A machine learning-based method for predicting groundwater microplastic pollution

Abstract

The application discloses a groundwater microplastic pollution prediction method based on machine learning, and relates to the field of ecological environment, which comprises the following steps: carrying out unified gridding and standardization processing on global-scale microplastic observation data and multi-source environmental factor data; constructing and training a regression model and a classification model to respectively predict the logarithmic value of the microplastic concentration and the probability of exceeding a preset threshold value; performing feature importance analysis on the regression model and the classification model to determine the main control environmental factors affecting the microplastic pollution; determining a risk threshold value according to the output probability of the classification model and combining a diagnostic criterion to divide a high-risk area; applying a spatial mask to the high-risk area; coupling and calculating the microplastic pollution risk probability, population distribution data and groundwater dependence data to evaluate the number of potentially exposed populations in a grid; and quantifying the uncertainty of the evaluation result of the number of potentially exposed populations. The application realizes the spatial distribution prediction of the groundwater microplastic pollution.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of ecological environment, and particularly relates to a groundwater microplastic pollution prediction method based on machine learning. BACKGROUND

[0002] In recent years, with the acceleration of industrialization and urbanization, microplastics (fibers, debris, etc.) can enter the soil and groundwater through various ways such as atmospheric deposition, irrigation recharge, sludge and solid waste disposal. The soil-groundwater interface gradually becomes an important "sink" for microplastics. Microplastic particles are retained under the action of pore size screening and surface attachment in the soil, and are suspended and migrated again under the condition of strong groundwater recharge or human pumping disturbance. This process crosses the time and space scales and has a complex mechanism. After entering the aquifer, it may carry or release additives, causing potential ecological and health risks. Therefore, it is necessary to finely characterize the distribution and risk division of microplastic particles on a global scale to provide scientific basis for environmental governance and control.

[0003] Existing prediction methods mostly follow traditional statistics and empirical models, which have three limitations in the context of particulate microplastics. Firstly, they generally assume spatial continuity and stationarity, which makes it difficult to characterize the patchy and jump distribution driven by climate, recharge intensity, land use and human activities. Secondly, they lack compatibility of data sources and types, making it difficult to integrate remote sensing, socio-economic and hydrogeological information. The sparse monitoring sites and different diameters weaken the generalization and comparability of the model. Thirdly, most studies focus on concentration prediction, and lack of exposure assessment and uncertainty characterization by coupling the results with population distribution and groundwater dependence.

[0004] In the new research trend, machine learning methods (such as random forest, support vector machine, XGBoost, etc.) are gradually introduced into groundwater pollution risk prediction. These methods can handle high-dimensional, multi-source and nonlinear input variables, and are more suitable for the complex migration characteristics of microplastics in theory. However, existing researches mostly focus on heavy metals or persistent organic pollutants, and the application of microplastics is still blank. At the same time, these methods also have some shortcomings: some models ignore the explanation of the interaction between variables, resulting in a certain accuracy of prediction but lack of interpretability; some studies use limited number of training data points, which lack spatial representativeness and have poor model generalization ability; in addition, most existing methods stay at the level of concentration prediction, and lack of exposure assessment framework combined with population distribution and groundwater dependence.

[0005] Therefore, it is urgent to propose a groundwater microplastic pollution prediction method based on machine learning. SUMMARY

[0006] To solve the above technical problems, the application provides a groundwater microplastic pollution prediction method based on machine learning, which effectively breaks through the limitations of traditional methods and can realize spatial distribution prediction and risk exposure assessment of groundwater microplastic pollution at regional and even global scales, thereby providing scientific support and technical tools for environmental management, drinking water safety guarantee and policy making.

[0007] To achieve the above object, the application provides a groundwater microplastic pollution prediction method based on machine learning, comprising:

[0008] The global scale microplastic observation data and multi-source environmental factor data are uniformly gridded and standardized to obtain training data;

[0009] The training data is used to build and train a regression model and a classification model to respectively predict the logarithmic value of the microplastic concentration and the probability of exceeding the preset threshold value;

[0010] The SHAP explainability analysis is used to analyze the feature importance of the regression model and the classification model to determine the main control environmental factors affecting the microplastic pollution;

[0011] According to the output probability of the classification model, the risk threshold is determined in combination with the diagnostic criteria, and the high-risk area is divided;

[0012] The high-risk area is subjected to spatial masking to exclude invalid areas and limit the hydrologically active land;

[0013] In the hydrologically active land, the microplastic pollution risk probability is coupled with the population distribution data and the groundwater dependence data for coupled calculation, the number of potentially exposed populations is evaluated grid by grid, and is aggregated according to administrative units;

[0014] The Monte Carlo simulation method is used to quantify the uncertainty of the evaluation results of the number of potentially exposed populations.

