Heavy metal pollution space risk prediction method for retired electroplating heavy metal site

Abstract

The invention discloses a decommissioned electroplating site heavy metal pollution depth space risk prediction method, and aims to provide an efficient, accurate and economic prediction scheme, and three innovation points of fusing a Stacking ensemble learning model, predicting deep pollution by surface data, and considering space risk evaluation of sensitive receptors. The problems that a traditional method depends on a large amount of drilling, cost is high, and ecological and health risks cannot be comprehensively evaluated are solved. The method comprises the following steps: 1, acquiring site historical survey, surface environment characteristics and key point vertical profile multi-source data; 2, grading the surface soil risk by using a pollution load index method, and quantifying the vertical profile risk by using a geoaccumulation index method; 3, constructing a two-stage hybrid model, predicting surface risk distribution by Stacking, and predicting deep vertical pollution by combining a BP neural network with the surface risk; and 4, generating a three-dimensional risk distribution map, and accurately identifying an early-warning high-risk area. According to the method, horizontal and vertical prediction is combined, the three-dimensional risk is comprehensively and accurately evaluated, and a scientific basis is provided for site restoration, management and control and redevelopment.

Description

Technical Field

[0001] The technical field of spatial prediction and methods for heavy metal pollution in decommissioned electroplating sites involves environmental pollution prevention and control, soil remediation, and heavy metal pollution monitoring and assessment. In particular, it focuses on the research and application of methods for spatial distribution and risk prediction of heavy metal pollution in decommissioned electroplating sites. Background Technology

[0002] Decommissioned electroplating sites may pose serious threats to the environment and health due to historical emissions of heavy metal pollutants such as zinc, cadmium, and lead. Existing contaminated site assessment methods typically rely on manual monitoring and simple statistics, making it difficult to capture the global characteristics of complex contaminated sites. With the rise of intelligent and big data technologies, machine learning-based risk prediction methods have gradually become a research hotspot, but specific solutions applicable to decommissioned electroplating sites are still lacking.

[0003] Traditional studies on the vertical distribution of heavy metals in soil from decommissioned electroplating sites typically rely on large-scale sampling, leading to high costs, low efficiency, and widespread data gaps. Furthermore, the relationships between soil spatial location and heavy metal content, as well as between pollution levels and heavy metal content, are complex and non-linear. Particularly in decommissioned electroplating sites, heavy metal pollution is significantly influenced by a combination of historical site activities and natural conditions. Simultaneously, the physicochemical properties of the soil and their interactions with heavy metals play a crucial role in its vertical distribution, making it difficult for traditional models to accurately predict distribution patterns. Therefore, there is an urgent need to develop a method for predicting the risk of heavy metal pollution in deep soil from decommissioned electroplating sites based on surface environmental parameters, to effectively address the problems of low accuracy, low efficiency, and high cost inherent in traditional methods. Summary of the Invention

[0004] This invention provides a method for predicting the spatial risk of heavy metal pollution from decommissioned electroplating sites, thereby overcoming the shortcomings of the prior art.

[0005] To achieve the above-mentioned technical objectives, the present invention adopts the following technical solution.

[0006] The method for predicting the spatial risk of heavy metal pollution at decommissioned electroplating sites includes the following steps:

[0007] Step 1 involves acquiring multi-source data for the site, including data from the decommissioned electroplating plant to be tested. This includes data on environmental factors, historical environmental survey data, and heavy metal content. Furthermore, remote sensing imagery and on-site surveys using geographic information systems can be used to obtain surface soil data. To prevent insufficient data, additional drilling is required at key sampling points to obtain vertical profile data at these points, thus determining the heavy metal content at different soil depths.

[0008] Step 2: Quantifying Pollution Risk. The soil pollution risk in the site needs to be quantified to establish a unified target variable for subsequent models. This invention classifies and quantifies soil site pollution risk. For surface soil sampling points, based on national standards such as the "Soil Environmental Quality Standard for Construction Land Soil Pollution Risk Control," and combined with soil background values, the Pollution Load Index (PLI) method is used to calculate the comprehensive pollution index, classifying it into four levels: "no risk," "low risk," "medium risk," and "high risk." Surface data is used as the target variable for horizontal spatial prediction. For vertical profile sampling points, the geocumulative index method is used to calculate the pollution risk indicators of various heavy metals in the soil at each depth. Vertical profile prediction data is used as the target variable.

