A soybean producing area soil micro-ecological risk assessment method and system based on complex pollution

Abstract

The application discloses a soybean origin soil composite pollution micro-ecological risk assessment method and system, and belongs to the technical field of agricultural environment monitoring. By acquiring soil pollutant data and microbial community data, risk entropy is calculated, a micro-ecological health index is constructed, a first result and a second result are respectively obtained by using a risk assessment neural network and a comprehensive model combining principal component analysis, a clustering algorithm and a predator optimization algorithm, and finally a risk grade and a high-risk pollutant are determined after verification by a key factor matching and a weighted correction mechanism. The application solves the problems of single evaluation dimension and insufficient dynamic adaptability, realizes accurate diagnosis and risk early warning of soil composite pollution through multi-source data fusion and multi-method cooperation, improves evaluation accuracy and decision efficiency, and provides technical support for soybean origin soil micro-ecological risk control and sustainable agricultural development.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural environment monitoring, and more particularly to a soybean production soil micro-ecological risk assessment method and system based on composite pollution. BACKGROUND

[0002] At present, the research on agricultural soil composite pollution is still relatively lagging behind, and the existing researches are mostly focused on the toxicological effects and degradation mechanisms of single or a few pollutants or pollutants of the same type, and lack of systematic and comprehensive analysis of the long-term composite exposure of multiple and multiple types of pollutants.

[0003] Therefore, how to propose a soybean production soil micro-ecological risk assessment method and system based on composite pollution, and comprehensively analyze multi-dimensional pollutants to improve the accuracy of risk assessment and decision-making efficiency, is a problem to be solved by the person skilled in the art. SUMMARY

[0004] Therefore, the present application provides a soybean production soil micro-ecological risk assessment method and system based on composite pollution, which solves the problems of single evaluation dimension, insufficient dynamic adaptability and data processing lag in the prior art, and realizes accurate diagnosis and risk warning of soybean production soil composite pollution.

[0005] In order to achieve the above purpose, the present application adopts the following technical scheme:

[0006] On the one hand, the present application provides a soybean production soil micro-ecological risk assessment method based on composite pollution, comprising the following steps:

[0007] Obtain the pollutant data and microbial community data of the soybean production soil;

[0008] Calculate the risk entropy of the pollutant data; construct a micro-ecological health index according to the microbial community data;

[0009] Construct a risk assessment neural network, train the risk assessment neural network based on the pollutant data and the microbial community data, and obtain a first result by using the trained risk assessment neural network, wherein the first result includes a first risk prediction result and a key pollution factor combination;

[0010] Obtain a second result by combining the risk entropy and the micro-ecological health index, wherein the second result includes a second risk prediction result and a priority control factor;

[0011] According to the first result and the second result, the soil micro-ecological risk assessment result is verified, and the final soil micro-ecological risk grade is obtained, and the high-risk pollutants are output.

[0012] Preferably, the formula for calculating the risk entropy RQ is as follows:

[0013] ;

[0014] In the formula, MEC is the measured concentration of a certain pollutant, PNEC is the predicted no-effect concentration of a certain pollutant, wherein the PNEC value is determined by the ratio between the results of short-term / long-term toxicity tests and an additional safety factor.

[0015] Preferably, the microecological health index is constructed according to the microbial community data, including:

[0016] The microbial health indicators are calculated based on the microbial community data, including but not limited to microbial diversity indicators, functional gene abundance, and pathogenic bacteria load.

[0017] The microbial health indicators are normalized.

[0018] The normalized microbial health indicators are weighted and summed to form a microecological health index.

[0019] Preferably, the second result is obtained by combining the risk entropy and the microecological health index, including:

[0020] The risk entropy and the microecological health index are integrated based on principal component analysis to define a pollution load index, a functional damage index, and a microbial diversity attenuation index.

[0021] The pollution load index, the functional damage index, and the microbial diversity attenuation index are used as coordinate axes, and a clustering algorithm is used to divide risk levels to obtain the second risk prediction result.

[0022] According to the clustering results, the priority control factors are determined in combination with key pollution factor combinations.