[0015] Optionally, the process of uniform gridding and standardization processing comprises:

[0016] The microplastic observation data and the multi-source environmental factor data are unified to the same coordinate system and the target grid of the resolution;

[0017] The microplastic observation data is logarithmically transformed;

[0018] The geometric mean value of multiple microplastic observation data in the same grid is taken as the representative value of the grid; according to the preset concentration threshold, the microplastic observation data of the grid is converted into a binary classification label;

[0019] The continuous environmental factor data is standardized, and the categorical variable environmental factor data is one-hot encoded.

[0020] Optionally, the process of constructing and training the regression model and the classification model comprises:

[0021] The regression model is constructed using an XGBoost algorithm, and the objective function is to minimize the regularized squared error loss;

[0022] The classification model is constructed using an XGBoost algorithm, and the objective function is to minimize the weighted log loss function;

[0023] The regression model and the classification model are trained using a stratified K-fold cross-validation strategy.

[0024] Optionally, the process of SHAP explainability analysis comprises:

[0025] The SHAP value of each environmental factor on the model prediction result is calculated;

[0026] According to the average SHAP absolute value of each environmental factor, the top preset number of environmental factors with the highest contribution are selected as the main control environmental factors.

[0027] Optionally, the process of determining the risk threshold value in combination with the diagnostic criteria comprises:

[0028] The ROC curve of the classification model is drawn and the Youden index is calculated;

[0029] The probability value that maximizes the Youden index is taken as the first risk threshold value, or the probability value when the sensitivity is equal to the specificity is taken as the second risk threshold value.

[0030] Optionally, the process of applying a spatial mask comprises:

[0031] The high-risk area is processed in sequence by applying a marine mask, a desert mask, and a permafrost mask;

[0032] After mask processing, only the grid with a population greater than zero is retained to obtain the hydrologically active land.

[0033] Optionally, the process of coupling calculation comprises:

[0034] The microplastic pollution risk probability, population, and groundwater dependence ratio of each grid are multiplied to obtain the potential exposure population of the grid.

[0035] Optionally, the process of quantifying uncertainty comprises:

[0036] Random perturbations are applied to the risk probability, population data, and groundwater dependence data;

[0037] Based on the perturbed data, the coupling calculation and aggregation steps are repeated a preset number of times.

[0038] According to the results of all repeated calculations, the mean and confidence interval of the number of the potential exposed population are statistically obtained.

[0039] The technical effect of the present application: the present application discloses a groundwater microplastic pollution prediction method based on machine learning, which adopts the strategy of ocean / sand / permafrost area mask and unified resolution grid, eliminates invalid samples and areas with low groundwater content, and uniformly sets the grid to NaN value to display as gray; Spatial K-fold is introduced in the construction of the sample to suppress spatial leakage in the model prediction process; the regression model and the classification model of microplastic prediction are carried out at the same time to obtain the prediction abundance of microplastics in groundwater and the risk probability; both kinds of models have SHAP explainability analysis; finally, the microplastic prediction risk probability is overlapped with the world map population grid and the groundwater dependence rate to calculate the exposed population between countries, and the confidence interval is given by Monte Carlo. Thus, the evaluation accuracy, robustness and policy usability are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0040] The accompanying drawings, which form a part of this application, are intended to provide further understanding of the application and are incorporated herein in their entirety, and the description of the illustrative embodiments thereof and are used to explain the present application, and do not constitute improper limitations on the present application. In the drawings:

[0041] Figure 1 A flowchart of a groundwater microplastic pollution prediction method based on machine learning according to an embodiment of the present application. DETAILED DESCRIPTION

[0042] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0043] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0044] As shown in Figure 1 The present embodiment provides a groundwater microplastic pollution prediction method based on machine learning, which comprises:

[0045] The global scale microplastic observation data and multi-source environmental factor data are uniformly gridded and standardized to obtain training data;

[0046] The training data is used to build and train a regression model and a classification model to predict the logarithmic value of microplastic concentration and the probability of exceeding a predetermined threshold, respectively.

