Data optimization and intelligent decision support system in cultivated land resource early warning

Abstract

The invention relates to the technical field of cultivated land resource management, and discloses a data optimization and intelligent decision support system in cultivated land resource early warning, which comprises a cultivated land data acquisition module, a soil index analysis module, a degradation identification module, a land force deduction module, a threshold adaptation module and a decision generation module. The cultivated land data acquisition module acquires cultivated land multi-dimensional attribute data in real time and partitions the data; the soil index analysis module divides soil fertility units and fuses soil indexes to generate resource state vectors; the degradation identification module extracts a core degradation factor, establishes a risk layering model and obtains a data correction coefficient; the land force deduction module identifies the association rule and the dynamic correction factor and calculates the land force attenuation gradient; the threshold adaptation module derives an optimal early warning threshold and generates a risk difference sequence; and the decision generation module analyzes the sequence and integrates the sequence into a regulation and control decision scheme. According to the system, intelligent analysis of cultivated land data, dynamic early warning of degradation risks and accurate decision making are realized, and scientificity and efficiency of cultivated land resource management are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cultivated land resource management, specifically a data optimization and intelligent decision-making support system for cultivated land resource early warning. Background Art

[0002] In the prior art, the cultivated land resource early warning system generally has the defect of insufficient data processing capacity. Traditional data collection methods are often limited to the monitoring of single-dimensional indicators. For example, only soil fertility or soil moisture data are obtained, making it difficult to comprehensively reflect the multi-dimensional attribute characteristics of cultivated land plots. This one-sidedness of data collection leads to a lack of systematicness in subsequent soil fertility evaluation, and it is impossible to accurately identify the soil fertility differences and degradation risks within different plot partitions. At the same time, for the analysis of soil indicators, simple threshold judgment or empirical classification methods are mostly used, without fully considering the complex correlations between soil characteristics, resulting in inaccurate division of soil fertility units and unable to effectively capture the dynamic evolution rules of soil indicators.

[0003] In terms of degradation identification and risk assessment, most existing systems rely on fixed risk stratification models, lacking a dynamic tracking and data correction mechanism for core degradation factors. When the state of cultivated land resources changes, traditional models are difficult to adjust the risk assessment results in real time, resulting in lagging early warning information and unable to timely reflect the actual degradation status of cultivated land. In addition, the evolution correlation rules between data correction coefficients and degradation factors are often ignored during the soil fertility deduction process, making it difficult to scientifically predict the soil fertility attenuation trend under different correction strategies, and the formulation of decision-making plans lacks pertinence and foresight.

[0004] The setting of early warning thresholds is one of the core links of the cultivated land resource early warning system. However, in the prior art, the determination of thresholds is mostly based on historical experience or fixed standards, without fully combining the spatial distribution characteristics and dynamic change rules of the current soil fertility attenuation gradient. This static threshold setting method leads to a deviation between the early warning results and the actual risk situation, which may cause misjudgment or missed judgment, affecting the effective implementation of cultivated land protection policies. At the same time, in the decision-making generation link, traditional systems lack the ability to deeply analyze the risk difference sequence and are difficult to integrate scattered risk information into a systematic regulation decision-making plan, resulting in a lack of scientificity and coordination in the decision-making process.

[0005] With the rapid development of information technology, the application of technologies such as artificial intelligence and big data analysis in the agricultural field has brought new opportunities for cultivated land resource management. By constructing an intelligent early warning system for cultivated land resources and realizing the full-process optimization of data collection, analysis, and decision-making, it has become an inevitable trend to improve the efficiency of cultivated land protection. This patent aims to address the deficiencies in the existing technology and provide a cultivated land resource early warning system based on data optimization and intelligent decision-making. Through breakthroughs in key technologies such as multi-dimensional data collection, feature fusion analysis, dynamic risk modeling, and intelligent deduction prediction, it solves the bottleneck problems of traditional systems in data processing, risk identification, decision support, etc., and provides a scientific basis and technical support for the refined management and sustainable utilization of cultivated land resources.

[0006] The patented technology obtains the multi-dimensional attribute data of cultivated land plots in real time through the cultivated land data collection module and divides the plots into zones based on the early warning level, laying a foundation for subsequent precise analysis. The soil index analysis module realizes the feature fusion of soil indicators and the precise division of soil fertility units by constructing a soil feature library and soil fertility clustering groups, improving the scientificity and accuracy of soil fertility evaluation. The degradation identification module realizes the real-time monitoring and dynamic assessment of the cultivated land degradation risk by extracting the core degradation factors and dynamically updating the risk stratification model. The soil fertility deduction module scientifically predicts the soil fertility attenuation gradient under different correction strategies based on the evolution correlation rules and plot weight matrix, providing data support for the formulation of decision-making plans. The threshold adaptation module improves the adaptability and accuracy of the early warning system by calculating the dynamic reference value and matching the optimal early warning threshold. The decision generation module generates a systematic and targeted cultivated land regulation decision-making plan through in-depth analysis and conflict reconciliation of the risk difference sequence, realizing the full-process intelligent management from data collection to decision execution. Summary of the Invention

[0007] The purpose of the present invention is to provide a data optimization and intelligent decision-making support system for cultivated land resource early warning to solve the problems raised in the above background technology.

[0008] To achieve the above purpose, the present invention provides the following technical solutions: A data optimization and intelligent decision-making support system for cultivated land resource early warning, the system includes: A cultivated land data collection module, used to obtain the multi-dimensional attribute data of cultivated land plots in real time and set the plot zoning rules corresponding to the early warning level; A soil index analysis module, used to divide multiple soil fertility units within the plot zone, perform feature fusion processing on the soil indicators of each soil fertility unit, and generate a resource status vector corresponding to the soil fertility unit; A degradation identification module, used to extract the core degradation factors from the resource status vector, establish a risk stratification model associated with the soil fertility unit, and obtain the data correction coefficient corresponding to the model; The soil fertility deduction module is used to identify the evolution correlation rules in the data correction coefficient, dynamically correct the core degradation factors according to the correlation rules, and calculate the soil fertility attenuation gradient of each soil fertility unit under different correction strategies; The threshold adaptation module is used to deduce the optimal warning threshold based on the soil fertility attenuation gradient, and generate a risk difference sequence by comparing the current resource index value with the optimal warning threshold; The decision-making generation module is used to analyze the risk difference sequence, and integrate the risk difference sequence into a cultivated land regulation decision-making scheme based on the soil fertility attenuation trend of the soil fertility unit.

[0009] Preferably, the implementation method of the soil index analysis module includes: constructing a soil characteristic library corresponding to the soil fertility unit, and the soil characteristic library contains soil indexes and a resource parameter set corresponding to the characteristic fusion; Perform similar feature matching on the resource parameter set, and divide the resource parameter set into soil fertility clustering groups according to the matching results; extract the feature focus domain of the soil index from the soil fertility clustering groups, and set the focus domain as the soil fertility unit.

[0010] Preferably, dividing the soil fertility clustering groups of the resource parameter set further includes: Extract the index distribution, anomaly polarity and resource density parameters according to the plot attributes and degradation intensity in the resource parameter set, and generate a soil characteristic identifier based on the above parameters; Associate the soil characteristic identifier with the resource parameter set, and screen the parameter sets with similarity higher than the preset warning threshold through calculating the index similarity between the characteristic identifiers to form a soil fertility clustering group.