[0023] Preferably, in the process of dividing risk levels using a clustering algorithm, the weights of the pollution load index, the functional damage index, and the microbial diversity attenuation index are adjusted adaptively using a predator optimization algorithm.

[0024] Preferably, the clustering algorithm is used to divide risk levels to obtain the second risk prediction result, including:

[0025] Initialize the population to generate N random weight vectors ; Let the weight of the pollution load index be w1, Let the weight of the functional damage index be w2, Let the weight of the microbial diversity attenuation index be w3.

[0026] For each weight vector , calculate the pollution load index, the functional damage index and the microbial diversity attenuation index, perform a clustering operation, and calculate an adaptive value based on the clustering result , the formula for calculating the adaptive value is as follows:

[0027] ;

[0028] In the formula, represents , represents the Davison-Burting index;

[0029] According to the current adaptive value, the positions of the predator and the prey are iteratively updated;

[0030] After each iteration, the position with the optimal adaptive value is selected as the current optimal solution, and the process is repeated until the adaptive value converges or the maximum number of iterations is reached, and the optimal weight vector is output ;

[0031] The optimal weight vector is used to calculate a comprehensive risk index;

[0032] The comprehensive risk index is used as input, and the pollution load index, the functional damage index and the microbial diversity attenuation index are combined for clustering to divide the risk level.

[0033] Preferably, the soil micro-ecological risk assessment result verification is performed according to the first result and the second result, and the final soil micro-ecological risk level is obtained, including:

[0034] The key pollution factor combination in the first result is intersected with the priority control factor in the second result to screen out high-risk pollutants;

[0035] When the first risk prediction result and the second risk prediction result are consistent, the final soil micro-ecological risk level is directly determined;

[0036] When the first risk prediction result and the second risk prediction result have a level difference, a key factor weighting correction mechanism is triggered;

[0037] According to the weight difference of the high-risk pollutants in the first result and the second result, a correction coefficient is given to the difference dimension, and the final level is determined by weighted average.

[0038] On the other hand, the application also proposes a soybean production soil micro-ecological risk assessment system based on composite pollution, which is used to realize the above-mentioned soybean production soil micro-ecological risk assessment method based on composite pollution, including:

[0039] The data acquisition module is configured to acquire pollutant data and microbial community data of the soybean production soil;

[0040] The parameter calculation module is configured to calculate risk entropy of the pollutant data and construct a micro-ecological health index according to the microbial community data;

[0041] The first result prediction module is configured to construct a risk assessment neural network, train the risk assessment neural network based on the pollutant data and the microbial community data, and obtain a first result by using the trained risk assessment neural network, wherein the first result comprises a first risk prediction result and a key pollution factor combination;

[0042] The second result prediction module is configured to obtain a second result by combining the risk entropy and the micro-ecological health index, wherein the second result comprises a second risk prediction result and a priority control factor;

[0043] The risk level and high-risk pollutant module is configured to perform soil micro-ecological risk assessment result verification according to the first result and the second result, obtain a final soil micro-ecological risk level, and output a high-risk pollutant.

[0044] According to the above technical solution, compared with the prior art, the present disclosure provides a soybean production soil micro-ecological risk assessment method and system based on complex pollution, which acquires soil pollutant data and microbial community data, calculates risk entropy, constructs a micro-ecological health index, uses a risk assessment neural network and a comprehensive model combining principal component analysis, K-means clustering and predator optimization algorithm to obtain first and second results, respectively, and determines the final risk level and high-risk pollutants after verification by a key factor matching and weighting correction mechanism. The present disclosure solves the problems of single evaluation dimension and insufficient dynamic adaptability in the prior art, realizes accurate diagnosis and risk warning of soil complex pollution through multi-source data fusion and multi-method cooperation, improves evaluation accuracy and decision-making efficiency, and provides technical support for soybean production soil micro-ecological risk control and sustainable agricultural development. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only embodiments of the present application, and those skilled in the art can obtain other drawings according to the provided drawings without creative labor.

