Early warning method and system for soil secondary salinization degree

Abstract

The invention discloses a soil secondary salinization degree early warning method and system, and the method comprises the steps: obtaining and preprocessing soil multi-source heterogeneous data, and obtaining driving factor data; based on the driving factor data, screening key driving factors of the soil secondary salinization key driving factors by adopting a feature importance evaluation method; analyzing the key driving factor by adopting a correlation analysis method, determining a salinization threshold of the driving factor by adopting a threshold segmentation technology according to an analysis result, if the driving factor exceeds the salinization threshold, judging that a salinization risk exists, and establishing a salinization early warning model through a random forest algorithm, the salinization development situation is simulated, the influence degree is evaluated, and an evaluation result is obtained; and constructing an early warning information issuing platform, uploading an evaluation result to the early warning information issuing platform in real time, carrying out visual display on the risk level and the key driving factor, and pushing early warning information through a mobile terminal. According to the invention, precise early warning of soil secondary salinization is realized.

Description

Technical Field

[0001] The invention belongs to the field of information technology, and in particular relates to a method and system for early warning of soil secondary salinization degree. Background Art

[0002] Soil secondary salinization is a complex dynamic process involving water and salt migration, changes in soil physical and chemical properties, and other aspects. To achieve early warning of salinization, it is first necessary to obtain the spatial distribution pattern of the current status of salinization. However, due to the high heterogeneity of soil salinity, it is difficult to fully reflect the salinization status of the entire region through conventional point sampling. At the same time, there are many factors that affect soil salinization, such as groundwater level, irrigation methods, crop types, meteorological conditions, etc. These factors often have strong spatiotemporal variability, resulting in obvious stage and regional differences in the occurrence of salinization. In addition, some saline soil remediation measures (such as improvement and flushing) will also change the soil salinity status within a certain period of time, bringing complexity to early warning. The acquisition of salinization monitoring data also faces the problem of scale conversion. Satellite remote sensing can provide soil salinity information in a large area, but its inversion accuracy is limited by spatial resolution and temporal distribution; although ground monitoring has high accuracy, the data obtained is relatively discrete in space. The effective fusion of monitoring data at different scales is the key to the analysis of spatiotemporal characteristics of salinization. In addition, due to the lack of long-term soil salinity dynamic monitoring data, there is still a lack of reliable data support for the prediction of salinization development trends. In short, the heterogeneity of soil secondary salinization, complex driving mechanisms, and multi-source data fusion have brought great challenges to the accurate monitoring, characterization, and evolution law analysis of salinization, which are key technical issues that need to be solved urgently. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a method and system for early warning of the degree of secondary salinization of soil. Among them, a method for early warning of the degree of secondary salinization of soil includes:

[0004] Obtain soil multi-source heterogeneous data and preprocess them to obtain preprocessed driving factor data;

[0005] Based on the preprocessed driving factor data, the key driving factors of the key driving factors of soil secondary salinization are screened by using the characteristic importance evaluation method;

[0006] The key driving factors are analyzed by correlation analysis method, and the salinization threshold of the driving factors is determined by threshold segmentation technology according to the analysis results. If the driving factors exceed the salinization threshold, it is judged that there is a salinization risk. A salinization early warning model is established by random forest algorithm, the development trend of salinization is simulated, the impact degree is evaluated, and the evaluation results are obtained;

[0007] Construct an early warning information release platform, upload the assessment results to the early warning information release platform in real time, visualize the risk level and key driving factors, and push early warning information through mobile terminals.

[0008] Preferably, the process of acquiring multi-source heterogeneous soil data and preprocessing to obtain preprocessed driving factor data includes:

[0009] Obtain multi-source heterogeneous soil data, and based on the key driving factors of soil secondary salinization, obtain the time scale and spatial scale of the key driving factors;

[0010] Preprocessing the soil multi-source heterogeneous data using a data standardization method to obtain standardized data;

[0011] Based on the standardized data, the key driving factor characteristics of soil secondary salinization are extracted to obtain preprocessed driving factor data.

[0012] Preferably, the soil multi-source heterogeneous data includes but is not limited to soil conductivity, groundwater level, and meteorological conditions;

[0013] The key driving factors of soil secondary salinization include but are not limited to soil salinity, groundwater mineralization, and precipitation;

[0014] The process of preprocessing the soil multi-source heterogeneous data by using a data standardization method to obtain standardized data includes:

[0015] The maximum and minimum normalization method is used to preprocess the soil multi-source heterogeneous data, and the data are uniformly mapped to the interval [0,1] to obtain preprocessed standardized data.

