Soil cadmium pollution risk identification method based on spatial information enhancement algorithm

Abstract

The invention discloses a soil cadmium pollution risk identification method based on a spatial information enhancement algorithm, and relates to the field of soil cadmium pollution risk prediction and risk identification, and the method comprises the steps: predicting a constructed soil cadmium pollution data set through employing a machine learning algorithm, and comparing the prediction precision of the machine learning algorithm, selecting an optimal model as a soil cadmium concentration prediction model; dividing a target area into grid units, utilizing a soil cadmium concentration prediction model to perform prediction, outputting soil cadmium concentration spatial distribution data, utilizing a Bayesian information criterion BIC to determine the optimal clustering number of soil cadmium pollution, adopting a Gaussian mixture model to perform clustering on the soil cadmium concentration, obtaining soil cadmium clustering features, and determining the soil cadmium pollution. Identifying and evaluating a soil cadmium pollution risk area; and carrying out overlay analysis on the risk receptor and the soil cadmium pollution risk area, and verifying the accuracy of identifying the risk area by the soil cadmium concentration prediction model. According to the method, accurate identification and reliable evaluation of the soil cadmium pollution risk are realized.

Description

Technical Field

[0001] The present invention relates to the field of soil cadmium pollution risk prediction and risk identification, and in particular to a soil cadmium pollution risk identification method based on a spatial information enhancement algorithm. Background Art

[0002] With the rapid development of economy and society, the problem of heavy metal pollution in soil is becoming increasingly serious, especially showing obvious regional characteristics in different regions, and has become an important factor threatening the safety of national agricultural products and human living environment. Since the causes and risk levels of heavy metal pollution in soil vary greatly in different regions, classifying and grading the risk of heavy metal pollution in soil, and then implementing risk control area management, has become an effective way to improve the level of regional soil pollution prevention and control management.

[0003] Accurately predicting the spatial distribution of regional soil heavy metals is the basis for identifying soil heavy metal pollution risk areas. However, although traditional spatial prediction methods, such as statistical methods such as Kriging, are often used in such problems, their prediction principles are only based on spatially adjacent observations and response variables, ignoring the relationship between soil heavy metals and environmental covariates, resulting in large deviations in prediction results. With the advancement of science and technology and detection technology, it is possible to obtain a variety of environmental covariates. In particular, machine learning has been widely used in spatial prediction due to its advantages in processing high-dimensional data and categorical variables. Machine learning spatial information enhancement algorithms usually provide more accurate results in spatial prediction by introducing spatial information covariates that are highly correlated with the predicted variables. However, most of the current machine learning spatial information enhancement algorithms only consider the local heterogeneity of space and fail to fully consider the spatial hierarchical heterogeneity, which limits their accuracy in practical applications. Therefore, incorporating spatial hierarchical heterogeneity into the spatial information enhancement algorithm is expected to further improve the accuracy of the prediction model.

[0004] In addition, selecting appropriate risk assessment standards for risk classification of soil heavy metal pollution is also the key to identifying pollution risk areas. Traditional methods mostly rely on point-by-point measured data of soil heavy metals, and classify and grade risk assessments based on national or provincial standards combined with human experience, and then determine risk areas in combination with geostatistical analysis. However, this method has two problems: first, the risk assessment results are easily affected by the selection of standards, which may lead to over- or under-grading; second, the uncertainty of risk occurrence is not considered, making the scope of the risk area delineated too absolute. As an autonomous learning method based on the distribution of probability density functions, the Gaussian mixture model (GMM) can accurately cluster heavy metal elements with high sensitivity at low thresholds, and provide thresholds between different clusters, thereby eliminating the influence of human experience and statistical errors, and enhancing the flexibility of risk area delineation. Therefore, how to combine the spatial information enhancement algorithm with the Gaussian mixture model to improve the identification accuracy of regional soil heavy metal pollution risk areas has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] In view of the following problems in the existing methods for identifying soil heavy metal pollution risk areas: the traditional spatial prediction method fails to fully consider the relationship between soil heavy metals and environmental covariates, resulting in large deviations in prediction results; and the existing risk classification method fails to effectively deal with the uncertainty of risk occurrence, resulting in the risk area delineation being too absolute and lacking in flexibility, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is how to incorporate spatial hierarchical heterogeneity into the machine learning spatial information enhancement algorithm, and combine it with the Gaussian mixture model to improve the accuracy of identifying regional soil heavy metal pollution risk areas, so as to more effectively manage soil pollution risks.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In the first aspect, an embodiment of the present invention provides a soil cadmium pollution risk identification method based on a spatial information enhancement algorithm, which includes obtaining soil data of a target area and constructing a soil cadmium (Cd) pollution data set; using a machine learning algorithm to predict the soil cadmium (Cd) concentration of the soil cadmium (Cd) pollution data set, and selecting the optimal model as the soil cadmium (Cd) concentration prediction model by comparing the prediction accuracy of the machine learning algorithm, wherein the machine learning algorithm includes a random forest RF algorithm, a gradient boosting decision tree GBDT algorithm, and an extreme gradient boosting XGB algorithm; dividing the target area into grid units, and using the soil cadmium (Cd) concentration prediction model to predict the soil cadmium (Cd) concentration of the grid units. The cadmium (Cd) concentration is predicted and the spatial distribution data of the soil cadmium (Cd) concentration is output; based on the spatial distribution data of the soil cadmium (Cd) concentration, the Bayesian Information Criterion (BIC) is used to determine the optimal number of clusters for the soil cadmium (Cd) pollution, and a Gaussian mixture model is used to cluster the soil cadmium (Cd) concentration, and the concentration thresholds between different clusters are calculated to obtain the soil cadmium (Cd) clustering characteristics; based on the clustering results of the soil cadmium (Cd) concentration and the concentration thresholds, the soil cadmium (Cd) pollution risk areas are identified and evaluated, and the probability of the risk area is calculated; the risk receptors of the target area and the soil cadmium (Cd) pollution risk areas are superimposed and analyzed to verify the accuracy of the soil cadmium (Cd) concentration prediction model in identifying risk areas.