[0046] Figure 1 The method flowchart provided by the present application is shown in the figure;

[0047] Figure 2 The system architecture schematic diagram provided by the present application is shown in the figure. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] On the one hand, embodiments of the present invention propose a method for assessing the soil microecological risk in soybean producing areas based on compound pollution, such as... Figure 1 As shown, it includes the following steps:

[0050] S1. Obtain pollutant data and microbial community data from the soil in soybean producing areas.

[0051] In this embodiment, the pollutant data includes field-detected concentrations of pollutants such as heavy metals, herbicides, pesticides, and phthalates, as well as multi-source data such as spatial coordinates, soil physicochemical properties, climatic factors, different crop rotation systems for grains and legumes, and continuous cropping years. The pollutant data is comprehensively summarized and a spatially visualized multidimensional database of the pollutant concentration-spatial coordinate-physicochemical property correlation matrix is ​​constructed using HTML. Electronic charts (Echarts) are used to visualize and interactively display the various types of pollution at each location, facilitating data overview and risk assessment analysis.

[0052] Microbial community data include: species annotation information of prokaryotes and fungi obtained through 16S rRNA / ITS sequencing (used to calculate α diversity indices, such as Shannon index, ACE index, and Pielou index); functional gene abundance analyzed using metagenomic sequencing (KEGG annotation of metabolic potential and C / N / P / S cycle-related genes); and pathogen load (such as the average abundance of Fusarium pathogenic genes FgSLPs and Fmk1).

[0053] S2. Calculate the risk entropy of pollutant data; construct a microecological health index based on microbial community data.

[0054] The formula for calculating risk entropy RQ is as follows:

[0055] ;

[0056] In the formula, MEC represents the measured concentration of a certain pollutant. The predicted no effect concentration (PNEC) for a certain pollutant. The PNEC value for a single pollutant compound is determined by the ratio between the results of short-term / long-term toxicity tests (including no observed effect concentration (NOEC), lowest observed effect concentration (LOEC), half maximal inhibitory concentration (IC50), half maximal lethal concentration (LC50) or half maximal effective concentration (EC50) values) and an additional safety factor (AF). According to the recommendations of the European Food Safety Authority and the European Chemicals Agency, the AF value for acute toxicity data (mainly EC50 / LC50) for terrestrial organisms is 1000, and the AF value for long-term or subacute toxicity data (mainly NOEC) is 100. Toxicity data for all target pollutants were found in the ECOTOX knowledge base, the Pesticide Properties Data Base (PPDB), and the database QSARToolbox4.6 developed jointly by the Organization for Economic Cooperation and Development and the European Chemicals Agency (OECD&ECHA, 2023). In the absence of measured toxicology data, assessment data were obtained from the prediction software of the United States Environmental Protection Agency (Ecosar2.2, USEPA).

[0057] The microecological health index is constructed according to the microbial community data, including:

[0058] The microbial health indicators are calculated based on the microbial community data, including but not limited to microbial diversity indicators (Shannon index (community richness), Pielou index (evenness), ACE index (species coverage)), functional gene abundance (KEGG annotated metabolic pathway genes (such as carbohydrate metabolism, nitrogen cycle genes), C / N / P / S cycle functional gene relative abundance), pathogenic bacteria load (average abundance of target pathogenic genes (such as FgSLPs, Hog1)), etc.

[0059] The possibility of retaining the optimization indicators in the subsequent research process of the present embodiment is also provided, and other indicators such as biomarker abundance, etc. may be added or reduced in the future after application verification.