[0016] Preferably, the process of extracting the key driving factor characteristics of soil secondary salinization based on the standardized data and obtaining the preprocessed driving factor data includes:

[0017] The principal component analysis method is used to reduce the dimension of the key driving factor characteristics to obtain the main driving factors;

[0018] According to the main driving factors, a machine learning prediction model for soil secondary salinization is constructed, and a support vector machine algorithm is used for training to obtain a trained prediction model;

[0019] By using the trained prediction model and combining the temporal and spatial variation trends of the main driving factors, the development trend of soil secondary salinization in the future is predicted to obtain a prediction result;

[0020] According to the prediction results, data visualization technology is used to generate a spatiotemporal distribution map of soil secondary salinization.

[0021] Preferably, based on the preprocessed driving factor data, the process of screening the key driving factors of the key driving factors of soil secondary salinization by using the feature importance evaluation method comprises:

[0022] Based on the preprocessed driving factor data, a random forest algorithm is used to evaluate the characteristic importance of the driving factor;

[0023] A feature selection criterion function is constructed, and a sequential forward selection algorithm is used to search for the optimal feature subset from the driving factor set. A soil secondary salinization prediction model is constructed using a support vector machine algorithm, and the optimal feature subset is used as the model input. The preprocessed driving factor data is screened through a feature subset search strategy to obtain key driving factors that have a significant impact on soil secondary salinization.

[0024] Preferably, the process of analyzing the key driving factors using a correlation analysis method includes:

[0025] Based on the data of the key driving factors and the degree of soil salinization, the Granger causality test causal inference method is used to judge the significance of the causal relationship between the key driving factors and the degree of soil salinization; if the significance of the causal relationship is greater than the preset threshold, it is determined that there is a significant causal relationship between the key driving factors and the degree of soil salinization; based on the analysis results of the spatiotemporal distribution characteristics, the interaction network and the causal relationship structural equation, the intrinsic relationship between the key driving factors and the degree of soil salinization is obtained.

[0026] Preferably, a salinization early warning model is established by using a random forest algorithm to simulate the development trend of salinization and evaluate the impact. The process of obtaining the evaluation results includes:

[0027] Based on the salinization early warning model, driving factor change scenarios are set, and the salinization development trends under different scenarios are simulated through scenario analysis. The fuzzy comprehensive evaluation method is used to evaluate the degree of salinization impact of the scenarios, and the optimal early warning strategy is determined in combination with prevention and control costs and ecological benefits.

[0028] Preferably, the process of determining the optimal early warning strategy includes:

[0029] Obtain salinization-related driving factors and establish a driving factor database; set change scenarios according to the driving factors to form a driving factor change scenario matrix;

[0030] The driving factor change scenario matrix is ​​input into the salinization early warning model to obtain the salinization development prediction result; if the prediction result exceeds the preset threshold, a multi-objective optimization algorithm is used to determine the optimal early warning strategy with the goal of minimizing prevention and control costs and maximizing ecological benefits, combined with prevention and control measures under different scenarios; the optimal early warning strategy is input into the salinization early warning model for verification, and the early warning strategy is optimized and adjusted according to the verification results to obtain the final early warning strategy.

[0031] Preferably, the process of constructing an early warning information release platform, uploading the assessment results to the early warning information release platform in real time, visually displaying the risk level and key driving factors, and pushing the early warning information through a mobile terminal includes:

[0032] Acquire early warning information according to the final optimal early warning strategy, and analyze and process the early warning information to obtain salinization risk level and key driving factor data;

[0033] Use WebGIS technology to visualize salinization risk levels and key driving factor data to generate a visual risk map;

[0034] According to user attributes, extract regional risk information related to the user from the visual risk map; obtain the user's location coordinates through the mobile terminal application to determine the salinization risk level of the user's area;

[0035] If the salinization risk level in the user's area exceeds the preset threshold, the early warning information and key driving factor information of the area will be pushed to the user; the visual information of the salinization risk level and key driving factors in the user's area will be displayed in the mobile terminal application; according to the user's interactive operations, the visual information of the salinization risk level and key driving factors will be dynamically updated to provide information query services.

[0036] The present invention also provides a soil secondary salinization degree early warning system, comprising:

[0037] The data preprocessing module is used to obtain and preprocess multi-source heterogeneous soil data to obtain preprocessed driving factor data;

[0038] A feature selection module, for screening key driving factors of key driving factors of soil secondary salinization by using a feature importance evaluation method based on the preprocessed driving factor data;

[0039] A driving factor analysis module is used to analyze the key driving factors by using a correlation analysis method, and determine the salinization threshold of the driving factors by using a threshold segmentation technology according to the analysis results;

[0040] A salinization early warning module is used to determine whether there is a salinization risk according to whether the driving factor exceeds the salinization threshold;

[0041] The scenario analysis module is used to establish a salinization early warning model through the random forest algorithm, simulate the development trend of salinization, evaluate the impact degree, and obtain the evaluation results;

[0042] The early warning information release module is used to build an early warning information release platform, upload the assessment results to the early warning information release platform in real time, visualize the risk level and key driving factors, and push the early warning information through the mobile terminal.