[0011] Preferably, the implementation method of generating the resource state vector corresponding to the soil fertility unit includes: For each soil fertility unit, obtain the index fluctuation data of the unit within a preset period according to the spatial position of the unit in the plot partition, and calculate the index fluctuation coefficient of the unit; When the index fluctuation coefficient exceeds the first warning threshold, mark the unit as a high-risk unit, and extract its soil indexes to form a resource state vector; when the index fluctuation coefficient is lower than the first warning threshold, mark the unit as a stable unit, and perform data aggregation on the soil indexes of the adjacent units of the unit, and reconstruct the aggregated data into a resource state vector.

[0012] Preferably, the implementation method of the degradation identification module includes: Separate the fertility anomaly parameter, soil moisture anomaly parameter and texture variation parameter from the resource state vector, and generate a risk stratification model of the soil fertility unit based on the fertility anomaly parameter, soil moisture anomaly parameter and texture variation parameter; If the number of soil fertility units covered by the current risk stratification model is less than the preset warning threshold, traverse the resource status vectors of adjacent soil fertility units, and add the index factors not included in the stratification model of adjacent units to the current model.

[0013] Preferably, the implementation method of the soil fertility deduction module includes: obtaining the time series factor of the index change frequency and the attenuation factor of the soil fertility jump amplitude in the evolution association rule; Construct a plot weight matrix associated with the time series factor and the attenuation factor, and determine the soil fertility attenuation gradient under different correction strategies according to the distribution probability of each element in the matrix.

[0014] Preferably, constructing the plot weight matrix further includes: Identify the evolution cycle characteristics of the time series factor. If the current cycle characteristics completely match the preset index cycle, set the time series factor as the starting node of the plot weight matrix; Calculate the evolution matching degree between the time series factor and the attenuation factor, and generate the intermediate node and the termination node of the plot weight matrix in descending order of the matching degree; Perform path backtracking on the termination node. When the matching degree of the termination node is lower than the preset matching threshold, output it as the final distribution of the plot weight matrix.

[0015] Preferably, the implementation method of calculating the soil fertility attenuation gradient includes: Statistically analyze the mean value of the time series factor and the range of the attenuation factor of each termination node in the plot weight matrix, and calculate the global variance of all node factors; Subtract the mean value of the time series factor of a single termination node from the mean value of the time series factor of the adjacent node, and divide by the global variance to obtain the time series attenuation coefficient; at the same time, calculate the ratio of the range of the attenuation factor to the global variance, and take the weighted sum of the two as the soil fertility attenuation gradient of this node.

[0016] Preferably, the implementation method of deriving the optimal warning threshold includes: Extract the degradation mode in the historical data that is closest to the current soil fertility attenuation gradient, and calculate the cosine similarity of their spatial distributions as the first dynamic reference value; Statistically analyze the difference in the number of abnormal peaks between the current soil fertility attenuation gradient and the historical degradation mode, and take the difference number as the second dynamic reference value; Based on the linear combination of the first dynamic reference value and the second dynamic reference value, match the optimal warning threshold in the preset warning threshold table.

[0017] Preferably, the implementation method of the decision-making generation module includes: dividing the positive attenuation interval and the negative attenuation interval according to the soil fertility attenuation direction of each unit in the risk difference sequence; Extract the convergence rate of the risk difference in the positive decay interval and the diffusion rate of the risk difference in the negative decay interval, and reconcile the two according to the spatial weights of the soil fertility units to generate the adjustment parameters for the cultivated land regulation decision-making scheme.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: At the level of data collection and analysis, the cultivated land data collection module realizes the real-time acquisition of multi-dimensional attribute data of cultivated land plots, and lays a foundation for subsequent refined analysis by setting plot zoning rules corresponding to the warning levels. The soil index analysis module transforms complex soil indexes into spatially correlated soil fertility units by constructing a soil feature library, performing similar feature matching, and dividing soil fertility clustering groups. It not only realizes the feature fusion processing of soil indexes, but also improves the data concentration and analysis efficiency by extracting the feature focus area. The processing method of the resource parameter set by this module fully considers multi-dimensional parameters such as plot attributes and degradation intensity. The generated soil feature identifiers and soil fertility clustering groups can accurately reflect the spatial distribution law of soil indexes, providing a reliable basis for generating the resource state vector of soil fertility units.

[0019] In terms of degradation identification and soil fertility deduction, the degradation identification module separates key abnormal parameters from the resource state vector and generates a risk stratification model. At the same time, by dynamically expanding the model coverage, it ensures the comprehensive identification of cultivated land degradation risks. The soil fertility deduction module realizes the dynamic correction of core degradation factors and the quantitative prediction of soil fertility decay trends by mining the time series factors and decay factors in the evolution correlation rules, constructing a plot weight matrix, and calculating the soil fertility decay gradient. This analysis method based on spatio-temporal characteristics and data correlation can accurately capture the dynamic evolution law of cultivated land degradation, provide a scientific basis for the effect evaluation of different correction strategies, enable the system to predict the soil fertility change trend in advance, and provide sufficient response time for early warning and decision-making.

[0020] In the link of warning threshold adaptation and decision-making generation, the threshold adaptation module matches the optimal warning threshold based on the dynamic reference value by combining the historical degradation mode and the current soil fertility decay gradient, breaking the limitation of the traditional fixed threshold and making the warning result more in line with the actual risk situation. The decision-making generation module generates a cultivated land regulation decision-making scheme with stronger spatial pertinence and dynamic adaptability by dividing the decay interval, analyzing the convergence and diffusion rates of risk differences, and reconciling conflicts by combining the spatial weights of soil fertility units. This full-process intelligent processing from data collection to decision-making generation realizes the transformation of cultivated land resource management from "experience-driven" to "data-driven", and significantly improves the scientificity and accuracy of decision-making. Description of the Drawings

[0021] Figure 1This is the working principle diagram of the data optimization and intelligent decision-making support system for cultivated land resource early warning described in the present invention; Figure 2 This is the flow chart for generating the resource status vector; Figure 3 This is the design diagram for dividing the soil fertility clustering groups; Figure 4 This is the working principle diagram of the degradation identification module. Specific implementation manners

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0023] Please refer to Figures 1-4 , the data optimization and intelligent decision-making support system for cultivated land resource early warning involved in the present invention, and this system includes: Cultivated land data acquisition module: Through a sensor network deployed on cultivated land plots (such as soil moisture sensors, fertility detectors, meteorological monitoring devices, etc.), multi-dimensional attribute data of cultivated land plots are obtained in real time, including soil fertility indicators (such as nitrogen, phosphorus, potassium contents), soil moisture indicators (water content, temperature), plot spatial attributes (geographical location, area), utilization types (paddy fields, dry lands), etc. At the same time, plot zoning rules are set according to the early warning level. For example, the early warning level is divided into three levels: high, medium, and low, corresponding to dividing the cultivated land into a red early warning area, a yellow early warning area, and a green safety area. The zoning rules are established based on parameters such as historical degradation data, soil types, and tillage intensity.

[0024] Soil index analysis module: Within the plot zoning, according to the spatial heterogeneity of soil characteristics, each zoning is divided into multiple soil fertility units. The division of soil fertility units is based on the similarity of soil indicators. For example, areas with similar soil attributes are divided into the same unit through a clustering algorithm for refined analysis. Feature fusion processing is performed on the soil indicators (such as fertility, soil moisture, texture) of each soil fertility unit, and the fusion methods include principal component analysis (PCA), neural network feature extraction, etc., to generate a vector that can comprehensively reflect the resource status of the soil fertility unit. This vector contains the feature weights and coupling relationships of each indicator.