[0060] In actual implementation, the prokaryotic and fungal species annotation information is obtained through 16S and ITS sequences, and the influence of different pollution conditions on the characteristics of microbial community is studied. The Shannon index (Shannon index) representing the species richness, the ACE index (Abundance-based coverage estimator metric), and the Pielou index (Pielou's evenness index) representing the uniformity of species distribution are selected as the α diversity measurement indexes. The FAPROTAX tool is used to associate the community structure information with metabolic functions, predict the role of microorganisms in the ecosystem, evaluate the functional diversity of microorganisms in the sample, and combine with structural equation modeling (SEM) for in-depth analysis. The microbial community of samples under different pollution conditions is analyzed by co-occurrence network, and the influence degree of microbial interaction is observed through network characteristic values such as edge number, modularity number, and positive and negative connection number. Based on metagenomic data, the metabolic potential of microorganisms is evaluated by KEGG and VB12, and the nutrient cycling potential is evaluated by calculating the abundance of C / N / P / S genes. The potential pathogenicity is evaluated by calculating the average abundance of pathogenic target genes (such as FgSLPs, Fmk1, Hog1, Mpk1 genes, etc.) of pathogenic bacteria in farmland soil.

[0061] The microbial health indicators are normalized, and in this embodiment, the minimum-maximum normalization method is used, wherein the pathogenic bacteria load is a reverse indicator (the higher the value, the higher the health risk, and the inverse is taken after normalization).

[0062] The normalized microbial health indicators are weighted and summed to form the microecological health index MEHI:

[0063] ;

[0064] is the index weight, which is determined by the analytic hierarchy process or entropy weight method; is the normalized biological health indicator.

[0065] S3. Constructing a risk assessment neural network, training the risk assessment neural network based on the pollutant data and the microbial community data, using the trained risk assessment neural network to obtain a first result, the first result including a first risk prediction result and a key pollution factor combination.

[0066] The input layer of the risk assessment neural network is the pollutant concentration (RQ value), the microbial health indicator (normalized data), and the soil physical and chemical parameters (m-dimensional features in total), the hidden layer uses the ReLU activation function, and the output layer is the risk level (classification task) and the key pollution factor weight (regression task).

[0067] The training process includes using a labeled dataset (correspondence between pollution level and microbial response), optimizing the loss function through stochastic gradient descent (cross-entropy + L2 regularization), and screening key pollution factor combinations by ranking features by importance.

[0068] S4. Combining risk entropy and microecological health index, a second result is obtained. The second result includes a second risk prediction result and priority control factors, including:

[0069] S41. Principal component analysis (PCA) was used to reduce the dimensionality and extract features of the constituent variables of risk entropy (RQ) and microbial health index (MEHI) to obtain principal components, which were defined as pollution load index (PLI), functional impairment index (FDI) and microbial diversity decline index (MDI), respectively.

[0070] Calculate the covariance matrix of the component variables of the standardized Risk Entropy (RQ) and Microbial Health Index (MEHI). Extracting feature values Lambda l and eigenvectors e l ( l =1,2,…,k, where k is the number of principal components, typically taken as a cumulative variance contribution rate ≥85%.

[0071] No. i The sample at the th l The scores on each principal component are: Where p represents the total number of indicators (pollution dimension + microbial dimension).

[0072] The Pollution Load Index (PLI) reflects the cumulative intensity of complex pollution, focusing on the pollution dimensions. , Principal component scores based solely on pollutant RQ values ​​(previous) k 1 principal component). For the first l The variance contribution rate (weight) of each principal component.

[0073] The Functional Impairment Index (FDI) quantifies the degree of ecological function impairment, focusing on the microbial function dimension. , Principal component scores (first k2 principal components) are based on functional gene abundance and metabolic potential indicators.

[0074] The Microbial Diversity Decline Index (MDI) characterizes the trend of community diversity degradation, focusing on the microbial structure dimension. , Principal component scores (first k3 principal components) based on diversity indicators.

[0075] S42. Using the pollution load index, functional damage index, and microbial diversity decline index as coordinate axes, a clustering algorithm is used to classify the risk levels and obtain the second risk prediction results.

[0076] In practice, risk level classification is not limited to clustering algorithms; other relevant methods can also be used depending on the data.

