Groundwater vulnerability indices conditioned by Supervised Intelligence Committee Machine (SICM)

Enhancing groundwater vulnerability assessment: Comparative study of three machine learning models and five classification schemes for Cuddalore district

Most of the groundwater vulnerability assessment methods using machine learning are binary classification. This study attempts multi-class classification models to map the groundwater vulnerability against Nitrate contamination. Further, the significance of the number of classes used in the multi-class classification is studied by considering three and five classes. Three machine learning models, namely Random Forest, Extreme Gradient Boosting and CART, with two classification schemes, were developed for the present study. The parameters used in the conventional DRASTIC method and with an additional parameter, Landuse, have been employed for the study. Evaluation metrics such as Accuracy, Kappa, Positive Predictive Value, Negative Predictive Value, and Area Under the Curve of the Receiver Operating Characteristic (AUC-ROC) were compared among all six models to select the optimal one. Based on the model evaluation metrics and consistent distribution of area among the classes Random Forest model with a three-class classification with an AUC of 0.95 is considered optimum for the selected objective. This study highlights the importance of the data classification process and the selection of the number of classes for ML model prediction in assessing groundwater vulnerability. Leveraging the effectiveness of the Geographic Information system and advanced machine learning techniques, the proposed approach offers valuable insights for enhanced groundwater management and contamination mitigation strategies.

Assessing vulnerability of coastal aquifer to seawater intrusion using Convolutional Neural Network

This study examined coastal aquifer vulnerability to seawater intrusion (SWI) in the Shiramin area in northwest Iran. Here, six types of hydrogeological data layers existing in the traditional GALDIT framework (TGF) were used to build one vulnerability map. Moreover, a modified traditional GALDIT framework (mod-TGF) was prepared by eliminating the data layer of aquifer type from the GALDIT model and adding the data layers of aquifer media and well density. To the best of our knowledge, there is a research gap to improve the TGF using deep learning algorithms. Therefore, this research adopted the Convolutional Neural Network (CNN) as a new deep learning algorithm to improve the mod-TGF framework for assessing the coastal aquifer vulnerability. Based on the findings, the CNN model could increase the performance of the mod-TGF by >30 %. This research can be a reference for further aquifer vulnerability studies.

Implementation and evaluation of different techniques to modify DRASTIC method for groundwater vulnerability assessment: a case study from Bouficha aquifer, Tunisia

Groundwater vulnerability assessment has nowadays evolved into an essential tool towards proper groundwater protection and management, while the DRASTIC method is included among the most widely applied vulnerability assessment methods. However, the high uncertainty of the DRASTIC method mainly associated with the subjectivity in assigning parameters ratings and weights has driven many researchers to apply various methods for improving its efficiency. In this context, in the present study, different techniques were implemented with the aim of modifying the DRASTIC framework and thus enhancing its performance for groundwater vulnerability assessment in the Bouficha aquifer, Tunisia. In a first stage, the land use type (L) was incorporated as an additional parameter in the typical DRASTIC framework, thus taking into consideration the impact of anthropogenic activities on groundwater vulnerability. Subsequently, the rating and weighting systems of the developed DRASTIC-L framework were modified through the application of statistical methods (DRASTIC-L-SA) and genetic algorithms (GA) (DRASTIC-L-GA) in an attempt to investigate and compare both linear and nonlinear modifications. To evaluate the various vulnerability frameworks, correlation between vulnerability values and nitrate concentrations, expressed as Spearman's rank correlation coefficient (ρ) and Correlation Index (CI), was examined. The results revealed that the DRASTIC-L-GA framework developed by applying a fully GA-based optimization procedure provided the highest values in terms of the performance metrics used, making it the most suitable for the study area. In addition, the aquifer under study was found to be less vulnerable to pollution when employing the typical DRASTIC framework instead of the modified ones, leading to the conclusion that the former substantially underestimates pollution potential in the study area.

