Cultivated land soil quality dynamic diagnosis method based on multi-source remote sensing collaborative inversion

Abstract

The invention discloses a farmland soil quality dynamic diagnosis method based on multi-source remote sensing collaborative inversion, and relates to the technical field of agricultural information, and the method comprises the steps: collecting physical and chemical parameters of soil temperature, humidity, pH value, organic matter content, nitrogen phosphorus and potassium concentration and the like in real time through a multi-source sensor network; removing abnormal values by using a data preprocessing algorithm, performing standardization processing, generating a standardized soil parameter set for subsequent climate and management factor correlation analysis, automatically generating a risk early warning signal, pushing the risk early warning signal to an agricultural management decision platform, and completing full-process analysis from data acquisition to decision support; the farmland soil quality dynamic diagnosis method based on multi-source remote sensing collaborative inversion can provide a scientific basis for agricultural management decision, effectively guide farmland management practice, and improve soil quality and agricultural production efficiency.

Description

Technical Field

[0001] The present invention relates to the field of agricultural information technology, and in particular to a method for dynamic diagnosis of cultivated land soil quality based on multi-source remote sensing collaborative inversion. Background Art

[0002] At present, soil quality assessment mainly relies on traditional field monitoring and empirical judgment, which are inadequate in the face of complex and changing climate conditions. Existing research is often limited to single factor analysis and lacks an in-depth understanding of the interaction of multiple environmental factors, making it difficult to provide a scientific and reliable prediction basis for agricultural management decisions.

[0003] Traditional assessment methods are time-consuming and costly, and cannot meet the urgent needs of modern agriculture for real-time dynamic monitoring. Soil quality change is a complex dynamic process affected by multiple factors, and there is an intricate interaction between different climatic conditions and agricultural management measures. Climate factors such as changes in precipitation patterns, temperature fluctuations, and extreme weather events will change the fertility level and structural stability of the soil by affecting the physical and chemical properties of the soil. This multi-factor coupling effect makes it extremely difficult to accurately simulate the evolution of soil quality. What is more complicated is that different farming methods such as no-tillage, deep plowing, crop rotation and other agricultural management measures will have a cumulative impact on the soil over a long period of time, and this long-term effect often has hysteresis and nonlinear characteristics. When extreme climate events are superimposed on specific agricultural management measures, the extent and duration of their impact on soil quality are more difficult to predict, and traditional methods cannot effectively capture this complex spatiotemporal change law. Summary of the invention

[0004] The purpose of the present invention is to provide a dynamic diagnosis method for cultivated land soil quality based on multi-source remote sensing collaborative inversion, to construct a dynamic soil quality simulation system that can comprehensively consider climate change, agricultural management measures and their interactions, to accurately predict the potential impact of extreme climate events on cultivated land soil, and to scientifically evaluate the long-term cumulative effects of different farming methods.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for dynamic diagnosis of cultivated land soil quality based on multi-source remote sensing collaborative inversion, the method comprising: Through a multi-source sensor network, physical and chemical parameters such as soil temperature, humidity, pH, organic matter content, and nitrogen, phosphorus, and potassium concentrations are collected in real time. Data preprocessing algorithms are used to remove outliers and perform standardization to generate a standardized soil parameter set for subsequent climate and management factor correlation analysis. Generate soil quality dynamic simulation data based on the standardized soil parameter set. Conduct uncertainty analysis on the soil quality dynamic simulation data. Generate one thousand parameter combinations through random sampling, and calculate the probability distribution interval of soil quality changes. If the prediction accuracy, evaluated based on the ratio of the mean square error to the actual observed value, is lower than the preset threshold of 85%, then return to the previous step to readjust the weight parameters of each factor in the neural network framework and generate soil quality change trend prediction data; Construct an early warning mechanism based on the soil quality change trend prediction data. If the predicted value of the soil fertility level is continuously lower than the preset safety threshold, i.e., 80% of the historical average, based on the comprehensive index of nitrogen, phosphorus, and potassium for three consecutive months, or the structural stability index drops by more than the preset threshold of 20%, then automatically generate a risk warning signal and push it to the agricultural management decision-making platform to complete the full-process analysis from data collection to decision support.

[0006] Preferably, the generation of soil quality dynamic simulation data based on the standardized soil parameter set includes: For the standardized soil parameter set, combine the historical records of meteorological stations and satellite remote sensing data to obtain climate change indicators such as precipitation, temperature fluctuations, humidity changes, and wind speed. If the precipitation exceeds twice the historical average of the same period calculated based on the local ten-year average for seven consecutive days, or the daily average temperature fluctuation exceeds the preset threshold of ten degrees Celsius, then it is judged as an extreme weather event and the time node is marked to generate a climate change and extreme weather mark set.

[0007] Preferably, the generation of soil quality dynamic simulation data based on the standardized soil parameter set further includes: Obtain agricultural management records such as tillage methods, rotation cycles, fertilization frequencies, and irrigation patterns from the farmland management database, construct a numerical coding system for tillage methods, and determine the quantitative weight coefficients of different management measures based on historical record analysis to generate an agricultural management quantitative parameter set for subsequent soil quality impact assessment.

[0008] Preferably, the generation of soil quality dynamic simulation data based on the standardized soil parameter set further includes: Use the standardized soil parameter set and the climate change and extreme weather mark set, and adopt the multiple regression analysis method to construct an association model between climate change factors and soil physical and chemical properties. If the monthly change rate of precipitation exceeds the preset threshold of 30%, then adjust the soil moisture impact coefficient. If the daily average temperature fluctuation amplitude exceeds ten degrees Celsius, then correct the organic matter decomposition rate parameter. If an extreme weather mark is detected, then add a short-term impact factor to generate a climate-soil coupling association matrix.

[0009] Preferably, the generation of soil quality dynamic simulation data based on the standardized soil parameter set further includes: Based on the set of agricultural management quantification parameters, the time series analysis method is applied to construct a cumulative effect model of tillage methods on soil quality. According to the historical management records, the cumulative impact weights of different tillage measures are calculated on monthly, quarterly, and annual time scales, and a long-term effect prediction model of agricultural management is generated for multi-factor comprehensive simulation.

[0010] Preferably, the generating of the dynamic simulation data of soil quality based on the set of standardized soil parameters further includes: Fusing the climate-soil coupling correlation matrix and the long-term effect prediction model of agricultural management, a dynamic simulation framework of soil quality with multi-factor interaction is constructed by using the neural network algorithm. If an extreme weather mark is detected, the emergency response mechanism is activated to adjust the weight of the soil moisture and organic matter decomposition rate parameters, and the dynamic simulation data of soil quality is generated.