[0077] This embodiment uses K-means clustering as an example. In the process of classifying risk levels using K-means clustering, the predator optimization algorithm is used to adaptively adjust the weights of the pollution load index, functional impairment index, and microbial diversity decline index, as follows:

[0078] S421. Initialize the population and generate N random weight vectors. ; As the weight of the pollution load index, As the weight of the functional impairment index, Weights for the microbial biodiversity decline index;

[0079] S422. For each weight vector The pollution load index, functional impairment index, and microbial diversity decline index were calculated. K-means clustering was performed, and the fitness value was calculated based on the clustering results. The formula for calculating the fitness value is as follows:

[0080] ;

[0081] In the formula, express Cluster compactness and separation index; the closer the value is to 1, the higher the cluster quality. This represents the Davidson-Bourdin index, an inter-class similarity index; the smaller the value, the better the clustering effect.

[0082] S423. Iteratively update the positions of predators and prey based on fitness values.

[0083] The formula for updating prey position (moving towards the optimal predator) is: ;

[0084] In the formula, Let t be the prey position (i.e., the weight vector) in the t-th iteration. The current optimal predator position (the weight vector corresponding to the historical highest fitness); The numbers are uniformly distributed random numbers; For dynamic step size, , , .

[0085] The formula for the predator position update (global search + local fine-tuning) is:

[0086]

[0087] is a random number; is a predator movement coefficient, , , ; is a random disturbance vector; is a disturbance intensity, with an initial value of 0.5 and decreasing with iterations.

[0088] S424. At the end of each iteration, the position with the optimal fitness value is selected as the current optimal solution, until the fitness value converges or the maximum number of iterations is reached, and the optimal weight vector is output ;

[0089] S425. The optimal weight vector is used to calculate the comprehensive risk index I as follows:

[0090] ;

[0091] S426. The comprehensive risk index is input, combined with the pollution load index, the functional damage index, and the microbial diversity attenuation index for clustering to divide the risk levels.

[0092] The comprehensive risk index (I) is combined with the three-dimensional index (PLI, FDI, MDI) to form a four-dimensional feature vector [PLI, FDI, MDI, I] as input data for K-means clustering.

[0093] The number of risk levels is preset, and the elbow rule or silhouette coefficient is used to verify the optimal number of clusters K.

[0094] The K-means++ algorithm is used to optimize the initial centroid distribution.

[0095] The Euclidean distance of each sample to each centroid is calculated, and it is assigned to the nearest cluster;

[0096] The centroid of each cluster is recalculated until the centroid position is stable or the maximum number of iterations is reached;

[0097] The clustering results are labeled to form the second risk prediction results.

[0098] S43. According to the clustering results combined with the key pollution factor combination, the priority control factor is determined.

[0099] The PLI, FDI, and MDI mean values of each risk level (cluster) are calculated, and the dominant characteristics of pollution load, functional damage, and diversity degradation are analyzed.

[0100] Extract the main pollutants corresponding to each risk level from the first result (key pollution factor combination output by the neural network), cross-verify with the second result (pollution load characteristics in the clustering result), and screen out pollutants that are significantly associated in multiple dimensions (PLI, FDI, MDI).

[0101] According to the weight of the pollutant in the neural network (key pollution factor combination) and its contribution to PLI, FDI, and MDI, calculate the comprehensive priority score:

[0102] ;

[0103] Where α, β are adjustment coefficients.

[0104] Set the priority score threshold, and screen out pollutants with priority score greater than the priority score threshold as pollutants that need to be prioritized.

[0105] S5. According to the first result and the second result, verify the soil micro-ecological risk assessment result, and obtain the final soil micro-ecological risk level, including:

[0106] S51. Match the key pollution factor combination in the first result with the priority control factor in the second result, and screen out high-risk pollutants.

[0107] Match the "key pollution factor combination" of the first result (neural network) with the "priority control factor" of the second result (clustering analysis), screen out "high-risk pollutants" that meet both high data-driven importance and clustering high-risk association, and further filter through the risk entropy threshold (such as RQ≥1).

[0108] S52. When the first risk prediction result and the second risk prediction result are consistent, directly determine the final soil micro-ecological risk level.

[0109] S53. When the first risk prediction result and the second risk prediction result have level differences, trigger the key factor weighting correction mechanism.