Quantifying the groundwater total contamination risk using an inclusive multi-level modelling strategy

This paper investigates aggregated risks in aquifers, where risk exposures may originate from different contaminants e.g. nitrate-N (NO₃-N), arsenic (As), boron (B), fluoride (F), and aluminium (Al). The main goal is to develop a new concept for the total risk problem under sparse data as an efficient planning tool for management through the following methodology: (i) mapping aquifer vulnerability by DRASTIC and SPECTR frameworks; (ii) mapping risk indices to anthropogenic and geogenic contaminants by unsupervised methods; (iii) improving the anthropogenic and geogenic risks by a multi-level modelling strategy at three levels: Level 1 includes Artificial Neural Networks (ANN) and Support Vector Machines (SVM) models, Level 2 combines the outputs of Level 1 by unsupervised Entropy Model Averaging (EMA), and Level 3 integrates the risk maps of various contaminants (nitrate-N, arsenic, boron, fluoride, and aluminium) modelled at Level 2. The methodology offers new data layers to transform vulnerability indices into risk indices and thereby integrates risks by a heuristic scheme but without any learning as no measured values are available for the integrated risk. The results reveal that the risk indexing methodology is fit-for-purpose. According to the integrated risk map, there are hotspots at the study area and exposed to a number of contaminants (nitrate-N, arsenic, boron, fluoride, and aluminium).

Groundwater level response identification by hybrid wavelet-machine learning conjunction models using meteorological data

Due to its heterogeneous and complex nature, groundwater modeling needs great effort to quantify the aquifer, a crucial tool for policymakers and hydrogeologists to understand the variations in groundwater levels (GWL). This study proposed a set of supervised machine learning (ML) models to delineate the GWL changes in the Zarand-Saveh complex aquifer in Iran using 15-year (2005-2020) monthly dataset. The wavelet transform (WT) procedure was also used to improve the GWL prediction ability of ML models for 3-month horizons using input datasets of precipitation, evapotranspiration, temperature, and GWL. The four well-accepted standalone ML methods, i.e., artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), group method of data handling (GMDH), and least square support vector machine (LSSVM), were implemented and compared with the hybrid wavelet conjunction models. The methods were compared based on root mean square error (RMSE), mean absolute error (MAE), correlation coefficient (R), and Nash-Sutcliffe efficiency (NSE). Comparison outcomes showed that the hybrid wavelet-ML considerably improved the standalone model results. The wavelet transform-least square support vector machine (WT-LSSVM) model was superior to other standalone and hybrid wavelet-ML methods to predict GWL. The best GWL predictions were acquired from the WT-LSSVM model with input scenario 5 involving all influential variables, and this model produced RMSE, MAE, R, and NSE as 0.05, 0.04, 0.99, and 0.99 for 1 month ahead of GWL prediction, while the corresponding values were obtained as 0.18, 0.14, 0.95, and 0.90 for 3 months ahead of GWL prediction, respectively.

Novel machine learning algorithms to predict the groundwater vulnerability index to nitrate pollution at two levels of modeling

The accurate mapping and assessment of groundwater vulnerability index are crucial for the preservation of groundwater resources from the possible contamination. In this research, novel intelligent predictive Machine Learning (ML) regression models of k-Neighborhood (KNN), ensemble Extremely Randomized Trees (ERT), and ensemble Bagging regression (BA) at two levels of modeling were utilized to improve DRASTIC-LU model in the Miryang aquifer located in South Korea. The predicted outputs from level 1 (KNN and ERT models) were used as inputs for ensemble bagging (BA) in level 2. The predictive groundwater pollution vulnerability index (GPVI), derived from DRASTIC-LU model was adjusted by NO₃-N data and was utilized as the target data of the ML models. Hyperparameters for all models were tuned using a Grid Searching approach to determine the best effective model structures. Various statistical metrics and graphical representations were used to evaluate the superior predictive performance among ML models. Ensemble BA model in level 2 was more precise than standalone KNN and ensemble ERT models in level 1 for predicting GPVI values. Furthermore, the ensemble BA model offered suitable outcomes for the unseen data that could subsequently prevent the overfitting issue in the testing phase. Therefore, ML modeling at two levels could be an excellent approach for the proactive management of groundwater resources against contamination.

A comparative analysis on groundwater vulnerability models-fuzzy DRASTIC and fuzzy DRASTIC-L

