultural overlapping area in the past 5 years. The kernel function of the SVR model is set to the Radial Basis Function (RBF), the penalty coefficient C is set to 100, and the gamma value is set to 0.01. After training, the area change trend of each agricultural overlapping area in the next 3 years is predicted. Furthermore, the soil nutrient content, organic matter content, and water accessibility data of each overlapping area are extracted from the land resource database. Based on the above prediction data and resource data, a linear programming method is used to construct a heterogeneous agricultural component interaction simulation model to simulate the nutrient flow, water competition, and spatial expansion trends under different crop boundary conditions, generating a heterogeneous agricultural component interaction dataset. The output format is GeoTIFF raster data.

[0031] Step S4: Construct an agricultural interaction effect field using data from interactions between different agricultural components; infer meteorological disaster mitigation capacity based on the agricultural interaction effect field and regional agricultural component data;

[0032] In this embodiment, during the construction of the agricultural interaction effect field using heterogeneous agricultural component interaction data, a three-dimensional spatial interpolation method is used. The water migration rate, nutrient migration rate, and biomass competition intensity indicators in the heterogeneous agricultural component interaction data are subdivided into spatial grids with a resolution of 10 meters × 10 meters. Natural Neighbor Interpolation is then used to generate a three-dimensional agricultural interaction effect field. Each grid cell in the effect field records the water flow direction, nutrient exchange rate, and competition pressure. Subsequently, the agricultural interaction effect field is overlaid and analyzed with regional agricultural component data. Through a Bayesian inference-based inference model, the meteorological disaster absorption capacity of each grid cell is inferred. In the inference process, the meteorological disaster absorption capacity is defined as the crop survival probability per unit area when a disaster occurs. The model sets the prior distribution to a Beta distribution, and the hyperparameters are α = 2 and β = 5.

[0033] Step S5: Based on the historical meteorological-agricultural disaster correlation model and the meteorological disaster absorption capacity, conduct composite agricultural disaster prediction and generate predicted agricultural meteorological disasters; construct a composite agricultural meteorological disaster monitoring index system based on the predicted agricultural meteorological disasters.

[0034] In this embodiment, during the prediction of composite agricultural disasters based on historical meteorological-agricultural disaster correlation patterns and meteorological disaster absorption capacity, the input features of historical meteorological-agricultural disaster correlation patterns and meteorological disaster absorption capacity parameters of each overlapping agricultural area are used. A time-series prediction model is established using a Long Short-Term Memory (LSTM) network. The LSTM network structure is set to three layers, with 128 hidden units in each layer. The input time step is 30 days, and the output is the probability of agricultural disasters occurring in the next 30 days and the intensity level of disaster impact. During training, the Adam optimizer is used, the learning rate is set to 0.001, and the loss function is set to the cross-entropy loss function. By predicting the probability and intensity of agricultural meteorological disasters, a composite agricultural meteorological disaster monitoring index system is constructed. The index system includes the probability of disaster occurrence, the percentage of disaster-affected area, the disaster severity index of key crops, and so on.

[0035] Preferably, step S1 includes the following steps:

[0036] Step S11: Obtain regional agricultural component data; slice the regional agricultural component data into plots to obtain sliced ​​agricultural components, wherein the area of ​​each plot slice unit is limited to 1-5 hectares;

[0037] Step S12: Perform spatial density mapping on the sliced ​​agricultural components to generate agricultural spatial density distribution data, with the spatial density resolution set to a 100m × 100m grid.

[0038] Step S13: Track the edge contours of agricultural spatial density distribution data to obtain the agricultural distribution boundary, wherein the minimum connected area threshold for contour extraction is set to 0.5 square kilometers;

[0039] Step S14: Cross-over data of agricultural distribution boundaries and sliced ​​agricultural components are performed to obtain overlapping agricultural areas.

[0040] In this embodiment, during the acquisition of regional agricultural component data and the granular slicing of plots, firstly, based on publicly available multi-source remote sensing image data, multispectral images with a spatial resolution better than 5 meters are selected as the basic data source. The remote sensing image preprocessing workflow, including radiometric correction, atmospheric correction, and topographic correction, is performed. The FLAASH (Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes) module in ENVI software is used to complete the correction operation. Then, an object-oriented image segmentation method is adopted, using the Multi-resolution Segmentation algorithm in eCognition software to slice the regional agricultural component data at the plot level. The slicing parameters are set as follows: scale factor 50, shape factor 0.3, and compactness factor 0.7. Area statistics are used to ensure that the area of ​​each plot unit is controlled between 1 hectare and 5 hectares. After slicing, the main agricultural component features of each unit are extracted, including the Normalized Difference Vegetation Index (NDVI) and the Normalized Difference Water Index (NDWI). Using the Normalized Difference Water Index (NDVI) and crop classification labels, the spatial density mapping of agricultural components in plots was performed. First, the centroid coordinates of each plot were extracted as point feature data. Spatial density modeling was then conducted using the Kernel Density tool in ArcGIS, with a normal distribution kernel, a Gaussian radius of 150 meters, and an output spatial density resolution of 100m × 100m. Density values ​​per unit area were weighted and superimposed based on the number of centroids and their associated NDVI values. Specifically, the weighting method was to multiply the contribution of each centroid point by the NDVI value normalized to the 0-1 range and a fixed coefficient of 0.8. The resulting agricultural spatial density distribution data was output in raster format using the WGS84 geographic coordinate system. Raster values ​​were linearly normalized to the 0-255 range to facilitate subsequent image processing. Finally, a 5×5 window size was applied to the raster map. Median filtering is used to remove isolated noise points. In tracking the edge contours of agricultural spatial density distribution data, the normalized density distribution raster data is first binarized using a fixed threshold of 180. Pixels with a density value greater than 180 are set to 1, and those less than or equal to 180 are set to 0. The OpenCV `findContours` function is used for edge contour extraction. The contour retrieval mode is set to `RETR_EXTERNAL` to extract the outer contour, and the contour approximation method is set to `CHAIN_APPROX_SIMPLE` to compress redundant points in the horizontal, vertical, and diagonal directions. Each extracted contour line is filtered based on its area attribute, with areas less than 0.Outlines with fewer than 500 raster cells corresponding to a 5 square kilometer area are directly discarded. The remaining outlines are further smoothed using the Douglas-Peucker algorithm with a tolerance parameter set to 5 meters to remove high-frequency noise and retain the main shape features. The final output is agricultural distribution boundary vector data in ESRI file format. The Shapefile maintains the same coordinate system as the original data and records the area, perimeter, and maximum internal density of each boundary. In the process of cross-overlaying agricultural distribution boundaries and sliced ​​agricultural component data to obtain agricultural overlapping areas, ArcGIS's Intersect tool is used to perform spatial overlay analysis. The input data are agricultural distribution boundary vector data and sliced ​​agricultural component vector data. The overlay operation type is set to geometric cross. Each pair of overlapping areas forms a new polygonal feature. The output new features inherit all attributes of the original two layers within the overlay area, including crop type, NDVI, NDWI, slice area, and boundary area. Further filtering is performed on the generated agricultural overlapping area features after overlay, removing small areas with an overlay area of ​​less than 0.2 hectares to ensure the stability and reliability of subsequent analysis data. The final agricultural overlapping area dataset is saved in GeoPackage format, with fields including overlapping area, overlapping crop type combination, mean overlay NDVI, and mean overlay NDWI.

[0041] Preferably, step S2 includes the following steps:

[0042] Step S21: Collect a collection of historical meteorological disasters; perform time-series subdivision on the collection of historical meteorological disasters to obtain the disaster occurrence time series, wherein the time subdivision step size is set to 7 days;

[0043] Step S22: Perform spatial raster coding on the historical meteorological disaster collection to generate disaster spatial distribution data, wherein the side length of the raster unit is set to 500 meters;

[0044] Step S23: Perform joint interpolation processing on the disaster occurrence time series and disaster spatial distribution data to generate meteorological disaster spatiotemporal series data, wherein the interpolation time window width is limited to ±3 days and the spatial interpolation radius is set to 1 kilometer;

[0045] Step S24: Map the spatiotemporal sequence of meteorological disasters to agricultural components based on regional agricultural component data to obtain meteorological-agricultural disaster matching data, wherein the matching tolerance time difference is set to ≤5 days;

[0046] Step S25: Extract interactive correlation patterns from meteorological-agricultural disaster matching data to generate historical meteorological-agricultural disaster correlation patterns.