[0009] As a preferred scheme of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm described in the present invention, the superposition analysis method is as follows: the first sensitive agricultural product in the target area is selected as the risk receptor; according to the national limit standard for pollutants in food, the cadmium exceeding the standard of the risk receptor is evaluated, and the cadmium Cd concentration of the risk receptor exceeding the standard and the soil cadmium Cd pollution risk area map of the target area are superimposed; according to the superimposed content, the spatial consistency of the cadmium exceeding standard position of the risk receptor and the soil cadmium Cd pollution risk area of ​​the target area is analyzed, and the accuracy of the soil cadmium Cd concentration prediction model in identifying risk areas is verified.

[0010] As a preferred solution of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm described in the present invention, the method for identifying the soil cadmium (Cd) pollution risk area is as follows: based on the soil cadmium (Cd) concentration threshold, the samples of the target area are divided into a soil cadmium (Cd) pollution safe area and a soil cadmium (Cd) pollution risk area; if the optimal number of clusters is in the expected group, the soil cadmium (Cd) pollution risk area is divided into different risk levels according to the soil cadmium pollution risk assessment standard; based on the different risk levels, the risk results are visualized using GIS technology to generate a distribution map of the soil cadmium pollution risk area in the target area; the minimum threshold of the soil cadmium (Cd) pollution safe area and the soil cadmium (Cd) pollution risk area is used as a standard to evaluate the probability of a certain area being identified as a risk area or a safe area, and obtain a probability distribution map of the soil cadmium (Cd) pollution risk.

[0011] As a preferred solution of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm described in the present invention, the method for establishing the soil cadmium pollution risk assessment standard is to calculate the optimal clustering number of soil cadmium Cd pollution using the Bayesian Information Criterion BIC, and the specific formula is as follows:

[0012]

[0013] in, is the log-likelihood value, M is the Gaussian mixture model, G is the number of mixture components, is the maximum likelihood estimate, n is the number of samples, and v is the maximum likelihood estimate The number of samples, x is the sample observation value.

[0014] Choose the Gaussian mixture model M that maximizes the BIC with G components using the Bayesian Information Criterion (BIC). M,G The component value at the time of is taken as the optimal number of clusters; based on the optimal number of clusters, the Gaussian mixture model M is used to perform cluster analysis on the samples in the target area, and the soil cadmium Cd concentration thresholds of different cluster components are calculated to establish a soil cadmium pollution risk assessment standard; the specific formula of the Gaussian mixture model M is as follows:

[0015]

[0016] Among them, f k (x i ,θ k ) is the i-th sample observation value x i The k-th mixture component cluster density, θ k is the parameter vector, G is the number of components, ψ is the mixing weight or probability, ψ={π1,···,π G-1 ,θ1,···,θ G}.

[0017] As a preferred solution of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm described in the present invention, it also includes: performing spatial interpolation on point variables to obtain surface grid data, and using the fishnet tool to create grid cells and corresponding grid points for the range of the target prediction area; using the grid points to extract the variable values ​​of the soil data in the grid cells; using the soil cadmium (Cd) concentration prediction model to predict the soil cadmium (Cd) concentration of the grid cells, outputting the spatial distribution data of the soil cadmium (Cd) concentration, and generating a soil cadmium (Cd) concentration spatial distribution prediction map of the target area.

[0018] As a preferred solution of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm described in the present invention, the soil cadmium Cd concentration prediction model is constructed by dividing the soil cadmium Cd pollution data set into an environmental covariate data set, a spatial neighbor covariate data set and a spatial regional covariate data set, and increasing the number of soil cadmium Cd pollution data sets in sequence; using the random forest RF algorithm, the gradient boosting decision tree GBDT algorithm and the extreme gradient boosting XGB algorithm respectively based on the environmental covariate data set, the environmental covariate data set + spatial neighbor covariate data set and the environmental covariate data set + spatial neighbor covariate data set + spatial regional covariate data set, to predict the soil cadmium Cd concentration in the target area; based on the prediction results, by comparing the root mean square R 2 The prediction accuracy and performance of different machine learning algorithms were evaluated and compared with the root mean square error (RMSE), and the optimal model was selected as the soil cadmium concentration prediction model.