Groundwater vulnerability assessment using the fuzzy logic technique is attempted in this study. A hierarchical fuzzy inference system is created to serve the selected objective. The parameters considered in this study are similar to the seven parameters used in conventional DRASTIC methods; however, the effect of land use and land cover is studied by including it as an additional parameter in a model. A hierarchy is created by comparing two input parameters, say (D and R), and the output of the same is paired as an input with the third parameter (A) and so on using the fuzzy toolbox in MATLAB. Thus, the final output of fuzzy inference systems six and seven (FI6 and FI7) is defuzzified and mapped using ArcGIS to obtain the groundwater vulnerability zones by fuzzy DRASTIC and fuzzy DRASTIC-L. Each map is grouped into five vulnerability classes: very high, high, moderate, low, and very low. Further, the results were validated using the observed nitrate concentration from 51 groundwater sampling points. The receiver operating curve (ROC) technique is adopted to determine the best suitable model for the selected study. From this, area under the curve is estimated and found to be 0.83 for fuzzy DRASTIC and 0.90 for fuzzy DRASTIC-L; the study concludes that fuzzy DRASTIC-L has a better value of AUC suits best for assessing the groundwater vulnerability in Thoothukudi District.

Formulating Convolutional Neural Network for mapping total aquifer vulnerability to pollution

Aquifer vulnerability mapping to pollution is topical research activity, and common frameworks such as the basic DRASTIC framework (BDF) suffer from the inherent subjectivity. This paper formulates an artificial intelligence modeling strategy based on Convolutional Neural Network (CNN) to decrease subjectivity. This formulation considers three definitions of intrinsic, specific, and total vulnerabilities. Accordingly, three CNN models are trained and tested to calculate IVI, SVI, and TVI, respectively referring to the intrinsic, specific, and total vulnerability indices. The formulation is applied in an unconfined aquifer northwest of Iran and delineates hotspots within the aquifer. The area under curve (AUC) values derived by the receiver operating curves evaluate the vulnerability indices versus nitrate concentrations. The AUC values for BDF, IVI, SVI, and TVI are 0.81, 0.91, 0.95, and 0.95, respectively. Therefore, CNNs significantly improve the results compared to BDF, but IVI, SVI, and TVI have approximately identical performances. However, the visual comparison between their results provides evidence that significant differences exist between the spatial patterns despite identical AUC values.

Nitrate concentration analysis and prediction in a shallow aquifer in central-eastern Tunisia using artificial neural network and time series modelling

Agricultural activities have become a major source of groundwater nitrate contamination. In this context, this study aims to analyse nitrate concentrations in a shallow aquifer of Mahdia-Kssour Essef in central-eastern Tunisia, identify the assignable sources, and predict the future levels using artificial neural network (ANN) and autoregressive integrated moving average (ARIMA) models. The results showed that nitrate concentrations measured in 21 pumping wells across the plain ranged from 17 to 521 mg L^-1. A total of 67% of the monitoring points greatly exceed the standard guideline value of 50 mg L^-1. The main relevant anthropogenic and natural factors, such as soil texture, land use, fertilizers application rates, livestock waste disposal, and groundwater table, are positively correlated with groundwater nitrate concentration. The ANN model showed good fitting between measured and simulated results with coefficient of determination (R²), root-mean-square error (RMSE), and mean absolute error (MAE) values of 0.88, 53.95, and 39.64, respectively. The ARIMA applied on annual average nitrate concentrations from 1998 to 2017 revealed that the best fitted model (p, d, q) is (1, 2, 1). The R² value is approximately 0.36, and the Theil inequality coefficient and bias proportion values are small and close to zero. These results proved the ARIMA model's adequacy in forecasting annual average nitrate concentrations of 116 mg L^-1 in 2025. These findings may be useful in making groundwater management decisions, particularly in rural and semi-arid areas, and the proposed ARIMA model could be used as a managed tool to monitor and reduce the nitrate intrusion into groundwater.

A study of uncertainties in groundwater vulnerability modelling using Bayesian model averaging (BMA)

Bayesian Model Averaging (BMA) is used to study inherent uncertainties using the Basic DRASTIC Framework (BDF) for assessing the groundwater vulnerability in a study area related to Lake Urmia. BMA is naturally an Inclusive Multiple Modelling (IMM) strategy at two levels; at Level 1 multiple models are constructed and the paper constructs three AI (Artificial Intelligence) models, which comprise ANN (Artificial Neural Network), GEP (Gene Expression Programming), and SVM (Support Vector Machines) but their outputs are fed to the next level model; at Level 2, BMA combines ANN, GEP and SVM (the Level 1 models) to quantify their inherent uncertainty in terms of within and in-between model errors. The model performance is tested by using the nitrate-N concentrations measured for the aquifer. The results show that in this particular study area, Level 1 models, even BDF, are quite accurate, but the above modelling strategy maximises the extracted information from the local data and BMA reveals that the higher uncertainties occur at areas with higher vulnerability; whereas lower uncertainties are observed at areas with lower vulnerabilities.