[0009] Step 3: Machine learning model training and validation. This is the core of the invention.

[0010] The first stage is horizontal spatial risk prediction. The goal of this stage is to predict the pollution risk level of the surface soil across the entire site. The model employs a Stacking ensemble learning framework, which effectively integrates the advantages of multiple heterogeneous models, improving prediction accuracy and robustness. Specifically, XGBoost and LightGBM are selected as base learners, and Random Forest (RF) is used as the meta-learner. The model input is the surface environmental feature data of each point, and the output is the surface pollution risk level at that point. By performing predictions on a high-resolution grid across the entire site, a surface pollution risk level distribution map covering the entire site is finally obtained.

[0011] The second stage is vertical pollution risk prediction. The goal of this stage is to predict the pollution status of deep soil layers based on surface information. The model employs a backpropagation (BP) neural network, whose powerful nonlinear fitting capability effectively captures the complex relationship between surface features and deep pollution. The model's input includes not only surface environmental feature data but also the "surface pollution risk level" predicted by the first-stage model as a new, highly informative feature. This design allows the vertical prediction model to perceive the relative risk level of that point within the entire site, thereby improving prediction accuracy. The model's output is the heavy metal pollution risk index at different depths below the point.

[0012] Step 4 generates a 3D site contamination risk distribution map. This is combined with the surface risk distribution map generated in the first stage and the predicted vertical contamination profiles at each point in the second stage. This allows for the construction of a 3D spatial risk distribution cloud map of heavy metal contamination across the entire site. This map clearly shows the spatial location, shape, and depth of the core contamination source. It also provides precise data support for remediation scope and technologies. The system's risk thresholds can detect whether risk values ​​exceed limits; if so, it will automatically mark and output alarm information, achieving intelligent risk management. Attached Figure Description

[0013] Figure 1 Flowchart of Spatial Risk Prediction Method for Heavy Metal Pollution

[0014] Figure 2 A schematic diagram of the predictive model.

[0015] Figure 3 A three-dimensional schematic diagram of the risk distribution of heavy metal pollution in the space of the invention for the prevention and treatment of heavy metal pollution in decommissioned electroplating sites. Detailed Implementation

[0016] The embodiments of the present invention will be described in detail below. These embodiments are based on the technical solutions of the present invention and provide detailed implementation methods and specific operation processes to further explain the technical solutions of the present invention.

[0017] The scope of protection of this invention covers all other embodiments obtained by those skilled in the art without inventive effort, which are based on the embodiments of this invention. Obviously, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings; however, the described embodiments are only a part of the embodiments of this invention, not all of them.

[0018] The technical solution of the present invention is the basis for this embodiment. This embodiment provides a detailed description of it, and gives detailed implementation methods and specific operation processes, thereby further explaining the technical solution of the present invention.

[0019] Taking a closed and decommissioned electroplating industrial site in a certain city as an example, this paper details the specific implementation process of the spatial risk prediction method for heavy metal pollution in decommissioned electroplating heavy metal sites described in this invention. The implementation steps are as follows:

[0020] Step 1: Obtain multi-source data. As shown in the figure, this step forms the data foundation for building the predictive model and requires comprehensive and representative data.

[0021] Historical Site Survey Data: First, historical data about the site was systematically collected, including but not limited to environmental impact assessment reports, preliminary site survey reports, detailed site survey reports, environmental monitoring data, and regulatory records from local environmental protection departments. Through this method, the geographical coordinates (latitude and longitude) of 47 sampling points (samples) within the site were compiled, along with the measured concentrations of major characteristic pollutants such as Pb, chromium (Cr), arsenic (As), cadmium (Cd), mercury (Hg), zinc (Zn), copper (Cu), cobalt (Co), and nickel (Ni), Zn, and Pb in the surface soil (0-20cm) at these points.

[0022] Site surface environmental characteristics data: These are the independent variables for constructing the model and require environmental data closely related to the migration and transformation of heavy metals.