[0025] Degradation Identification Module: Extract core degradation factors from the resource status vector. For example, factors highly correlated with cultivated land degradation are screened through correlation analysis (such as the decrease in organic matter content, the increase in soil salinization index, etc.). Establish a risk stratification model associated with soil fertility units. The model adopts a hierarchical structure and divides soil fertility units into different risk levels according to the severity of degradation factors (such as high-risk degradation, moderate degradation, potential degradation). At the same time, calibrate the model through historical data to obtain data correction coefficients for adjusting the accuracy of real-time monitoring data.

[0026] Soil Fertility Deduction Module: Analyze the evolution association rules in the data correction coefficients. For example, mine the laws of degradation factor changes over time through time series analysis (such as seasonal fertility fluctuations, long-term soil moisture trends). Dynamically correct the core degradation factors according to the association rules. Use time series prediction models (such as ARIMA, LSTM) to simulate the soil fertility attenuation process under different correction strategies, and calculate the soil fertility attenuation gradients of each soil fertility unit under different strategies (such as fertilization adjustment, irrigation optimization). The gradient value reflects the speed and trend of soil fertility degradation.

[0027] Threshold Adaptation Module: Deduce the optimal early warning threshold based on the soil fertility attenuation gradient. By comparing the historical degradation pattern with the current data, adopt dynamic reference value calculation methods (such as cosine similarity analysis of spatial distribution differences, abnormal peak quantity statistics) to match the preset threshold table to determine the optimal threshold in the current scenario. Compare the current resource index value with the optimal threshold to generate a risk difference sequence, which includes the degree of index deviation and risk level of each unit.

[0028] Decision Generation Module: Analyze the risk difference sequence to identify the soil fertility attenuation trend (such as positive attenuation indicates aggravated degradation, negative attenuation indicates a recovery trend). Based on the spatial position and attenuation direction of the soil fertility unit, integrate the sequence into a cultivated land regulation decision plan, which includes zoning governance measures (such as fallow in high-risk areas, crop rotation in medium-risk areas), parameter adjustment suggestions (such as adjusted fertilization amount, irrigation frequency), etc.

[0029] The present invention will be further described below in conjunction with Embodiments 1 to 5: Embodiment 1: The specific implementation of the soil index analysis module includes: Construct a soil feature library corresponding to the soil force unit. The soil feature library is the core carrier for storing soil-related data, and its content covers a set of resource parameters after the fusion of various soil indicators and characteristics. Soil indicators include, but are not limited to, soil fertility indicators (such as total nitrogen, total phosphorus, total potassium content, organic matter content), soil moisture indicators (soil water content, temperature, pH value), physical property indicators (soil bulk density, porosity, texture composition), and trace element indicators (such as iron, manganese, zinc, copper content), etc. These indicators are obtained through real-time monitoring by sensors, laboratory tests, etc., to ensure the accuracy and timeliness of the data. The resource parameter set is a data set formed after statistical analysis and feature extraction of the original soil indicators. For example, for each sampling period or spatial region, statistical quantities such as the mean, variance, maximum value, minimum value, and coefficient of variation of each soil indicator are calculated, and at the same time, methods such as principal component analysis and factor analysis are used to extract the principal component factors or common factors that can represent the comprehensive characteristics of the soil, forming a parameter set containing feature weights and coupling relationships. The construction of the soil feature library needs to follow the principle of data standardization, and normalize indicators with different dimensions and types for subsequent feature matching and clustering analysis.

[0030] Perform similar feature matching on the resource parameter set to divide the soil force clustering groups. The core of similar feature matching is to measure the similarity degree between resource parameter sets through a suitable algorithm, and then classify the parameter sets with higher similarity into the same category. In specific implementation, algorithms such as the fuzzy C-means clustering algorithm, K-means clustering algorithm, or DBSCAN density clustering algorithm can be used. Taking the fuzzy C-means clustering algorithm as an example, this algorithm allows a data point to belong to multiple clusters, and through iterative optimization of the objective function (such as minimizing the sum of squared within-class distances), divides the resource parameter set into several soil force clustering groups. During the clustering process, first set the number of clusters (i.e., the number of soil force clustering groups), and this number can be preset according to factors such as the size of the plot partition and the degree of soil heterogeneity, or automatically determined through indicators such as the silhouette coefficient. Then, calculate the distance between each resource parameter set and each cluster center (such as Euclidean distance, Manhattan distance, or cosine similarity), and determine the membership degree of the parameter set belonging to each cluster according to the distance size. By continuously updating the cluster center and membership degree matrix until the algorithm converges. The clustering result makes the resource parameter sets within the same soil force clustering group have higher similarity, indicating that the soil characteristics of the cultivated land spatial area corresponding to this group are relatively uniform, while there are significant differences between the resource parameter sets of different clustering groups, reflecting the spatial heterogeneity of the soil characteristics.

[0031] When dividing the soil fertility clustering groups, it is also necessary to consider the plot attributes and degradation intensity in the resource parameter set to further improve the accuracy and practicality of clustering. Plot attributes include the spatial location of the plot (latitude and longitude coordinates), topographical and geomorphological features (such as slope, elevation, slope aspect), irrigation conditions (irrigation type, irrigation frequency), tillage methods (such as plowing depth, fertilization habits), etc. These attributes directly affect the formation and evolution of soil characteristics. The degradation intensity is measured by parameters such as crop yield changes and soil degradation indicators (such as soil erosion modulus, salinization degree), reflecting the degree of decline in cultivated land quality. In specific operations, first extract the parameters related to plot attributes and degradation intensity from the resource parameter set, such as calculating the correlation coefficient between the average slope, irrigation frequency and crop yield of each plot, or counting the proportion of soil salinization indicators exceeding the critical value. Then, based on these parameters, generate index distribution, anomaly polarity and resource density parameters. The index distribution parameter is used to describe the distribution characteristics of soil indicators in space or time. For example, calculate the Moran index through spatial autocorrelation analysis to measure the spatial aggregation of soil fertility indicators; the anomaly polarity parameter is used to judge the direction and degree of deviation of soil indicators from the normal range. For example, if the soil moisture content of a plot is lower than the lower limit of the suitable interval, it is defined as a negative anomaly, and the degree of deviation is the difference between the actual value and the lower limit value; the resource density parameter represents the quantity of effective soil indicator data per unit area or unit time, reflecting the richness and reliability of the data.

[0032] Based on the above parameters, generate soil characteristic identifiers. The soil characteristic identifier is a multi-dimensional feature vector that contains information such as plot attributes, degradation intensity, index distribution, anomaly polarity and resource density, and is used to comprehensively characterize the comprehensive characteristics of cultivated land soil. For example, a soil characteristic identifier may contain the following elements: [slope 15°, irrigation frequency 2 times a week, crop yield decline 10%, Moran index of fertility index 0.6, negative anomaly of water content -5%, data density 80%]. Associate the soil characteristic identifier with the resource parameter set, that is, each resource parameter set corresponds to a unique soil characteristic identifier, so as to facilitate subsequent calculation of feature similarity. By calculating the index similarity between different soil characteristic identifiers, select the parameter sets with similarity higher than the preset warning threshold to form soil fertility clustering groups. The calculation methods of index similarity can adopt Euclidean distance, Mahalanobis distance or Jaccard coefficient, etc. The preset warning threshold is set according to the actual application scenario and data characteristics. For example, when using the Jaccard coefficient to measure the feature overlap degree, the threshold can be set to 0.7, that is, when the proportion of the same features in two soil characteristic identifiers exceeds 70%, it is considered that the corresponding resource parameter sets belong to the same soil fertility clustering group.