[0011] Preferably, the specific formula for the uncertainty analysis is: ; Wherein, represents the predicted soil quality evolution function, represents the number of sampling times, represents the th entropy weight factor corresponding to the th sample, represents the non-linear model function for soil state evaluation, represents the th group of input data sets, represents the model structure parameters, represents the perturbation sensitivity coefficient corresponding to the th sample, represents the variance of the input variable set of the th sample,

[0012] Preferably, the calculation formula of the is: ; Wherein, represents the th entropy weight factor corresponding to the th sample, represents the weighted normalization factor of all samples, represents the input variable dimension, represents the th fuzzy membership degree of the th sample on the th index, represents the normalized value of the th sample on the The average value of the in the full sample, represents the sample number,

[0013] Preferably, the is calculated as follows: ; Among them, represents the perturbation sensitivity coefficient corresponding to the th sample, represents the total number of soil input characteristics, represents the dimension number of the input variable being traversed currently, represents the partial derivative of the model function B with respect to the input variable , represents the input variable, represents the th sample's degree of fluctuation on the th input variable, represents the maximum fluctuation value of all samples on the th variable.

[0014] Preferably, the early warning mechanism uses color - level classification to display the risk level, which is divided into three levels: normal, mild warning, and severe warning, and synchronizes the early warning results to the plot view of the agricultural management platform in the form of a graphical interface.

[0015] From the above - mentioned technical solutions, the present invention has the following beneficial effects: This dynamic diagnosis method of cultivated land soil quality based on multi - source remote sensing collaborative inversion collects soil physical and chemical parameters, climate change indicators, and agricultural management data, constructs a climate - soil coupling relationship matrix and a long - term effect prediction function of agricultural management. Using the neural network algorithm to fuse the above models, a dynamic simulation model of soil quality with multi - factor interaction is established, and combined with the extreme weather event response mechanism, an accurate prediction of the soil quality change trend is realized. The present invention also uses the Monte Carlo method for uncertainty analysis to improve the prediction accuracy and establishes an early warning mechanism to automatically generate a risk warning signal when the prediction result is lower than the safety threshold. This method can provide a scientific basis for agricultural management decisions, effectively guide farmland management practices, and improve soil quality and agricultural production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is the flow chart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] like Figure 1 As shown, the present invention provides a technical solution: a method for dynamic diagnosis of cultivated land soil quality based on multi-source remote sensing collaborative inversion, the method comprising: Through a multi-source sensor network, physical and chemical parameters such as soil temperature, humidity, pH, organic matter content, and nitrogen, phosphorus, and potassium concentrations are collected in real time. Data preprocessing algorithms are used to remove outliers and perform standardization to generate a standardized soil parameter set for subsequent climate and management factor correlation analysis. Generate soil quality dynamic simulation data based on the standardized soil parameter set, conduct uncertainty analysis on the soil quality dynamic simulation data, generate one thousand sets of parameter combinations through random sampling, calculate the probability distribution interval of soil quality change, and if the prediction accuracy based on the ratio of mean square error to actual observation value is lower than the preset threshold of 85%, return to the previous link to readjust the weight parameters of each factor in the neural network framework to generate soil quality change trend prediction data; Based on the predicted data on soil quality change trends, an early warning mechanism is established. If the predicted value of the soil fertility level based on the comprehensive nitrogen, phosphorus and potassium index is lower than the preset safety threshold, which is 80% of the historical average, for three consecutive months, or the structural stability index drops by more than 20% of the preset threshold, a risk warning signal will be automatically generated and pushed to the agricultural management decision-making platform, completing the full process analysis from data collection to decision support.

[0019] In this embodiment, by constructing a data acquisition system based on the fusion of multi-source remote sensing and ground sensors, problems such as high spatial heterogeneity, strong time-variability, and discontinuous information acquisition of cultivated land soil quality are solved. In terms of system structure, a multi-source information input end is formed by combining unmanned aerial vehicle (UAV) remote sensing, satellite images, and ground fixed-point sensors to achieve real-time and high-density acquisition of key physicochemical parameters such as soil temperature, humidity, pH value, organic matter content, and nitrogen, phosphorus, and potassium concentrations. The collected data is first preliminarily processed by an edge computing unit, and outlier values are removed using the IQR method, 3σ method, or time series difference detection technology, and standardized using normalization or Z-score methods to ensure parameter scale consistency and data comparability. In the data modeling stage, a deep learning model based on the LSTM (Long Short-Term Memory) or GRU (Gated Recurrent Unit) structure is introduced to establish a dynamic prediction model for the standardized soil parameter set. This model can effectively capture the temporal variation law of soil properties, construct a multi-dimensional input matrix using a time window, and learn the change trend of soil quality through model training. On this basis, Monte Carlo simulation is used to randomly sample the input parameter space, generating one thousand groups of parameter combinations, calculating the corresponding soil quality change prediction value set through model forward propagation, and then constructing a probability distribution interval and extracting statistical indicators such as confidence intervals and quantiles for uncertainty assessment. If the error indicators (such as mean squared error MSE, root mean squared error RMSE, coefficient of determination R², etc.) between the predicted output and the actual observed data are lower than the set accuracy threshold (such as 85%), the model reverse optimization process is triggered, and the weight parameters of each factor (such as climate variables, fertilization methods, tillage habits, etc.) in the neural network are adjusted, and the model is retrained to iteratively optimize the prediction results. The finally output soil quality change trend data is input into the early warning module, which sets multiple index triggering mechanisms, mainly using the comprehensive nitrogen, phosphorus, and potassium index and structural stability as the early warning benchmarks. If the comprehensive fertility index is continuously lower than 80% of the historical average for three consecutive months, or the structural stability drops by more than 20% compared to the benchmark, the system automatically generates a risk signal and pushes it to the manager's terminal through the agricultural Internet of Things platform, including graphical interface prompts, alarm record archiving, and generation of auxiliary decision-making suggestions, thus completing the full-chain intelligent analysis system from data acquisition, processing, prediction to intelligent decision-making.

[0020] This embodiment constructs a dynamic diagnosis system for cultivated land soil quality based on a data acquisition architecture that fuses multi-source remote sensing data and ground sensor networks. This system solves problems such as strong spatial heterogeneity, insufficient temporal data, and analysis lag in current soil monitoring. Its core technical path includes five steps: real-time data acquisition, standardized processing, deep learning prediction, uncertainty assessment, and risk early warning mechanism.