[0047] In this embodiment, historical records are synchronized through the system database. Data types include six categories of meteorological disasters: rainstorms, droughts, hail, high temperatures, low temperatures, and strong winds. Collected fields include the disaster occurrence date, duration (in days), disaster center latitude and longitude, disaster impact range, and disaster level. All data is uniformly stored in a PostgreSQL database format. Data fields are standardized, and the time field format is set to YYYY-MM-DD. Then, the data is divided into time series segments according to a 7-day partitioning step, with each 7-day segment representing a time interval. The data slides continuously from the earliest collected record, with boundaries aligned to the natural week. During the partitioning process, isolated disaster events with a time span of less than one day are removed. Simultaneously, disaster events spanning multiple time periods are also analyzed. Disaster events are weighted and their impact is distributed across different time periods according to their duration, resulting in a disaster occurrence time series. This time series is stored as an independent time index table, where each record includes start and end dates, disaster type, time period code, and event number. In generating spatial distribution data by spatially rasterizing historical meteorological disaster collections, the disaster impact range data is first vectorized. For disaster records with a center point and radius, a buffer method is used to generate polygons representing the corresponding impact range. For records with existing polygon impact ranges, the original boundaries are used directly. The projection coordinate system is unified to UTM (Universal Transverse Mercator) Zone 50N. Then, a standard grid is established across the entire area, with grid cell side lengths set to 500 meters. ArcGIS's Fishnet tool is used to generate a regular grid, with raster numbering starting from the southwest corner, increasing eastward, and advancing northward. Finally, Spatial... The Join (spatial connection) method spatially overlays each disaster-affected polygon with a raster cell. If the area overlap between the raster and the disaster-affected polygon exceeds 25%, the raster number and disaster type are recorded. The final output disaster spatial distribution data is in the form of a point raster. Each raster records the number of disasters and the set of disaster categories. The data is saved as a GeoTIFF raster file and the corresponding attribute table. In the process of jointly interpolating the disaster occurrence time series and disaster spatial distribution data to generate meteorological disaster spatiotemporal sequence data, the interpolation time window is first determined based on the disaster occurrence time series. The width of the interpolation time window is limited to ±3 days. A step-by-step approach is used to time-complement all disaster events within each time segment. A Gaussian weight function is set, with a weight of 1 when the time distance is 0 days and a weight decaying to 0 when the distance is 3 days.1. After temporal interpolation, spatial interpolation is performed with a radius of 1 km. Inverse Distance Weighting (IDW) is used for spatial extrapolation, with the weight formula set to 1 / squared distance. For each target grid point, the search radius contains a maximum of 15 disaster records. If there are fewer than 3 disaster records within the search radius, the grid point is interpolated as null. After interpolation, a spatiotemporal sequence of meteorological disasters is generated and stored as a four-dimensional matrix. The dimensions are time period, spatial row number, spatial column number, and disaster type number. The matrix values ​​are standardized disaster impact intensity values. The standardization process maps disaster levels to a 0-1 range, assigning the highest weight to severe disasters such as strong winds and heavy rain, and appropriately decreasing weight to chronic disasters such as drought and frost. This is based on regional agriculture. In the process of mapping agricultural components to the spatiotemporal series of meteorological disasters to obtain meteorological-agricultural disaster matching data, the first step is to align the plot vector boundaries of the regional agricultural component data with the raster boundaries in the spatiotemporal series of meteorological disasters. Through spatial overlay analysis, each agricultural plot is matched one-to-one with the disaster raster it covers. The overlay operation requires that the overlap area between the agricultural plot and the disaster raster exceeds 30% to be considered a valid match. Then, for each agricultural plot, time matching is performed according to the disaster time label and the agricultural growth cycle timetable. The time difference tolerance for matching is set to no more than 5 days. During matching, the disaster event with the smallest time difference is selected as the associated disaster. If multiple disasters match simultaneously, the matching is not performed. The data is sorted by disaster impact intensity and the highest value is selected. After matching, a meteorological-agricultural disaster matching data table is generated. Record fields include agricultural plot number, crop type, corresponding disaster type, disaster intensity value, time difference, and number of overlapping rasters. In the process of extracting interaction association patterns from the meteorological-agricultural disaster matching data to generate historical meteorological-agricultural disaster association patterns, the Apriori frequent itemset mining algorithm (association rule mining method) is first used to analyze the matching data table. The minimum support threshold is set to 0.01, and the minimum confidence threshold is set to 0.7. The co-occurrence relationship between crop type and disaster type is analyzed to mine high-frequency association rules. During the mining process, crop... Using disaster type as a prerequisite and disaster type as the result, all rules meeting the conditions are extracted, and the lift index is calculated. Rules with a lift greater than 1.2 are considered valid strongly correlated rules. Furthermore, the K-means clustering algorithm is used to cluster agricultural plots according to their disaster impact characteristics. The initial number of cluster centers is set to 8, and the maximum number of iterations is set to 300. After clustering, the disaster type distribution and crop response characteristics of each cluster center are statistically analyzed to form a meteorological-agricultural disaster interaction pattern cluster. The final output is a CSV file and a relational database table, with fields including cluster number, main disaster type, corresponding crop category, average disaster impact intensity, and number of typical cases.

[0048] Preferably, step S3, which involves predicting the development data of the composite agricultural boundary based on overlapping agricultural areas, includes:

[0049] The overlapping agricultural areas are subdivided into raster layers to generate an agricultural overlapping grid.

[0050] Extracting agricultural boundary gradients based on overlapping agricultural grids;

[0051] The boundary profile of overlapping agricultural regions is determined by agricultural boundary gradient;

[0052] Morphological refinement of the interface contours was performed, and an agricultural interface topology network was constructed.

[0053] Analyze the connectivity of the agricultural interface topology network to identify the interface connected components;

[0054] Regional growth prediction is performed based on interface connectivity to generate predicted agricultural interface blocks.

[0055] Data on the development of complex agricultural boundaries is determined based on predicted agricultural interface blocks.

[0056] In this embodiment, during the process of subdividing the agricultural overlapping areas into raster subdivisions to generate agricultural overlapping meshes, the previously generated vector data of the agricultural overlapping areas is first imported and uniformly projected onto WGS-84 (World Geodetic System). The 1984 World Geodetic System (WGMS) coordinate system was used to ensure spatial consistency across data layers. The Rasterize tool from the GDAL library was then used to convert vector polygons into raster data, with an output resolution of 20 meters. The raster encoding rule was based on agricultural type, using the crop category with the largest proportion in agricultural plots as the raster encoding attribute. The fill value for blank areas was set to -1, and the output data format was GeoTIFF. During subdivision, a raster filling strategy was used to ensure that each original agricultural overlapping area was divided into at least 300 independent raster units. If a small agricultural overlapping area resulted in fewer than 300 raster units, the resolution was dynamically adjusted to 10 meters. After subdivision, an agricultural overlapping grid dataset was generated, along with a raster index file. The index file records the original agricultural plot number and overlapping area identifier for each raster. In the process of extracting agricultural boundary gradients based on the agricultural overlapping grid, the Sobel operator from the OpenCV library was first used to extract the first derivative of the agricultural overlapping grid. The gradient images in the horizontal (X-axis) and vertical (Y-axis) directions are extracted separately. A 3×3 kernel matrix is ​​used for gradient calculation, with the convolution kernel set to [-101; -202; -101] and its transpose. During the calculation, the gradient value of each grid cell is taken as the square root of the sum of the squares of the horizontal and vertical gradients as the overall gradient strength. Then, a threshold is set based on the gradient strength, and grid cells with gradient values ​​greater than the set threshold of 0.6 are selected as preliminary boundary points to generate an agricultural boundary gradient map. The boundary gradient map is stored in grayscale format, with grayscale values ​​ranging from 0 to 255. A larger grayscale value indicates a stronger gradient. Gradient direction information for each grid cell is also included, expressed in degrees ranging from 0 to 360 degrees. In determining the interface contour of overlapping agricultural areas using the agricultural boundary gradient, the Canny edge detection algorithm is used to binarize the agricultural boundary gradient map, setting a low threshold of 50 and a high threshold of 150. The gradient direction information is used for non-maximum suppression. The process involves a suppression step to refine the boundary, followed by the Hough Transform method to detect linear features in the binary image. The detection parameters are set to a minimum line segment length of 30 meters and a maximum interval of 5 meters. The extracted set of line segments constitutes the preliminary interface contour of the agricultural overlapping area. To ensure the continuity of the interface, the extracted line segments are connected using a connection threshold of 8 meters. If the distance between the endpoints of adjacent line segments is less than the threshold, the two line segments are merged into one long line segment, forming complete interface contour data. This data is stored as a vector polyline, with an attribute table recording the start and end coordinates and length attributes.In the process of morphologically refining the interface contour and constructing the agricultural interface topology network, the interface contour lines are first subjected to morphological erosion. The erosion structure element is set to a 3×3 cross-shaped convolution kernel, and the erosion operation is iterated twice to eliminate isolated short line segments. Then, morphological closing operation is applied, with the structure element set to a 5×5 square convolution kernel, to further fill small breaks in the interface. After refinement, the interface contour is transformed into a topology network using the NetworkX library. Nodes are the endpoints of line segments, and edges are the connected line segments. Each edge is assigned a length attribute, and each node is assigned a degree (i.e., the number of connected edges) attribute. Isolated nodes and corresponding edges with a degree less than 2 are removed, finally forming a complete agricultural interface topology network. The network structure is output in GraphML format, with nodes and edges storing coordinates and attribute information, respectively. In the process of analyzing the connectivity of the agricultural interface topology network to identify the interface connected components, a depth-first search algorithm is first used. A Direct-First Search (DFS) algorithm traverses the topological network to identify all connected subgraphs. Each connected subgraph corresponds to an interface connected region. A minimum connected region area threshold of 2000 square meters is set, calculated using the minimum convex hull formed by the node coordinates. Connected regions smaller than the threshold are discarded, retaining only valid ones. The connected region attribute table records the connected region number, number of contained nodes, number of contained edges, connected region area, and principal direction angle. The principal direction angle is estimated using a least-squares fitting method. Based on the characteristics of each connected region, the dominant agricultural overlap type is labeled. In the process of generating predicted agricultural interface blocks based on interface connected regions, each connected region is first used as an initial seed, and a Markov random field (Markov)-based algorithm is employed. This paper extends the region growing algorithm of the RandomField (MRF) model. During the growth iteration process, agricultural plot crop similarity and meteorological disaster impact data are used as growth potential inputs. The growth rules are set such that region expansion is allowed when the similarity weight is greater than 0.8 and the difference in disaster vulnerability is less than 0.2. The maximum expansion distance for each expansion is no more than 30 meters, and the maximum expansion area is no more than 1.5 times the area of ​​the initial connected domain. The growth ends when there is no effective expansion for three consecutive rounds or the area limit is reached. After growth, predicted agricultural interface block data is generated, and the blocks are output in vector polygon form. Each block is assigned attributes such as the connected domain number from which the growth originated, the expansion area, and the growth round. In the process of determining the development data of composite agricultural boundaries based on the predicted agricultural interface blocks, all predicted agricultural interface blocks are first classified according to the main crop type, overlap intensity level, and growth expansion trend. The classification criteria are that the main crop types are consistent, the difference in overlap intensity level is no more than 1 level, and the difference in expansion trend angle is less than 15 degrees. Then, internal statistics are performed on each group of blocks to calculate the total area, average overlap intensity, and dominant growth direction, generating a composite agricultural boundary development data table.The data table fields include block group number, total area, main crop type, dominant growth direction, and average overlap intensity. The final output of the mixed agricultural boundary development data is presented as a Shapefile and attribute table.