[0043] Compared with the prior art, the present invention has the following advantages and technical effects:

[0044] The present invention discloses a method for early warning of secondary soil salinization. The method first obtains multi-source heterogeneous data, and eliminates the differences in driving factors through data standardization preprocessing. Then, feature importance evaluation and subset search strategies are used to screen out key driving factors. The spatiotemporal distribution characteristics and interaction studies of these factors are performed to reveal their intrinsic mechanisms with the degree of salinization. Based on this, the present invention determines the salinization threshold of the driving factor and establishes a random forest early warning model. Through scenario analysis and fuzzy comprehensive evaluation, the development trend of salinization under different scenarios is evaluated to determine the optimal early warning strategy. Finally, the present invention constructs an early warning information release platform, uses WebGIS technology to realize the visual display of risk levels and key factors, and pushes early warning information through mobile terminals. The present invention realizes accurate early warning of secondary soil salinization, providing strong support for government decision-making and farmer management. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0046] Figure 1 A schematic diagram of a method flow of an embodiment of the present invention;

[0047] Figure 2 Schematic diagram of the system structure of an embodiment of the present invention. DETAILED DESCRIPTION

[0048] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0049] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0050] Embodiment 1

[0051] like Figure 1 As shown, this embodiment provides a method for early warning of soil secondary salinization degree, including:

[0052] Obtain soil multi-source heterogeneous data and preprocess them to obtain preprocessed driving factor data;

[0053] Based on the preprocessed driving factor data, the key driving factors of soil secondary salinization were screened using the feature importance assessment method;

[0054] The key driving factors are analyzed by correlation analysis method. According to the analysis results, the salinization threshold of the driving factors is determined by threshold segmentation technology. If the driving factors exceed the salinization threshold, it is judged that there is a salinization risk. The salinization early warning model is established by random forest algorithm to simulate the development trend of salinization and evaluate the impact degree to obtain the evaluation results.

[0055] Build an early warning information release platform, upload the assessment results to the early warning information release platform in real time, visualize the risk level and key driving factors, and push early warning information through mobile terminals.

[0056] Furthermore, the process of obtaining soil multi-source heterogeneous data and preprocessing it to obtain the preprocessed driving factor data includes:

[0057] Obtain multi-source heterogeneous soil data, and based on the key driving factors of soil secondary salinization, obtain the time scale and spatial scale of the key driving factors; the multi-source heterogeneous soil data include but are not limited to soil conductivity, groundwater level, and meteorological conditions; the key driving factors of soil secondary salinization include but are not limited to soil salt content, groundwater mineralization, and precipitation;

[0058] The data standardization method is used to preprocess the multi-source heterogeneous soil data, eliminate the impact of different data types and quality differences, and obtain the preprocessed standardized data;

[0059] The key driving factor characteristics of soil secondary salinization were extracted based on standardized data to obtain preprocessed driving factor data.

[0060] A further optimization scheme uses data standardization methods to preprocess multi-source heterogeneous soil data. The process of obtaining standardized data includes:

[0061] The maximum and minimum normalization method is used to preprocess the multi-source heterogeneous soil data, and the data are uniformly mapped to the interval [0,1] to obtain the preprocessed standardized data.

[0062] Furthermore, the key driving factor characteristics of soil secondary salinization are extracted based on the standardized data. The process of obtaining the preprocessed driving factor data includes:

[0063] The principal component analysis method is used to reduce the dimension of the key driving factor characteristics and obtain the main driving factors;

[0064] According to the main driving factors, a machine learning prediction model for soil secondary salinization was constructed and trained using the support vector machine algorithm to obtain a trained prediction model.

[0065] Through the trained prediction model, combined with the temporal and spatial variation trends of the main driving factors, the development trend of soil secondary salinization in the future is predicted and the prediction results are obtained;

[0066] According to the prediction results, data visualization technology is used to generate the spatiotemporal distribution map of soil secondary salinization.