[0021] First, the system obtains key parameters such as temperature, humidity, pH value, organic matter content, and nitrogen, phosphorus, and potassium through sensing nodes distributed in different areas of cultivated land. These data come from ground-fixed sensors, UAV hyperspectral images, and satellite remote sensing images. To improve the accuracy of analysis, the system standardizes the collected raw data. Specifically, for each type of parameter, the historical mean and standard deviation are first calculated, and then the current value is subtracted from the mean and divided by the standard deviation, so as to unify all parameters under a comparable standard scale.

[0022] Next, the system uses a long short-term memory neural network to train a dynamic prediction model based on the standardized parameter sequence. This model learns the change rules of soil properties in the time dimension and predicts the soil quality trend in the future time period. During the training process, the system calculates the error between the prediction result and the true observed value, and measures the accuracy of the model based on this. If the error is large, the system will automatically adjust the internal parameters of the model, such as reallocating the weights of each input factor, to improve the prediction accuracy.

[0023] Based on the model prediction, the system also introduces an uncertainty analysis mechanism. Using the Monte Carlo method, a thousand random samplings are performed on the input parameter set to form a thousand groups of simulation data. Each group of simulation data is processed by the prediction model to obtain the corresponding set of soil quality prediction values. The system performs statistical processing on these results and extracts representative values such as the fifth percentile and the ninety-fifth percentile from them, so as to construct a confidence interval describing the range of variation of the prediction results. This process helps to judge the stability and reliability of the prediction results.

[0024] To achieve the risk management and early warning functions, the system designs a judgment standard with the soil fertility index and the structure stability index as the core. The soil fertility index is calculated by weighted summation of the concentrations of nitrogen, phosphorus, and potassium, where the weight of nitrogen is forty percent, and phosphorus and potassium are each thirty percent. When the soil fertility index of a certain cultivated land is lower than eighty percent of the historical average level of this region for three consecutive months, the system will consider that there is a risk of fertility decline. At the same time, if the structure stability index drops by more than twenty percent compared to the average level of the past year, it will also be regarded as a sign of potential structural damage. If any of these situations occur, the system will automatically generate a risk warning signal.

[0025] The early warning results will be pushed to relevant agricultural management terminals in the form of a data report through the agricultural Internet of Things platform. The report not only includes the early warning type, level, and triggering reasons, but also includes charts of soil quality change trends, time series statistical values of key parameters, and recommended intervention measures, assisting managers to formulate scientific response strategies to ensure the sustainable and healthy use of cultivated land resources.

[0026] The prediction accuracy threshold is used to judge whether the model prediction result is accurate enough. The "prediction accuracy percentage" is calculated by the ratio of the mean square error to the variance of the observed values. Referring to the relevant literature on agricultural soil quality prediction, an accuracy of 85% can meet the requirements of most management decisions. If it is lower than this threshold, it indicates that the prediction error is large, and the system needs to readjust the neural network model.

[0027] The fertility warning threshold (80% of the historical average) is used to judge whether the current soil fertility is at a low level and trigger a warning. The comprehensive index calculated by weighting the concentrations of nitrogen, phosphorus, and potassium is used as the fertility level indicator. By statistically analyzing the historical fertility indices of the same cultivated land in the past 5 - 10 years and calculating their average value, if the current value is lower than 80% of this average value for three consecutive months, it is considered to have a risk of decline. 80% is a commonly used "low - value alarm line" in meteorological and soil environment monitoring and has practical reference basis.

[0028] The soil structure stability threshold (a decrease of more than 20%) is used to evaluate the degree of deterioration of the soil aggregate structure. Monitor the 12 - month moving average of the structural stability in a certain area (such as the content of water - stable aggregates). If the current index drops by more than 20% compared to this average value, it is judged that the structure has deteriorated. Soil structure changes usually lag behind chemical indicators, so a 20% decrease is regarded as a significant fluctuation, referring to the empirical data of the long - term fixed - point monitoring project of farmland ecosystems.

[0029] This method significantly improves the timeliness and spatial coverage of soil quality monitoring by integrating multi - source remote sensing data and ground sensor data. The reliability of the data is enhanced by using standardized processing and outlier rejection algorithms, and the accuracy and stability of soil quality prediction are improved by combining deep learning models with uncertainty analysis. The early warning mechanism is based on the judgment of continuous data trends, ensuring the timeliness and accuracy of risk response, which helps agricultural managers intervene in advance and formulate scientific farming strategies. The entire system has functions such as data - driven, adaptive parameter adjustment, and automatic early warning, realizing a closed - loop management from data perception to intelligent decision - making.

[0030] According to the standardized soil parameter set, dynamic soil quality simulation data is generated. For the standardized soil parameter set, combined with the historical records of meteorological stations and satellite remote sensing data, climate change indicators such as precipitation, temperature fluctuation, humidity change, and wind speed are obtained. If the precipitation exceeds twice the historical average value calculated based on the local ten - year mean for seven consecutive days, or the daily temperature fluctuation exceeds the preset threshold of ten degrees Celsius, it is judged as an extreme weather event and the time node is marked, generating a set of climate change and extreme weather marks.

[0031] Based on the dynamic soil quality diagnosis, this embodiment further integrates climate change factors and extreme weather event recognition mechanisms, thereby enhancing the sensitivity and prediction ability of the model to external disturbances and improving the integrity and timeliness of dynamic soil quality simulation.

[0032] First, after the construction of the standardized soil parameter set is completed, an interface of the external meteorological database and the high spatio-temporal resolution remote sensing platform is called to obtain historical and current meteorological observation data corresponding to the geographical coordinates of the target area. The main meteorological factors include daily precipitation, average daily temperature, relative humidity, and wind speed. By establishing a meteorological index fluctuation model, these parameters are used as important external driving factors affecting the evolution of soil quality.

[0033] Extreme precipitation event judgment rule First, the system calculates the seven-day average cumulative precipitation on the same date (such as May 10th) in the ten years before the target date in this area, which is set as the historical average precipitation benchmark value. If the total precipitation for seven consecutive days within the target time period exceeds twice the historical benchmark value, then this time period is defined as an extreme precipitation event.

[0034] In the specific implementation, after statistically analyzing the ten-year historical data, its mean value is calculated, that is, the daily precipitation from May 4th to May 10th each year in this area is summed and averaged. Let this average value be the "seven-day precipitation mean value in the same period of history". Then, it is compared with the actual cumulative precipitation value from May 4th to May 10th in the current year. If the current value is greater than twice the historical mean value, the system marks this time period as an "extreme precipitation event".