[0019] As a preferred solution of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm described in the present invention, the soil data of the target area is obtained, and a soil cadmium (Cd) pollution data set is constructed, including: the soil data includes soil sampling points, environmental covariates, spatial neighbor covariates and spatial regional covariates; the spatial neighbor covariates are calculated by the neighbor analysis tool; the soil sampling points include soil cadmium (Cd) concentration, soil pH, soil organic matter, soil cation exchange capacity, soil conductivity, bulk density and soil texture data at different sampling points; the environmental covariates include topography, temperature, precipitation and NDVI, soil type, parent material type, land use type, population density, GDP, road density, enterprise density, agricultural activity intensity, distance from enterprise points and atmospheric cadmium (Cd) dry and wet deposition flux data.

[0020] The method for obtaining the spatial regional covariate is to interpolate the soil cadmium (Cd) concentration based on inverse distance weighting (IDW), ordinary kriging (OK) and trend surface interpolation (Tr); based on the interpolation result, the natural breakpoint method is used for reclassification, the target area is divided into several cadmium (Cd) concentration gradient sub-areas, the raster data of the cadmium (Cd) concentration gradient sub-areas are calculated, and the average value is taken as the spatial regional covariate.

[0021] In a second aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the method for identifying soil cadmium pollution risks based on a spatial information enhancement algorithm as described in the first aspect of the present invention are implemented.

[0022] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the method for identifying soil cadmium pollution risks based on a spatial information enhancement algorithm as described in the first aspect of the present invention are implemented.

[0023] The beneficial effects of the present invention are as follows: the present invention effectively overcomes the problem of insufficient information of a traditional single data source by constructing a comprehensive data set including multi-source environmental covariates, spatial neighbor covariates and spatial regional covariates; improves the accuracy of soil cadmium concentration prediction by applying three advanced machine learning algorithms in parallel and optimizing the best model; adopts grid processing to achieve accurate conversion from discrete sampling points to continuous spatial distribution, overcoming the limitations of traditional interpolation methods in dealing with nonlinear relationships; introduces Bayesian information criterion and Gaussian mixture model for risk clustering, avoiding the subjectivity of artificially set thresholds; quantifies the probability of risk occurrence through a probability assessment mechanism, providing a more comprehensive basis for pollution prevention and control decisions; adopts a risk receptor verification method to ensure the reliability and practicality of the identification results, realizes accurate identification and reliable evaluation of soil cadmium pollution risks, and provides important technical support for the accurate prevention and control of soil heavy metal pollution. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0025] Figure 1 This is a flow chart of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm of Example 1.

[0026] Figure 2 This is a comparison of the prediction accuracy of soil Cd concentration by RF, GBDT and XGB models of the soil cadmium pollution risk identification method based on spatial information enhancement algorithm in Example 2 on different variable data sets.

[0027] Figure 3 This is the data distribution of the training set and test set of the optimal prediction model of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm in Example 2.

[0028] Figure 4 This is the spatial prediction distribution map of soil Cd concentration of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm in Example 2.

[0029] Figure 5This is the optimal clustering component number of soil Cd concentration based on BIC in the soil cadmium pollution risk identification method based on spatial information enhancement algorithm in Example 2.

[0030] Figure 6 This is the cluster distribution of soil Cd concentration and its concentration threshold based on the Gaussian mixture model of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm in Example 2.

[0031] Figure 7 Example 2 is the identification of risk areas and accuracy verification of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm.

[0032] Figure 8 Example 2 is a probability distribution map of risk area identification based on the soil cadmium pollution risk identification method of the spatial information enhancement algorithm. DETAILED DESCRIPTION

[0033] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0034] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0035] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0036] Example 1

[0037] Reference Figure 1 , which is the first embodiment of the present invention, and provides a soil cadmium pollution risk identification method based on a spatial information enhancement algorithm, comprising:

[0038] S1: Obtain soil data of the target area and construct a soil cadmium (Cd) pollution dataset.

[0039] Specifically, soil data include soil sampling points, environmental covariates, spatial neighbor covariates, and spatial regional covariates. Spatial neighbor covariates are calculated using neighbor analysis tools; soil sampling points include soil cadmium (Cd) concentration, soil pH, soil organic matter, soil cation exchange capacity, soil conductivity, bulk density, and soil texture data at different sampling points; environmental covariates include topography, temperature, precipitation, NDVI, soil type, parent material type, land use type, population density, GDP, road density, enterprise density, agricultural activity intensity, distance from enterprise points, and atmospheric cadmium (Cd) dry and wet deposition flux data.

[0040] Furthermore, the method for acquiring soil sampling point data is to refer to the "Specifications for Geochemical Evaluation of Land Quality" (DZ / T 0295-2016), lay out and collect surface soil samples (0-20cm) of cultivated land, collect corresponding rice sample grains at some sampling points, and record their geographical coordinates using GPS positioning and bring them back to the laboratory for pre-treatment. Soil cadmium Cd and rice cadmium Cd are determined by inductively coupled plasma mass spectrometry (ICP-MS) after microwave digestion; soil pH and EC are determined by electrode method, SOM is determined by potassium dichromate heating method, CEC is determined by barium chloride-sulfuric acid forced exchange method, soil texture is determined by Malvern laser particle size analyzer, and soil bulk density is determined by ring knife method.