[0067] Exemplarily, multi-source heterogeneous data such as soil conductivity, groundwater level, and meteorological conditions are obtained, and the key driving factors of soil secondary salinization, such as soil salinity, groundwater mineralization, and precipitation, are obtained on a daily, monthly, and annual basis, and a spatial scale of 10 meters × 10 meters. The maximum and minimum normalization method is used for preprocessing the multi-source heterogeneous raw data, and the data is uniformly mapped to the [0,1] interval to eliminate the impact of different data types and quality differences, and obtain the preprocessed standardized data. Based on the preprocessed standardized data, the key driving factor characteristics of soil secondary salinization are extracted, including the mean soil salt content of 5g / kg and the coefficient of variation of 3; the mean groundwater mineralization is 0g / L and the coefficient of variation is 4; the mean annual precipitation is 400mm and the coefficient of variation is 2. The principal component analysis method was used to reduce the dimension of the key driving factor characteristics. By calculating the eigenvalues ​​and eigenvectors, the first three principal components were selected, and the cumulative variance contribution rate reached more than 85%. The main driving factors were soil salinity, groundwater mineralization and precipitation. According to the main driving factors, a support vector machine prediction model for soil secondary salinization was constructed. The radial basis kernel function was used, the penalty factor C = 0, the kernel function parameter γ = 5, the ratio of training set to test set was 7:3, and 10-fold cross validation was performed to obtain a trained prediction model with an average accuracy of more than 90%. Using the trained prediction model, combined with the spatiotemporal variation trend of the main driving factors, such as the annual increase of 1g / kg in soil salinity, the annual increase of 2g / L in groundwater mineralization, and the annual decrease of 20mm in annual precipitation, the development trend of soil secondary salinization in the next 10, 20, and 30 years was predicted, and the prediction results showed that the soil secondary salinization in the study area showed a trend of gradual intensification. According to the prediction results, ArcGIS spatial analysis and mapping functions were used to generate spatial distribution maps of soil secondary salinization degrees at different periods, visually displaying its spatiotemporal dynamic changes.

[0068] Furthermore, based on the preprocessed driving factor data, the process of screening the key driving factors of soil secondary salinization using the feature importance evaluation method includes:

[0069] Based on the preprocessed driving factor data, the random forest algorithm is used to evaluate the characteristic importance of the driving factors;

[0070] A feature selection criterion function was constructed, and the sequential forward selection algorithm was used to search for the optimal feature subset from the driving factor set. The support vector machine algorithm was used to construct a soil secondary salinization prediction model, and the optimal feature subset was used as the model input. The preprocessed driving factor data was screened through the feature subset search strategy to obtain the key driving factors that significantly affected soil secondary salinization.

[0071] Exemplarily, the pre-processed data of driving factors such as soil salinity, groundwater mineralization, and precipitation are obtained, and the feature importance of the driving factors is evaluated by the random forest algorithm. The importance score of each driving factor is obtained by calculating the mean impurity reduction (MeanDecrease Impurity), where the soil salinity score is 35, the groundwater mineralization score is 28, and the precipitation score is 22, indicating that soil salinity is the most important driving factor. Construct a feature selection criterion function f(X) = α × accuracy + β × number of features, where α = 8, β = -2, and use the sequential forward selection algorithm to search for the optimal feature subset from all driving factor sets. By iteratively calculating the criterion function value, when the criterion function value reaches the maximum, the optimal feature subset is obtained as {soil salinity, groundwater mineralization, precipitation, soil texture}. A soil secondary salinization prediction model was constructed using the support vector machine algorithm. The optimal feature subset was used as the model input, and the hyperparameters of the support vector machine were optimized by the grid search method. The optimal parameter combination was searched in the parameter space of penalty factor C∈{1,10,100} and kernel function parameter γ∈{0.1,1,10}. The model performance was evaluated using 5-fold cross validation, and the optimized support vector machine prediction model was obtained. The optimal parameters were C=1 and γ=1, and the average accuracy reached 92%, indicating that the model can better predict the occurrence and development of soil secondary salinization.

[0072] Furthermore, the process of analyzing key driving factors using correlation analysis method includes:

[0073] Based on the data of key driving factors and soil salinization degree, the Granger causality test causal inference method is used to judge the significance of the causal relationship between the key driving factors and the degree of soil salinization; if the significance of the causal relationship is greater than the preset threshold, it is determined that there is a significant causal relationship between the key driving factors and the degree of soil salinization; the analysis results of the spatiotemporal distribution characteristics, interaction network and causal relationship structural equation are combined to obtain the intrinsic relationship between the key driving factors and the degree of soil salinization.

[0074] For example, the key driving factors related to soil salinization are screened out from multi-source heterogeneous data, including soil salinity, groundwater mineralization, precipitation, evaporation, groundwater depth, etc. The Granger causality test method is used to perform causal inference analysis on these key driving factors and soil salinization degree data. The significance level α is set to 0.5, and the F statistic and P value of the Granger causality test are calculated to determine the significance of the causal relationship between each driving factor and the degree of soil salinization. The results show that the P values ​​of soil salinity, groundwater mineralization, precipitation and soil salinization are 02, 03, and 04, respectively, which are all less than the significance level of 05. Therefore, it is determined that there is a significant causal relationship between these three key driving factors and the degree of soil salinization. Further analysis of the spatiotemporal distribution characteristics of these key driving factors shows that soil salinity and groundwater mineralization show a decreasing trend from northwest to southeast in space, while precipitation shows the opposite distribution pattern. By constructing the interaction network of driving factors, it is revealed that there is a significant positive correlation between soil salinity and groundwater mineralization, with a correlation coefficient of 78, while precipitation is negatively correlated with both. Finally, the structural equation model is used to quantitatively analyze the influence mechanism of driving factors on soil salinization. The direct path coefficients of soil salinity, groundwater mineralization, and precipitation on soil salinization are 52, 35, and -28, respectively, indicating that soil salinity is the most important factor affecting soil salinization, while precipitation plays a certain mitigating role. Based on the above analysis results, it is revealed that key driving factors affect the occurrence and development of soil salinization through direct and indirect action pathways, providing a scientific basis for the prevention and control of soil salinization.