[0033] After the formation of the soil fertility clustering groups, it is necessary to extract the characteristic focus areas of soil indicators from each clustering group and set the focus areas as soil fertility units. The characteristic focus area refers to the area within the clustering group where the soil indicators show high consistency or significant characteristics, and this area can represent the core soil characteristics of the clustering group. The methods for extracting the characteristic focus areas include statistical analysis, spatial interpolation, etc. For example, for a soil fertility clustering group, by calculating the mean and standard deviation of each soil indicator, the core interval of the indicator (such as the area within the range of mean ± 1 times the standard deviation) is determined, and this interval is the characteristic focus area. The spatial area corresponding to the characteristic focus area is delimited as a soil fertility unit, realizing further subdivision of the cultivated land parcel zoning. Each soil fertility unit has a clear spatial boundary and unique soil characteristics, providing a refined analysis unit for subsequent soil indicator characteristic fusion, degradation identification, and soil fertility deduction.

[0034] During the implementation of the entire soil indicator analysis module, it is necessary to ensure the coherence and accuracy of data processing. From the construction of the soil characteristic library to the determination of soil fertility units, each link depends on the data quality and the rationality of the algorithm. For example, in the data collection stage, it is necessary to ensure that the layout density of sensors can fully reflect the spatial variation of soil characteristics, avoiding deviations in clustering results caused by missing or sparse data; in the process of feature matching and clustering, it is necessary to select appropriate algorithms and parameters according to the data characteristics to ensure that the soil fertility clustering groups can truly reflect the similarities and differences of soil characteristics; when extracting the characteristic focus areas and dividing the soil fertility units, it is necessary to comprehensively consider the spatial scale and agricultural management requirements to make the size and shape of the soil fertility units convenient for actual cultivated land protection and utilization decisions.

[0035] Example 2: The process of generating the resource status vector corresponding to the soil fertility unit is as follows: For each soil fertility unit, according to its spatial position in the parcel zoning (such as the longitude and latitude coordinates obtained through the geographic information system), the index fluctuation data within the unit is obtained at a preset cycle (such as daily, weekly, or monthly). The index fluctuation data is the deviation of the real-time monitoring value of the soil indicator from the historical mean. For example, the total nitrogen content of the soil in a certain soil fertility unit is 1.2 g / kg during a certain monitoring cycle, and its historical mean is 1.0 g / kg, then the fluctuation data is +0.2 g / kg; the real-time value of soil moisture content is 20%, and the historical mean is 25%, and the fluctuation data is -5%. These data are collected in real time through the sensor network deployed within the soil fertility unit and are calculated by comparing with the long-term statistical values in the historical database, reflecting the dynamic changes of soil indicators in the time dimension.

[0036] Calculate the index fluctuation coefficient of the soil fertility unit. The fluctuation coefficient is used to measure the fluctuation amplitude of soil indicators within the preset cycle, and it is calculated using the coefficient of variation or the standard deviation method. Here, taking the standard deviation method as an example, the calculation formula is: , Among them, the "standard deviation" is the standard deviation of the soil index monitoring values of the soil force unit within the preset period, representing the degree of data dispersion; the "mean value" is the arithmetic mean of the monitoring values in the corresponding period, reflecting the average level of the index. The fluctuation coefficient is processed by dimensionless normalization, eliminating the influence of different index dimensions on the measurement of the fluctuation amplitude, and facilitating the horizontal comparison between different types of indexes (such as fertility indexes and soil moisture indexes).

[0037] According to the calculated index fluctuation coefficient, the soil force units are divided into two categories: high-risk units and stable units. When the fluctuation coefficient exceeds the first warning threshold, mark this unit as a high-risk unit. The first warning threshold is a critical value preset according to the natural variation range of soil indexes and agricultural production experience. For example, for soil fertility indexes, the threshold can be set at 20%, that is, when the fluctuation coefficient is greater than 20%, it indicates that the soil indexes of this unit have significant abnormal changes and there is a relatively high risk of degradation. At this time, directly extract the soil index composition of this unit as the resource state vector. The dimension of the resource state vector is the same as the number of monitored soil indexes. For example, if 6 indexes including total nitrogen, total phosphorus, total potassium, organic matter, water content, and pH value are monitored, the vector is a six-dimensional vector, and each element is the normalized index value. The normalization method can adopt min-max standardization, and the formula is: , Among them, is the original index value, and are respectively the minimum and maximum values of this index in historical data, is the normalized index value, and its value range is [0, 1]. Through normalization processing, indexes with different dimensions are made comparable, facilitating subsequent model analysis and calculation.

[0038] When the fluctuation coefficient is lower than the first warning threshold, mark this unit as a stable unit, indicating that the fluctuation of its soil indexes is within the normal range and the degradation risk is relatively low. For stable units, it is necessary to aggregate the soil index data of their adjacent units to reflect the overall soil force state of the region. The definition of adjacent units is based on the spatial position relationship. For example, taking the current unit as the center, all soil force units within a circular area with a radius of R or a square area with a side length of L are used as adjacent units. The values of R and L are determined according to the soil spatial autocorrelation range and agricultural management scale. For example, at the small field scale, R = 50 meters can be set, and at the large field scale, R = 200 meters can be set. The data aggregation method uses spatial weighted average, and the weight is determined based on the reciprocal of the distance between the adjacent unit and the current unit, that is, the closer the unit, the greater the weight. The calculation formula is: , Among them, is the aggregated index value, is the soil index value of the th adjacent unit, is the th adjacent unit's weight, , is the th adjacent unit's spatial distance from the current unit, is the number of adjacent units. Through spatial weighted averaging, the data deviation caused by local accidental fluctuations in stable units can be effectively reduced, enhancing the stability and representativeness of the data.

[0039] After completing data aggregation, the aggregated data needs to be reconstructed into a resource state vector. The reconstruction process also adopts normalization to ensure the dimensional consistency of vector elements. Different from the resource state vector of high-risk units, the vector elements of stable units do not directly come from the monitoring data of this unit, but are obtained through data aggregation of adjacent units. Therefore, it can better reflect the overall soil characteristics of the area where the unit is located. For example, if the soil moisture content of the current stable unit fluctuates little, but the moisture content of multiple adjacent units shows a downward trend, data aggregation can make the moisture content index in the resource state vector of this unit reflect the regional drought trend, avoiding ignoring the potential overall degradation risk due to the stability of single-unit data.

[0040] During the implementation process, attention should be paid to the reasonable selection of the preset period. If the preset period is too short, the fluctuation coefficient may be greatly affected by accidental factors, resulting in misjudgment; if the period is too long, it may lag in reflecting the true changes of soil indicators. In practical applications, different preset periods can be set for different types of indicators according to the change rate of soil indicators and early warning requirements. For example, for soil moisture indicators that change rapidly (such as moisture content, temperature), a daily or weekly period can be set, and for fertility indicators that change slowly (such as organic matter content), a monthly or quarterly period can be set.