[0035] Temperature fluctuation anomaly identification method For the identification of temperature fluctuation anomalies, the system statistically calculates the difference between the highest temperature and the lowest temperature every day as the daily temperature difference. If the average daily temperature difference within any consecutive seven days exceeds 10 degrees Celsius, which is the preset extreme fluctuation threshold in the system, it is considered that the current climate is in an unstable high-fluctuation state, which may significantly affect soil water evaporation, organic matter decomposition, and microbial activity.

[0036] The preset threshold is 10 degrees Celsius, which is formulated according to the standards of the Agricultural Climate Data Center. The adjustable range is set to 8 to 15 degrees Celsius, and it is fine-tuned according to the climate resilience of different regions. The temperature data comes from national meteorological stations or ground automatic weather stations.

[0037] Construction of the climate change and extreme event marking set In the above detection process, each time period that meets the definitions of "extreme precipitation event" or "high-temperature fluctuation event" generated by the system will be recorded as an "extreme climate event node". Each node contains the following fields: event type, start and end dates, multiple or amplitude exceeding the benchmark value, and the category of soil parameters corresponding to the impact (such as humidity, organic matter content, etc.), forming a "climate change and extreme weather marking set".

[0038] This set serves as an additional input variable for the soil quality dynamic simulation model, assisting the model training phase in identifying abnormal disturbance responses in soil quality changes and improving the model's generalization ability under conditions of climate change.

[0039] Data fusion mechanism and parameter optimization path Finally, the climate factor set and the standardized soil parameter set are input into the neural network model, and the climate input channel is expanded in the input layer of the network to form a time-synchronized input matrix. The model will learn the interaction between soil response and climate events, optimize the weight distribution through the back-propagation mechanism, and enable the model to adapt to the nonlinear effects brought about by extreme weather disturbances.

[0040] Through this integrated approach, not only can the scenario completeness of soil quality simulation be improved, but climate risk factors that may lead to soil degradation can also be identified in advance, providing more accurate and dynamic forecasting support for agricultural management.

[0041] The precipitation anomaly threshold (twice the historical average for the same period) is the standard for judging extreme precipitation events. The daily precipitation cumulative data of the target area in the same period of the previous ten years (for example, the seven days including May 10) is averaged. If the total precipitation in the same period of the current year exceeds twice this average, it is marked as "extreme precipitation". "Twice the historical average" can reflect the significant impact of sudden heavy precipitation on soil permeability and fertilizer loss.

[0042] The temperature fluctuation threshold (daily average temperature difference exceeds 10°C) is an indicator used to identify strong temperature fluctuations. The difference between the highest and lowest temperatures of each day is calculated to obtain the "daily temperature difference". If the average daily temperature difference for seven consecutive days exceeds 10 degrees Celsius, it is marked as a "high temperature fluctuation event". A large amount of agricultural meteorological data shows that a continuous temperature difference of 10°C will significantly interfere with the surface water evaporation rate, microbial activity and fertility absorption, which is the standard limit value for abnormal climate fluctuations in the agricultural field.

[0043] This implementation method effectively identifies and labels extreme climate events (such as rainstorms and drastic temperature fluctuations) by introducing historical meteorological data and remote sensing observation data, combined with a standardized set of soil parameters, and enhances the stability and prediction accuracy of the soil quality dynamic diagnosis model in response to sudden environmental disturbances. By setting scientific climate threshold standards, potential risks can be identified in advance, effectively improving the timeliness of early warnings and the pertinence of agricultural management, while also having good adaptability and regional promotion capabilities.

[0044] Generating dynamic simulation data of soil quality based on a standardized set of soil parameters also includes obtaining agricultural management records such as tillage methods, rotation cycles, fertilization frequencies, and irrigation patterns from a farmland management database, constructing a numerical coding system for tillage methods, and analyzing historical records to determine the quantitative weight coefficients of different management measures, generating a set of agricultural management quantitative parameters for subsequent soil quality impact assessment.

[0045] Based on the soil quality diagnosis model established with remote sensing and sensor networks, this embodiment further incorporates agricultural management factors. By structuring the coding and modeling of historical farming behaviors, the model's ability to simulate the impact of human farming interventions is enhanced. First, the system extracts multiple agricultural management parameters from the farmland management database, including tillage methods, rotation cycles, fertilization frequencies, and irrigation patterns. For ease of modeling, these management information is uniformly encoded as numerical values. For example, in tillage methods, no-till is encoded as 1, rotary tillage as 2, and deep tillage as 3; rotation cycles are classified by years, such as one-year rotation encoded as 1, two to three years as 2, and four years or more as 3; fertilization frequencies are divided into 1 time, 2 to 3 times, and more than 4 times per year, corresponding to encodings 1, 2, and 3; irrigation patterns are also encoded according to whether it is manual, automatic, or no irrigation. Then, the system uses a large amount of existing historical arable land sample data to analyze the actual impact of the above management behaviors on the improvement of soil quality. The analysis process uses regression modeling techniques, that is, by comparing the relationship between the management behavior encoding and the change in soil quality, calculating the positive or negative impact degree of each management method on soil improvement, and thus obtaining the corresponding impact weight. For example, if historical data shows that the improvement effect of soil organic matter content in arable land managed by no-till is significantly better than that of deep tillage, the system will assign a higher positive impact weight to no-till. The system multiplies the current management behavior encoding of each arable land by the weight obtained from the above regression model and sums them to generate a comprehensive score value, called the agricultural management score. This score value reflects the potential impact intensity of the current management strategy on soil quality. To keep the scale of this parameter consistent with other input features, the system also standardizes this score value, that is, subtracting the historical mean and then dividing by the standard deviation, thus converting it into a dimensionless standardized score. Finally, the agricultural management score, the standardized set of soil physical and chemical index processed previously, and the climate change marker set together constitute a complete input feature set, which is input into the deep prediction model. The model can comprehensively evaluate the evolution trend of future soil quality based on the current soil state, external climate conditions, and management behaviors of the arable land, providing prediction support that is more in line with the actual farming conditions. The definition of management behavior categories is based on agricultural technical standards and local agricultural bureau guidelines, the weight coefficients are obtained through training with historical plot monitoring data over 5 years, and the model accuracy is comprehensively evaluated through cross-validation, mean square error, and determination coefficient to ensure the interpretability and prediction effectiveness of the management score parameters.