[0110] According to the weight difference of high-risk pollutants in the first result and the second result, give a correction coefficient to the difference dimension, and determine the final level by weighted average.

[0111] Obtain the weight of high-risk pollutants in the "key pollution factor combination" from the neural network model (connection weight of the input layer of the neural network); from the results of the predator optimization algorithm, obtain the contribution weight of high-risk pollutants to the three-dimensional index (PLI / FDI / MDI).

[0112] Define the weight difference coefficient .

[0113] According to the pollution impact dimension (pollution load / functional damage / diversity attenuation), a differentiated correction coefficient is given. If the pollutant i mainly affects PLI, the correction coefficient is ; if it mainly affects FDI, the correction coefficient is ; if it mainly affects MDI, the correction coefficient is . 、 、 Set according to experience.

[0114] The correction coefficients of all high-risk pollutants are averaged to obtain the final correction coefficient , M is the number of high-risk pollutants.

[0115] After the risk level is numerized (mapping the risk level to a numerical value: "low risk" = 1, "medium risk" = 2, "high risk" = 3, "extremely high risk" = 4), the final grade is determined by weighted average (R final =q R1+(1-q) R2), which is converted into the final risk by rounding or interval division (1.5-2.4 is classified as "medium risk").

[0116] The final output of this step is the risk level and high-risk pollutants.

[0117] On the other hand, the embodiments of the present application also propose a soybean production soil micro-ecological risk assessment system based on complex pollution, which is used to realize the above-mentioned soybean production soil micro-ecological risk assessment method based on complex pollution, as shown in Figure 2 , the system comprises:

[0118] A data acquisition module is configured to acquire pollutant data and microbial community data of soybean production soil.

[0119] A parameter calculation module is configured to calculate the risk entropy of the pollutant data, and construct a micro-ecological health index based on the microbial community data.

[0120] A first result prediction module is configured to construct a risk assessment neural network, train the risk assessment neural network based on the pollutant data and the microbial community data, and obtain a first result using the trained risk assessment neural network, wherein the first result includes a first risk prediction result and a key pollution factor combination.

[0121] A second result prediction module is configured to obtain a second result by combining the risk entropy and the micro-ecological health index, wherein the second result includes a second risk prediction result and a priority control factor.

[0122] The risk grade and high-risk pollutant output module is used for verifying the soil micro-ecological risk assessment result according to the first result and the second result, obtaining the final soil micro-ecological risk grade, and outputting the high-risk pollutant.

[0123] The various embodiments are described in the specification by way of progression, each building on the last to facilitate ease of understanding. Likewise, the same reference numerals are used throughout the drawings and specification to refer to same or like parts. Not all of the features to which the embodiments are directed are required, and individual feature can be implemented or enabled independently of others described.

[0124] The foregoing description of the disclosed embodiments enables a person skilled in the art to carry out or use the application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for assessing the microecological risk of soybean-producing soil based on compound pollution, characterized in that, Includes the following steps: Obtain pollutant data and microbial community data from soybean-producing soils; Calculate the risk entropy of the pollutant data; A microbial community health index was constructed based on the aforementioned microbial community data; A risk assessment neural network is constructed, and the risk assessment neural network is trained based on the pollutant data and the microbial community data. The trained risk assessment neural network is used to obtain a first result, which includes a first risk prediction result and a combination of key pollutants. The second result is obtained by combining the risk entropy and the microecological health index. The second result includes a second risk prediction result and a priority control factor. Based on the first and second results, the soil microecological risk assessment results are verified to obtain the final soil microecological risk level, and high-risk pollutants are output, including: By matching the intersection of the key pollutant combination in the first result with the priority control factors in the second result, high-risk pollutants are screened out. When the first risk prediction result and the second risk prediction result are consistent, they are directly determined as the final soil micro-ecological risk level; When there is a difference in the level between the first risk prediction result and the second risk prediction result, the key factor weighted correction mechanism is triggered. Based on the weight difference of the high-risk pollutants in the first and second results, a correction coefficient is assigned to the difference dimension, and the final level is determined by weighted averaging.