[0041] Furthermore, the method for obtaining environmental covariates is that the environmental covariate data set includes the above-mentioned measured soil physical and chemical properties. In this embodiment, the annual average temperature and annual average precipitation are from meteorological monitoring data, the land use type data is downloaded from the National Glacier Frozen Soil and Desert Science Data Center, the population density and GDP data are from the China Statistical Yearbook 2020, the industrial point of interest data is from the AutoNavi Map Open Platform, the road and river data are from OpenStreetMap, and the atmospheric deposition flux data is from the regional air quality long-term monitoring data. Among them, the distance to the road / river / enterprise point is calculated by the Nearest neighbor analysis tool of ArcGIS 10.8, and the intensity of agricultural activities is obtained by gridding the target study area and then calculating the proportion of farmland area in the target sample grid.

[0042] Furthermore, the method for obtaining spatial neighbor covariates and spatial regional covariates is as follows: the steps for obtaining spatial neighbor covariates are: using the neighbor table generated by the Generate Near Table tool in ArcGIS10.8; matching the distance and observation sample ID information to the target sample layer through the Table Join tool; based on the ID of the observation sample, extracting the cadmium Cd concentration of the three nearest observation sample points in the neighbor table or the original observation sample point attribute table, and finally obtaining the attribute table of the target sample point (Cadmium Cd1, Cadmium Cd2, Cadmium Cd3: Cadmium Cd concentration of the three nearest observation sample points; Dist1, Dist2, Dist3: distance to the three nearest observation sample points).

[0043] Specifically, the steps for obtaining spatial regional covariates are as follows: interpolating soil cadmium Cd concentration based on inverse distance weight IDW, ordinary kriging OK and trend surface interpolation Tr; since the interpolation result is continuous data, it is necessary to convert it into graded data according to the natural breakpoint method or artificial gradient division. The graded data can be adjusted according to the study area and research objectives. In this embodiment, the data is divided into 5 categories; based on the interpolation results, the natural breakpoint method is used for reclassification, the target area is divided into several cadmium Cd concentration gradient sub-areas, the raster data of the cadmium Cd concentration gradient sub-areas are calculated, and the average value is taken as the spatial regional covariate. The role of the spatial regional covariate is that when the true value is greater than or less than the predicted value, the variable can reduce errors and outliers by affecting the average value; when the modeled sampling point is an outlier, the variable can limit its influence by dividing it into sub-areas to achieve the effect of improving the prediction accuracy.

[0044] It should be noted that the variables required by the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm include three categories: environmental covariates, spatial neighbor covariates and spatial regional covariates, which are specifically subdivided into the following variable factors, as shown in Table 1.

[0045] Table 1 Variable factor table

[0046]

[0047]

[0048] S2: Use a machine learning algorithm to predict the soil cadmium (Cd) concentration of the soil cadmium (Cd) pollution data set, and select the optimal model as the soil cadmium (Cd) concentration prediction model by comparing the prediction accuracy of the machine learning algorithm, wherein the machine learning algorithm includes a random forest RF algorithm, a gradient boosting decision tree GBDT algorithm, and an extreme gradient boosting XGB algorithm.

[0049] Specifically, the method for constructing the soil cadmium (Cd) concentration prediction model is to divide the soil cadmium (Cd) pollution dataset into an environmental covariate dataset, a spatial neighbor covariate dataset and a spatial region covariate dataset, and increase the number of soil cadmium (Cd) pollution dataset groups in sequence.

[0050] Furthermore, the random forest RF algorithm, gradient boosting decision tree GBDT algorithm and extreme gradient boosting XGB algorithm were used respectively to predict the soil cadmium Cd concentration in the target area based on the environmental covariate dataset, environmental covariate dataset + spatial neighbor covariate dataset and environmental covariate dataset + spatial neighbor covariate dataset + spatial regional covariate dataset.

[0051] Furthermore, based on the prediction results, by comparing the root mean square R 2 The prediction accuracy and performance of different machine learning algorithms were evaluated and compared with the root mean square error (RMSE), and the optimal model was selected as the soil cadmium concentration prediction model.

[0052] Specifically, the machine learning algorithm modeling and optimization were carried out. The soil cadmium pollution data set was divided into a training set and a test set according to the ratio of 80% and 20%. The three machine learning algorithms, namely, the forest RF algorithm, the gradient boosting decision tree GBDT algorithm and the extreme gradient boosting XGB algorithm, were used for prediction. The prediction accuracy (R 2 The model with the best accuracy was selected as the soil cadmium concentration prediction model in this embodiment.

[0053] Furthermore, the optimal parameters of the forest RF algorithm are set as: n_estimators = 100, max_depth = 3, max_features = 3. A 5-fold cross validation is performed on the training set, and the number of random seeds for both the training and test data sets is 5. The gradient boosting decision tree GBDT algorithm replaces the residual direction with the negative gradient direction of the loss function. When the loss function is not a square loss, the global optimal direction is replaced by the local optimal direction of each iteration, which is constructed by the GradientBoostingRegressor software package.