[0041] In addition, the setting of the first early warning threshold needs to be adjusted differentially in combination with factors such as soil type, climate conditions, and farming systems. For example, in sandy soil areas, due to the poor water and fertilizer retention capacity of the soil, the natural fluctuations of fertility indicators may be relatively large, and the first early warning threshold can be appropriately increased; while in clay soil areas, the soil properties are relatively stable, and a lower threshold can be set. By dynamically adjusting the threshold, the adaptability of the system to different cultivated land environments can be improved, reducing the occurrence of false alarms or missed alarms.

[0042] After generating the resource status vector corresponding to the soil fertility unit, this vector will be used as the input data for the degradation identification module to extract the core degradation factors and establish a risk stratification model. For high-risk units, the vector directly reflects their significantly abnormal soil indicators, facilitating the rapid identification of local degradation risks; for stable units, the vector aggregates data from adjacent units to capture the soil fertility change trend at the regional scale and achieve early warning of potential degradation risks.

[0043] Embodiment 3: The degradation identification module realizes the accurate identification and hierarchical management of the degradation risks of soil fertility units by separating the core degradation factors, constructing a risk stratification model, and dynamically expanding the model index factors.

[0044] Separate the fertility anomaly parameters, soil moisture anomaly parameters, and texture variation parameters from the resource status vector of the soil fertility unit. The resource status vector is a multi-dimensional vector composed of normalized soil indicators and contains key parameters reflecting soil quality. The fertility anomaly parameters mainly include the values of nutrient indicators such as organic matter content, nitrogen, phosphorus, and potassium that deviate from the normal range. For example, when the soil organic matter content is lower than the critical value of the regional soil fertility grade standard, it is determined as a fertility anomaly; the soil moisture anomaly parameters involve abnormal fluctuations in indicators such as soil water content, temperature, and pH value. For example, when the water content is lower than the lower limit or higher than the upper limit of the suitable range for crop growth, or the pH value exceeds the normal soil acid-base range; the texture variation parameters characterize the changes in soil particle composition. For example, an increase in sand content leads to a decrease in soil water retention capacity, or a change in clay content affects soil aeration. The process of separating these parameters needs to combine professional knowledge and historical data in the field of soil science, and identify the abnormal parameters beyond the range by setting the normal threshold range for each indicator.

[0045] Based on the separated fertility anomaly parameters, soil moisture anomaly parameters, and texture variation parameters, establish a risk stratification model for the soil fertility unit. The risk stratification model adopts a hierarchical structure and aims to divide the soil fertility units into different risk levels according to the combination and severity of the degradation factors. Machine learning algorithms such as decision trees, random forests, or logistic regression can be used for model construction. Taking the decision tree algorithm as an example, first select the parameter with the greatest impact on the degradation risk as the root node, such as the soil organic matter content, and divide the soil fertility units into different branches according to its numerical value; then continue to select the second most important parameter (such as soil water content) as the sub-node in each branch to further divide the risk levels until the preset tree depth or node purity requirement is reached. The finally formed risk stratification model can divide the soil fertility units into levels such as high-risk degradation, moderate degradation, and potential degradation. For example, when the fertility anomaly parameters exceed the critical value and the soil moisture anomaly parameters also exceed the standard at the same time, it is determined as high-risk degradation; if only the fertility anomaly parameters are close to the critical value and the soil moisture indicators are normal, it is determined as potential degradation.

[0046] After establishing the risk stratification model, it is necessary to verify whether the coverage of the model meets the early warning requirements. Specifically, if the number of soil fertility units covered by the current risk stratification model is less than the preset early warning threshold (such as set to 5% of the total number of soil fertility units), it indicates that the model may not effectively capture enough degraded risk units, and there is a possibility of missing key degradation factors. At this time, it is necessary to traverse the resource status vectors of adjacent soil fertility units, analyze the index factors included in the risk stratification models of adjacent units, and identify the index factors not included in the current model. For example, the current model is only established based on fertility and soil moisture parameters, while the models of adjacent units include soil texture variation parameters or heavy metal content indicators, which may have important impacts on degradation risks but are not included in the current model.

[0047] To ensure the integrity of the model, it is necessary to add the index factors not included in the adjacent unit model to the current model. For example, if it is found that the degradation risk of adjacent units is closely related to soil compaction (a parameter representing texture variation) and the current model does not consider this factor, then soil compaction is introduced as a new input variable into the model. After introducing the new factor, it is necessary to retrain the model and adjust the node division rules and risk level determination criteria so that the model can comprehensively consider more dimensions of degradation factors. This process may require multiple iterations until the number of soil fertility units covered by the model reaches the preset threshold, or it is confirmed through feature importance analysis (such as Gini importance score based on random forest) that all factors with significant impacts on degradation risks have been included.

[0048] During the implementation of the degradation identification module, the following key points need to be noted: Setting of abnormal parameter thresholds: The thresholds need to be determined in combination with the regional soil background values, crop growth requirements, and agricultural production practice experience. For example, the normal ranges of organic matter content in different soil types are different, and the suitable moisture content intervals for sandy loam and clay loam are also different. It is necessary to formulate differentiated thresholds according to the actual situation to avoid misjudgments caused by a unified threshold.

[0049] Interpretability of the risk stratification model: Use algorithms with strong interpretability (such as decision trees) to build the model, which is convenient for agricultural practitioners to understand the driving factors of degradation risks. The hierarchical division of the model should conform to the ecological principles of soil degradation. For example, the high-risk level should correspond to irreversible or difficult-to-recover degradation states, the moderate level corresponds to degradation that can be improved through agronomic measures, and the potential level corresponds to early warning signals.

[0050] Definition criteria for adjacent units: The scope of adjacent units needs to be determined based on the spatial autocorrelation of the soil. The spatial variation range of soil indicators can be analyzed through geostatistical methods (such as semivariograms) to delimit a reasonable neighborhood radius. For example, if the spatial autocorrelation range of soil fertility indicators is 200 meters, then all soil fertility units within 200 meters of the current unit are regarded as adjacent units.

[0051] Dynamic update mechanism: The process of soil degradation has timeliness. As climate conditions and farming methods change, the importance of degradation factors may change. Therefore, the degradation identification module needs to retrain the model regularly (such as annually) based on the latest data, update the abnormal parameter thresholds and risk stratification rules to ensure the timeliness and accuracy of the model.

[0052] The core objective of the degradation identification module is to achieve precise classification and early warning of the degradation risk of soil fertility units through multi-dimensional factor analysis and dynamic model adjustment. For example, in the cultivated land of a certain area, through this module, it can be identified that some plots are suffering from soil salinization due to long-term excessive fertilization (the combined effect of fertility abnormal parameters and texture variation parameters), and are classified as high-risk degradation units; some other plots are suffering from continuously low soil moisture content due to insufficient irrigation (dominated by soil moisture abnormal parameters), and are classified as moderately degraded units; while some plots, although there are no significant abnormalities, show a downward trend in soil organic matter content (fertility abnormal parameters are close to the critical value), and are classified as potential degradation units. These classification results provide clear target objects for subsequent soil fertility deduction and decision-making generation, enabling targeted implementation of cultivated land protection measures.