[0057] Preferably, step S3, which involves identifying agriculturally available resources in overlapping agricultural areas using data on the development of composite agriculture, and simulating interactions between different components based on these resources, includes:

[0058] Density kernel estimation was performed on the data of the development of mixed agriculture at the boundary to obtain the density distribution of the boundary area.

[0059] Hotspot analysis was conducted on the boundary area density distribution of composite agriculture boundary development data, and hotspot areas of the boundary were divided based on the hotspot data.

[0060] Based on the hotspot areas at the interface, environmental gradients are superimposed on the overlapping agricultural areas to obtain the agricultural ecological gradient;

[0061] Estimation of species composition data in the boundary zone based on agricultural ecological gradient and regional agricultural component data;

[0062] Identify agriculturally available resources in overlapping agricultural areas by using species composition data from the boundary zone;

[0063] Based on available agricultural resources, agricultural resource competition is mapped in overlapping agricultural areas to generate an agricultural resource competition pattern.

[0064] By simulating the interaction of different agricultural components using agricultural resource competition patterns and regional agricultural component data, interaction data of different agricultural components is generated.

[0065] In this embodiment, during the process of performing density kernel estimation on the composite agricultural boundary development data to obtain the boundary area density distribution, the vector layer of the composite agricultural boundary development data is first loaded. The KernelDensity tool in ArcGIS is used for kernel density estimation. The input parameters are set to the area field of each vector polygon as the weight value, the search radius is set to 300 meters, the kernel function is a quartic kernel, the output raster resolution is set to 10 meters, and normalization is set to True to ensure comparability of density distributions in different areas. After kernel density estimation, a boundary area density distribution raster map is obtained, with raster values ​​in square meters per square kilometer. In the process of performing hotspot analysis on the composite agricultural boundary development data based on the boundary area density distribution and dividing the boundary hotspot areas based on the hotspot data, the Hot Spot tool in ArcGIS is first used. The Analysis tool performs hotspot detection analysis on the boundary area density distribution. The spatial weight matrix is ​​set to a fixed distance weight, the threshold distance is set to 500 meters, and the significance level is selected as 0.05. The analysis results output the Z-Score and P-Value of each grid cell. Then, hotspot regions are divided according to the Z-Score threshold, where the Z-Score is greater than 2.Area 58 (corresponding to a 99% confidence interval) was identified as a strong hotspot region. Vector surface data of these hotspot regions was generated, and each hotspot region was appended with fields for area, average boundary area density, and maximum boundary area density. In the process of obtaining the agricultural ecological gradient by overlaying environmental gradients of overlapping agricultural areas based on these hotspot regions, the following steps were performed: First, the raster data of the overlapping agricultural areas and the vector data of the hotspot regions were loaded. The GDALWarp tool was used for spatial clipping, cropping the raster data of the overlapping agricultural areas to the range of each hotspot region. Then, the environmental factor dataset was loaded, including four factors: annual mean temperature, annual mean precipitation, slope, and soil organic matter content. All environmental factor data were uniformly resampled to 10-meter resolution and spatially referenced to WGS-84. Finally, principal component analysis was performed within the hotspot regions using the environmental factor values ​​of all raster cells within each hotspot region. Component Analysis (PCA) was used to extract the first two principal components as the comprehensive environmental gradient, generating agricultural ecological gradient raster data. The raster values ​​were standardized to the [0,1] interval. In estimating the species composition data of the boundary zone based on the agricultural ecological gradient and regional agricultural component data, the agricultural ecological gradient raster data and the regional agricultural component data were first registered at the pixel level. Nearest Neighbor resampling was performed using the Resample tool in ArcGIS to ensure that the raster size and geographic alignment of the two datasets were completely consistent. Then, for each boundary hotspot area, the composition of each agricultural component within that area was calculated. The area distribution ratio within different agricultural ecological gradient intervals (segmented in steps of 0.1) was used to obtain distribution histograms of each species component under different ecological gradients. The distribution histograms were then used to perform weighted estimation of each component, and finally, a boundary zone species component data table was generated for each hotspot area. The fields included agricultural component ID, ecological gradient interval, area proportion, and average ecological gradient value. In the process of identifying agriculturally available resources in agricultural overlapping areas through boundary zone species component data, resource identification threshold standards were first set, including a species ecological adaptation interval width greater than 0.6, an area proportion greater than 5% of the total area of ​​the hotspot area, and a species ecological stability index (inversely calculated using the ecological gradient standard deviation) less than 0.2. Species components are used as agriculturally available resources. All boundary zone species component data tables are traversed, and agricultural components meeting the above conditions are selected. Identified agriculturally available resources are recorded in vector polygon data form. Each agriculturally available resource unit includes attributes such as species ID, area, dominant ecological gradient value, and ecological stability index, and is output as a GeoPackage format file. Simultaneously, a spatial distribution map of available resources is generated, with the map colored according to agricultural component type. In the process of mapping agricultural resource competition in overlapping agricultural areas based on agriculturally available resources to generate an agricultural resource competition pattern, the spatial distribution of agriculturally available resources is first used as a basis, employing Voronoi diagrams... The Tessellation method is used to divide the spatial region, using the centroid of each agricultural resource unit as the seed point to generate a Voronoi diagram. This diagram is then overlaid onto the raster data of overlapping agricultural areas, with pixel-based assignment. Subsequently, the resource competition pressure value of each agricultural resource unit within its neighboring areas is calculated. This pressure value is calculated as a weighted average of the differences in area ratio between adjacent units and the differences in ecological adaptability of agricultural components, yielding agricultural resource competition pattern data. This pattern data is output as vector polygons, with each polygon accompanied by a competition pressure value field ranging from 0 to 1, where higher values ​​indicate more intense competition. This data is then analyzed using the agricultural resource competition pattern and regional agricultural components... In the process of generating heterogeneous agricultural component interaction data through data simulation, the agricultural resource competition pattern is first divided into units based on the competition pressure value into three categories: low (less than 0.3), medium (0.3 to 0.6), and high (greater than 0.6). Within each unit, a Lotka-Volterra competition model is established based on the regional agricultural component data. The model parameters are set as follows: the initial relative density of each agricultural component; the intraspecific competition coefficient is set to a random distribution between 0.1 and 0.2; and the interspecific competition coefficient is mapped according to the differences in ecological adaptability between components, with a higher interspecific competition coefficient for larger differences, initially set to a range of 0.2 to 0.8. A one-year timescale is simulated using a numerical iteration method (Euler method, step size set to 0.01), recording the density changes of each component within each time step. Finally, the final density data of each component in each competitive unit after one year is output, generating heterogeneous agricultural component interaction data in CSV format. The fields include unit ID, agricultural component ID, and final relative density. Spatially, the spatial distribution of different component densities is displayed using a raster overlay method.