[0054] Furthermore, a regularization term is added to the extreme gradient boosting XGB algorithm. Using the XGboost package, the optimal parameters of Grid Search CV are as follows: learning rate η is equal to 1, minimum child weight is 1, maximum depth is 2, and number of epochs is 500. The loss function is set to linear and the slope top is set to tree.

[0055] S3: Divide the target area into grid units, and use the soil cadmium (Cd) concentration prediction model to predict the soil cadmium (Cd) concentration of the grid units, and output the spatial distribution data of the soil cadmium (Cd) concentration.

[0056] Specifically, spatial interpolation is performed on point variables to obtain planar raster data, and the fishnet tool is used to create grid cells and corresponding grid points for the range of the target prediction area.

[0057] It should be noted that the target prediction area was gridded using the fishnet tool according to the size of 500m×500m, and the environmental covariates and spatial covariates in each grid were counted.

[0058] Furthermore, the variable values ​​of the soil data in the grid cells are extracted using the grid points; the soil cadmium (Cd) concentration prediction model is used to predict the soil cadmium (Cd) concentration of the grid cells, the spatial distribution data of the soil cadmium (Cd) concentration is output, and a prediction map of the spatial distribution of the soil cadmium (Cd) concentration in the target area is generated.

[0059] S4: Based on the spatial distribution data of soil cadmium Cd concentration, the Bayesian Information Criterion (BIC) is used to determine the optimal number of clusters for soil cadmium Cd pollution. At the same time, a Gaussian mixture model is used to cluster the soil cadmium Cd concentration, and the concentration thresholds between different clusters are calculated to obtain the soil cadmium Cd clustering characteristics.

[0060] Specifically, the Bayesian information criterion BIC is used to calculate the optimal number of clusters for soil cadmium Cd pollution. The specific formula is as follows:

[0061]

[0062] in, is the log-likelihood value, M is the Gaussian mixture model, G is the number of mixture components, is the maximum likelihood estimate, n is the number of samples, and v is the maximum likelihood estimate The number of samples, x is the sample observation value.

[0063] It should be noted that the Bayesian information criterion BIC is selected to maximize the BIC of the Gaussian mixture model M under G components. M,G The component value at is taken as the optimal number of clusters;

[0064] Furthermore, based on the optimal number of clusters, a Gaussian mixture model M is used to perform cluster analysis on samples in the target area, and soil cadmium Cd concentration thresholds of different cluster components are calculated to establish a soil cadmium pollution risk assessment standard;

[0065] Furthermore, the specific formula of the Gaussian mixture model M is as follows:

[0066]

[0067] Among them, f k (x i ,θk ) is the i-th sample observation value x i The k-th mixture component cluster density, θ k is the parameter vector, G is the number of components, ψ is the mixing weight or probability, ψ={π1,···,π G-1 ,θ1,···,θ G}.

[0068] S5: Based on the clustering results of soil cadmium (Cd) concentration and the concentration threshold, identify and evaluate the soil cadmium (Cd) pollution risk areas, and calculate the probability of risk areas.

[0069] Specifically, the method for identifying soil cadmium (Cd) pollution risk areas is to divide the samples in the target area into soil cadmium (Cd) pollution safe areas and soil cadmium (Cd) pollution risk areas based on the soil cadmium (Cd) concentration threshold; if the optimal number of clusters is in the expected group, the soil cadmium (Cd) pollution risk areas are divided into different risk levels according to the soil cadmium pollution risk assessment standard.

[0070] Furthermore, based on the different risk levels, GIS technology is used to visualize the risk results and generate a soil cadmium pollution risk area distribution map for the target area. The minimum thresholds of the soil cadmium (Cd) pollution safe area and the soil cadmium (Cd) pollution risk area are used as standards to evaluate the probability of an area being identified as a risk area or a safe area, and obtain a probability distribution map of the soil cadmium (Cd) pollution risk.

[0071] It should be noted that if the optimal number of clusters exceeds two, the risk level can be subdivided into high, medium and low risk levels. With the help of GIS technology, the risk assessment results can be visualized to identify different levels of soil cadmium Cd risk areas; the threshold between cluster 1 and cluster 2, that is, the critical value between safety and risk, is selected as the risk assessment standard to evaluate the probability of risk exceeding the threshold, thereby increasing the flexibility of risk area delineation and implementation.

[0072] S6: Overlay analysis of the risk receptors in the target area - the distribution of excessive cadmium (Cd) in agricultural products and the soil cadmium (Cd) pollution risk areas to verify the accuracy of risk area identification.

[0073] Specifically, the method of superposition analysis is to select the first sensitive agricultural product in the target area as the risk receptor; according to the national standard for limiting contaminants in food, evaluate the cadmium exceeding the standard in the risk receptor, and superimpose the cadmium (Cd) concentration of the risk receptor that exceeds the standard and the soil cadmium (Cd) pollution risk area map of the target area; based on the superposition content, analyze the spatial consistency between the cadmium exceeding standard position of the risk receptor and the soil cadmium (Cd) pollution risk area of ​​the target area, and verify the accuracy of the soil cadmium (Cd) concentration prediction model in identifying risk areas.