[0046] In this embodiment, by introducing agricultural management records, management factors such as tillage methods, fertilization, irrigation, and crop rotation are structurally encoded and quantitatively modeled, enhancing the agronomic relevance of the soil quality prediction model. By using regression analysis to extract the influence weights of different management measures, the model can accurately reflect the long-term impact of human tillage behavior on soil evolution, improving the scientificity, pertinence, and management guidance value of the simulation. At the same time, the integration application of agricultural data and remote sensing soil information is realized, expanding the application depth of the system in the field of precision agriculture.

[0047] Generating soil quality dynamic simulation data based on the standardized soil parameter set further includes using the standardized soil parameter set and the climate change and extreme weather marker set, and constructing an association model between climate change factors and soil physical and chemical properties by using the multiple regression analysis method. If the monthly change rate of precipitation exceeds the preset threshold of 30%, the soil moisture influence coefficient is adjusted. If the daily average temperature fluctuation range exceeds 10 degrees Celsius, the organic matter decomposition rate parameter is corrected. If an extreme weather marker is detected, a short-term impact factor is added to generate a climate-soil coupling association matrix.

[0048] In this embodiment, in order to further improve the accuracy and flexibility of soil quality prediction, the system constructs a dynamic climate-soil coupling association matrix by introducing climate change factors and combining the standardized soil parameter set with climate change data. The model can adjust the key parameters of soil quality according to real-time climate change and extreme weather conditions, so as to achieve more accurate dynamic simulation of soil quality.

[0049] Data preprocessing and input feature construction First, the standardized soil parameter set used by the system and the climate change marker set together constitute the input features. These parameters include: key physical and chemical indicators such as soil temperature, humidity, pH value, organic matter content, nitrogen, phosphorus, and potassium concentration; climate change factors, including the monthly change rate of precipitation, the daily average temperature fluctuation range, etc.; the extreme weather marker set, which marks the occurrence time of events such as extreme precipitation and temperature fluctuation.

[0050] Before all these data are input into the regression analysis model, they are standardized to ensure that they have the same dimension and scale.

[0051] Multiple regression analysis and association model construction The system uses the multiple regression analysis method to construct an association model between climate change factors and soil physical and chemical properties. Specifically, the goal of the regression model is to establish a mathematical relationship between climate factors (such as precipitation, temperature fluctuation) and soil properties (such as humidity, organic matter content, etc.). The form of the regression model is: ; Where, represents the predicted value of soil quality (such as humidity, organic matter, etc.), represents climate-related variables (such as precipitation, temperature fluctuations, etc.), is the regression coefficient, representing the weight of the influence of each climate factor on soil properties, is the error term in the regression model.

[0052] The regression coefficient is obtained through training with historical data and is used to deduce the specific impact of climate change on soil parameters.

[0053] Correction and adjustment mechanism In the process of generating the climate-soil coupling correlation matrix, the system sets multiple thresholds to correct the parameters in the soil quality model to cope with the immediate impact brought by climate change. The specific correction mechanism is as follows: Monthly change rate correction of precipitation: If the monthly change rate of precipitation exceeds the preset threshold (30%), that is, there is a large change in precipitation compared with the previous month, the system will automatically adjust the influence coefficient of soil humidity. Specifically, if the precipitation increases, the soil humidity will increase accordingly, so the humidity influence coefficient needs to be corrected positively according to the changed precipitation, and vice versa.

[0054] Daily average temperature fluctuation correction: If the fluctuation range of the daily average temperature exceeds the set threshold (10°C), that is, the temperature changes violently, the system will correct the parameters of the soil organic matter decomposition rate. A large temperature fluctuation usually leads to a change in the soil organic matter decomposition rate, and the system will adjust the predicted value of the organic matter decomposition rate to reflect the impact of this temperature fluctuation on soil microbial activities.

[0055] Extreme weather mark correction: When the system detects extreme weather marks (such as extreme precipitation or temperature fluctuations), the system will attach a short-term shock factor to immediately correct the change trend of soil quality. For example, extreme precipitation may cause soil erosion or nutrient loss, and the system will increase the sensitivity of soil quality change in the short term and adjust the prediction results of soil texture and nutrient status.

[0056] Generating the climate-soil coupling correlation matrix Finally, all climate factors, soil parameters, and the corrected influence coefficients are combined into a climate-soil coupling correlation matrix. This matrix describes the coupling relationship between climate factors and soil physical and chemical properties, and can dynamically adjust the soil quality prediction results under different climate scenarios, so as to provide more accurate decision-making support for agricultural managers.

[0057] In this embodiment, by introducing climate change factors and extreme weather event markers, the dynamic relationship between climate factors and soil quality is closely combined, thereby improving the accuracy and real-time performance of soil quality prediction. The impact of climate change on soil quality, especially the short-term impact under extreme weather conditions, is reflected by modifying key parameters such as soil moisture and the rate of organic matter decomposition, enabling the model to respond promptly to environmental changes and ensuring more accurate and reliable prediction results. This climate-soil coupling correlation model not only enhances the model's adaptability to climate disturbances but also provides a more accurate climate impact assessment for agricultural management, contributing to improving the risk resistance of agricultural production and optimizing farming management and decision-making processes.

[0058] Generating soil quality dynamic simulation data based on the standardized soil parameter set also includes constructing a cumulative effect model of tillage methods on soil quality using time series analysis methods based on the agricultural management quantification parameter set, calculating the cumulative impact weights of different tillage measures on monthly, quarterly, and annual time scales according to historical management records, and generating a long-term effect prediction model for agricultural management for multi-factor comprehensive simulation.

[0059] In this embodiment, to improve the accuracy and adaptability of soil quality dynamic simulation, the system introduces a cumulative effect model of tillage methods based on the agricultural management quantification parameter set and uses time series analysis methods to capture and predict the long-term cumulative impacts of different tillage measures on monthly, quarterly, and annual time scales. This method helps the system better understand the long-term impacts of different tillage measures on soil quality and further optimize soil quality prediction through multi-factor comprehensive simulation.

[0060] First, the system obtains agricultural management parameters including tillage methods, fertilization frequency, rotation cycle, irrigation mode, etc. from the farmland management database and constructs an agricultural management quantification parameter set based on these parameters. Each management behavior (such as tillage methods like no-till, rotary tillage, deep tillage, etc.) is quantified into a numerical code, and these codes are recorded in the system over time to form a time series data set, reflecting the continuous impacts of agricultural management measures on different time scales (such as monthly, quarterly, annual).