[0066] Preferably, the interaction simulation of heterogeneous agricultural components is conducted using agricultural resource competition patterns and regional agricultural component data to generate heterogeneous agricultural component interaction data, including:

[0067] Based on regional agricultural component data, component hierarchical stripping is performed to generate an agricultural component hierarchical structure;

[0068] Intra-group supply inversion is performed based on the hierarchical structure of agricultural components to obtain component functional indicators;

[0069] Fuzzy clustering and merging of component functional indicators are performed to obtain component functional groups;

[0070] Evolutionary simulations are performed based on component functional groups to generate component evolution trends;

[0071] Inferring the intra-component support effect based on the component evolution trend, and superimposing the intra-component support effects to generate intra-component support gain data;

[0072] By modifying the component evolution trend through the agricultural resource competition pattern and determining the interactive equilibrium point based on the modified evolution trend, the interactive steady-state equilibrium point is obtained.

[0073] The elastic interaction range is defined based on the interactive steady-state equilibrium point;

[0074] The intensity of competition among different components is analyzed by considering the elastic interaction interval and the competition pattern of agricultural resources, and then quantified as the intensity coefficient of competition among different components.

[0075] Interactive response simulations were performed based on intra-component support gain data and heterogeneous component competition intensity coefficients to obtain heterogeneous agricultural component interaction data.

[0076] In this embodiment, regional agricultural component data is structurally encoded using agricultural units as the basic spatial unit. The data source is rasterized agricultural component classification data obtained through multi-source remote sensing interpretation and ground surveys, with a resolution set to 10 meters. Agricultural components are functionally divided into three main categories: primary production (e.g., food crops, vegetables), secondary production (e.g., cash crops, forage grasses), and auxiliary support (e.g., green manure, shade plants). Within each main category, functional subcategories are further subdivided based on crop systems and ecological adaptability. For example, primary production is divided into short-term perennial and perennial subcategories. By constructing an agricultural component attribute tree structure, a multi-level directed graph is built using NetworkX in Python. Each component node is assigned a unique identifier and its hierarchical position in the system. The hierarchical structure data is output and saved in JSON format, which is then converted into an attribute structured table. Each row contains the component ID, its main category, functional subcategory, node level, and parent node. In the process of inverting intra-group supply based on the hierarchical structure of agricultural components to obtain component functional indicators,... First, using a structured table of agricultural component attributes as input, historical output data for each component under different ecological types were extracted. Data sources included the National Agricultural Remote Sensing Monitoring Platform and field observation records from typical sample areas, covering four categories of indicators: crop yield, water requirement per unit area, soil nitrogen uptake per unit area, and labor input per unit area. For each functional indicator, least squares fitting was performed within the same ecological zone (divided into 1 km × 1 km grids) to construct a crop supply capacity model, with output as the dependent variable and resource input as the independent variable. The linear regression module in Scikit-learn was used to fit regression coefficients for each component. After fitting, a functional indicator table was output for each agricultural component, with fields including unit yield, unit water resource requirement, unit nutrient uptake, and unit labor requirement. In the process of fuzzy clustering of the component functional indicators to obtain component functional groups, the functional indicator table was first standardized. Z-score standardization was used to transform the four types of indicator data to a standard distribution based on the mean and standard deviation, respectively. Then, Fuzzy clustering was performed... The C-Means algorithm (fuzzy C-means clustering algorithm) is used for clustering. The number of clusters is set to 5, the fuzziness coefficient is set to 2, the maximum number of iterations is set to 1000, and the stopping error threshold is set to 1e-5. The skfuzzy module in Python is used to perform the clustering task. Each agricultural component is assigned a membership degree to five cluster centers. Then, the final functional group is determined based on the center with the highest membership degree. The clustering results are output with agricultural component ID and group ID as a pair in CSV format. A functional group distribution map is constructed, using colors to identify different group types, and overlaid on the agricultural area distribution map for visualization. In the process of simulating the evolution trend of components based on functional groups, a Markov chain model is used to simulate the evolution state of functional groups over time.First, a time-series raster overlay is performed on the spatial distribution maps of functional groups generated every five years. Each 10m × 10m pixel corresponds to a time-series state vector. Then, a state transition matrix is ​​constructed with a 5×5 dimension. Each element represents the probability of transitioning from functional group i to functional group j, calculated from the statistical transition frequency in historical time-series raster data. Based on the current year's functional group distribution map and the state transition matrix, state projections for future time points are performed, outputting a predicted functional group map for those future time points. Simultaneously, the area change trend of each functional group over time is recorded, plotted with time on the horizontal axis and area on the vertical axis. The output format is a PNG image and a trend data table (CSV). In the process of inferring intra-component support effects based on the component evolution trend and overlaying these effects to generate intra-component support gain data, the growth factor is first calculated based on the rate of area change of functional groups in the evolution trend map. The factor is the change in area divided by the length of the time period. Then, based on the average unit supply index within the functional group, the intra-group output growth caused by the increase in unit area is calculated. This involves multiplying the growth factor by the unit supply index one by one to generate the support gain value for each functional group in each time period. Subsequently, the support gain values ​​belonging to multiple functional groups at the same spatial pixel are superimposed to obtain pixel-level intra-group support gain data. This data is output in 10-meter resolution raster format, with a 32-bit floating-point data type. The raster value represents the support gain per unit area, with the unit being "unit supply per year". In the process of correcting the component evolution trend through resource constraints using the agricultural resource competition pattern and locating the interaction equilibrium point based on the corrected evolution trend to obtain the interaction steady-state equilibrium point, the agricultural resource competition pattern raster data and the component evolution trend map are first spatially resampled and overlaid using ArcGIS Raster. The Calculator tool uses resource competition pressure as a correction coefficient, defining the correction function as f(x) = x*(1-p), where x is the component area growth factor and p is the competition pressure value (between 0 and 1). This function spatially corrects the original growth factor, which is then reapplied to the state transition matrix to adjust the Markov state matrix. The evolution simulation process is then rerun until the state stabilizes and no longer undergoes transition changes. The state of the functional groups at this time point is recorded and defined as the interaction steady-state equilibrium point. The output includes the equilibrium time point and the spatial distribution map of each functional group at the equilibrium point. During the process of defining the elastic interaction interval based on the interaction steady-state equilibrium point, a baseline state is set based on the functional group area value at the interaction steady-state equilibrium point. A fluctuation range of ±10% is set upwards and downwards from the equilibrium area value of each functional group, defining this range as the elastic interaction interval. The RasterMapAlgebra method is then used to numerically transform the equilibrium point distribution map, assigning the area deviation of each pixel's functional group to determine whether it falls within the elastic interaction interval.If a region falls within the defined interaction zone, it is marked as a stable interaction zone; if it exceeds the defined interaction zone, it is marked as a critical interaction zone. A flexible interaction zone mask is generated and stored in raster form, with pixel values ​​of 0 (inelastic zone), 1 (stable elastic zone), and 2 (critical elastic zone). In the process of analyzing the competition intensity of dissimilar components and quantifying it into a dissimilar component competition intensity coefficient through the flexible interaction zone and agricultural resource competition pattern, the boundaries between the stable and critical elastic zones are first extracted based on the flexible interaction zone mask. The GDALPolygonize tool is used to convert the raster into vector boundary lines. Then, the agricultural resource competition pattern map is overlaid. Statistics are performed based on the functional group combinations on both sides of the boundary line, and the difference in resource competition pressure values ​​between the two components within a unit length on the boundary line is calculated, defined as the competition gradient. A larger gradient indicates higher competition intensity. Based on this, the dissimilar component competition intensity coefficient K is constructed. ij The calculation method is K ij =ΔP ij / ΔA ij , where ΔP ij ΔA represents the difference in competitive pressure between components i and j. ij For the difference in the area of ​​the elastic interaction region between the two, K is the K of all component pairs. ij The values ​​are organized into a heterogeneous competition coefficient matrix, and the output is in CSV format, with fields including component i, component j, and competition intensity coefficient K. ij In the process of obtaining interaction data of heterogeneous agricultural components by simulating the interaction response based on the support gain data within the components and the competition intensity coefficient of heterogeneous components, a simulation framework based on the Differential Response Function is adopted. The response function model is constructed in Python, and the model is defined as Ri(t)=Gi(t)-∑Kij*Rj(t), where Ri(t) is the response intensity of component i at time t, Gi(t) is its support gain value, Kij is the heterogeneous competition intensity coefficient, and Rj(t) is the response value of other components at time t. The response values ​​of all components are updated synchronously at each time step through iteration. The time step is set to 0.1, the total iteration time is 5 years, and a total of 50 steps are executed. The trend of the response value of each component at each time step is recorded. Finally, the spatial distribution of the response intensity of each component at the final time point is output as a 10-meter resolution raster map. The data value unit is unit response intensity, the layer format is GeoTIFF, and a response trend map and corresponding CSV data table are attached.

[0077] Preferably, the construction of the agricultural interaction effect field through the interaction data of heterogeneous agricultural components in step S4 includes:

[0078] Multi-component situation mapping is performed on heterogeneous agricultural interaction data to generate agricultural component situation mapping data.

[0079] The agricultural component situation mapping data is converted into heterogeneous component tensors to obtain component tensor converted data.