[0023] Natural factor data: Digital elevation model (DEM) data of the region was acquired at this resolution, and the elevation and slope information of each sampling point was extracted using geographic information software such as ArcGIS. Local hydrogeological survey maps were consulted to obtain hydrogeological data such as groundwater depth, aquifer lithology, and main seepage directions for the site area. Simultaneously, key soil physicochemical properties from 47 historical sampling points were tested in the laboratory, including: determining soil pH using the potentiostatic method; determining soil organic matter (SOM) content using the low-temperature external heating potassium dichromate oxidation-colorimetric method; determining cation exchange capacity (CEC) using the ammonium acetate method; and determining soil texture (percentage content of sand, silt, and clay) using a hydrometer method.

[0024] Social factors data: Based on the latest high-resolution remote sensing imagery and land use map of the site, land use types in the site area were interpreted and vectorized. Simultaneously, the spatial analysis capabilities of GIS were utilized to calculate the Euclidean distance between each sampling point and key potential pollution sources such as historical workshops, wastewater treatment plants, and sensitive receptors, serving as a feature to quantify the spatial location impact.

[0025] To process soil vertical profile data and train data to predict the vertical migration depth of heavy metal pollution, it is necessary to select 47 representative vertical profiles in the highest-risk areas based on historical production layout maps, using borehole sampling. The sampling depth needs to be determined according to…

[0026] For samples with high heavy metal content, inductively coupled plasma mass spectrometry (ICP-MS) is the preferred method. Samples are freeze-dried, ground, and sieved through a 200-mesh sieve. They are then microwave-digested with HF + HNO3, followed by evaporation and the addition of HCl + HNO3. The processed samples are analyzed for heavy metal content using ICP-MS. Stratified sampling is performed at depths of 0-0.5m, 0.5-1.5m, 1.5-3.0m, and 3.0-5.0m. All soil samples are pretreated using the HNO3-HCl-H2O2 digestion method, and ICP-MS is used for precise determination of Pb, chromium (Cr), arsenic (As), cadmium (Cd), mercury (Hg), zinc (Zn), copper (Cu), cobalt (Co), and nickel (Ni).

[0027] Step 2: Pollution risk quantification. This step is mainly to establish a unified system for subsequent machine learning models and to clarify the target dependent variable for prediction.

[0028] For surface risk classification, a comprehensive risk assessment was conducted using the pollution load index method based on data from 47 soil surface sampling points. The calculation formula for each heavy metal pollutant was as follows:

[0029]

[0030] Among them, C i C represents the measured concentration of heavy metal i. 0i The soil background value for heavy metal i is given. These surface sample sites are labeled with risk levels of 0, 1, 2, and 3, serving as training targets for the first-stage spatial prediction model.

[0031] Calculate the pollution risk assessment table corresponding to the pollution load index at each sampling point:

[0032]

[0033] Vertical risk quantification: Soil samples from vertical profile points in the target area are selected, and the geocumulative index method is used to more accurately quantify the pollution level of a single heavy metal. The calculation formula is as follows:

[0034] For the risk indicator of the i-th heavy metal, For the detection content of the i-th heavy metal, Let α be the soil background content of the i-th heavy metal, and α be the natural fluctuation coefficient of the soil background content.

[0035] Calculated The value will serve as a continuous dependent variable in the second-stage vertical pollution model.

[0036] Step 3 involves data preprocessing. To eliminate dimensional differences between different feature variables and improve model training efficiency and convergence speed, all continuous numerical features are standardized using the Z-score method. For classification features, one-hot encoding is performed.

[0037] Data preprocessing is necessary to improve the model's generalization ability and avoid unnecessary redundancy and dimensionality errors. A random forest method was used to calculate data for each important heavy metal element. The data was then sorted, outliers and low values ​​were removed, ensuring approximately 10 effective key features. Soil physicochemical properties also need to be included. This process aims to identify the factors that have the greatest impact on the distribution of heavy metal pollution.

[0038] The features obtained after data filtering need to undergo independence testing to exclude highly correlated features and ensure that the features input into the model have good independence.

[0039] Step four involves model construction and training, such as... Figure 2As shown, this is the core technical aspect of the present invention. It mainly consists of several stages:

[0040] Phase 1: Horizontal spatial risk prediction model. Topsoil sample data from 47 sampling points were divided into sub-samples at a ratio of 3:1. These sub-samples were then used as a training set.