[0061] To model the cumulative effects of tillage methods, the system uses time series analysis methods such as autoregressive models or exponential smoothing models to study the long-term impacts of each tillage method on soil quality. These tillage methods have different degrees of impact on soil quality changes through certain weight coefficients. Time series analysis methods can help the system analyze and capture the cumulative effects of tillage behaviors on different time scales. For example, on a monthly time scale, the impact of tillage methods on soil quality is relatively short-term, while on a quarterly or annual time scale, the impact of tillage methods is more significant and lasts longer.

[0062] The system calculates the cumulative impact of each farming practice on different time scales based on historical data. For example, the system evaluates how management practices each month, each quarter, and each year affect key soil parameters such as soil moisture, organic matter content, and soil pH. By reviewing historical data, the system analyzes the long-term effects of these farming methods on soil quality and generates impact weights. These weight coefficients help the system reflect the long-term impact of different farming methods on soil quality at different time scales. After calculating the monthly, quarterly, and annual impacts of each farming method, the system integrates this data into a long-term effect prediction model for agricultural management. This model can predict how future changes in farming methods will affect soil quality, especially the cumulative effects of these management practices at different time scales.

[0063] In addition, the system also combines the quantified parameters of agricultural management with other soil quality impact factors (such as climate change, soil physical and chemical properties, etc.) for multi-factor comprehensive simulation, further enhancing the comprehensiveness and accuracy of soil quality prediction. This multi-factor comprehensive simulation can help agricultural managers better understand the combined impact of different agricultural management measures, climate change, and soil characteristics on soil quality, and thus provide more accurate decision-making support for agricultural management.

[0064] In this embodiment, by combining the quantified parameters of agricultural management with time series analysis, a cumulative effect model of farming methods on soil quality is constructed, which can comprehensively capture the impact of agricultural management behaviors on different time scales. Through monthly, quarterly, and annual time scale analysis, this method makes soil quality prediction more refined and long-term, and can accurately simulate the cumulative impact of different farming measures on soil quality, thereby improving the operability and scientific nature of soil quality management. Combining multiple factors such as agricultural management and climate change, the system can provide more reliable and long-term decision-making support for agricultural managers, helping to enhance the sustainability and efficiency of agricultural production.

[0065] Generating soil quality dynamic simulation data according to the standardized soil parameter set also includes integrating the climate-soil coupling correlation matrix and the long-term effect prediction model of agricultural management, constructing a soil quality dynamic simulation framework with multi-factor interaction using neural network algorithms. If an extreme weather mark is detected, the emergency response mechanism is activated to adjust the weight of the soil moisture and organic matter decomposition rate parameters, and generate soil quality dynamic simulation data.

[0066] In this embodiment, the system will integrate the climate-soil coupling correlation matrix and the long-term effect prediction model of agricultural management, and use the neural network algorithm to construct a dynamic simulation framework for soil quality with multi-factor interactions. This framework can comprehensively consider the interactions among multiple factors such as climate, soil properties, and agricultural management measures to predict the changing trend of soil quality in real time and accurately. In addition, when extreme weather events are detected, the system will activate the emergency response mechanism, and further optimize the simulation results by adjusting the weights of key soil quality parameters (such as soil moisture and organic matter decomposition rate) to cope with the impact of climate anomalies on soil quality.

[0067] Integrate the climate-soil coupling correlation matrix and the long-term effect prediction model of agricultural management First, the system combines the climate-soil coupling correlation matrix with the long-term effect prediction model of agricultural management. The climate-soil coupling correlation matrix is obtained by modeling the correlation between climate change factors (such as precipitation, temperature fluctuations, etc.) and soil physical and chemical properties (such as moisture, organic matter content, etc.). The long-term effect prediction model of agricultural management analyzes the long-term impact of agricultural management measures such as tillage methods, fertilization frequencies, and rotation cycles on soil quality.

[0068] These two models are integrated to form a multi-dimensional input feature set, which contains dynamic change information of various factors such as climate, soil, and agricultural management. This feature set will provide a multi-level and all-round basis for soil quality prediction by the neural network.

[0069] Use the neural network algorithm to construct a dynamic simulation framework for soil quality with multi-factor interactions After integrating the climate-soil coupling correlation matrix and the long-term effect prediction model of agricultural management, the system uses the neural network algorithm to construct a dynamic simulation framework for soil quality with multi-factor interactions. The neural network model will receive data inputs from multiple sources, including soil parameters, climate change data, agricultural management behaviors, etc., and learn the complex interaction relationships among these factors through architectures such as multi-layer perceptrons or recurrent neural networks.

[0070] The training process of the neural network includes the following steps: Input layer: Receive standardized soil parameters, climate factors, agricultural management data, etc.; Hidden layer: Through the calculations of multiple layers of neurons, simulate the interaction and coupling effects among different factors; Output layer: Generate dynamic simulation data of soil quality, including predicted values such as soil moisture, organic matter content, nitrogen, phosphorus, and potassium concentrations.

[0071] Through the backpropagation algorithm, the neural network continuously optimizes the weight coefficients to minimize the error between the actual observed values and the predicted values, thereby achieving efficient and accurate dynamic simulation of soil quality.

[0072] Extreme weather emergency response mechanism When the system detects extreme weather events (such as heavy rain, extreme temperature fluctuations, etc.), the model will automatically activate the emergency response mechanism. Extreme weather events usually have a sudden impact on soil quality, especially on changes in soil moisture and the rate of organic matter decomposition. Therefore, the system will adjust the weights of these parameters to respond in a timely manner to the impact of climate anomalies.

[0073] Specifically, if the precipitation anomaly increases or the daily temperature difference fluctuation exceeds the preset threshold, the system will make adjustments in the following ways: Soil moisture adjustment: In the case of extreme precipitation, the system will increase the influence weight of soil moisture to reflect the short-term impact of increased soil moisture on soil quality; Organic matter decomposition rate adjustment: If the temperature fluctuates greatly, the system will adjust the weight of the organic matter decomposition rate parameter to increase the impact of temperature changes on microbial activity and organic matter decomposition.

[0074] These adjustments will take effect in real time in the model, so as to ensure that the soil quality simulation data can reflect the real changes in the current environment.

[0075] Generate dynamic simulation data of soil quality Through the integrated multi-factor interaction and emergency response mechanism, the dynamic simulation data of soil quality finally generated by the system can comprehensively reflect the impact of different factors on soil quality and provide an accurate prediction of the soil quality change trend. These data can not only help agricultural managers understand the soil status in real time, but also provide a scientific basis for future soil management strategies.