[0080] Interactive spectral domain translation based on component tensor reduced data is used to generate agricultural component interactive spectra;

[0081] Local interaction weaving processing was performed on the interaction spectrum of agricultural components to obtain local interaction weaving data;

[0082] Based on local interactive weaving data, the interaction effects of different agricultural components were determined, and an agricultural interaction effect field was constructed.

[0083] In this embodiment, when performing multi-component situational mapping on heterogeneous agricultural interaction data, it is necessary to first extract features from the interaction data based on the spatiotemporal distribution characteristics of each component in the heterogeneous agricultural interaction data by constructing a 3D Convolutional Neural Network (3D-CNN) model. The convolution kernel size is set to 3×3×3, the stride is set to 1, and zero-padding is used. The extracted spatial features are then mapped to a unified standard spatial grid system. The grid system size is set to 1 grid point per kilometer, and the time granularity is set to 1 record per hour. During the mapping process, linear interpolation is used to fill in irregular spatiotemporal points, and the maximum error threshold of the filled data is controlled within 1%, thereby generating agricultural component situational mapping data. During the situational mapping process, different identification codes are assigned to different crop types, such as rice being assigned a value of 1, corn being assigned a value of 2, soybeans being assigned a value of 3, and other crops being assigned values ​​incrementally. When performing heterogeneous component tensor reduction on situational mapping data, the Tucker decomposition algorithm from higher-order tensor decomposition methods needs to be used. First, the situational mapping data is constructed into a third-order tensor according to time, space, and component type, with tensor size set to T×X×Y, where T represents the number of time steps, and X and Y represent the number of spatial grid points, respectively. Tucker decomposition is then used to decompose the original tensor into a core tensor and three factor matrices. The core tensor size is set to T'×X'×Y', with a compression ratio of 50%, meaning T', X', and Y' are each half of T, X, and Y, respectively. The factor matrices correspond to low-dimensional representations of time, space, and component features, respectively. Alternating least squares is used during the reduction process. Iterative optimization using Least Squares (ALS) was performed, with a maximum of 500 iterations. The process terminated when the convergence error was less than 10^-4, ultimately yielding component tensor reduced data. For interactive spectral domain translation based on this data, a Fast Fourier Transform (FFT) was used to transform the reduced core tensor in the frequency domain. First, the time dimension was transformed to extract the temporal frequency features of the component interactions. The sampling frequency was set to 1 Hz, and the frequency resolution to 0.01 Hz. Then, a two-dimensional FFT was performed on the spatial dimension, with a spatial sampling interval of 1 km and a frequency range controlled between 0 and 0.At a speed of 5 cycles / km, a spectral energy threshold of 1% was set to remove noise frequency components, ultimately generating an agricultural component interaction spectrum. This spectrum includes the dominant frequency, frequency energy distribution, and their correspondence with component types. All frequency feature data are stored in a matrix, where each row corresponds to a dominant frequency and each column corresponds to a component type. During local weaving of the agricultural component interaction spectrum, it is divided into blocks based on geographical regions, with each block set at a scale of 10km × 10km. For the dominant frequency components within each block, local weaving is performed using a method based on Local Multiscale Analysis (LMA). Wavelet transform is employed during the LMA process. The Transform method is used for decomposition, employing the Daubechies-8 wavelet basis as the mother wavelet. The wavelet decomposition level is set to 3 levels. Local interaction energy features are extracted at each scale, and then these features are synthesized into local interaction-weaved data through a weighted feature stacking method. Weighting coefficients are allocated according to the proportion of interaction energy, with higher proportions assigned higher weights. The weight values ​​are controlled between 0 and 1. All local interaction-weaved data are stored in GeoTIFF raster format, with each raster corresponding to the interaction intensity value after local weaving. When determining the interaction effects of heterogeneous agricultural components based on the local interaction-weaved data, a method based on cluster analysis and Support Vector Regression (SVR) is used for effect extraction. First, density-based spatial clustering of applications with... A noise-based DBSCAN algorithm was used, with the parameter eps set to 0.5 and the minimum sample size set to 5. This algorithm identified high-density and low-density interaction regions. For each cluster, the main interaction feature vectors were extracted. Then, an SVR model was constructed, using the interaction feature vectors as input and the interaction intensity change as output. The radial basis function (RBF) kernel was used, with a penalty factor C set to 1 and a kernel width γ set to 0.1. The SVR model was trained to fit the interaction effects. Finally, the interaction effect data fitted from all clusters were fused to construct a complete agricultural interaction effect field. The interaction effect field was stored in a three-dimensional array, where the first dimension was the time step, and the second and third dimensions were the spatial grid locations. Each location stored the corresponding time step interaction effect value. All data were uniformly projected onto the WGS84 geographic coordinate system.

[0084] Preferably, step S4, which involves inferring meteorological disaster mitigation capacity based on agricultural interaction field and regional agricultural component data, includes:

[0085] The agricultural interaction field is reassigned to its components and divided into component grid cells;

[0086] Interactively connect and interleave data from component grid cells and regional agricultural components to generate interactive interconnected cells;

[0087] The interaction effect gradient of the interaction overlapping unit is determined based on the agricultural interaction effect field;

[0088] Based on the interaction effect gradient, the interaction effect intensity is divided into strong and weak partitions to generate interaction effect intensity blocks.

[0089] The viability strength of each component in the regional agricultural component data is defined based on the interaction effect intensity block.

[0090] The absorption capacity is mapped based on the survival strength of each component, thereby obtaining the meteorological disaster absorption capacity.

[0091] In this embodiment, when the agricultural interaction field is componentized and divided into component grid cells, the interaction effect values ​​of different crop types are first aggregated into their respective independent data layers based on the crop component identification information recorded in the interaction field. During the componentization process, a classification method based on crop numbers is used, with the numbers consistent with the identification codes in the aforementioned agricultural component situation mapping. The componentized data is stored in the form of a three-dimensional matrix for each crop, with the matrix dimensions corresponding to the time step, latitude grid points, and longitude grid points, respectively. After componentization, a fixed grid division strategy is used to divide the data layer of each crop into grid cells, with the grid resolution set to 5 km × 5 km. The data within each grid cell is reduced by spatial weighted averaging, with the weights set according to the reciprocal of the square of the distance from the grid center point, the weight of the farthest point being 0 and the weight of the nearest point being 1. The data of all grid cells are uniformly projected onto the EPSG:4326 coordinate reference system. When performing interactive overlay processing on the component grid cells and regional agricultural component data, a Geographic Information System (GIS) is used. The vector overlay analysis function in a GIS (System.GIS) first spatially resamples the regional agricultural component data into identical 5km × 5km grids. During resampling, the Nearest Neighbor Interpolation method is used to ensure that each component grid cell contains corresponding agricultural component feature records. After resampling, spatial overlay is used to overlay the realigned agricultural interaction effect grid with the regional agricultural component grid. For each spatially overlapping cell, its corresponding crop type identifier, spatial location number, interaction effect value, and crop feature value are extracted to generate an interaction overlap cell. Each interaction overlap cell is structured and stored in JSON format, with fields including cell number, crop type, interaction effect value, agricultural component index value, and spatial index. When determining the interaction effect gradient of the interaction overlap cell based on the agricultural interaction effect field, a five-point finite difference (Five-point Finite) method is used based on the interaction effect value within the interaction overlap cell. The Difference method calculates the spatial interaction effect gradient for each overlapping unit. The gradient calculation formula is to subtract the average effect values ​​of the four surrounding units from the effect value of each unit, and then divide by the average distance between the center point of the unit and the center points of its neighbors. The gradient value is set to the change in interaction effect per unit distance. The gradient data is stored as a floating-point array, with the array index corresponding to the number of the overlapping interaction unit. Units with gradient values ​​less than a preset threshold (set to 0.01) are considered low-gradient regions, and units with gradient values ​​greater than or equal to 0.01 are considered high-gradient regions. The high-gradient regions are further subdivided, and strong and weak partitions are formed according to the interaction effect gradient. When generating interaction effect intensity blocks, a mean-shift clustering method is used, with the bandwidth parameter set to 0.5. During clustering, the interaction effect gradient value and spatial location are used as feature inputs to partition all overlapping interaction units, generating several spatially continuous blocks with similar interaction effect intensity characteristics. Each clustered block is numbered sequentially from low to high based on its gradient mean. The boundaries of the interaction effect intensity blocks are generated using a vectorization module in GIS. Each block records its corresponding gradient mean, area, and a list of included crop types and their numbers. When defining the viability intensity of each component in the regional agricultural component data based on the interaction effect intensity blocks, a weighted index comprehensive scoring method is used to calculate the viability intensity of each component for each interaction effect intensity block. The formula for calculating viability intensity is set as the interaction effect gradient mean multiplied by the agricultural component stability index plus the average crop health value in the overlapping interaction units. The agricultural component stability index is derived from historical agricultural census data, and its specific value is quantified according to crop stress resistance indicators, such as 0.8 for rice, 0.7 for corn, and 0.6 for soybeans. Crop health is taken from the Normalized Difference Vegetation Index (NDI) of each unit. The mean of the index (NDVI) is standardized to values ​​between 0 and 1. The overall score is standardized within the range of 0 to 1. The final output is the survivability strength value of each component within each block. The results are stored in tabular form, with columns including block number, crop type, and survivability strength. Based on the survivability strength of each component, a mitigation capacity mapping is performed. When obtaining the meteorological disaster mitigation capacity, a spatial interpolation method is used to map the survivability strength values ​​to the entire region. The interpolation method uses inverse distance weighting (IDW) interpolation, with a weighting index set to 2 and a maximum interpolation radius set to 15 kilometers. Each interpolation point is weighted and averaged based on the survivability strength of the five nearest blocks to generate a continuous mitigation capacity field. The mitigation capacity field is saved in GeoTIFF raster data format with a spatial resolution of 1 kilometer. The value of each grid point represents the meteorological disaster mitigation capacity strength of the agricultural system at that location, ranging from 0 to 1. All results are uniformly processed using min-max normalization, ultimately forming a complete distribution map of meteorological disaster mitigation capacity.