[0053] In addition, the module expands the model by introducing the index factors of adjacent units, effectively solving the problem of data limitation of a single unit. For example, when the monitoring data of a certain soil fertility unit itself does not show obvious abnormalities, but soil heavy metal exceedance occurs in multiple surrounding units, by traversing the resource state vectors of adjacent units, the heavy metal content can be incorporated into the risk stratification model of the current unit to early warn of potential pollution diffusion risks and avoid missed risk judgments caused by isolated data analysis.

[0054] Example 4 The specific implementation of the soil fertility deduction module is as follows: Obtain the time-series factor and decay factor in the evolution association rule. The time-series factor characterizes the periodic or trend characteristics of the degradation factor changing over time. For example, the seasonal fluctuations of soil fertility indicators during the crop growth cycle (such as the increase in nitrogen content after fertilization at the sowing stage and the decrease after harvest due to crop absorption), or the annual decreasing trend of organic matter content caused by long-term farming. The decay factor reflects the speed and amplitude of soil fertility degradation. For example, the decrease value of the soil fertility index per unit time, the decreasing rate of crop yield, etc. These factors are obtained through time-series analysis of historical monitoring data. Methods such as moving average, exponential smoothing, or time series decomposition can be used to separate the trend term, seasonal term, and random term from the noisy data, and then extract the evolution cycle characteristics (such as annual cycle, monthly cycle) of the time-series factor and the quantitative indicators (such as annual average decay rate) of the decay factor.

[0055] Construct a plot weight matrix associated with temporal factors and decay factors. The plot weight matrix is a two-dimensional matrix. Its row dimension corresponds to time steps (such as monitoring periods divided by days, weeks, or months), and its column dimension corresponds to core degradation factors (such as organic matter content, soil water content). The matrix element values represent the influence weights of each factor on soil fertility degradation at the corresponding time points. The process of constructing the matrix is as follows: Determine the starting node: Identify the evolution cycle characteristics of the temporal factor. For example, detect its main cycle components through Fourier transform or autocorrelation analysis. If the current cycle characteristics exactly match the preset index cycle (such as the crop growth cycle, irrigation cycle) (for example, the main cycle of the temporal factor is 12 months, which is consistent with the natural annual cycle), then set the starting point of the current cycle of the temporal factor as the starting node of the plot weight matrix, serving as the reference point of the time axis to ensure that the time series of the matrix is synchronized with the actual agricultural production cycle.

[0056] Generate intermediate nodes and termination nodes: Calculate the evolution matching degree between the temporal factor and the decay factor. The matching degree measures the synchrony and correlation between the two in the time series, and can be calculated using the dynamic time warping (DTW) algorithm or the Pearson correlation coefficient. For example, if the correlation coefficient between the temporal factor (such as the seasonal fluctuation of soil water content) and the decay factor (such as the quarterly decrease in crop yield) is 0.8, it indicates a high degree of matching between the two. Arrange the factor combinations at different time points in descending order of the matching degree to generate the intermediate nodes and termination nodes of the matrix in sequence. The intermediate nodes represent the key states in the factor evolution process (such as the corresponding point between the water content peak and the yield trough), and the termination node represents the end state of the current analysis cycle (such as the last time point of the annual monitoring cycle).

[0057] Path backtracking and matrix optimization: Conduct path backtracking on the termination node to check whether the factor evolution path from the starting node to the termination node conforms to logical rules (such as whether the change of the decay factor lags behind the fluctuation of the temporal factor). If the matching degree of the termination node is lower than the preset matching threshold (such as set to 0.6), it indicates that there are abnormalities or data noises in the factor combination of this node, and its weight value needs to be adjusted or the node needs to be reselected until the matching degrees of all nodes meet the requirements, and then output the stable distribution of the plot weight matrix.

[0058] After constructing the plot weight matrix, it is necessary to calculate the soil fertility decay gradient under different correction strategies. The soil fertility decay gradient is a quantitative index to measure the speed and trend of soil fertility degradation. The calculation process is as follows: Statistical factor parameters: For each terminal node in the plot weight matrix, calculate the mean value of the time series factor and the range of the attenuation factor. The mean value of the time series factor is the average value of the time series factor within the corresponding time step of the node, reflecting the overall level of the factor in this stage; the range of the attenuation factor is the difference between the maximum and minimum values of the attenuation factor within the corresponding time step of the node, reflecting the amplitude of the factor change. At the same time, calculate the global variance of all node factors (including time series factors and attenuation factors), and the global variance characterizes the overall dispersion degree of factor fluctuations in the entire matrix.

[0059] Calculate the ratio of the time series attenuation coefficient to the range: Subtract the mean value of the time series factor of a single terminal node from the mean value of the time series factor of the adjacent node to obtain the difference in the mean value of the time series factor, and divide it by the global variance to obtain the time series attenuation coefficient. The formula is: , This coefficient reflects the impact of the change rate of the time series factor in adjacent time stages on soil fertility degradation. At the same time, calculate the ratio of the range of the attenuation factor to the global variance. The formula is: , This ratio reflects the contribution degree of the fluctuation amplitude of the attenuation factor to soil fertility degradation.

[0060] 3. Obtain the soil fertility attenuation gradient by weighted summation: Weight and sum the ratio of the time series attenuation coefficient to the range according to the preset weight as the soil fertility attenuation gradient of this node. The preset weight is calibrated and determined according to the physical meaning of the factor and historical data. For example, set the weight of the time series attenuation coefficient to 0.6 and the weight of the range ratio to 0.4. The formula is: , The larger the value of the soil fertility attenuation gradient, the faster the soil fertility degradation rate, and it is necessary to take control measures first.

[0061] During the implementation process, the following key issues need to be noted: Accuracy of factor extraction: The extraction of time series factors and attenuation factors depends on high-quality historical data and appropriate analysis methods. For example, for factors with significant seasonal fluctuations (such as soil water content affected by rainfall), a seasonal decomposition model (such as STL decomposition) needs to be used to separate the trend term and the seasonal term to avoid misjudging periodic fluctuations as long-term degradation trends.

[0062] Time scale of matrix construction: The time step of the plot weight matrix needs to match the warning period. If the system needs to achieve monthly warning, the time step is set to 1 month; if it is used for annual trend analysis, the step is set to 1 year. The choice of time scale directly affects the recognition accuracy of factor evolution laws and the practicality of the matrix.

[0063] Scenario Adaptation of Correction Strategies: Different correction strategies (such as fertilization adjustment, irrigation optimization, and rotation system change) correspond to different factor influence paths. For example, fertilization strategies mainly affect the temporal factors related to soil fertility, while irrigation strategies mainly affect the temporal factors related to soil moisture. When constructing the matrix, factor weights need to be set separately for different strategies to ensure the pertinence of the decay gradient calculation.

[0064] Dynamic Nature of Parameter Calibration: Parameters such as preset matching thresholds and weight coefficients need to be dynamically adjusted according to changes in regional agricultural production conditions. For example, when introducing new farming techniques or when factor correlations change due to climate change, parameters need to be recalibrated to ensure the timeliness of the model.

[0065] The core role of the soil fertility deduction module is to provide a scientific basis for cultivated land regulation by quantitatively analyzing the evolution law and degradation rate of factors. For example, in a certain soil fertility unit, through the plot weight matrix, it is found that the temporal factor of soil organic matter content has shown a continuous downward trend in the past three years (average annual decay rate of 1.5%), and its matching degree with the crop yield decline rate (decay factor) is 0.75. The constructed matrix shows that the temporal decay coefficient of the current termination node is 0.8, the range ratio is 0.5, and the weighted soil fertility decay gradient is 0.68, indicating that the soil fertility degradation rate of this unit is relatively fast, and correction strategies such as increasing the application of organic fertilizers need to be implemented first.