[0074] It should be noted that the agricultural products of risk receptors are mainly rice. Therefore, the excessive concentration of cadmium Cd in rice samples was selected to verify the risk area identification results. According to the National Standard for Limits of Contaminants in Food (GB2762-2017), if the cadmium Cd concentration of rice samples is ≥ 0.2 mg / kg, the cadmium Cd concentration in rice samples is considered to be excessive; the cadmium Cd concentration of rice samples and the soil cadmium Cd pollution risk zoning map of the target area are superimposed. If 85% of the rice cadmium Cd samples exceeding the standard are distributed in the soil cadmium Cd risk area, the risk area identified by the soil cadmium Cd concentration prediction model is considered to be valid.

[0075] Furthermore, this embodiment also provides a computer device, which is suitable for the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm proposed in the above embodiment.

[0076] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0077] The present embodiment also provides a storage medium having a computer program stored thereon, which implements the following steps when executed by a processor: obtaining soil data of a target area and constructing a soil cadmium (Cd) pollution data set; predicting the soil cadmium (Cd) concentration of the soil cadmium (Cd) pollution data set using a machine learning algorithm, and selecting an optimal model as a soil cadmium (Cd) concentration prediction model by comparing the prediction accuracy of the machine learning algorithm, wherein the machine learning algorithm includes a random forest RF algorithm, a gradient boosting decision tree GBDT algorithm, and an extreme gradient boosting XGB algorithm; dividing the target area into grid units, and using the soil cadmium (Cd) concentration prediction model to predict the soil cadmium (Cd) concentration of the grid units. The soil cadmium Cd concentration is predicted and the spatial distribution data of the soil cadmium Cd concentration is output; according to the spatial distribution data of the soil cadmium Cd concentration, the Bayesian Information Criterion BIC is used to determine the optimal clustering number of the soil cadmium Cd pollution, and the Gaussian mixture model is used to cluster the soil cadmium Cd concentration, and the concentration thresholds between different clusters are calculated to obtain the soil cadmium Cd clustering characteristics; based on the clustering results and concentration thresholds of the soil cadmium Cd concentration, the soil cadmium Cd pollution risk areas are identified and evaluated, and the probability of the risk area is calculated; the risk receptors of the target area and the soil cadmium Cd pollution risk areas are superimposed and analyzed to verify the accuracy of the soil cadmium Cd concentration prediction model in identifying risk areas.

[0078] In summary, the present invention effectively overcomes the problem of insufficient information of traditional single data source by constructing a comprehensive data set including multi-source environmental covariates, spatial neighbor covariates and spatial regional covariates; improves the accuracy of soil cadmium concentration prediction by applying three advanced machine learning algorithms in parallel and optimizing the best model; adopts grid processing to realize the accurate conversion from discrete sampling points to continuous spatial distribution, overcoming the limitations of traditional interpolation methods in dealing with nonlinear relationships; introduces Bayesian information criterion and Gaussian mixture model for risk clustering to avoid the subjectivity of artificially set thresholds; quantifies the probability of risk occurrence through a probability assessment mechanism to provide a more comprehensive basis for pollution prevention and control decision-making; adopts risk receptor verification method to ensure the reliability and practicality of identification results, realizes accurate identification and reliable assessment of soil cadmium pollution risk, and provides important technical support for the accurate prevention and control of soil heavy metal pollution.

[0079] Example 2

[0080] Reference Figure 2 to Figure 8 , which is the second embodiment of the present invention, and this embodiment provides a soil cadmium pollution risk identification method based on a spatial information enhancement algorithm. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0081] Specifically, according to the three variable data sets provided in Table 1, the three combinations of environmental covariate data set, environmental covariate data set + spatial neighbor covariate data set, and environmental covariate data set + spatial neighbor covariate data set + spatial regional covariate data set were used to predict the soil cadmium Cd concentration in Jiande City using three machine learning algorithms: random forest RF algorithm, gradient boosting decision tree GBDT algorithm, and extreme gradient boosting XGB algorithm. The determination coefficient R 2 , the mean square error RSME coefficient determines and verifies the simulation effect of the model.

[0082] Further, such as Figure 2 As shown in Figure 2, all models have poor prediction accuracy for soil cadmium Cd in the environmental covariate dataset and the environmental covariate dataset + spatial neighbor covariate dataset. 2 Between 0.1 and 0.3, it is far from meeting the requirements of spatial prediction. On the contrary, when the environmental covariate dataset + spatial neighbor covariate dataset + spatial region covariate dataset are used, the performance of the three models is significantly improved, and the prediction R of the best model is 2 This proves that spatial covariates can make up for the lack of spatial information in environmental covariates and neighboring covariates. Figure 3 As shown in the figure, the gradient boosting decision tree GBDT algorithm showed the lowest RMSE (0.005 mg / kg) and the highest R2 (0.93), and had accurate and stable prediction capabilities. Therefore, based on the environmental covariate dataset + spatial neighbor covariate dataset + spatial regional covariate dataset, the gradient boosting decision tree GBDT algorithm was selected as the optimal model for spatial prediction of soil cadmium Cd.