[0092] Preferably, step S5, which involves composite agricultural disaster prediction based on historical meteorological-agricultural disaster correlation patterns and meteorological disaster mitigation capacity, includes:

[0093] Map agricultural disaster trajectories based on historical meteorological-agricultural disaster correlation data;

[0094] Composite disturbance field is generated by imprinting composite disturbance field based on agricultural disaster trajectory;

[0095] The composite disturbance field and the meteorological disaster absorption capacity are correlated by absorption potential difference, and response simulation is performed based on absorption potential difference data to generate absorption potential difference response data.

[0096] The response superposition simulation is performed on the potential difference response data to obtain the superimposed response data;

[0097] Based on historical meteorological-agricultural disaster correlation patterns and superimposed response data, composite agricultural disaster prediction is performed to generate predicted agricultural meteorological disasters.

[0098] In this embodiment, when mapping agricultural disaster trajectories based on historical meteorological-agricultural disaster correlation pattern data, the spatial and temporal characteristics of different types of meteorological disasters and agricultural disaster records are first extracted based on the dataset of the correspondence between agricultural disasters and meteorological conditions over the past twenty years. Data sources include ground observation data and disaster survey data from the Ministry of Agriculture and Rural Affairs. For each disaster type, such as drought, high temperature, and flood, a Support Vector Regression (SVR) model is used for fitting. The model input is a daily sequence of meteorological variables, and the output is the corresponding curve showing the change in affected area. The trajectory mapping process uses a monthly time step, with a spatial resolution of 10 km × 10 km. The trajectories are recorded in GeoJSON format, with fields including disaster type, occurrence time, starting location, movement path, and expansion radius. Each trajectory is mapped based on the Standardized Precipitation Index (SPI) of the starting point. The NDVI index (SPI) value and the corresponding crop sensitivity period determination results are used to determine the composite disturbance field based on the agricultural disaster trajectory. When generating the composite disturbance field, an impact buffer zone is first generated by extending 15 kilometers on each side of the trajectory centerline, following the agricultural disaster trajectory path. Within the buffer zone, disturbance weights are set based on the decrease in NDVI index and the increase in surface temperature during historical disaster periods. The disturbance weight value is defined as the decrease in NDVI index multiplied by the normalized value of the increase in surface temperature, with the weight range set between 0 and 1. A two-dimensional Gaussian kernel density estimation (Kernel Density) is used. The KDE (Knowledge, Determination, and Estimation) method spatially expands the disturbance weights, setting the bandwidth parameter to 10 km. The generated disturbance field is output in GeoTIFF format with a resolution of 1 km. A larger disturbance field value indicates a higher probability of agricultural system damage. After superimposing the disturbance fields corresponding to all trajectories, the final composite disturbance field is generated according to the principle of merging the maximum disturbance weight values. The composite disturbance field and the meteorological disaster absorption capacity are correlated using absorption potential difference processing, and response simulation is performed based on the absorption potential difference data. When generating absorption potential difference response data, the composite disturbance field and the meteorological disaster absorption capacity field are first grid-registered, with a unified spatial resolution of 1 km and a unified coordinate system of WGS84. For each grid point, the difference between the disturbance value and the absorption capacity value is calculated and defined as the absorption potential difference. A absorption potential difference greater than zero indicates that the disturbance is greater than the absorption capacity, while a absorption potential difference less than zero indicates that the absorption capacity can withstand the disturbance. Based on the absorption potential difference data, a locally adaptive dynamic evolution (Locally Adaptive Dynamic Evolution) method is adopted. The Evolution model is used for response simulation. The model sets the rate of state change at each grid point to be proportional to the potential difference, the simulation step size to 6 hours, and the evolution period to 30 days. The time series of response intensity for each grid point is output, ultimately generating a raster file of the potential difference response data. Response superposition simulation is then performed on this data to obtain the superimposed response data.First, based on the crop distribution map, the potential difference response data is grouped and processed according to crop type. For each crop type, a weighted overlay method is used to generate overlay response data. The weights are set according to the crop biomass sensitivity coefficient, which is obtained from the crop biomass response database published by FAO (Food and Agriculture Organization). For example, the coefficient is set to 0.85 for rice, 0.75 for corn, and 0.65 for soybeans. During the overlay process, the final response value of each grid point is obtained by multiplying the potential difference response value of the corresponding crop type by the biomass sensitivity coefficient and then summing them. The output is a multi-layer raster dataset, with each layer corresponding to one crop type. The spatial resolution is maintained at 1 km, and all grids are stored using 32-bit floating-point data. Based on the historical meteorological-agricultural disaster correlation pattern and the overlay response data, composite agricultural disaster prediction is performed. When generating predictions for agricultural meteorological disasters, a Bayesian inference model is constructed based on the previously extracted historical meteorological-agricultural disaster trajectory pattern and the current overlay response data. The model takes superimposed response values ​​and anomalous indicators from current weather forecast data (including temperature anomalies, precipitation anomalies, and wind speed anomalies) as input variables, and outputs the probability of agricultural disasters occurring within the next 30 days. A Bayesian prior distribution is generated by fitting historical trajectory data. The likelihood function is set such that the probability of a disaster is greater than 0.6 when the superimposed response value is greater than 0.5, and linearly decreases to 0.2 when it is less than 0.5. Finally, the probability of disaster occurrence is calculated for each grid point, and classification is performed based on a threshold of 0.5 to generate a raster map predicting agricultural meteorological disasters. The prediction results are output in GeoTIFF format with an attached probability attribute layer, and the spatial resolution is 1 kilometer.

[0099] Of particular importance is the correlation processing between the composite disturbance field and the meteorological disaster absorption capacity using absorption potential difference, and the response simulation based on the absorption potential difference data, including:

[0100] Extract the disturbance intensity data of the composite disturbance field;

[0101] Mapping the meteorological disaster absorption capacity to the absorption capacity to generate absorption capacity data;

[0102] The disturbance intensity data and absorption capacity data are subjected to potential difference fitting to obtain absorption potential difference data.

[0103] Based on the potential difference data, a disturbance-triggered response simulation is performed to generate potential difference response data.

[0104] In this embodiment, when extracting the disturbance intensity data of the composite disturbance field, the Raster Calculator module in ArcGIS is first used to extract pixel values ​​from the composite disturbance field raster file. For each pixel, its disturbance weight value is read. This weight value represents the intensity of the impact of the composite meteorological disaster on the agricultural system per unit area. The resolution is set to 1 kilometer during extraction, and each pixel value is between 0 and 1. After reading, a preliminary disturbance intensity dataset is formed. Subsequently, the dataset is normalized, mapping the minimum value to 0 and the maximum value to 1. The Min-Max normalization formula is used for calculation, that is, each disturbance value minus the minimum disturbance value and then divided by the difference between the maximum and minimum disturbance values. After normalization, the final disturbance intensity data is generated. The meteorological disaster absorption capacity is mapped, and the absorption capacity data is generated. First, the raster data on meteorological disaster absorption capacity was read into the Python environment and processed using the rasterio library. For each grid point, a specific capacity value was assigned based on the absorption capacity level of the agricultural component. The absorption capacity levels were divided into five levels: Level 1 with a capacity value of 0.9, Level 2 with 0.7, Level 3 with 0.5, Level 4 with 0.3, and Level 5 with 0.1. The capacity value represents the maximum carrying capacity of the agricultural system under a unit disturbance intensity. Each grid point was assigned a capacity value according to its corresponding absorption capacity level, resulting in preliminary absorption capacity data. To ensure spatial consistency, bilinear interpolation was used for the absorption capacity data. The Interpolation method is used for resampling, with a target resolution of 1 km. Potential difference fitting is performed on the disturbance intensity data and the absorption capacity data to obtain the absorption potential difference data. First, based on the NumPy library, corresponding elements of the disturbance intensity data array and the absorption capacity data array are subtracted. The absorption potential difference is defined as the disturbance intensity minus the absorption capacity, calculated as D = SC, where D is the absorption potential difference, S is the disturbance intensity, and C is the absorption capacity. When the absorption potential difference D is greater than 0, it indicates that the disturbance exceeds the absorption capacity; when the absorption potential difference D is less than 0, it indicates that the disturbance exceeds the absorption capacity. A value of 0 indicates that the absorption capacity is sufficient to offset the disturbance. Batch calculations are performed for all grid points. After the potential difference is fitted, the absorption potential difference array is pruned. The maximum value of the absorption potential difference is set to 1, and the minimum value is -1. Values ​​exceeding the range are directly truncated to the boundary value. Based on the absorption potential difference data, disturbance trigger response simulation is performed. When generating absorption potential difference response data, a disturbance trigger threshold is first constructed based on the absorption potential difference data. Grid points with absorption potential differences greater than 0.2 are set as trigger units. The discrete grid point automatic growth algorithm (CellularAutomata Growth Model) is used for simulation. Each trigger unit propagates to its four neighboring grids according to the set rules. The propagation probability is defined as the absorption potential difference value of the current grid point multiplied by the basic propagation coefficient 0.8. The propagation process is set to a maximum of 48 steps, with each step representing 1 hour. Propagation is blocked if the potential difference between adjacent grid points is less than -0.1. During the simulation, the cumulative intensity of the disturbance's impact on each grid point is recorded. The cumulative intensity is defined as the maximum potential difference value propagated to that grid point. After completing all simulation steps, a complete dataset of potential difference response is generated.