[0041] The second stage involves training the base learners, using XGBoost and LightGBM as the base learners. A 5-fold cross-validation method is employed on the training set. That is, the training set is divided into five parts, with four parts used alternately as the base training set and one part as the base validation set. This process is repeated five times, ensuring that each sample has one chance to be predicted, ultimately yielding complete prediction results for the entire training set, forming a new feature set Tr_Data. Simultaneously, this process produces five trained XGBoost models and five LightGBM models.

[0042] The third stage, meta-learner training, uses the prediction results Tr_Data generated on the training set of the base learners as the meta-learner. Random Forest (RF) is used as the input data, with the original risk level labels as the output, to train the meta-learner RF. A grid search method is employed to optimize the hyperparameters of RF to obtain the best model performance.

[0043] Model Testing and Evaluation: First, the five trained XGBoost models are used to predict on the test set, and the average (or voting result) of their predictions is taken as the final prediction of the XGBoost model. Similarly, the final prediction of LightGBM is obtained. Then, these two final prediction results are used as input to the meta-learner RF to obtain the final prediction result Ts_Pred of the entire Stacking model.

[0044] Model validation: Precision, recall, and F1-score were used as model evaluation metrics, and their calculation formulas are as follows:

[0045]

[0046]

[0047]

[0048] TP represents true positives, FP represents false positives, and FN represents false negatives. Calculations show that the Stacking model in this embodiment achieves an F1-score of 0.91 on the test set, significantly outperforming any single base learner, demonstrating the effectiveness of ensemble learning.

[0049] Phase IV Vertical Soil Pollution Risk Prediction Model

[0050] Dataset Construction: The vertical profile data used will serve as the training data for this stage. The input features for each sample are the surface environmental features selected in step three, and innovatively, the surface risk level predicted by the Stacking model in the first stage is used as the input value. The output is the Igeo values ​​of key heavy metals such as Cr, Ni, Cu, and Zn in this soil layer.

[0051] The fifth stage involves model construction and training: constructing a backpropagation (BP) neural network model. The network structure is 9-10-10-3, consisting of one input layer (9 neurons), two hidden layers (10 neurons each), and one output layer (3 neurons). The Rectifier function (ReLU) is used as the activation function for the hidden layers. A genetic algorithm is employed to optimize the network's hyperparameters, such as the learning rate and number of iterations.

[0052] Phase 6 Model Evaluation: Similarly, the 15 profile points were proportionally divided into training and test sets for validation. Evaluation metrics included the coefficient of determination, root mean square error (RMSE), and mean absolute error (MAE). The formula is as follows:

[0053]

[0054]

[0055]

[0056] Among them, R 2 The coefficient of determination is denoted by N, where N is the sample size and Obs is the number of samples. i Pred is the measured or experimental value of the i-th sample. i The predicted value of the i-th sample. The average of all measured or experimental values. In this embodiment, the model achieves an R² of 0.85 on the test set, with low RMSE and MAE levels, indicating that the model can accurately predict the risk of deep contamination based on surface information.

[0057] Step five, the three-dimensional risk generation and early warning, mainly consists of several stages.

[0058] The entire decommissioned electroplating plant is divided into fine grids using GIS software. For each grid center point, the corresponding surface soil environmental characteristics are extracted.

[0059] A two-stage prediction process is performed. The features of the grid points are input into a pre-trained first-stage Stacking model to predict the surface pollution risk level of each point. Then, the surface features and the predicted risk level of each point are used as input to a second-stage BP neural network model to predict the Igeo values ​​of important heavy metal elements such as Cr, Ni, Cu, Zn, and Pb at four depths of 0.5m, 1.5m, 3.0m, and 5.0m.

[0060] 3D visualization and early warning: Import the prediction results of all grid points into 3D visualization software to generate a model such as... Figure 3 The image shows a 3D spatial risk distribution map of heavy metal pollution at the site. This image can intuitively and three-dimensionally display the spatial location, morphology, range, and depth of high-concentration pollution plumes. Simultaneously, the system backend can set warning thresholds; for example, when the predicted Igeo value of any heavy metal at any depth exceeds a certain number, the grid point is marked as a "high-risk warning point," and a report containing its geographic coordinates and risk value is automatically generated, providing decision support for subsequent precise remediation and risk management.