2. The method for assessing the soil microecological risk in soybean producing areas based on compound pollution as described in claim 1, characterized in that, The formula for calculating the risk entropy RQ is as follows: ； In the formula, MEC represents the measured concentration of a certain pollutant. The predicted ineffective concentration for a certain pollutant, where the PNEC value is determined by the ratio between the results of short-term / long-term toxicity tests and an additional safety factor.

3. The method for assessing the soil microecological risk in soybean producing areas based on compound pollution as described in claim 1, characterized in that, A microbial community health index is constructed based on the aforementioned microbial community data, including: Microbial health indicators are calculated based on the microbial community data, including but not limited to microbial diversity indicators, functional gene abundance, and pathogen load. The microbial health indicators were normalized; The normalized microbial health indicators are weighted and summed to form a microecological health index.

4. The method for assessing the soil microecological risk in soybean producing areas based on compound pollution as described in claim 1, characterized in that, The second result is obtained by combining the risk entropy and the microecological health index, including: Based on principal component analysis, the risk entropy and the microecological health index are integrated to define the pollution load index, functional damage index and microbial diversity decline index. Using the pollution load index, the functional damage index, and the microbial diversity decline index as coordinate axes, a clustering algorithm is used to classify the risk levels, and the second risk prediction result is obtained. Priority control factors are determined based on the clustering results and combinations of key pollution factors.

5. The method for assessing the soil microecological risk in soybean producing areas based on compound pollution as described in claim 4, characterized in that, In the process of classifying risk levels using clustering algorithms, the predator optimization algorithm is used to adaptively adjust the weights of the pollution load index, the functional impairment index, and the microbial diversity decline index.

6. The method for assessing the soil microecological risk in soybean producing areas based on compound pollution as described in claim 5, characterized in that, The risk levels are divided using a clustering algorithm to obtain the second risk prediction result, including: Initialize the population and generate N random weight vectors. ; As the weight of the pollution load index, As the weight of the functional impairment index, Weights for the microbial biodiversity decline index; For each weight vector The pollution load index, the functional impairment index, and the microbial diversity decline index are calculated, a clustering operation is performed, and the fitness value is calculated based on the clustering results. The formula for calculating the fitness value is as follows: ； In the formula, express , This represents the Davidson-Bourdin index; The positions of predators and prey are iteratively updated based on the current fitness values; After each iteration, the position with the optimal fitness value is selected as the current optimal solution. This process continues until the fitness value converges or the maximum number of iterations is reached, at which point the optimal weight vector is output. ; Using the optimal weight vector Calculate the comprehensive risk index; Using the comprehensive risk index as input, clustering is performed by combining the pollution load index, the functional damage index, and the microbial diversity decline index to classify risk levels.

7. A soil microecological risk assessment system for soybean producing areas based on compound pollution, characterized in that, include: The data acquisition module is used to acquire pollutant data and microbial community data of the soil in soybean producing areas; The parameter calculation module is used to calculate the risk entropy of the pollutant data; A microbial community health index was constructed based on the aforementioned microbial community data; The first result prediction module is used to construct a risk assessment neural network, train the risk assessment neural network based on the pollutant data and the microbial community data, and obtain a first result using the trained risk assessment neural network. The first result includes a first risk prediction result and a combination of key pollutants. The second result prediction module is used to combine the risk entropy and the microecological health index to obtain a second result, which includes a second risk prediction result and a priority control factor. The risk level and high-risk pollutant output module is used to verify the soil microecological risk assessment results based on the first and second results, obtain the final soil microecological risk level, and output high-risk pollutants, including: By matching the intersection of the key pollutant combination in the first result with the priority control factors in the second result, high-risk pollutants are screened out. When the first risk prediction result and the second risk prediction result are consistent, they are directly determined as the final soil micro-ecological risk level; When there is a difference in the level between the first risk prediction result and the second risk prediction result, the key factor weighted correction mechanism is triggered. Based on the weight difference of the high-risk pollutants in the first and second results, a correction coefficient is assigned to the difference dimension, and the final level is determined by weighted averaging.

Images