Comparative study of machine learning models for evaluating groundwater vulnerability to nitrate contamination

The accurate evaluation of groundwater contamination vulnerability is essential for the management and prevention of groundwater contamination in the watershed. In this study, advanced multiple machine learning (ML) models of Radial Basis Neural Networks (RBNN), Support Vector Regression (SVR), and ensemble Random Forest Regression (RFR) were applied to determine the most accurate performance for the evaluation of groundwater contamination vulnerability. Eight vulnerability factors of DRASTIC-L were rated based on the modified DRASTIC model (MDM) and were used as input data. The adjusted vulnerability index (AVI) with nitrate values was used as output data for the modeling process. The performance of three models was verified using the statistical performance criteria of MAE, RMSE, r², and ROC/AUC values. The ensemble RFR model showed the highest performance in comparison with standalone SVR and RBNN models. Specifically, ensemble RFR kept all promising solutions during the model performance due to its flexibility and robustness, and the vulnerability map obtained by the RFR model was more accurate for predicting the most vulnerable areas to contamination. It was concluded that ensemble RFR was a robust tool to enhance the evaluation of groundwater contamination vulnerability, and that it could contribute to environmental safety against groundwater contamination.

Modeling of aquifer vulnerability index using deep learning neural networks coupling with optimization algorithms

A reliable assessment of the aquifer contamination vulnerability is essential for the conservation and management of groundwater resources. In this study, a recent technique in artificial intelligence modeling and computational optimization algorithms have been adopted to enhance the groundwater contamination vulnerability assessment. The original DRASTIC model (ODM) suffers from the inherited subjectivity and a lack of robustness to assess the final aquifer vulnerability to nitrate contamination. To overcome the drawbacks of the ODM, and to maximize the accuracy of the final contamination vulnerability index, two levels of modeling strategy were proposed. The first modeling strategy used particle swarm optimization (PSO) and differential evolution (DE) algorithms to determine the effective weights of DRASTIC parameters and to produce new indices of ODVI-PSO and ODVI-DE based on the ODM formula. For strategy-2, a deep learning neural networks (DLNN) model used two indices resulting from strategy-1 as the input data. The adjusted vulnerability index in strategy-2 using the DLNN model showed more superior performance compared to the other index models when it was validated for nitrate values. Study results affirmed the capability of the DLNN model in strategy-2 to extract the further information from ODVI-PSO and ODVI-DE indices. This research concluded that strategy-2 provided higher accuracy for modeling the aquifer contamination vulnerability in the study area and established the efficient applicability for the aquifer contamination vulnerability modeling.

An investigation into time-variant subsidence potentials using inclusive multiple modelling strategies

Groundwater over-abstraction due to the absence of an effective management plan is often one of the main reasons for land subsidence in aquifer areas. This paper investigates this environmental problem at Salmas plain, Iran, by using the ALPRIFT framework, an acronym of a set of seven general-purpose data layers, introduced recently by the authors. It is capable of mapping Subsidence Vulnerability Indices (SVI) and the paper investigates an innovation to transform it into Time-variant SVI (TSVI) mapping capabilities through a three module strategy: Module 1: maps SVI; Module 2: develops a predictive model for Groundwater Levels (GWL); Module 3: combines both modules to produces TSVI maps. Modules 1 and 2 employ Inclusive Multiple Modelling (IMM) practices, which promote learning from multiple models, as opposed to their ranking and selecting a 'superior' one. IMM is implemented through the same single modelling strategy for both Modules 1 and 2 at two levels: at Level 1, multiple models are constructed by three Fuzzy Logic (FL) models: Sugeno FL (SFL), Mamdani FL (MFL) and Larsen FL (LFL). (ii) At Level 2, FL models at Level 1 are reused by Support Vector Machine (SVM) as the combiner model. The results show that (i) the models at Level 1 are fit-for-purpose; (ii) the models at Level 2 are defensible owing to IMM strategies focussed on enhancing their accuracy and investigating their residuals; and (iii) according to TSVI maps, the north of the plain is vulnerable to hotspot areas and is exposed to subsidence risks due to unplanned over-abstraction of groundwater from the aquifer at Salmas plain.