[0105] Of particular importance is the simulation of response superposition based on the potential difference response data, which includes:

[0106] The potential difference response data is then processed by regional merging to obtain the regional merged response data.

[0107] The regional merged response data is overlaid with time windows to generate time-series overlaid response data;

[0108] Anomaly amplification simulation is performed on the time-series superimposed response data to obtain anomalously amplified response data;

[0109] Intensity redistribution is performed based on the amplified response data, and data reconstruction is performed based on the redistributed data to generate superimposed response data.

[0110] In this embodiment, when performing regional merging processing on the potential difference response data, the `label` function from the scikit-image library in Python is first used to mark the connected regions of the potential difference response raster data. Eight-neighbor connectivity is defined (i.e., adjacent grid points in the top, bottom, left, right, and four diagonal directions are considered connected). After extracting the connected regions, the average value of the potential difference response values ​​of all grid points within each connected region is calculated. This average value is used as the uniform response intensity value for that region. Simultaneously, the coordinates of the boundary rectangle of each connected region are recorded, including the minimum row number, maximum row number, minimum column number, and maximum column number. After assigning the average value of the connected regions, the regional merged response data is generated and output as a GeoTIFF format file. The spatial reference frame is set to WGS84, and the resolution is 1 kilometer. When performing time window overlay on the regional merged response data, the regional merged response data is first arranged according to a time series. A time window width of 6 hours is selected. Within each time window, all merged response raster data within the corresponding time period are overlaid. The overlay method adopts the maximum value superposition rule, i.e., the time value is taken for the same grid point. The maximum value among all time-series response values ​​within the window is used as the overlay result. During processing, the Python xarray library is used, with a three-dimensional array defined as time, latitude, and longitude. The maximum value is extracted by rolling the window along the time dimension, with a rolling step of 1 hour. After overlay, time-series overlay response data is generated. When simulating anomalous expansion of the time-series overlay response data, outliers are first identified across all grid points. The anomaly standard is defined as a time-series overlay response value exceeding the mean plus twice the standard deviation of all grid points. The mean and standard deviation are calculated based on the response values ​​of all grid points in the entire region. The identified outlier grid points serve as expansion seeds. A numerical simulation method based on the diffusion equation is used to expand the anomalous region. The diffusion radius is set according to the anomaly intensity: 4 km for response values ​​greater than 0.8, 3 km for response values ​​between 0.6 and 0.8, and 2 km for response values ​​between 0.4 and 0.6. Spatial expansion is achieved using the gaussian filter from the SciPy library, with the standard deviation parameter set to the corresponding diffusion radius divided by 1.5. After completing the anomalous expansion, anomalous amplified response data is generated. When redistributing intensity based on this data and reconstructing the data, the anomalous amplified response data is first normalized. A new maximum response value of 1 and a new minimum response value of 0 are set. A linear stretching method is used to adjust the response value distribution of all grid points. The stretching formula is (current value - minimum value) divided by (maximum value - minimum value), then multiplied by the new maximum value to obtain the normalized intensity. Next, local smoothing is performed based on the neighborhood characteristics of each grid point in space. The local smoothing method uses a 3x3 window mean filter. The final value of each grid point is the average of the response values ​​of its nine surrounding grid points. This is implemented using the `uniform_filter` function in SciPy. After smoothing, the reconstructed superimposed response data is obtained.

[0111] Preferably, the construction of a composite agricultural meteorological disaster monitoring index system based on predicted agricultural meteorological disasters in step S5 includes:

[0112] Extracting disaster fragment data from the prediction of agricultural meteorological disasters;

[0113] Disaster trajectory data is generated by reconstructing the time trajectory based on disaster fragment data.

[0114] The disaster trajectory data is divided into impact intensity zones to obtain impact zone data;

[0115] Based on preset monitoring indicators and impact zone data, indicator response matching is performed to generate indicator response data;

[0116] The indicator system is coded based on the indicator response data, and a composite agricultural meteorological disaster monitoring indicator system is constructed.

[0117] In this embodiment, when extracting disaster fragment data from the predicted agricultural meteorological disasters, the spatial regions where the disaster intensity exceeds a specific threshold are first extracted from the superimposed response data of the predicted agricultural meteorological disasters. The intensity threshold is set as the 75th percentile of all grid point values ​​in the superimposed response data, which serves as the criterion for determining the occurrence of a disaster. Using the NumPy library in Python, Boolean filtering operations are used to extract grid points that meet the criteria. The extracted grid point set is aggregated according to geographical location, and clustering is performed using DBSCAN (density-based spatial clustering algorithm) from the scikit-learn library. The minimum number of samples is set to 5, and the maximum neighborhood radius is set to 2 kilometers. Each cluster represents a disaster fragment, and each disaster fragment is assigned a unique identifier. The center point coordinates, spatial range, average disaster intensity, and duration information of the disaster fragment are recorded. The fields include fragment ID, fragment centroid latitude and longitude, fragment average intensity, and fragment duration. Each disaster fragment is sorted in chronological order, with a time step of 1 hour, to track the spatial movement trajectory of the disaster fragments. The trajectory is calculated using Euclidean distance to determine the spatial positional changes of adjacent segments within a continuous time step. If the centroid distance between segments within two consecutive time steps is less than a set threshold of 3 kilometers, they are considered continuous segments of the same trajectory. A connection graph between segments is constructed using Python's NetworkX library, where nodes represent segments and edges represent connections between segments at consecutive time points with sufficient spatial distance. A depth-first search algorithm is used to extract the complete trajectory sequence. Each trajectory includes a start time, an end time, a sequence of segments along the trajectory path, and their corresponding time labels. The trajectory path is rasterized according to spatial range, with a raster resolution of 1 kilometer. The `gdal.RasterizeLayer` function from the GDAL library is used to convert the trajectory lines into raster data. The raster value is taken as the average disaster intensity of the corresponding segments on the trajectory. Intensity levels are classified based on the raster value: an intensity value greater than 0.8 indicates an extremely strong impact area; an intensity value between 0.6 and 0.8 indicates a strong impact area; an intensity value between 0.4 and 0.6 indicates a moderate impact area; and an intensity value less than 0 indicates a severe impact area.Area 4 is designated as the weak impact zone. NumPy in Python is used for conditional checks and assignment to generate impact zone data. The output format is a GeoTIFF file, using WGS84 as the spatial reference system. The unit is a disaster impact level value, ranging from 1 to 4, representing weak, moderate, strong, and extremely strong, respectively. When matching the indicator response based on the preset monitoring indicators and impact zone data, the preset agricultural meteorological disaster monitoring indicator table is first loaded. The table format is a CSV file, containing the indicator name, applicable disaster type, applicable intensity level, and corresponding threshold. The system iterates through the impact zone data cell by cell, matching the corresponding indicator rules based on the impact level of each cell. For example, the indicators matched for extremely strong impact zones are crop lodging rate, soil erosion index, and vegetation cover decline rate. For each matched indicator, the actual response value is calculated from basic meteorological data and crop growth monitoring data according to the response formula set in the indicator table. For example, the crop lodging rate is estimated by the number of times the wind speed is greater than 25 meters per second and the cumulative duration is greater than 2 hours, and the soil erosion index is estimated by the number of days with cumulative precipitation greater than 100 millimeters. Finally, the indicators are generated. The response data is output as vector data with attribute tables. Each record contains a grid ID, impact level, matching indicator name, and indicator response value. When constructing a composite agricultural meteorological disaster monitoring indicator system based on the indicator response data, a unique code is first assigned to each indicator response data item. The coding rule is set as a four-segment coding system: the first segment is the disaster type number (two digits), the second segment is the impact level number (one digit), the third segment is the indicator category number (three digits), and the fourth segment is the sequence number (three digits). For example, "01-3-005-002" represents the second data item of the fifth indicator corresponding to the third level of impact under the first type of disaster. A Python script is used to automatically generate the code and assign it to the indicator response data table. Subsequently, a tree structure of the indicator system is constructed based on all indicator response data. The root node is the disaster type, the secondary nodes are the impact levels, and the child nodes are the specific indicators. This is organized using a DataFrame structure from the pandas library, outputting a complete composite agricultural meteorological disaster monitoring indicator system with a clear structure indicating the hierarchical relationships between each level. .