[0047] performing feature importance analysis on the regression model and the classification model by SHAP interpretability analysis to determine the main control environmental factors affecting microplastic pollution;

[0048] determining a risk threshold according to the output probability of the classification model in combination with diagnostic criteria, and dividing a high-risk area;

[0049] applying a spatial mask to the high-risk area to exclude invalid areas and define a hydrologically active land area;

[0050] In the hydrologically active land area, the microplastic pollution risk probability is coupled with population distribution data and groundwater dependency data for calculation, potential exposed population is assessed grid by grid, and is aggregated according to administrative units;

[0051] The Monte Carlo simulation method is used to quantify the uncertainty of the assessment results of the potential exposed population.

[0052] Further, the process of unifying gridding and standardization includes:

[0053] unifying the microplastic observation data and the multi-source environmental factor data to the same coordinate system and target grid resolution;

[0054] performing logarithmic transformation on the microplastic observation data;

[0055] taking the geometric mean of multiple microplastic observation data in the same grid as the representative value of the grid; according to a preset concentration threshold, the microplastic observation data of the grid is converted into a binary classification label;

[0056] standardizing the continuous environmental factor data and one-hot encoding the categorical variable environmental factor data.

[0057] Further, the process of constructing and training the regression model and the classification model includes:

[0058] the XGBoost algorithm is used to construct the regression model, and the objective function is to minimize the regularized squared error loss;

[0059] the XGBoost algorithm is used to construct the classification model, and the objective function is to minimize the weighted log loss function;

[0060] The regression model and the classification model are trained by using the stratified K-fold cross-validation strategy.

[0061] Further, the process of using SHAP interpretability analysis includes:

[0062] calculating the SHAP value of each environmental factor on the model prediction result;

[0063] According to the average SHAP absolute value of each environmental factor, the top preset number of environmental factors with the highest contribution degree are screened out as the master environmental factors.

[0064] Further, the process of determining the risk threshold value in combination with the diagnostic criteria includes:

[0065] Draw the ROC curve of the classification model and calculate the Youden index;

[0066] The probability value that maximizes the Youden index is taken as the first risk threshold value, or the probability value at which the sensitivity is equal to the specificity is taken as the second risk threshold value.

[0067] Further, the process of applying the spatial mask includes:

[0068] The high-risk area is processed by sequentially applying the marine mask, the desert mask and the permafrost mask;

[0069] After the mask processing, only the grids with a population greater than zero are retained to obtain the hydrologically active land.

[0070] Further, the process of coupling calculation includes:

[0071] The microplastic pollution risk probability, the population and the groundwater dependence ratio of each grid are multiplied to obtain the potential exposed population of the grid.

[0072] Further, the process of uncertainty quantification includes:

[0073] Random perturbation is applied to the risk probability, population data and groundwater dependence data;

[0074] Based on the perturbed data, the coupling calculation and the aggregation steps are repeatedly executed for a preset number of times;

[0075] According to the results of all repeated calculations, the mean and the confidence interval of the potential exposed population are statistically obtained.

[0076] Specifically, the implementation manner of the embodiment includes:

[0077] The present embodiment takes each 0.1° grid as a basic operation unit, integrates observation data and environmental factors such as climate, hydrogeology, land use, etc., and completes data alignment, standardization and feature construction under the same resolution and coordinate system; adopts a regression-classification dual-channel learning framework to output the logarithmic prediction of the concentration and the threshold probability, respectively, and determines the main control factors and checks the robustness according to the model interpretability; determines the risk threshold according to the diagnostic criteria, and limits the effective domain by using spatial masks such as oceans, deserts and permafrost; in the effective domain, the risk probability is coupled with the population grid and the groundwater dependence degree pixel by pixel, and the exposure is aggregated according to the administrative unit; the uncertainty is quantified by means such as Monte Carlo simulation to form reviewable and traceable evaluation results. The pipeline includes the following steps:

[0078] Step (1) data standardization and gridding:

[0079] Unit and scale unification: determine the target grid as 0.1°, unify to the same coordinate system and resolution; perform robust logarithmic transformation on the microplastic concentration of groundwater:

[0080] (1);

[0081] wherein, C represents the original microplastic concentration data in the grid; y reg represents the concentration data of the microplastic after logarithmic transformation.