[0076] In this embodiment, by integrating the climate-soil coupling correlation matrix with the long-term effect prediction model of agricultural management, a dynamic simulation framework of soil quality with multi-factor interaction is constructed, which can more comprehensively and accurately predict the changes in soil quality. The neural network algorithm enables the system to identify the complex relationships among climate, soil and agricultural management factors, thus improving the accuracy of soil quality prediction. The introduction of the emergency response mechanism enables the system to timely adjust key parameters (such as soil moisture and organic matter decomposition rate) when extreme weather events occur, thus avoiding the impact of climate anomalies on the soil quality simulation results. Through this method, agricultural managers can obtain more accurate and real-time soil quality prediction data, providing more scientific and reliable support for decision-making and enhancing the sustainability and risk resistance of agricultural production.

[0077] The specific formula for conducting uncertainty analysis is: ; Among them, represents the predicted soil quality evolution function, represents the number of sampling times, represents the entropy weight factor corresponding to the represents a non - linear model function for soil state assessment, represents the th group of input data sets, represents the model structure parameters, represents the perturbation sensitivity coefficient corresponding to the th sample, represents the variance of the input variable set of the th sample,

[0078] By fusing multi - source remote sensing data and sample data and adopting uncertainty analysis techniques, the dynamic inversion and prediction of cultivated land soil quality are carried out. First, the method combines data from multiple sensors, evaluates the soil state through a non - linear model, and adjusts it according to the weight and influence degree of each sample. Through the attenuation function, the system considers the attenuation effect of soil quality over time and adjusts the model according to the sensitivity of different input changes to more accurately predict the change of soil quality. This process combines multi - source data and sample information, uses uncertainty analysis to improve the accuracy of soil quality prediction, and enhances the adaptability and reliability of the model. Finally, the generated dynamic simulation data of soil quality can provide more accurate decision - making support for agricultural management, helping to effectively respond to the spatio - temporal changes of soil quality and the influence of external factors.

[0079] Utilizing the synergistic effect of multi - source data and uncertainty analysis effectively improves the accuracy and reliability of soil quality prediction, provides strong technical support for precision agriculture, and has better adaptability and operability, especially in the face of complex soil quality changes and extreme climate changes.

[0080] The calculation formula of is: where, represents the entropy weight factor corresponding to the th sample, represents the weighted normalization factor of all samples, represents the input variable dimension, represents the th sample's fuzzy membership degree on the th index, represents the th sample's normalized value on the th index, represents the average value of the th index in the whole sample, represents the sample number, represents the dimension number index.

[0081] By calculating the weight coefficient of each sample, its contribution to the soil quality prediction model is dynamically adjusted. First, the system evaluates each sample based on its performance in different soil quality-related indicators, and calculates the weight coefficient by combining the standardized value of each sample in each indicator with the average value of all samples. Specifically, by weighting the fuzzy coefficients of each sample in multiple dimensions, considering its relative importance in each indicator. During the calculation process, the system uses a normalization factor to standardize the samples to ensure that all samples have a consistent scale when weighted. Finally, the weight coefficient of the sample reflects its influence on the overall soil quality prediction result, and can dynamically adjust the weight according to the performance of the sample in different indicators, thereby improving the accuracy and adaptability of the model prediction. In this way, the system can more accurately evaluate the contribution of different samples to soil quality and achieve a more reasonable weight allocation in soil quality prediction.

[0082] The calculation formula is: ; Where represents the perturbation sensitivity coefficient corresponding to the th sample, represents the total number of soil input characteristics, represents the dimension number of the input variable being traversed currently, represents the partial derivative of the model function B with respect to the input variable , represents the input variable, represents the th sample, represents the fluctuation degree of the th input variable for the th sample,

[0083] The dynamic changes in soil quality are evaluated by calculating the sensitivity of each sample to perturbations. Specifically, the system evaluates the sensitivity of each sample under different characteristic changes by analyzing the response of the soil model to different input characteristics. First, the system calculates the partial derivative of each sample with respect to different input characteristics for each sample, which reflects the influence degree of the input characteristics on soil quality prediction. Then, by combining the relative change amplitude of the sample in these characteristics with the maximum fluctuation range, its contribution to the overall prediction is adjusted. In this way, the system can identify which input characteristics have a greater impact on soil quality changes, and adjust the weight of the model according to the sensitivity of these characteristics, thereby improving the accuracy and adaptability of soil quality prediction. Finally, the system comprehensively considers the sensitivity of each sample to dynamically adjust the input of the soil quality simulation model to ensure that the model can accurately reflect the impact of different input changes on soil quality.

[0084] The early warning mechanism uses color - level classification to display the risk level, which is divided into three levels: normal, mild warning, and severe warning. And the early warning results are synchronized to the plot view of the agricultural management platform in the form of a graphical interface.

[0085] First, the arable land soil quality is monitored in real - time through multi - source remote sensing and ground sensor networks, collecting key soil parameters (such as humidity, organic matter content, pH value, nitrogen - phosphorus - potassium concentration, etc.) and climate change data (such as temperature, precipitation, etc.), and then performing data analysis and processing. Based on these data, the system calculates the change trend of soil quality and evaluates the soil health status.

[0086] During the analysis process, the system evaluates the change of soil quality according to the set thresholds. According to the change range of soil quality, the system classifies the risk levels into three categories: Normal: When the soil quality is in a stable state or the change range is within the safe range, the system determines it as normal and displays green, indicating that the soil quality is healthy and no special measures are required.

[0087] Mild warning: When there are slight changes in soil quality and these changes may pose potential risks to agricultural production, the system issues a mild warning and displays yellow, prompting agricultural managers to pay attention to the soil conditions of this plot and may need to take appropriate management measures, such as applying fertilizers in a timely manner or adjusting the irrigation method.

[0088] Severe warning: When there is a significant decline in soil quality or major risks (such as a substantial reduction in soil fertility or acidification), the system will issue a severe warning and display red, reminding agricultural managers to immediately take effective measures, such as adjusting the tillage method, strengthening fertilization management, or conducting soil remediation, to avoid long - term damage to crop yields and soil ecology.

[0089] All early warning results are real - time synchronized to the plot view of the agricultural management platform through a graphical interface. This view shows the risk status of each plot, enabling agricultural managers to clearly and intuitively view the soil quality conditions of each plot and formulate corresponding countermeasures for different risk levels. The system may also provide specific operation suggestions for different risk levels, such as soil improvement suggestions for mild warnings and emergency repair plans for severe warnings.