[0118] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0119] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for constructing a composite agricultural meteorological disaster monitoring index system, characterized in that, Includes the following steps: Step S1: Obtain regional agricultural component data; Spatial distribution analysis of regional agricultural component data was performed, and agricultural distribution boundaries were identified; Agricultural overlap areas are delineated based on agricultural distribution boundaries and regional agricultural component data. The process includes: Step S11: Obtaining regional agricultural component data; performing plot-level slicing on the regional agricultural component data to obtain sliced ​​agricultural components, wherein the area of ​​each plot slice unit is limited to 1-5 hectares; Step S12: Perform spatial density mapping on the sliced ​​agricultural components to generate agricultural spatial density distribution data, with the spatial density resolution set to a 100m × 100m grid. Step S13: Track the edge contours of agricultural spatial density distribution data to obtain the agricultural distribution boundary, wherein the minimum connected area threshold for contour extraction is set to 0.5 square kilometers; Step S14: Cross-over data of agricultural distribution boundaries and sliced ​​agricultural components are performed to obtain overlapping agricultural areas; Step S2: Collect a collection of historical meteorological disasters; discretize the historical meteorological disaster collection into a spatiotemporal sequence to generate a spatiotemporal sequence of meteorological disasters; identify the correlation pattern between historical meteorological and agricultural disasters based on the spatiotemporal sequence of meteorological disasters; Step S3: Predict the development data of the complex agricultural boundary based on the agricultural overlap area, including: The overlapping agricultural areas are subdivided into raster layers to generate an agricultural overlapping grid. Extracting agricultural boundary gradients based on overlapping agricultural grids; The boundary profile of overlapping agricultural regions is determined by agricultural boundary gradient; Morphological refinement of the interface contours was performed, and an agricultural interface topology network was constructed. Analyze the connectivity of the agricultural interface topology network to identify the interface connected components; Regional growth prediction is performed based on interface connectivity to generate predicted agricultural interface blocks. Based on the predicted agricultural interface blocks, composite agricultural boundary development data is determined; through the composite agricultural boundary development data, agriculturally available resources in the overlapping agricultural areas are identified, and based on the agriculturally available resources, heterogeneous component interaction simulations are performed to generate heterogeneous agricultural component interaction data, which includes: Based on regional agricultural component data, component hierarchical stripping is performed to generate an agricultural component hierarchical structure; Intra-group supply inversion is performed based on the hierarchical structure of agricultural components to obtain component functional indicators; Fuzzy clustering and merging of component functional indicators are performed to obtain component functional groups; Evolutionary simulations are performed based on component functional groups to generate component evolution trends; Inferring the intra-component support effect based on the component evolution trend, and superimposing the intra-component support effects to generate intra-component support gain data; By modifying the component evolution trend through the agricultural resource competition pattern and determining the interactive equilibrium point based on the modified evolution trend, the interactive steady-state equilibrium point is obtained. The elastic interaction range is defined based on the interactive steady-state equilibrium point; The intensity of competition among different components is analyzed by considering the elastic interaction interval and the competition pattern of agricultural resources, and then quantified as the intensity coefficient of competition among different components. Interactive response simulations were performed based on intra-component support gain data and heterogeneous component competition intensity coefficients to obtain heterogeneous agricultural component interaction data. Step S4: Construct an agricultural interaction effect field using interaction data of heterogeneous agricultural components, which includes: Multi-component situation mapping is performed on heterogeneous agricultural interaction data to generate agricultural component situation mapping data. The agricultural component situation mapping data is converted into heterogeneous component tensors to obtain component tensor converted data. Interactive spectral domain translation based on component tensor reduced data is used to generate agricultural component interactive spectra; Local interaction weaving processing was performed on the interaction spectrum of agricultural components to obtain local interaction weaving data; Based on local interactive weaving data, the interaction effects of different agricultural components are determined, and an agricultural interaction effect field is constructed; based on the agricultural interaction effect field and regional agricultural component data, the meteorological disaster absorption capacity is inferred. Step S5: Based on the historical meteorological-agricultural disaster correlation model and the meteorological disaster absorption capacity, conduct composite agricultural disaster prediction and generate predicted agricultural meteorological disasters; construct a composite agricultural meteorological disaster monitoring index system based on the predicted agricultural meteorological disasters.

2. The method for constructing a composite agricultural meteorological disaster monitoring index system according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Collect a collection of historical meteorological disasters; perform time-series subdivision on the collection of historical meteorological disasters to obtain the disaster occurrence time series, wherein the time subdivision step size is set to 7 days; Step S22: Perform spatial raster coding on the historical meteorological disaster collection to generate disaster spatial distribution data, wherein the side length of the raster unit is set to 500 meters; Step S23: Perform joint interpolation processing on the disaster occurrence time series and disaster spatial distribution data to generate meteorological disaster spatiotemporal series data, wherein the interpolation time window width is limited to ±3 days and the spatial interpolation radius is set to 1 kilometer; Step S24: Map the spatiotemporal sequence of meteorological disasters to agricultural components based on regional agricultural component data to obtain meteorological-agricultural disaster matching data, wherein the matching tolerance time difference is set to ≤5 days; Step S25: Extract interactive correlation patterns from meteorological-agricultural disaster matching data to generate historical meteorological-agricultural disaster correlation patterns.

3. The method for constructing a composite agricultural meteorological disaster monitoring index system according to claim 1, characterized in that, Step S3, which involves identifying agriculturally available resources in overlapping agricultural areas using data on the development of integrated agriculture, and simulating interactions between different components based on these resources, includes: Density kernel estimation was performed on the data of the development of mixed agriculture at the boundary to obtain the density distribution of the boundary area. Hotspot analysis was conducted on the boundary area density distribution of composite agriculture boundary development data, and hotspot areas of the boundary were divided based on the hotspot data. Based on the hotspot areas at the interface, environmental gradients are superimposed on the overlapping agricultural areas to obtain the agricultural ecological gradient; Estimation of species composition data in the boundary zone based on agricultural ecological gradient and regional agricultural component data; Identify agriculturally available resources in overlapping agricultural areas by using species composition data from the boundary zone; Based on available agricultural resources, agricultural resource competition is mapped in overlapping agricultural areas to generate an agricultural resource competition pattern. By simulating the interaction of different agricultural components using agricultural resource competition patterns and regional agricultural component data, interaction data of different agricultural components is generated.

4. The method for constructing a composite agricultural meteorological disaster monitoring index system according to claim 1, characterized in that, Step S4, which infers the meteorological disaster mitigation capacity based on agricultural interaction field and regional agricultural component data, includes: The agricultural interaction field is reassigned to its components and divided into component grid cells; Interactively connect and interleave data from component grid cells and regional agricultural components to generate interactive interconnected cells; The interaction effect gradient of the interaction overlapping unit is determined based on the agricultural interaction effect field; Based on the interaction effect gradient, the interaction effect intensity is divided into strong and weak partitions to generate interaction effect intensity blocks. The viability strength of each component in the regional agricultural component data is defined based on the interaction effect intensity block. The absorption capacity is mapped based on the survival strength of each component, thereby obtaining the meteorological disaster absorption capacity.

5. The method for constructing a composite agricultural meteorological disaster monitoring index system according to claim 1, characterized in that, Step S5, which involves composite agricultural disaster prediction based on historical meteorological-agricultural disaster correlation patterns and meteorological disaster absorption capacity, includes: Map agricultural disaster trajectories based on historical meteorological-agricultural disaster correlation data; Composite disturbance field is generated by imprinting composite disturbance field based on agricultural disaster trajectory; The composite disturbance field and the meteorological disaster absorption capacity are correlated by absorption potential difference, and response simulation is performed based on absorption potential difference data to generate absorption potential difference response data. The response superposition simulation is performed on the potential difference response data to obtain the superimposed response data; Based on historical meteorological-agricultural disaster correlation patterns and superimposed response data, composite agricultural disaster prediction is performed to generate predicted agricultural meteorological disasters.

6. The method for constructing a composite agricultural meteorological disaster monitoring index system according to claim 1, characterized in that, Step S5, which involves constructing a composite agricultural meteorological disaster monitoring indicator system based on predicted agricultural meteorological disasters, includes: Extracting disaster fragment data from the prediction of agricultural meteorological disasters; Disaster trajectory data is generated by reconstructing the time trajectory based on disaster fragment data. The disaster trajectory data is divided into impact intensity zones to obtain impact zone data; Based on preset monitoring indicators and impact zone data, indicator response matching is performed to generate indicator response data; The indicator system is coded based on the indicator response data, and a composite agricultural meteorological disaster monitoring indicator system is constructed.

Images