[0082] Same grid aggregation: there are multiple observation data in the same grid, and the geometric mean represents the grid:

[0083] (2);

[0084] wherein, represents the geometric mean of the multiple observation data in the same grid; K represents the number of microplastic observation data in the grid; represents the original concentration observation data of the kth microplastic in the same grid.

[0085] Classification label construction: the threshold value is configurable, and the present embodiment selects 10.4 items / L as the grid label threshold value,

[0086] (3);

[0087] wherein, y cls represents the grid label, C represents the original microplastic concentration data in the grid; C th represents the grid label threshold value.

[0088] Environmental factor registration: continuous variables such as climate, soil, terrain, groundwater, land cover, plastic emissions, and socioeconomics are resampled to the target grid and spatially registered with the sample points; categorical variables maintain coding consistency. Missing values are imputed using KNN; continuous features are standardized by Z-score, and categorical features are one-hot encoded to avoid sparse high-dimensional data.

[0089] Step (2) regression and classification integrated modeling (mainly XGBoost):

[0090] Based on the aligned data obtained in step (1), two machine learning models are established: one is a regression model for predicting the log-transformed value of groundwater microplastic concentration; the other is a classification model for estimating the probability of the microplastic observation value exceeding the preset threshold. Through comparison and selection, XGBoost is preferred as the core learner.

[0091] Regression model (concentration log value prediction): XGBoost regressor (objective=reg:squarederror) is used to minimize the regularized squared error:

[0092] (4);

[0093] where, L reg represents the regularized squared error loss function; n represents the number of training samples; represents the true label of the i th sample; the predicted value of the model for the i th sample; represents the regularization term, which is used to punish the model complexity and prevent overfitting; Gamma represents the penalty coefficient of the number of leaf nodes; T represents the total number of leaf nodes in all trees in the model; Lambda represents the L2 regularization coefficient; represents the L2 norm of the weight vector, i.e., the "sum of squares" of all leaf node weights.

[0094] Classification model (probability prediction of exceeding threshold): XGBoost binary classifier (objective=binary:logistic) is used, set:

[0095] (5);

[0096] where, represents the probability of the pixel element exceeding the set microplastic threshold; x i represents thei Feature vectors of each sample; f ( x i ) represents the prediction function of the XGBoost binary classifier.

[0097] For class imbalance, the iterative loss of the model is logarithmic loss plus regularization:

[0098] (6);

[0099] in, L cls Represents the classification loss function of the classification model; n Indicates the number of training samples; Indicates the first i The true label of each sample; This indicates the probability that a pixel exceeds a set microplastic threshold. This represents a regularization term used to penalize model complexity and prevent overfitting.

[0100] Training strategy: The collected microplastic data were uniformly divided into training and test sets in an 8:2 ratio; hierarchical k-fold cross-validation and randomized grids were used with Bayesian search for optimal parameters, including max_depth, learning_rate, n_estimators, subsample, colsample_bytree, min_child_weight, reg_alpha, reg_lambda, etc.

[0101] Step (3) Interpretability and Feature Selection (SHAP):

[0102] The Shapley Additive Interpretation (SHAP) method is used to perform additivity decomposition on the output of the trained microplastic pollution prediction model to quantify the marginal contribution of each feature to microplastic pollution and select the main control factors accordingly. Model prediction writing:

[0103] (7);

[0104] in, This represents the predicted value of the microplastic pollution prediction model for sample x; This represents the global average predicted value; m represents the total number of features used by the model. Indicates the first j Each feature for the sample x Individual contributions to the prediction results. The top-10 non-zero variables that make a substantial contribution to the model are retained by sorting by the absolute value of the average SHAP, for robust modeling and mechanism interpretation.

[0105] Step (4) Determination of diagnostic thresholds and delineation of high-risk areas:

[0106] ROC and Youden Index:

[0107] (8);

[0108] (9);

[0109] Sensitivity = Specificity threshold:

[0110] (10);

[0111] in, This is the Youden Index. t The decision threshold for the classification model; TPR ( t The true positive rate, also known as sensitivity, is the percentage of positive results. TNR ( t The true negative rate, also known as specificity, is the percentage of true negatives. To maximize the Youden index J ( t The optimal threshold obtained by ")"; The threshold for making "sensitivity = specificity".