[0066] In addition, through the path backtracking mechanism, abnormal nodes in the factor evolution process can be identified to avoid misjudgment caused by data mutation or noise.

[0067] Example 5: This example involves the specific implementation of the threshold adaptation module and the decision-making generation module. Through dynamic reference value calculation, optimal warning threshold matching, and risk difference sequence analysis, the quantitative warning of cultivated land degradation risk and the intelligent generation of regulation decisions are realized. The specific process is as follows: In the threshold adaptation module, the core of deriving the optimal warning threshold is to achieve the adaptive adjustment of the threshold through a dual dynamic reference value by combining the current soil fertility decay gradient and historical degradation patterns. First, extract the degradation pattern in the historical data that is closest to the current soil fertility decay gradient. The historical degradation pattern refers to typical cases where the quality of cultivated land has declined due to soil degradation over a certain period of time (such as the past 5 years). Each pattern includes the spatio-temporal distribution characteristics, evolution process, and corresponding warning thresholds of the degradation factors. The first dynamic reference value is determined by calculating the cosine similarity between the current gradient and the historical pattern in terms of spatial distribution. Cosine similarity is used to measure the directional similarity of two vectors, and the formula is: , where, is the eigenvector of the current soil fertility decay gradient, including the decay rate and spatial distribution parameters of the core degradation factors; They are the feature vectors of the historical degradation mode, with the same dimension, and the influence of the dimension is eliminated through normalization. The value range of cosine similarity is [-1, 1], and the larger the value, the more similar the spatial distribution is.

[0068] Secondly, count the difference in the number of abnormal peaks between the current soil fertility attenuation gradient and the historical degradation mode as the second dynamic reference value. The number of abnormal peaks refers to the number of times the soil index exceeds the warning threshold within the preset analysis period (such as 1 year). For example, if the number of abnormal peaks corresponding to the current gradient is 8 times, and the average number of abnormal peaks in the historical mode is 5 times, the difference is +3 times, indicating that the current degradation pressure is greater. This reference value reflects the difference in the intensity of the degradation process and makes up for the deficiency of the spatial similarity analysis in terms of time dimension characteristics.

[0069] Based on the linear combination of the first dynamic reference value and the second dynamic reference value, match the optimal threshold in the preset warning threshold table. The preset warning threshold table is a lookup table obtained through training with historical data, storing the warning thresholds corresponding to different combinations of reference values. The weights of the linear combination are determined according to the characteristics of regional cultivated land degradation. For example, in areas sensitive to soil degradation, the weight of the first dynamic reference value can be set to 70% and the weight of the second dynamic reference value to 30% to highlight the leading role of spatial distribution similarity; in areas significantly affected by climate change, the weights can be adjusted to 50% each to take into account both spatial and temporal characteristics. The matching process uses nearest neighbor interpolation or a regression model to dynamically find or calculate the corresponding threshold according to the combination of reference values, ensuring the accurate adaptation of the threshold to the current degradation state.

[0070] In the decision-making generation module, the key to analyzing the risk difference sequence lies in identifying the direction of soil fertility attenuation and spatial conflicts and generating a differential regulation decision-making plan. According to the direction of soil fertility attenuation of each soil unit in the risk difference sequence (positive attenuation means that the index value deviates from the threshold in the direction of increasing degradation, and negative attenuation means approaching the threshold in the recovery direction), all units are divided into a positive attenuation interval and a negative attenuation interval. For example, if the soil organic matter content of a unit is lower than the threshold and continues to decline, it is divided into the positive attenuation interval; if the organic matter content of another unit increases after treatment, it is divided into the negative attenuation interval.

[0071] Extract the convergence rate of the risk difference in the positive attenuation interval and the diffusion rate of the risk difference in the negative attenuation interval. The convergence rate refers to the rate of decrease in the risk difference value (the difference between the current index value and the threshold) per unit time, reflecting the degree of mitigation of the degradation trend; the diffusion rate refers to the rate of increase in the risk difference value per unit time, reflecting the degree of intensification of the degradation trend. For example, if the risk difference value of a unit in the positive attenuation interval expands from -5% (5% lower than the threshold) to -7% per month, the diffusion rate is +2% / month; if the risk difference value of a unit in the negative attenuation interval shrinks from +3% (3% higher than the threshold) to +1% per month, the convergence rate is -2% / month.

[0072] Reconcile the conflict between the convergence rate and the diffusion rate according to the spatial weights of the soil fertility units. The spatial weights are determined based on the distance between units and the degree of mutual influence. The closer the units are, the higher the weight. For example, using the inverse distance weighting method, the weight formula is: , where, is the weight of the th adjacent unit, and is the spatial distance between this unit and the target unit. The purpose of conflict reconciliation is to balance the contradiction of degradation trends between adjacent units and avoid exacerbating the overall regional risk due to local optimization. For example, if a positive decay unit is adjacent to a negative decay unit, the diffusion and convergence rates of the two may affect each other, and it is necessary to calculate the comprehensive adjustment parameter through the spatial weight to ensure the coordination of the control measures at the regional scale.

[0073] Generate the adjustment parameters of the cultivated land control decision-making plan according to the reconciled rate parameters. The adjustment parameters include the type of treatment measures for the soil fertility units (such as fallow, rotation, fertilization), the intensity of the measures (such as the amount of fertilizer applied, irrigation frequency), and the implementation time window. For example, for a positive decay unit with a high diffusion rate and a high spatial weight, set high-intensity treatment measures (such as immediate fallow and application of soil amendments); for a negative decay unit with a stable convergence rate, set maintenance measures (such as regular fertilization management). The decision-making plan needs to be presented in a spatial visualization form (such as a GIS thematic map), marking the risk level, measure type, and expected effect of each unit, which is convenient for the agricultural management department to implement overall.

[0074] During the implementation process, the following key links need to be noted: Historical data needs to cover degradation scenarios under different climate conditions and farming systems to ensure the comprehensiveness of the preset early warning threshold table. The data collection period should be no less than 5 years to capture long-term degradation trends and periodic fluctuation characteristics.

[0075] Accidental peaks caused by short-term natural disasters (such as heavy rain, drought) need to be excluded, and only persistent anomalies caused by long-term farming or soil evolution are counted to avoid misjudgment of the threshold.

[0076] When the land use type of the soil fertility unit changes (such as from cultivated land to orchard), or after implementing major ecological projects, it is necessary to recalculate the spatial weights to reflect the changes in the interaction between units.

[0077] The adjustment parameters need to be connected with the actual agricultural production. For example, the adjustment of the amount of fertilizer applied needs to refer to the fertilizer requirement law of crops and the fertilizer supply capacity to avoid idealized suggestions that are divorced from production conditions.