[0075] Furthermore, based on the analysis of key driving factors, the threshold segmentation technology is used to determine the salinization threshold of the driving factor. If the driving factor exceeds the threshold, it is judged that there is a salinization risk. The salinization early warning model is established through the random forest algorithm, and the feature selection results are used to explain the model early warning mechanism.

[0076] According to the salinization-related data, the key driving factor analysis method is used to determine the key driving factors affecting salinization. For each key driving factor, the salinization threshold of the factor is obtained through the threshold segmentation technology. If the key driving factor value of a certain area exceeds the preset salinization threshold, it is judged that there is a salinization risk in the area. The historical salinization data and key driving factor data are obtained, and the salinization early warning model is trained using the random forest algorithm. During the training process of the random forest model, the driving factors that play a key role in salinization early warning are determined through the feature selection method. Using the results of feature selection, the early warning mechanism of the random forest early warning model is analyzed and explained, revealing the intrinsic connection between the key driving factors and the salinization early warning. The key driving factor data of the new area is input into the trained random forest early warning model to determine whether salinization may occur in the area in the future, thereby realizing salinization early warning.

[0077] Exemplarily, salinization-related data, including soil salinity, groundwater mineralization, precipitation, evaporation, etc., are obtained from multi-source heterogeneous data. The Pearson correlation coefficient method is used to analyze the correlation between each factor and the degree of soil salinization, and factors with correlation coefficients greater than 6 are selected as key driving factors. For each key driving factor, the maximum entropy threshold segmentation algorithm is used to obtain the salinization threshold, for example, the salinization threshold of soil salinity is 5g / kg, and the threshold of groundwater mineralization is 5g / L. If the soil salinity in a certain area exceeds 5g / kg and the groundwater mineralization exceeds 5g / L, it is judged that there is a salinization risk in the area. The historical salinization data and key driving factor data of the study area from 2010 to 2020 are collected, and the random forest algorithm is used to train the salinization early warning model. Feature selection is performed by the average impurity reduction method, and it is determined that soil salinity, groundwater mineralization, and precipitation are the three most critical factors affecting salinization early warning. The analysis found that the higher the soil salinity and groundwater mineralization, and the less precipitation, the greater the probability of salinization. The key driving factor data of the study area in 2021 were input into the trained random forest early warning model, and it was predicted that the probability of salinization in the area within the next year was 65%, and prevention and control measures needed to be strengthened. Comprehensive analysis shows that key driving factor analysis and random forest early warning models can effectively predict salinization risks and provide important decision-making basis for salinization prevention and control.

[0078] A salinization early warning model was established using the random forest algorithm, and the feature selection results were used to explain the model's early warning mechanism.

[0079] Obtain multi-source data related to salinization, such as environment, meteorology, and soil, pre-process high-dimensional data, and remove noise and outliers; use the random forest algorithm to build a salinization early warning model, and improve the generalization ability and robustness of the model through ensemble learning methods; during the model training process, introduce feature selection methods, such as tree-based feature importance evaluation or recursive feature elimination, to screen out key factors that have a greater impact on the early warning results; based on the feature selection results, analyze the intrinsic relationship between key factors and salinization, explain the decision-making mechanism of the early warning model, and enhance the interpretability of the model; optimize and adjust the parameters of the early warning model to improve the early warning accuracy and recall rate, and evaluate the model performance on the test set to ensure that it meets the actual application needs; deploy the early warning model to the production environment, regularly obtain the latest environmental, meteorological, soil and other data, and conduct real-time salinization early warning to provide a basis for prevention and control decisions; continuously collect salinization monitoring data and early warning feedback, iteratively update and optimize the early warning model, continuously improve the early warning performance and adaptability, and support the long-term mechanism of salinization prevention and control.