DRASTIC framework improvement using Stepwise Weight Assessment Ratio Analysis (SWARA) and combination of Genetic Algorithm and Entropy

Hybrid and integrated techniques can be used to investigate intrinsic uncertainties of the overlay and index groundwater vulnerability assessment methods. The development of a robust groundwater vulnerability assessment framework for precise identification of susceptible zones may contribute to more efficient policies and plans for sustainable managements. To achieve an overall view of the groundwater pollution potential, the DRASTIC framework (Depth to the water table, net Recharge, Aquifer media, Soil media, Topography, Impact of the vadose zone, and hydraulic Conductivity) can be used for intrinsic vulnerability assessment. However, the unreliability of this index is because of its inherent drawbacks, including the weight and rating assignment subjectivity. To modify the rating range, this study recommended a new DRASTIC modification using a recently introduced Multi-Criteria Decision-Making (MCDM) method, namely the Stepwise Weight Assessment Ratio Analysis (SWARA); in addition, the Entropy and Genetic Algorithm (GA) methods were employed to alter the relative weights of DRASTIC parameters. To improve the DRASTIC index, nitrate concentration data from 50 observation wells in the study site were used. To assess the models' overall performance, the datasets obtained from new observation wells, receiver operating characteristic (ROC) curve, and area under the ROC curve (AUC) were studied. The experiments were carried out in the aquifer of the Qazvin Plain in Iran. The results indicated the higher performance of the modified DRASTIC framework, manifested as an increase in the AUC value from 0.58 for generic DRASTIC to 0.68 for the SWARA-Ent framework and 0.74 for the SWARA-GA framework. The application of the SWARA technique, as an effective MCDM method, resulted in the DRASTIC rating system enhancement. The generic DRASTIC optimization by integrating SWARA and GA provided an effective framework to assess groundwater vulnerability to nitrate contamination in the Qazvin Plain.

A GIS-based groundwater pollution potential using DRASTIC, modified DRASTIC, and bivariate statistical models

The objective of the current study is groundwater vulnerability assessment using DRASTIC, modified DRASTIC, and three statistical bivariate models (frequency ratio (FR), evidential belief function (EBF), and weights-of-evidence (WOE)) for Sari-Behshahr plain, Iran. A total of 218 wells were sampled for nitrate concentration measurement in 2015. Datasets were generated using results from 109 wells having nitrate concentrations greater than 50 mg/L. The nitrate data were divided into two groups of 70% (76 locations as training dataset) for modeling and 30% (33 locations as a testing dataset) for model validation. Finally, five groundwater potential pollution (GPP) maps were produced by the training dataset and then evaluated using the testing dataset and receiver operating characteristic (ROC) method. Results of the ROC method showed that the WOE model had the highest predictive power, followed by EBF, FR, modified DRASTIC, and DRASTIC models. Results of the maps obtained revealed that high and very high pollution potential covered the southern part of the study areas, where big cities are located. Results of the present study can be replicated in other locations for identifying groundwater contaminant prone areas.

ANFIS-MOA models for the assessment of groundwater contamination vulnerability in a nitrate contaminated area

The enhanced assessment of groundwater contamination vulnerability is necessary for the management and conservation of groundwater resources because groundwater contamination has been much increased continuously in the world by anthropogenic origin. The purpose of this study is to determine the best model among three ANFIS-MOA models (the adaptive neuro-fuzzy inference system (ANFIS) combined with metaheuristic optimization algorithms (MOAs) such as genetic algorithm (GA), differential evolution algorithm (DE) and particle swarm optimization (PSO)) in assessing groundwater contamination vulnerability at a nitrate contaminated area. The Miryang City of South Korea was selected as the study area because the nitrate contamination was widespread in the city with two functions of urban and rural activities. Eight parameters (depth to water, net recharge, topographic slope, aquifer type, impact to vadose zone, hydraulic conductivity and landuse) were classified into the numerical ratings on basis of modified DRASTIC method (MDM) for the input variables of ANFIS-MOA models. The Original ANFIS, and 3 combined models of ANFIS-PSO, ANFIS-DE and, ANFIS-GA used 95 adjusted vulnerability indices (AVI) as the target data of training (70% data) and testing (30% data) processing. The performance of 4 models was evaluated by mean absolute errors (MAE), root mean square errors (RMSE), correlation coefficients (R), ROC/AUC curves and predicted AVI (PAVI) maps. The statistical results, spatial vulnerability maps and correlation coefficients between PAVIs and nitrate concentrations revealed that the order of model excellence was ANFIS-PSO, ANFIS-DE, ANFIS-GA, and Original ANFIS, and that ANFIS-PSO showed the highest performance in training and testing processing. The performance rates of ANFIS-MOA models were also compared with 10 recent popular worldwide models using the correlation coefficients between PVI and nitrate concentrations, and they were superior to other recent popular models. ANFIS-MOA models were also useful for resolving the subjectivity of physical and hydrogeological parameters in original DRASTIC method (ODM) and MDM. It is expected that ANFIS-PSO models will produce the excellent results in assessing groundwater contamination vulnerability and that they can greatly contribute to the groundwater security in other areas of the world as well as Miryang City of South Korea.