[0112] Indicator definitions (TP, TN, FP, FN are elements of the confusion matrix):

[0113] (11);

[0114] (12);

[0115] (13);

[0116] (14);

[0117] Among these, Sensitivity refers to the proportion of individuals who correctly identify a positive result; Specificity refers to the proportion of individuals who actually identify a negative result; PPV refers to the proportion of individuals who were correctly identified as positive; NPV refers to the proportion of individuals who were correctly identified as negative. TP indicates true positive; FN indicates false negative; TN indicates true negative; FP indicates false positive.

[0118] Select and High-risk areas can be designated; sensitivity analysis can be conducted in parallel with alternative concentration thresholds.

[0119] Step (5) Spatial mask (hydrologically active terrestrial area):

[0120] A composite mask is constructed on a unified grid, and three types of areas are successively removed: ocean, desert and permafrost. Only pixels with population > 0 are retained to obtain the "hydrologically active land pixel set", which is used to normalize the effective denominator of area and exposure statistics.

[0121] Step (6) Population Exposure Assessment (Pixel-level → National-level):

[0122] Pixel-level exposure count (combining urban / rural binary masking and groundwater dependence in various countries):

[0123] (15);

[0124] Among them, Risk ij The output probability is the classification channel, Population is the population size in the grid, and GWDep is the output probability. ij This represents the groundwater dependence ratio assigned by country and urban / rural category. After calculation, a national-level aggregation is performed: pixel-level results are topologically aggregated and sorted by country boundaries, outputting the total exposure for each country.

[0125] Step (7) Uncertainty Quantification (Monte Carlo):

[0126] Perform N (preferably N=2000) perturbation simulations to give the national average and 95% confidence interval:

[0127] (16);

[0128] (17);

[0129] (18);

[0130] (19);

[0131] in, This is the risk value after random perturbation; The original risk value; The random disturbance term of risk; refer to Follows a mean of 0 and a variance of The normal distribution; The population size after random perturbation; This represents the population size within the original grid. This refers to the random disturbance term in the population. For the random disturbance term of the population The vector follows a normal distribution with a mean of 0 and a variance of . ; is the disturbed groundwater dependency ratio; is the original groundwater dependency ratio assigned by the national and urban-rural categories; refers to the uniform distribution identifier; refers to the disturbance ratio parameter of groundwater dependency; is the number of exposed people after disturbance simulation.

[0132] Step (8) model evaluation:

[0133] Regression:

[0134] (20);

[0135] (21);

[0136] wherein, R 2 R-squared indicates the coefficient of determination, which measures the fitting effect of the model on the data; y i Yi represents the actual value of the i-th sample; i Yi represents the actual value of the i-th sample; Yi represents the model prediction value of the i-th sample; i Yi represents the model prediction value of the i-th sample; Yi represents the average value of all sample actual values; n Yi represents the total number of samples; RMSE represents the root mean square error, reflecting the average error size between the predicted value and the actual value.

[0137] Classification: ROC-AUC, accuracy, precision, recall, F1, wherein:

[0138] (22);

[0139] (23);

[0140] (24);

[0141] (25);

[0142] wherein, Accuracy represents the accuracy; Precision represents the precision; Recall represents the recall; F1 represents the F1 score, which is the harmonic mean of precision and recall; TP refers to true positive; FN refers to false negative; TN refers to true negative; FP refers to false positive.

[0143] The embodiment aims to propose a groundwater microplastic pollution prediction and risk assessment method based on machine learning, to overcome the shortcomings of the prior art in data scarcity, insufficient variable integration, limited prediction accuracy, and lack of population exposure quantification. Current research on groundwater microplastics is mostly limited to single-point or regional monitoring, lacking technical means to support large-scale or even global prediction. In the modeling process, the comprehensive influence of climate conditions, hydrogeological characteristics, land use patterns, and socio-economic factors is often not fully considered, resulting in insufficient representativeness and reliability of the prediction results. Therefore, under a unified data processing and modeling framework, the embodiment systematically integrates climate, hydrogeology, land use, and socio-economic environmental factors. Through advanced machine learning algorithms, the complex nonlinear rules of groundwater microplastic migration and spatial distribution are captured, thereby improving prediction accuracy and model generalization ability. At the same time, the embodiment introduces variable importance analysis, which can clearly determine the contribution of different environmental factors to groundwater microplastic pollution, thereby enhancing the model's interpretability and making up for the shortcomings of existing methods in the explanation and verification process. In addition, the embodiment combines model prediction results with population distribution data and groundwater usage ratio to realize the quantitative estimation of the number of potentially contaminated populations and risk levels. It not only reveals pollution hotspots but also provides a basis for public health risk assessment. Through the above technical means, the embodiment effectively breaks through the limitations of traditional methods and can realize spatial distribution prediction and risk exposure assessment of groundwater microplastic pollution at regional and even global scales, thereby providing scientific support and technical tools for environmental management, drinking water safety assurance, and policy making.