[0090] This early warning mechanism simplifies the complexity of soil quality monitoring through color - level classification display, enabling agricultural managers to quickly identify potential risks, make decisions in a timely manner, thereby improving soil management efficiency and ensuring the sustainability of agricultural production.

[0091] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A dynamic diagnosis method for cultivated land soil quality based on multi-source remote sensing collaborative inversion, characterized in that The method comprises: Through a multi-source sensor network, physical and chemical parameters such as soil temperature, humidity, pH, organic matter content, and nitrogen, phosphorus, and potassium concentrations are collected in real time. Data preprocessing algorithms are used to remove outliers and perform standardization to generate a standardized soil parameter set for subsequent climate and management factor correlation analysis. Generate soil quality dynamic simulation data based on the standardized soil parameter set, conduct uncertainty analysis on the soil quality dynamic simulation data, generate one thousand sets of parameter combinations through random sampling, calculate the probability distribution interval of soil quality change, and if the prediction accuracy based on the ratio of mean square error to actual observation value is lower than the preset threshold of 85%, return to the previous link to readjust the weight parameters of each factor in the neural network framework to generate soil quality change trend prediction data; Based on the predicted data on soil quality change trends, an early warning mechanism is established. If the predicted value of the soil fertility level based on the comprehensive nitrogen, phosphorus and potassium index is lower than the preset safety threshold, which is 80% of the historical average, for three consecutive months, or the structural stability index drops by more than 20% of the preset threshold, a risk warning signal will be automatically generated and pushed to the agricultural management decision-making platform, completing the full process analysis from data collection to decision support.

2. The dynamic diagnosis method for cultivated land soil quality based on multi-source remote sensing collaborative inversion according to claim 1, wherein: Generating soil quality dynamic simulation data according to the standardized soil parameter set includes: For the standardized soil parameter set, combined with the historical records of meteorological stations and satellite remote sensing data, we obtain climate change indicators such as precipitation, temperature fluctuations, humidity changes and wind speed. If the precipitation for seven consecutive days exceeds twice the historical average value for the same period calculated based on the local ten-year average, or the average daily temperature fluctuation exceeds the preset threshold of ten degrees Celsius, it is judged as an extreme weather event and the time node is marked to generate a climate change and extreme weather mark set.

3. The dynamic diagnosis method of cultivated land soil quality based on multi-source remote sensing collaborative inversion according to claim 2, characterized in that: The step of generating soil quality dynamic simulation data according to the standardized soil parameter set further comprises: Agricultural management records of tillage methods, crop rotation cycles, fertilization frequency and irrigation patterns are obtained from the farmland management database, and a numerical coding system for tillage methods is constructed. The quantitative weight coefficients of different management measures are determined based on historical record analysis, and a set of agricultural management quantitative parameters is generated for subsequent soil quality impact assessment.

4. The dynamic diagnosis method for cultivated land soil quality based on multi-source remote sensing collaborative inversion according to claim 3, characterized in that: The step of generating soil quality dynamic simulation data according to the standardized soil parameter set further comprises: Using a set of standardized soil parameters and a set of climate change and extreme weather markers, a multivariate regression analysis method was used to construct a correlation model between climate change factors and soil physical and chemical properties. If the monthly precipitation change rate exceeded the preset threshold of 30%, the soil moisture influence coefficient was adjusted. If the daily average temperature fluctuation exceeded ten degrees Celsius, the organic matter decomposition rate parameter was corrected. If extreme weather markers were detected, short-term impact factors were added to generate a climate-soil coupling correlation matrix.

5. The dynamic diagnosis method for cultivated land soil quality based on multi-source remote sensing collaborative inversion according to claim 4, characterized in that: The step of generating soil quality dynamic simulation data according to the standardized soil parameter set further comprises: Based on the agricultural management quantitative parameter set, apply the time series analysis method to construct a cumulative effect model of tillage methods on soil quality, calculate the cumulative impact weights of different tillage measures on monthly, quarterly, and annual time scales according to historical management records, and generate a long-term effect prediction model of agricultural management for multi-factor comprehensive simulation.

6. The dynamic diagnosis method of cultivated land soil quality based on multi-source remote sensing collaborative inversion according to claim 5, characterized in that: The generation of soil quality dynamic simulation data based on the standardized soil parameter set further includes: Fuse the climate-soil coupling correlation matrix and the long-term effect prediction model of agricultural management, use the neural network algorithm to construct a dynamic simulation framework of soil quality with multi-factor interaction. If an extreme weather mark is detected, activate the emergency response mechanism, adjust the weights of soil moisture and organic matter decomposition rate parameters, and generate soil quality dynamic simulation data.

7. The dynamic diagnosis method for cultivated land soil quality based on multi-source remote sensing collaborative inversion according to claim 1, characterized in that: The specific formula for the uncertainty analysis is as follows: ； Among them, represents the predicted soil quality evolution function, represents the number of sampling times, represents the entropy weight factor corresponding to the nonlinear model function for soil state evaluation, represents the input data set of the group, represents the model structure parameter, represents the perturbation sensitivity coefficient corresponding to the represents the variance of the input variable set of the sample number.

8. The dynamic diagnosis method for cultivated land soil quality based on multi-source remote sensing collaborative inversion according to claim 7, characterized in that: The said The calculation formula is as follows: ； Among them, represents the entropy weight factor corresponding to the th sample, represents the weighted normalization factor of all samples, represents the dimension of the input variable, represents the th sample's fuzzy membership degree on the th index, represents the th sample's normalized value on the th index, represents the average value of the th index in the full sample, represents the sample number, represents the dimension number index.

9. The dynamic diagnosis method for cultivated land soil quality based on multi-source remote sensing collaborative inversion according to claim 7, characterized in that: The said The calculation formula is as follows: ； Among them, represents the perturbation sensitivity coefficient corresponding to the th sample, represents the total number of soil input features, represents the dimension number of the input variable being traversed currently, represents the partial derivative of the model function B with respect to the input variable , represents the input variable, represents the th sample's degree of fluctuation on the th input variable, represents the maximum fluctuation value of all samples on the th variable.

10. The dynamic diagnosis method for cultivated land soil quality based on multi-source remote sensing collaborative inversion according to claim 1, characterized in that: The early warning mechanism uses color grade classification to display the risk level, which is divided into three levels: normal, mild warning, and severe warning, and synchronizes the warning results to the plot view of the agricultural management platform in the form of a graphical interface.

Images