New hybrid-based approach for improving the accuracy of coastal aquifer vulnerability assessment maps

Due to excessive exploitation, groundwater resources of coastal regions are exposed to seawater intrusion. Therefore, vulnerability assessments are essential for the quantitative and qualitative management of these resources. The GALDIT model is the most widely used approach for coastal aquifer vulnerability assessment, but suffers from subjectivity of the identification of rates and weights. This study aimes at developing a new hybrid framework for improving the accuracy of coastal aquifer vulnerability assessment using various statistical, metaheuristic, and Multi-Attribute Decision Making (MADM) methods to improve the GALDIT model. The Gharesoo-Gorgan Rood coastal aquifer in northern Iran is used as study site. In order to meet this aim, the Differential Evolution (DE) and Biogeography-Based Optimization (BBO) metaheuristic algorithms were employed to optimize the GALDIT weights. In addition, a novel MADM method, named Step-wise Weight Assessment Ratio Analysis (SWARA), and the bivariate statistical method called statistical index (SI) were used to modify the GALDIT ratings. Finally, correlation coefficients between the maps obtained from each method and Total Dissolved Solid (TDS) as an indicator of seawater intrusion were computed to evaluate the models' prediction power. Correlation coefficients of 0.72, 0.75, 0.76 and 0.78 were obtained for the GALDIT_SWARA-BBO, GALDIT_SI-BBO, GALDIT_SWARA-DE and GALDIT_SI-DE models, respectively. The results from the GALDIT_SI-DE model outperformed all other models at improving the accuracy of the vulnerability assessment. Moreover, the statistical-metaheuristic method yielded more accurate results than SWARA-metaheuristic hybrid models. The vulnerability map of the studied region indicates that the northwestern and western areas are very highly vulnerable. According to GALDIT_SI-DE model, 42%, 17%, 18% and 22% of the aquifer areas respectively have a low, medium, high and very high vulnerability to seawater intrusion. The research findings could be applied by regional authorities to manage and protect groundwater resources.

Comparison of multi-criteria and artificial intelligence models for land-subsidence susceptibility zonation

Land subsidence (LS) in arid and semi-arid areas, such as Iran, is a significant threat to sustainable land management. The purpose of this study is to predict the LS distribution by generating land subsidence susceptibility models (LSSMs) for the Shahroud plain in Iran using three different multi-criteria decision making (MCDM) and five different artificial intelligence (AI) models. The MCDM models we used are the VlseKriterijumska Optimizacija IKompromisno Resenje (VIKOR), Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) and Complex Proportional Assessment (COPRAS), and the AI models are the extreme gradient boosting (XGBoost), Cubist, Elasticnet, Bayesian multivariate adaptive regression spline (BMARS) and conditional random forest (Cforest) methods. We used the Receiver Operating Characteristic (ROC) curve, Area Under Curve (AUC) and different statistical indices,i.e. accuracy, sensitivity, specificity, F score, Kappa, Mean Absolute Error (MAE) and Nash-Sutcliffe Criteria (NSC)to validate and evaluate the methods. Based on the different validation techniques, the Cforest method yielded the best results with minimum and maximum values of 0.04 and 0.99, respectively. According to the Cforest model, 30.55% of the study area is extremely vulnerable to land subsidence. The results of our research will be of great help to planners and policy makers in the identification of the most vulnerable regions and the implementation of appropriate development strategies in this area.