[0061] In summary, the technical solution provided by this invention, through data fusion, risk quantification, and an innovative two-stage hybrid prediction model, can efficiently, accurately, and cost-effectively achieve a comprehensive assessment and prediction of the three-dimensional pollution risk of decommissioned electroplating sites. It overcomes the limitations of existing technologies and has high practical value and promising prospects for promotion.

[0062] Preferred embodiments of the present invention have been described in detail, but the present invention is not limited to the described embodiments. Any equivalent modifications or alternatives that can be made by those skilled in the art without departing from the spirit of the present invention should be considered as part of the present invention and included within the scope of protection of the claims of this application.

Claims

1. A method for predicting the spatial risk of heavy metal pollution at decommissioned electroplating sites, characterized in that, Includes the following steps: 1) Multi-source data acquisition and preprocessing: Sampling points were set up at the target decommissioned electroplating site according to the "Technical Guidelines for Soil Sampling in Soil Quality" (GB / T36197). Topsoil samples (0-20cm) and vertical profile soil samples at key points (layer depths of 0-0.5m, 0.5-1.5m, 1.5-3.0m, and 3.0-5.0m) were collected simultaneously. Historical survey data of the site (including the geographical coordinates of sampling points and the measured concentrations of heavy metals such as Pb, Cr, As, Cd, Hg, Zn, Cu, Co, and Ni in the topsoil) and surface environmental characteristic data were also acquired. The acquired data were standardized, missing value imputation was performed, outlier removal was performed, and feature dimensionality reduction was performed. 2) Pollution risk quantification: The pollution risk of the topsoil was classified using the Pollution Load Index (PLI) method, and the pollution risk of the vertically layered soil was quantified using the Geoaccumulation Index (Igeo) method. 3) Two-stage hybrid prediction model construction and training: The first stage is to build a Stacking ensemble learning model to predict the distribution of pollution risk levels in the surface soil of the entire site; The second stage involves constructing a BP neural network model, using surface environmental characteristic data and the surface pollution risk level predicted in the first stage as inputs to predict the vertical distribution of heavy metal pollution risk in deep soil. 4) Three-dimensional risk generation and early warning: By integrating the surface risk level distribution of the first stage and the vertical pollution risk prediction results of the second stage, a three-dimensional spatial risk distribution map of heavy metal pollution in the site is generated, and areas with risk values ​​exceeding the preset threshold are identified and early warning information is output.

2. The method for predicting the spatial risk of heavy metal pollution at decommissioned electroplating sites according to claim 1, characterized in that, The surface environmental characteristic data mentioned in step 1) includes natural factors and social factors; the natural factors include soil physicochemical properties (pH value, organic matter content, cation exchange capacity, soil texture), hydrogeological data (groundwater depth, aquifer lithology, main seepage direction), and topographic data (elevation and slope extracted from the digital elevation model DEM); the social factors include one or more of the following: soil type, pollution source type (historical workshop, sewage treatment pond, etc.), and distance between the sampling point and the sensitive receptor.

3. The method for predicting the spatial risk of heavy metal pollution at decommissioned electroplating sites according to claim 1, characterized in that, The specific data preprocessing operations in step 1) include: standardizing continuous numerical features using the Z-score method and encoding categorical features using one-hot encoding; handling missing values ​​using the mean imputation method and removing outliers using the 3σ principle; using random forest or principal component analysis in machine learning methods to reduce the dimensionality of the site soil environmental feature data, selecting key effective features (the number of which is controlled to about 10), and performing independence tests on the selected features to exclude highly correlated features.

4. The method for predicting the spatial risk of heavy metal pollution at decommissioned electroplating sites according to claim 1, characterized in that, In step 2), the Pollution Load Index (PLI) method calculates the nth root of the product of the ratios of the measured concentrations of each heavy metal to the corresponding soil background values ​​(where n is the number of heavy metal species being assessed) to classify the pollution risk of surface soil. Based on the PLI value, the risk of surface soil pollution is divided into four levels: PLI < 1 is no risk, 1 ≤ PLI < 2 is low risk, 2 ≤ PLI < 3 is medium risk, and PLI ≥ 3 is high risk.