[0083] Furthermore, the study area was divided into 9493 grid samples of 500 × 500 m in size, and based on the spatial information enhancement algorithm, e.g. Figure 3 As shown in the figure, the gradient boosting decision tree GBDT algorithm is used to accurately predict the concentration and spatial distribution of soil cadmium Cd. It can be found that the high concentration of soil cadmium Cd (1.35-16.5 mg / mg) is mainly distributed in the black rock series and carbonate rock-dominated geological high background areas in the northeast and west.

[0084] Specifically, the predicted soil cadmium concentrations in 9493 grids were clustered by combining the Gaussian mixture model to classify and assess the soil cadmium pollution risk. The BIC value first increases and then decreases with the number of classifications, such as Figure 5 As shown in , when the number of clusters is 4, the BIC value reaches the maximum, indicating that the number of classifications is optimal. Figure 6 As shown, GMM divides all samples into 4 clusters, and calculates the thresholds between different clusters from high to low as 0.305 mg / kg, 0.56 mg / kg and 1.61 mg / kg respectively.

[0085] Furthermore, according to the GMM clustering results and the threshold evaluation system, such as Figure 4 As shown, soil cadmium Cd is divided into safe areas, low-risk areas, medium-risk areas and high-risk areas. Among them, the high-risk areas are concentrated in the black rock coal mining areas in the northeast, with an average cadmium Cd concentration of 6.30 mg / kg, accounting for 1.65% of the total area of ​​the region; the medium-risk areas are located on the periphery of the high-risk areas, the northwest and central river valley areas, with an average soil cadmium Cd concentration of 0.82 mg / kg, accounting for 2.78% of the area; the low-risk areas are concentrated and continuous, with an average soil cadmium Cd concentration of 0.39 mg / kg, accounting for as much as 27.2% of the area; the soil cadmium Cd concentration in the safe area is the lowest (0.2 mg / kg), which is lower than the risk screening value of the soil environmental quality standard. Further, as Figure 7 As shown, by calculating the probability of exceeding the pollution concentration threshold of 0.30 mg / kg, the uncertainty analysis of the risk zone is performed. The probability of high-risk areas being identified is as high as 95%, the probability of medium-risk areas is 75%, and the probability of safe areas being accurately identified is 95%.

[0086] Furthermore, the accuracy of risk zone identification was verified. The cadmium concentration of 60 rice samples in the target area exceeded the standard by 21.67%. When added to the soil cadmium pollution risk area, 11 samples with excessive concentrations were located in the risk area, such as Figure 8 As shown, it accounts for 84.6% (≈85%) of the total samples exceeding the standard (n=13), indicating that the model of the present invention has a very good effect in identifying the risk area of ​​soil cadmium Cd pollution in the target area. In addition, with the increase of risk level, the overall rate of exceeding the standard of rice cadmium Cd also increases accordingly, from 10% to 50%. It can be seen that the identification method of soil cadmium Cd pollution risk area based on spatial information enhancement algorithm and Gaussian mixture model of the present invention can provide strong technical support for the precise prevention and control of regional soil heavy metal pollution.

[0087] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A soil cadmium pollution risk identification method based on spatial information enhancement algorithm, characterized by: include, Obtain soil data of the target area and construct a soil cadmium (Cd) pollution dataset; The soil cadmium (Cd) concentration of the soil cadmium (Cd) pollution data set is predicted using a machine learning algorithm, and the optimal model is selected as the soil cadmium (Cd) concentration prediction model by comparing the prediction accuracy of the machine learning algorithm, wherein the machine learning algorithm includes a random forest RF algorithm, a gradient boosting decision tree GBDT algorithm, and an extreme gradient boosting XGB algorithm; Divide the target area into grid units, and use the soil cadmium (Cd) concentration prediction model to predict the soil cadmium (Cd) concentration of the grid units, and output the spatial distribution data of the soil cadmium (Cd) concentration; According to the spatial distribution data of soil cadmium Cd concentration, the optimal number of clusters of soil cadmium Cd pollution is determined using the Bayesian Information Criterion (BIC), and the soil cadmium Cd concentration is clustered using a Gaussian mixture model to calculate the concentration thresholds between different clusters to obtain the soil cadmium Cd clustering characteristics; Based on the clustering results of soil cadmium Cd concentration and the concentration threshold, identify and evaluate the soil cadmium Cd pollution risk areas and calculate the probability of risk areas; The risk receptors in the target area and the soil cadmium (Cd) pollution risk areas were overlaid and analyzed to verify the accuracy of the soil cadmium (Cd) concentration prediction model in identifying risk areas.

2. The soil cadmium pollution risk identification method based on spatial information enhancement algorithm according to claim 1, characterized in that: The method of superposition analysis is Select the most sensitive agricultural product in the target area as the risk receptor; According to the national food contaminant limit standards, the cadmium excess of risk receptors is evaluated, and the excessive cadmium concentration of risk receptors is superimposed with the soil cadmium pollution risk area map of the target area; Based on the superimposed content, the spatial consistency between the cadmium-exceeding locations of the risk receptors and the soil cadmium (Cd) pollution risk areas in the target areas was analyzed to verify the accuracy of the soil cadmium (Cd) concentration prediction model in identifying risk areas.