[0080] For example, salinization-related data were collected from multiple departments such as meteorology, hydrology, and soil, including 50 indicators such as precipitation, evaporation, groundwater level, and soil salinity, forming a data set of up to 1 million records. The median filter and IQR method were used to remove noise and outliers in each indicator, and the data quality was improved by 15%. The salinization early warning model was constructed using the random forest algorithm. Through the integration of 500 decision trees, the average accuracy of the model reached 85%, which was 10% higher than that of the single decision tree model. During the model training process, the tree-based feature importance evaluation method was used to screen out 10 important features such as groundwater mineralization, soil salinity, and precipitation. After dimensionality reduction, the model training speed was increased by 20%. The analysis found that when the groundwater mineralization exceeded 3g / L, the soil salt content exceeded 5g / kg, and the annual precipitation was less than 400mm, the probability of salinization increased significantly, and the early warning mechanism of the model was consistent with the formation mechanism of salinization. The grid search and cross-validation methods were used to optimize the hyperparameters of the random forest model, and the warning accuracy and recall rate reached 80% and 85% respectively, meeting the requirements of practical applications. The model was deployed to the cloud platform, which automatically obtained the latest environmental, meteorological, and soil data every day, and generated a salinization warning map for the next week. It was found that the degree of consistency between the key warning areas and the actual salinization areas reached 90%. After one year of continuous operation, 5,000 salinization monitoring data and 500 warning feedback were collected. The model was fine-tuned based on the new data, and the warning accuracy was further improved by 2%, supporting the normalization of salinization prevention and control.

[0081] Furthermore, a salinization early warning model was established through the random forest algorithm to simulate the development trend of salinization and evaluate the impact. The process of obtaining the evaluation results includes:

[0082] Based on the salinization early warning model, scenarios of driving factor changes were set, and the salinization development trends under different scenarios were simulated through scenario analysis. The fuzzy comprehensive evaluation method was used to evaluate the degree of salinization impact of the scenarios, and the optimal early warning strategy was determined in combination with prevention and control costs and ecological benefits.

[0083] Furthermore, the process of determining the optimal early warning strategy includes:

[0084] Obtain salinization-related driving factors and establish a driving factor database; set change scenarios based on the driving factors to form a driving factor change scenario matrix;

[0085] The scenario matrix of driving factor changes is input into the salinization early warning model to obtain the salinization development forecast result; if the forecast result exceeds the preset threshold, a multi-objective optimization algorithm is used to determine the optimal early warning strategy with the goal of minimizing prevention and control costs and maximizing ecological benefits, combined with prevention and control measures under different scenarios; the optimal early warning strategy is input into the salinization early warning model for verification, and the early warning strategy is optimized and adjusted according to the verification results to obtain the final early warning strategy.

[0086] For example, salinization-related driving factors are obtained from multi-source heterogeneous data, including meteorological data (precipitation, evaporation, temperature, etc.), hydrological data (groundwater level, mineralization, etc.), soil data (soil texture, salt content, etc.), terrain data (elevation, slope, etc.) and remote sensing data (vegetation index, land use, etc.), to establish a driving factor database. For each driving factor, 3-5 change scenarios are set, such as a 10%, 20%, and 30% decrease in precipitation, and a 1m, 2m, and 3m decrease in groundwater level, to form a driving factor change scenario matrix containing 120 scenario combinations. The scenario matrix is ​​input into the salinization early warning model based on support vector machine to obtain the forecast results of salinization development in the next 10 years. If the forecast results show that the salinization area will increase by more than 20%, the multi-objective particle swarm optimization algorithm is used to minimize the prevention and control cost and maximize the ecological benefit as the optimization goal, and the optimal combination is selected from 10 prevention and control measures such as water diversion, afforestation, and engineering management to determine the optimal early warning strategy under different scenarios. The optimal early warning strategy is input into the salinization early warning model for verification. If the salinization area is reduced by less than 15%, the investment in prevention and control measures is appropriately increased according to the results of the sensitivity analysis, and the early warning strategy is optimized and adjusted to finally form an implementable early warning strategy. Through scenario analysis and multi-objective optimization, the early warning model can scientifically evaluate the salinization risk under the conditions of changes in different driving factors, and provide accurate and effective early warning strategies, providing intelligent decision-making support for salinization prevention and control.

[0087] Furthermore, a warning information release platform is constructed, the assessment results are uploaded to the warning information release platform in real time, the risk level and key driving factors are visualized, and the warning information is pushed through the mobile terminal. The process includes:

[0088] Obtain early warning information according to the final optimal early warning strategy, and analyze and process the early warning information to obtain salinization risk level and key driving factor data;

[0089] Use WebGIS technology to visualize salinization risk levels and key driving factor data to generate a visual risk map;

[0090] According to user attributes, regional risk information related to the user is extracted from the visual risk map; the user's location coordinates are obtained through the mobile terminal application to determine the salinization risk level of the user's area;

[0091] If the salinization risk level in the user's area exceeds the preset threshold, the early warning information and key driving factor information of the area will be pushed to the user; the visual information of the salinization risk level and key driving factors in the user's area will be displayed in the mobile terminal application; according to the user's interactive operations, the visual information of the salinization risk level and key driving factors will be dynamically updated to provide information query services.