Susceptibility Prediction of Groundwater Hardness Using Ensemble Machine Learning Models

Groundwater resources, unlike surface water, are more vulnerable to disturbances and contaminations, as they take a very long time and significant cost to recover. So, predictive modeling and prevention strategies can empower policymakers for efficient groundwater governance through informed decisions and recommendations. Due to the importance of groundwater quality modeling, the hardness susceptibility mapping using machine learning (ML) models has not been explored. For the first time, the current research aimed to predict groundwater hardness susceptibility using the ML models. The performance of two ensemble models of boosted regression trees (BRT) and random forest (RF) is investigated through the arrangement of a comparative study with multivariate discriminant analysis (MDA). According to the hardness values in 135 groundwater quality monitoring wells, the hard and soft water are determined; then, 11 predictor variables including distance from the sea (DFS), land use, elevation, distance from the river (DFR), depth to groundwater (DTGW), pH, precipitation (PCP), evaporation (E), groundwater level (GWL), curvature, and lithology are used for predicting the groundwater hardness susceptibility map. Results indicated that the variables of DFR, DTGW, elevation, and DFS had a higher contribution to the modeling process. So, the high harness areas are mostly related to low elevations, low DTGW, and proximity to river and sea, which facilitate the percolation conditions for minerals containing calcium or magnesium into groundwater.

Groundwater protection under water scarcity; from regional risk assessment to local wastewater treatment solutions in Jordan

The infiltration of untreated wastewater into aquifers highly endangers the availability of fresh-water for human consumption in semi-arid areas. This growing problem of potable water scarcity urgently requires solutions for groundwater protection. Decision support systems for local wastewater treatments in settlements already exist. However, the main challenge of implementing these for regional groundwater protection is to identify where wastewater treatments are most efficient for the whole region. In this paper, we addressed this scale-crossing problem with an interdisciplinary approach that combines regional risk assessment and assessment of local wastewater treatment scenarios. We analysed the impact of polluting the groundwater using vulnerability, hazard, and risk assessments. Thus, we identified the need for semi-arid and karst-related adjustments, defined more suitable standards for wastewater hazard values, and accounted for the groundwater dynamics beyond the vertical flow paths. Using a lateral groundwater flow model, we analysed the impact of the pollution sources and linked the regional and local scale successfully. Furthermore, we combined the geoscientific results with the urban water engineering methods of area and cost assessments for local wastewater scenarios. Based on the example of the Wadi al Arab aquifer in Jordan, we showed that implementing an adapted treatment solution in one of the heavily polluted suburban settlements could reduce 12% of the aquifer pollution, which affects 93% of the potential aquifer users. This novel method helps to identify settlements with significant pollution impact on the groundwater, as well as the users, and also gives specific guidelines to establish the most efficient locally tailored treatment solution.

Vulnerability Indexing to Saltwater Intrusion from Models at Two Levels using Artificial Intelligence Multiple Model (AIMM)

Unplanned groundwater exploitation in coastal aquifers results in water decline and consequently triggers saltwater intrusion (SWI). This study formulates a novel modeling strategy based on GALDIT method using Artificial Intelligence (AI) models for mapping the vulnerability to SWI. This AI-based modeling strategy is a two-level learning process, where vulnerability to SWI at Level 1 can be predicted by such models as Artificial Neural Network (ANN), Sugeno Fuzzy Logic (SFL), and Neuro-Fuzzy (NF); and their outputs serve as the input to the model at Level 2, such as Support Vector Machine (SVM). This model is applied to Urmia aquifer, west coast of Lake Urmia, where both are currently declining. The construction of the above four models both at Levels 1 and 2 provide tools for mapping the SWI vulnerability of the study area. Model performances in the paper are studied using RMSE and R² metrics, where the models at Level 1 are found to be fit-for-purpose and the SVM at Level 2 is improved particularly with respect to the reduced scale of scatters in the results. Evaluating the result and groundwater samples by Piper diagram confirms the correspondence of SWI status with vulnerability index.

Sequential treatment of paper and pulp industrial wastewater: Prediction of water quality parameters by Mamdani Fuzzy Logic model and phytotoxicity assessment