3. The soil cadmium pollution risk identification method based on spatial information enhancement algorithm according to claim 2, characterized in that: The method for identifying soil cadmium Cd pollution risk areas is: Based on the soil cadmium (Cd) concentration threshold, samples in the target area were divided into soil cadmium (Cd) pollution safe areas and soil cadmium (Cd) pollution risk areas; If the optimal number of clusters is in the expected group, the soil cadmium pollution risk area is divided into different risk levels according to the soil cadmium pollution risk assessment standard; Based on the different risk levels, GIS technology is used to visualize the risk results and generate a distribution map of soil cadmium pollution risk areas in the target area; Taking the minimum thresholds of soil cadmium (Cd) pollution safe areas and soil cadmium (Cd) pollution risk areas as standards, the probability of an area being identified as a risk area or a safe area was evaluated, and the probability distribution map of soil cadmium (Cd) pollution risk was obtained.

4. The soil cadmium pollution risk identification method based on spatial information enhancement algorithm according to claim 3, characterized in that: The method for establishing the soil cadmium pollution risk assessment standard is as follows: The Bayesian information criterion (BIC) was used to calculate the optimal number of clusters for soil cadmium pollution. The specific formula is as follows: in, is the log-likelihood value, M is the Gaussian mixture model, G is the number of mixture components, is the maximum likelihood estimate, n is the number of samples, and v is the maximum likelihood estimate The number of, x is the sample observation value; Choose the Gaussian mixture model M that maximizes the BIC with G components using the Bayesian Information Criterion (BIC). M,G The component value at is taken as the optimal number of clusters; Based on the optimal number of clusters, a Gaussian mixture model M is used to perform cluster analysis on samples in the target area, and soil cadmium Cd concentration thresholds of different cluster components are calculated to establish a soil cadmium pollution risk assessment standard; The specific formula of the Gaussian mixture model M is as follows: Among them, f k (x i ,θ k ) is the i-th sample observation value x i The k-th mixture component cluster density, θ k is the parameter vector, G is the number of components, ψ is the mixing weight or probability, ψ={π1,···,π G-1 ,θ1,···,θ G }.

5. The soil cadmium pollution risk identification method based on spatial information enhancement algorithm according to claim 4, characterized in that: Also includes, Perform spatial interpolation on point variables to obtain planar raster data, and use the fishnet tool to create grid cells and corresponding grid points for the target prediction area; Use grid points to extract variable values ​​of soil data within grid cells; The soil cadmium (Cd) concentration prediction model is used to predict the soil cadmium (Cd) concentration of the grid unit, output the spatial distribution data of the soil cadmium (Cd) concentration, and generate a soil cadmium (Cd) concentration spatial distribution prediction map of the target area.

6. The soil cadmium pollution risk identification method based on spatial information enhancement algorithm according to claim 5, characterized in that: The method for constructing the soil cadmium (Cd) concentration prediction model is as follows: The soil cadmium (Cd) pollution dataset is divided into an environmental covariate dataset, a spatial neighbor covariate dataset and a spatial region covariate dataset, and the number of soil cadmium (Cd) pollution dataset groups is increased in sequence; The random forest RF algorithm, gradient boosting decision tree GBDT algorithm and extreme gradient boosting XGB algorithm were used to predict the soil cadmium Cd concentration in the target area based on the environmental covariate dataset, environmental covariate dataset + spatial neighbor covariate dataset and environmental covariate dataset + spatial neighbor covariate dataset + spatial regional covariate dataset respectively. Based on the prediction results, by comparing the root mean square R 2 The prediction accuracy and performance of different machine learning algorithms were evaluated and compared with the root mean square error (RMSE), and the optimal model was selected as the soil cadmium concentration prediction model.

7. The soil cadmium pollution risk identification method based on spatial information enhancement algorithm according to claim 6, characterized in that: Obtain soil data of the target area and construct a soil cadmium pollution dataset, including: The soil data include soil sampling points, environmental covariates, spatial neighbor covariates and spatial regional covariates; the spatial neighbor covariates are calculated by neighbor analysis tools; the soil sampling points include soil cadmium (Cd) concentration, soil pH, soil organic matter, soil cation exchange capacity, soil conductivity, bulk density and soil texture data at different sampling points; the environmental covariates include topography, temperature, precipitation and NDVI, soil type, parent material type, land use type, population density, GDP, road density, enterprise density, agricultural activity intensity, distance from enterprise points and atmospheric cadmium (Cd) dry and wet deposition flux data.

8. The soil cadmium pollution risk identification method based on spatial information enhancement algorithm according to claim 7, characterized in that: The method for obtaining the spatial region covariate is: Soil cadmium Cd concentration was interpolated based on inverse distance weighting IDW, ordinary kriging OK and trend surface interpolation Tr; Based on the interpolation results, the natural breakpoint method was used for reclassification, and the target area was divided into several cadmium (Cd) concentration gradient sub-regions. The raster data of the cadmium (Cd) concentration gradient sub-regions were calculated, and the average value was taken as the spatial regional covariate.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the soil cadmium pollution risk identification method based on the spatial information enhancement algorithm described in any one of claims 1 to 8 are implemented.

Images