[0078] The collaborative operation of the threshold adaptation module and the decision generation module realizes the closed-loop management from risk quantification to decision implementation. For example, in a certain area of cultivated land, through the threshold adaptation module, it is found that the cosine similarity between the current soil fertility attenuation gradient and the historical "salinization caused by over-fertilization" pattern is 0.85, and the difference in the number of abnormal peaks is +2 times. The optimal warning threshold is matched to be 0.3% soil salt content (higher than the conventional threshold of 0.25%), indicating that increased vigilance is required. After the decision generation module analyzes the risk difference sequence, it is found that 20% of the units belong to the positive attenuation interval, and they are concentrated in the center of the irrigation area, with a diffusion rate of +1.5% / month. After being adjusted by the spatial weight, a rotation + leaching improvement plan for this area is generated, and the adjustment parameters include a rotation cycle of 2 years and a leaching frequency of 3 times per year. At the same time, monitoring and maintenance measures are implemented for the surrounding negative attenuation units.

[0079] In addition, this module supports the multi-scenario simulation function and can generate multiple sets of decision-making plans for different correction strategies (such as different fertilization amounts and irrigation methods) for managers to compare and select. For example, by adjusting the factor weights in the plot weight matrix, the changes in the soil fertility attenuation gradient under two strategies of "increasing organic fertilizer application" and "water-saving irrigation" are simulated, and the input-output ratio and environmental benefits of the plans are compared to achieve scientific and refined decision-making.

[0080] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0081] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A data optimization and intelligent decision support system in cultivated land resource early warning, characterized in that Including: A cultivated land data acquisition module, which is used to obtain multi-dimensional attribute data of cultivated land plots in real time and set plot zoning rules corresponding to warning levels; A soil index analysis module, which is used to divide multiple soil fertility units within the plot zoning, perform feature fusion processing on the soil indexes of each soil fertility unit, and generate a resource status vector corresponding to the soil fertility unit; A degradation identification module, which is used to extract core degradation factors from the resource status vector, establish a risk stratification model associated with the soil fertility unit, and obtain the data correction coefficient corresponding to the model; A soil fertility deduction module, which is used to identify the evolution association rules in the data correction coefficient, dynamically correct the core degradation factors according to the association rules, and calculate the soil fertility attenuation gradient of each soil fertility unit under different correction strategies; A threshold adaptation module, which is used to deduce the optimal warning threshold based on the soil fertility attenuation gradient, and generate a risk difference sequence by comparing the current resource index value with the optimal warning threshold; A decision-making generation module, which is used to analyze the risk difference sequence and integrate the risk difference sequence into a cultivated land regulation decision-making scheme based on the soil fertility attenuation trend of the soil fertility unit.

2. The data optimization and intelligent decision-making support system in the cultivated land resource early warning according to claim 1, characterized in that The implementation method of the soil index analysis module includes: constructing a soil feature library corresponding to the soil fertility unit, and the soil feature library contains soil indexes and a resource parameter set corresponding to feature fusion; Perform similar feature matching on the resource parameter set, divide the soil fertility clustering groups of the resource parameter set according to the matching results; extract the feature focus domain of the soil index from the soil fertility clustering groups, and set the focus domain as the soil fertility unit.

3. The data optimization and intelligent decision-making support system in the cultivated land resource early warning according to claim 2, characterized in that, Dividing the soil fertility clustering groups of the resource parameter set further includes: Extracting index distribution, anomaly polarity and resource density parameters according to the plot attributes and degradation intensity in the resource parameter set, and generating soil feature identifiers based on the above parameters; Associate the soil feature identifiers with the resource parameter set, and screen out the parameter sets with similarity higher than the preset warning threshold through calculating the index similarity between the feature identifiers to form soil fertility clustering groups.

4. The data optimization and intelligent decision-making support system in the cultivated land resource early warning according to claim 1, wherein, The implementation method of generating a resource status vector corresponding to the soil fertility unit includes: For each soil fertility unit, according to the spatial position of the unit in the plot zoning, obtain the index fluctuation data of the unit within a preset period, and calculate the index fluctuation coefficient of the unit; When the index fluctuation coefficient exceeds the first warning threshold, mark the unit as a high-risk unit, and extract its soil indexes to form a resource status vector; when the index fluctuation coefficient is lower than the first warning threshold, mark the unit as a stable unit, and perform data aggregation on the soil indexes of the adjacent units of the unit, and reconstruct the aggregated data into a resource status vector.

5. The data optimization and intelligent decision-making support system in the cultivated land resource early warning according to claim 1, wherein The implementation method of the degradation identification module includes: Separate the fertility anomaly parameter, soil moisture anomaly parameter and texture variation parameter from the resource status vector, and generate a risk stratification model of the soil fertility unit based on the fertility anomaly parameter, soil moisture anomaly parameter and texture variation parameter; If the number of soil fertility units covered by the current risk stratification model is less than the preset warning threshold, traverse the resource status vectors of adjacent soil fertility units, and add the index factors not included in the stratification model of the adjacent units to the current model.

6. The data optimization and intelligent decision-making support system in the cultivated land resource early warning according to claim 1, characterized in that, The implementation method of the soil fertility deduction module includes: obtaining the time series factor of the index change frequency and the attenuation factor of the soil fertility jump amplitude in the evolution association rule; Construct a plot weight matrix associated with temporal factors and decay factors, and determine the soil fertility decay gradient under different correction strategies according to the distribution probability of each element in the matrix.

7. The data optimization and intelligent decision-making support system in the arable land resource early warning according to claim 6, wherein Constructing the plot weight matrix also includes: Identifying the evolution period characteristics of temporal factors. If the current period characteristics exactly match the preset index period, set the temporal factor as the starting node of the plot weight matrix; Calculating the evolution matching degree between the temporal factor and the decay factor, and generating the intermediate node and the termination node of the plot weight matrix in descending order of the matching degree; Perform path backtracking on the termination node. When the matching degree of the termination node is lower than the preset matching threshold, output it as the final distribution of the plot weight matrix.

8. The data optimization and intelligent decision-making support system in the cultivated land resource early warning according to claim 7, characterized in that, The implementation method of calculating the soil fertility decay gradient includes: Statistical mean of the temporal factor and the range of the decay factor of each termination node in the plot weight matrix, and calculate the global variance of all node factors; Subtract the mean of the temporal factor of a single termination node from the mean of the temporal factor of the adjacent node, and divide by the global variance to obtain the temporal decay coefficient; at the same time, calculate the ratio of the range of the decay factor to the global variance, and weight and sum the two as the soil fertility decay gradient of this node.

9. The data optimization and intelligent decision-making support system in the cultivated land resource early warning according to claim 1, characterized in that, The implementation method of deriving the optimal early warning threshold includes: Extract the degradation pattern in the historical data that is closest to the current soil fertility decay gradient, and calculate the cosine similarity in the spatial distribution between the two as the first dynamic reference value; Statistical difference in the number of abnormal peaks between the current soil fertility decay gradient and the historical degradation pattern, and use the difference number as the second dynamic reference value; Based on the linear combination of the first dynamic reference value and the second dynamic reference value, match the optimal early warning threshold in the preset early warning threshold table.

10. The data optimization and intelligent decision-making support system in the cultivated land resource early warning according to claim 1, wherein The implementation method of the decision-making generation module includes: dividing the positive decay interval and the negative decay interval according to the soil fertility decay direction of each unit in the risk difference sequence; Extract the convergence rate of the risk difference in the positive decay interval and the diffusion rate of the risk difference in the negative decay interval, and reconcile the conflict between the two according to the spatial weight of the soil fertility unit to generate the adjustment parameter of the cultivated land regulation decision-making plan.

Images