[0092] For example, according to the early warning information obtained by the early warning strategy, the fuzzy comprehensive evaluation method is used to classify the salinization risk, and the salinization risk level is set to four levels: mild, moderate, severe and extremely severe, corresponding to membership values ​​of 2, 4, 6 and 8 respectively. The key driving factors are screened by the random forest algorithm, and the top five driving factors with the highest contribution rate are selected as the key driving factors, including annual average precipitation, groundwater mineralization, soil salinity, groundwater level depth and slope. The salinization risk level and key driving factor data are visualized and rendered using WebGIS technology to generate a rasterized risk map with a resolution of 30m. According to the user attributes of government departments and farmers, the regional risk information related to the user is extracted from the risk map using spatial query and overlay analysis methods. The GPS location coordinates of the user are obtained through the Android mobile terminal application, and the salinization risk level of the user's area is determined by spatial interpolation and inverse distance weighted method. If the salinization risk level of the user's area exceeds 6, the early warning information and key driving factor information of the area are pushed to the user, including risk level, main salt-causing factors, early warning measures, etc. In the mobile terminal application, the visualization information of the salinization risk level and key driving factors in the user's area is displayed in the form of bar charts, radar charts, etc. According to the user's interactive operations, such as zooming, panning, clicking, etc., the visualization information of the salinization risk level and key driving factors is dynamically updated using AJAX technology to provide information query services. Users can select salinization risk information at different times through the drop-down menu to view. The system uses time series analysis methods to generate salinization risk levels and key driving factor information at different times to achieve dynamic display.

[0093] Embodiment 2

[0094] like Figure 2 As shown, this embodiment also provides a soil secondary salinization degree early warning system, including:

[0095] The data preprocessing module is used to obtain and preprocess multi-source heterogeneous soil data to obtain preprocessed driving factor data;

[0096] The feature selection module is used to screen the key driving factors of soil secondary salinization based on the preprocessed driving factor data by using the feature importance evaluation method;

[0097] The driving factor analysis module is used to analyze the key driving factors using the correlation analysis method, and determine the salinization threshold of the driving factors using the threshold segmentation technology based on the analysis results;

[0098] The salinization early warning module is used to determine the salinization risk based on whether the driving factor exceeds the salinization threshold;

[0099] The scenario analysis module is used to establish a salinization early warning model through the random forest algorithm, simulate the development trend of salinization, evaluate the impact degree, and obtain the evaluation results;

[0100] The early warning information release module is used to build an early warning information release platform, upload the assessment results to the early warning information release platform in real time, visualize the risk level and key driving factors, and push early warning information through mobile terminals.

[0101] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for early warning of soil secondary salinization, characterized in that: include: Obtain soil multi-source heterogeneous data and preprocess them to obtain preprocessed driving factor data; Based on the preprocessed driving factor data, the key driving factors of the key driving factors of soil secondary salinization are screened by using the characteristic importance evaluation method; The key driving factors are analyzed by correlation analysis method, and the salinization threshold of the driving factors is determined by threshold segmentation technology according to the analysis results. If the driving factors exceed the salinization threshold, it is judged that there is a salinization risk. A salinization early warning model is established by random forest algorithm, the development trend of salinization is simulated, the impact degree is evaluated, and the evaluation results are obtained; Construct an early warning information release platform, upload the assessment results to the early warning information release platform in real time, visualize the risk level and key driving factors, and push early warning information through mobile terminals.

2. The method for early warning of soil secondary salinization according to claim 1, characterized in that: The process of obtaining soil multi-source heterogeneous data and preprocessing it to obtain preprocessed driving factor data includes: Obtain multi-source heterogeneous soil data, and based on the key driving factors of soil secondary salinization, obtain the time scale and spatial scale of the key driving factors; Preprocessing the soil multi-source heterogeneous data using a data standardization method to obtain standardized data; Based on the standardized data, the key driving factor characteristics of soil secondary salinization are extracted to obtain preprocessed driving factor data.

3. The method for early warning of soil secondary salinization according to claim 2, characterized in that: The soil multi-source heterogeneous data include but are not limited to soil conductivity, groundwater level, and meteorological conditions; The key driving factors of soil secondary salinization include but are not limited to soil salinity, groundwater mineralization, and precipitation; The process of preprocessing the soil multi-source heterogeneous data by using a data standardization method to obtain standardized data includes: The maximum and minimum normalization method is used to preprocess the soil multi-source heterogeneous data, and the data are uniformly mapped to the interval [0,1] to obtain preprocessed standardized data.

4. The method for early warning of soil secondary salinization according to claim 2, characterized in that: The process of extracting the key driving factor characteristics of soil secondary salinization based on the standardized data and obtaining the preprocessed driving factor data includes: The principal component analysis method is used to reduce the dimension of the key driving factor characteristics to obtain the main driving factors; According to the main driving factors, a machine learning prediction model for soil secondary salinization is constructed, and a support vector machine algorithm is used for training to obtain a trained prediction model; By using the trained prediction model and combining the temporal and spatial variation trends of the main driving factors, the development trend of soil secondary salinization in the future is predicted to obtain a prediction result; According to the prediction results, data visualization technology is used to generate a spatiotemporal distribution map of soil secondary salinization.