[0144] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or replacements within the technical scope disclosed by the present application can be easily thought of by those skilled in the art, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1.A method for predicting groundwater microplastic pollution based on machine learning, characterized in that, The method comprises the following steps: unifying and griding global microplastic observation data and multi-source environmental factor data to obtain training data; using the training data, constructing and training a regression model and a classification model to respectively predict the logarithm value of microplastic concentration and the probability of exceeding a preset threshold value; using SHAP explainability analysis to analyze the feature importance of the regression model and the classification model to determine the main control environmental factors affecting microplastic pollution; determining a risk threshold value according to the output probability of the classification model and combining diagnostic criteria, and dividing a high-risk area; applying a spatial mask to the high-risk area to exclude invalid areas and limit the hydrologically active land area; in the hydrologically active land area, coupling calculation of microplastic pollution risk probability, population distribution data and groundwater dependency data to evaluate the number of potentially exposed populations grid by grid, and aggregating by administrative units; using the Monte Carlo simulation method to quantify the uncertainty of the evaluation results of the number of potentially exposed populations; the process of unifying and standardizing griding comprises: unifying the microplastic observation data and the multi-source environmental factor data to the same coordinate system and target grid of resolution; logarithm transforming the microplastic observation data; taking the geometric mean of multiple microplastic observation data in the same grid as the representative value of the grid; according to a preset concentration threshold, converting the microplastic observation data of the grid into a binary classification label; standardizing continuous environmental factor data and one-hot encoding categorical variable environmental factor data; the process of constructing and training the regression model and the classification model comprises: using the XGBoost algorithm to construct the regression model, and the objective function is to minimize the regularized squared error loss; using the XGBoost algorithm to construct the classification model, and the objective function is to minimize the weighted log loss function; using a stratified K-fold cross-validation strategy to train the regression model and the classification model. 2.The machine learning-based groundwater microplastic pollution prediction method of claim 1, wherein, the process of using SHAP explainability analysis comprises: calculating the SHAP value of each environmental factor on the model prediction result; sorting according to the average SHAP absolute value of each environmental factor, and selecting the top preset number of environmental factors with the highest contribution as the main control environmental factors. 3.The machine learning-based groundwater microplastic pollution prediction method of claim 1, wherein, the process of determining the risk threshold value according to the diagnostic criteria comprises: drawing the ROC curve of the classification model and calculating the Youden index; taking the probability value that maximizes the Youden index as the first risk threshold value, or taking the probability value at which the sensitivity is equal to the specificity as the second risk threshold value. 4.The machine learning-based groundwater microplastic pollution prediction method of claim 1, wherein, the process of applying a spatial mask comprises: applying marine mask, desert mask and permafrost mask to the high-risk area in turn; after mask processing, only the grids with a population greater than zero are retained to obtain the hydrologically active land area. 5.The machine learning based groundwater microplastic pollution prediction method of claim 1, wherein, the process of coupling calculation comprises: multiplying the microplastic pollution risk probability, population and groundwater dependency ratio of each grid to obtain the number of potentially exposed populations in the grid. 6.The machine learning based groundwater microplastic pollution prediction method of claim 1, wherein, the process of quantifying uncertainty comprises: applying random perturbation to the risk probability, population data and groundwater dependency data; based on the perturbed data, repeating the coupling calculation and aggregation steps a preset number of times; From the results of all the replicates, the mean and confidence interval of the number of the potentially exposed population is statistically derived. The mean and confidence interval of the number of the potentially exposed population is statistically derived from the results of all the replicates.

Images