5. The method for predicting the spatial risk of heavy metal pollution at decommissioned electroplating sites according to claim 1, characterized in that, Step 2) The geoaccumulation index (Igeo) method quantifies the pollution level of a single heavy metal in vertically stratified soil by calculating the ratio of the detected content of a single heavy metal to the product of the background content of that heavy metal in the soil and the natural fluctuation coefficient (with a value of 1.5). The calculated Igeo value is used as a continuous dependent variable in the second-stage vertical pollution risk prediction model.

6. The method for predicting the spatial risk of heavy metal pollution at decommissioned electroplating sites according to claim 1, characterized in that, Step 3) involves the construction and training of the Stacking ensemble learning model, which includes: using the key surface environmental features selected in Step 1) as independent variables and the surface soil pollution risk level obtained in Step 2) as the dependent variable; selecting at least two of the following as base learners: augmented regression tree, extreme gradient boosting algorithm (XGBoost), and LightGBM; training the base learners using 5-fold cross-validation and outputting the training set prediction results to form a new feature set; using this new feature set as input and the surface soil pollution risk level as output, employing random forest as the meta-learner and optimizing the meta-learner hyperparameters using grid search; and using precision, recall, and F1-score as model evaluation metrics.

7. The method for predicting the spatial risk of heavy metal pollution at decommissioned electroplating sites according to claim 1, characterized in that, Step 3) involves the construction and training of the BP neural network model, which includes: using the key surface environmental feature data selected in Step 1) and the surface pollution risk level predicted by the first-stage Stacking model as independent variables, and the vertical soil Igeo value obtained in Step 2) as the dependent variable; constructing a BP neural network structure containing an input layer, at least two hidden layers, and an output layer (e.g., 9-10-10-3, i.e., 9 neurons in the input layer, 10 neurons in each of the two hidden layers, and 3 neurons in the output layer), with ReLU selected as the activation function for the hidden layers; optimizing the neural network hyperparameters (including the number of neural network layers, the number of neurons per layer, the activation function, the learning rate, and the number of iterations) using a genetic algorithm or grid search method; and using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE) as model evaluation indicators.

8. The method for predicting the spatial risk of heavy metal pollution at decommissioned electroplating sites according to claim 1, characterized in that, The method for detecting heavy metal content in soil samples in step 1) is as follows: the soil sample is freeze-dried, ground for 40 minutes and then passed through a 200-mesh sieve. The sieved sample is then microwave-digested with HF+HNO3. After digestion, the sample is evaporated to dryness and HCl+HNO3 is added. Finally, the heavy metal content of the processed sample is analyzed by inductively coupled plasma mass spectrometry (ICP-MS).

9. The method for predicting the spatial risk of heavy metal pollution at decommissioned electroplating sites according to claim 1, characterized in that, Step 4) involves generating a 3D pollution risk distribution map and early warning system. The specific operations include: dividing the target site into high-resolution grids in GIS software and extracting surface environmental feature data corresponding to the center point of each grid; inputting the grid feature data into a two-stage hybrid prediction model to obtain the surface pollution risk level of each grid point and the Igeo values ​​at different depths (0-0.5m, 0.5-1.5m, 1.5-3.0m, 3.0-5.0m); importing the prediction results into 3D visualization software to generate a 3D pollution risk distribution cloud map, visually displaying the spatial location, shape, range, and depth of the pollution plume; pre-setting risk thresholds, automatically marking grid points with risk values ​​exceeding the thresholds, and generating an early warning report containing the geographic coordinates of the point, the risk value, and the type of heavy metal exceeding the standard.

10. A spatial risk prediction system for heavy metal pollution at decommissioned electroplating sites implementing the method of any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire historical site survey data, surface environmental characteristic data, and surface and vertical profile soil sample data, and to perform soil sample pretreatment and heavy metal content detection; the risk assessment module is used to calculate the risk level of surface soil using the Pollution Load Index (PLI) method and to calculate the risk index of vertically stratified soil using the Geoaccumulation Index (Igeo) method, thereby quantifying the pollution risk of surface and vertical soils; the risk prediction module is used to build and train Stacking ensemble learning models and BP neural network models, and to perform two-stage pollution risk prediction by substituting input data into the trained models. Risk warning module: It is used to integrate the two-stage prediction results to generate a three-dimensional pollution risk distribution map, identify areas where the risk value exceeds the preset threshold, and output a warning report containing the geographical coordinates of the warning point, the risk value, and the types of heavy metals exceeding the standard.

Images