5. The method for early warning of soil secondary salinization according to claim 1, characterized in that: Based on the pre-processed driving factor data, the process of screening the key driving factors of soil secondary salinization by using the feature importance evaluation method includes: Based on the preprocessed driving factor data, a random forest algorithm is used to evaluate the characteristic importance of the driving factor; A feature selection criterion function is constructed, and a sequential forward selection algorithm is used to search for the optimal feature subset from the driving factor set. A soil secondary salinization prediction model is constructed using a support vector machine algorithm, and the optimal feature subset is used as the model input. The preprocessed driving factor data is screened through a feature subset search strategy to obtain key driving factors that have a significant impact on soil secondary salinization.

6. The method for early warning of soil secondary salinization according to claim 1, characterized in that: The process of analyzing the key driving factors using the correlation analysis method includes: Based on the data of the key driving factors and the degree of soil salinization, the Granger causality test causal inference method is used to judge the significance of the causal relationship between the key driving factors and the degree of soil salinization; if the significance of the causal relationship is greater than the preset threshold, it is determined that there is a significant causal relationship between the key driving factors and the degree of soil salinization; based on the analysis results of the spatiotemporal distribution characteristics, the interaction network and the causal relationship structural equation, the intrinsic relationship between the key driving factors and the degree of soil salinization is obtained.

7. The method for early warning of soil secondary salinization according to claim 1, characterized in that: A salinization early warning model was established through the random forest algorithm to simulate the development trend of salinization and evaluate the impact. The process of obtaining the evaluation results includes: Based on the salinization early warning model, driving factor change scenarios are set, and the salinization development trends under different scenarios are simulated through scenario analysis. The fuzzy comprehensive evaluation method is used to evaluate the degree of salinization impact of the scenarios, and the optimal early warning strategy is determined in combination with prevention and control costs and ecological benefits.

8. The method for early warning of soil secondary salinization according to claim 7, characterized in that: The process of determining the optimal early warning strategy includes: Obtain salinization-related driving factors and establish a driving factor database; set change scenarios according to the driving factors to form a driving factor change scenario matrix; The driving factor change scenario matrix is ​​input into the salinization early warning model to obtain the salinization development prediction result; if the prediction result exceeds the preset threshold, a multi-objective optimization algorithm is used to determine the optimal early warning strategy with the goal of minimizing prevention and control costs and maximizing ecological benefits, combined with prevention and control measures under different scenarios; the optimal early warning strategy is input into the salinization early warning model for verification, and the early warning strategy is optimized and adjusted according to the verification results to obtain the final early warning strategy.

9. The method for early warning of soil secondary salinization according to claim 1, characterized in that: The process of constructing an early warning information release platform, uploading the assessment results to the early warning information release platform in real time, visually displaying the risk level and key driving factors, and pushing the early warning information through a mobile terminal includes: Acquire early warning information according to the final optimal early warning strategy, and analyze and process the early warning information to obtain salinization risk level and key driving factor data; Use WebGIS technology to visualize salinization risk levels and key driving factor data to generate a visual risk map; According to user attributes, extract regional risk information related to the user from the visual risk map; obtain the user's location coordinates through the mobile terminal application to determine the salinization risk level of the user's area; If the salinization risk level in the user's area exceeds the preset threshold, the early warning information and key driving factor information of the area will be pushed to the user; the visual information of the salinization risk level and key driving factors in the user's area will be displayed in the mobile terminal application; according to the user's interactive operations, the visual information of the salinization risk level and key driving factors will be dynamically updated to provide information query services.

10. A soil secondary salinization degree early warning system, characterized in that: include: The data preprocessing module is used to obtain and preprocess multi-source heterogeneous soil data to obtain preprocessed driving factor data; A feature selection module, for screening key driving factors of key driving factors of soil secondary salinization based on the preprocessed driving factor data by using a feature importance evaluation method; A driving factor analysis module is used to analyze the key driving factors by using a correlation analysis method, and determine the salinization threshold of the driving factors by using a threshold segmentation technology according to the analysis results; A salinization early warning module is used to determine whether there is a salinization risk according to whether the driving factor exceeds the salinization threshold; The scenario analysis module is used to establish a salinization early warning model through the random forest algorithm, simulate the development trend of salinization, evaluate the impact degree, and obtain the evaluation results; The early warning information release module is used to build an early warning information release platform, upload the assessment results to the early warning information release platform in real time, visualize the risk level and key driving factors, and push the early warning information through the mobile terminal.

Images