Recycling of industrial wastewater meeting quality standards for agricultural and industrial demands is a viable option. In this study, paper and pulp industrial wastewater were treated with three biological treatments viz. aerobic, anaerobic and sequential (i.e. 20 days of anaerobic followed by 20 days of aerobic cycle), associated with simulation modeling by Mamdani Fuzzy Logic (MFL) model of some selected parameters. Electric air diffuser and minimal salt medium in sealed plastic bottles at control temperature were used for aerobic and anaerobic treatments, respectively. The significant reduction in chemical (COD: 81%) and biological oxygen demand (BOD: 71%), total suspended (TSS: 65%), dissolved solids (TDS: 60%) and turbidity (68%) was recorded during sequential treatment. The treated water was irrigated to determine its phytotoxic effects on seed germination, vigor and seedling growth of mustard (Brassica campestris). Sequential treatment greatly reduced phytotoxicity of wastewater and showed the highest germination percentage (90%) compared to aerobic (60%), anaerobic (70%) treatments and untreated wastewater (30%). Regression analysis also endorsed these findings (R² = 0.76-0.95 between seed germination, seedling growth and vigor). MFL technique was adopted to simulate sequential treatment process. The results support higher performance of MFL model to predict TDS, TSS, COD, and BOD based on the physico-chemical water quality parameters of raw wastewater, time of treatment and treatment type variation. Based on these findings, we conclude that the sequential treatment could be a more effective strategy for treatment of pulp and paper industrial wastewater with efficiency to be used for agricultural industry without toxic effects.

A new hybrid framework for optimization and modification of groundwater vulnerability in coastal aquifer

Effects of pollution caused by seawater intrusion into groundwater in coastal aquifers cannot be ignored. Identification of areas exposed to this pollution by preparing vulnerability maps is one way of preventing aquifer pollution. In its primary section, the present study compared three different index ranking methods of DRASTIC, GALDIT, and SINTACS to select an optimal model for determining vulnerability of the Gharesoo-Gorgan Rood coastal aquifer. Initial results led to selection of the GALDIT model for vulnerability assessment of the selected coastal aquifer. Since this type of models use a rating system, the model must be modified and optimized in various regions to show the vulnerable areas more accurately. In the next step, and for the first time, the ratings in this index were modified using the Wilcoxon nonparametric statistical method and its weights were optimized employing particle swarm optimization (PSO) and single-parameter sensitivity analysis (SPSA) methods. Finally, in order to select the best hybrid model, the total dissolved solids (TDS) parameter was used to determine correlation coefficients. Results indicated that the GALDT model modified by the Wilcoxon-PSO method has the strongest correlation (0.77) with the TDS parameter. Moreover, the correlations of the Wilcoxon-GALDIT and Wilcoxon-SPSA models were 0.66 and 0.73, respectively. Final results of the Wilcoxon-PSO model revealed that the northwestern and western areas of the study region needed considerable protection against pollution. In general, we can conclude that by combining statistical, mathematical, and metaheuristic methods, we can obtain more accurate results for preparing vulnerability maps.

Mapping specific vulnerability of multiple confined and unconfined aquifers by using artificial intelligence to learn from multiple DRASTIC frameworks

An investigation is presented to improve on the performances of the Basic DRASTIC Framework (BDF) and its variation by the Fuzzy-Catastrophe Framework (FCF), both of which provide an estimate of intrinsic aquifer vulnerabilities to anthropogenic contamination. BDF prescribes rates and weights for 7 data layers but FCF is an unsupervised learning framework based on a multicriteria decision theory by integrating fuzzy membership function and catastrophe theory. The challenges in the paper include: (i) the study area comprises confined and unconfined aquifers and (ii) Artificial Intelligence (AI) is used to run Multiple Framework (AIMF) in order to map specific vulnerability due to a specific contaminant. Predicted results by AIMF are referred to as Specific Vulnerability Indices, as the learned VIs are referenced to site-specific nitrate-N. The results show that correlation coefficient between BDF or FCF with nitrate-N is lower than 0.7 but the AIMF strategy improves it to greater than 0.95. The results are evidence for the proof-of-concept for transforming frameworks to models capable of further learning.

A comparison study of DRASTIC methods with various objective methods for groundwater vulnerability assessment

Groundwater vulnerability assessment is a measure of potential groundwater contamination for areas of interest. The main objective of this study is to modify original DRASTIC model using four objective methods, Weights-of-Evidence (WOE), Shannon Entropy (SE), Logistic Model Tree (LMT), and Bootstrap Aggregating (BA) to create a map of groundwater vulnerability for the Sari-Behshahr plain, Iran. The study also investigated impact of addition of eight additional factors (distance to fault, fault density, distance to river, river density, land-use, soil order, geological time scale, and altitude) to improve groundwater vulnerability assessment. A total of 109 nitrate concentration data points were used for modeling and validation purposes. The efficacy of the four methods was evaluated quantitatively using the Area Under the Receiver Operating Characteristic (ROC) Curve (AUC). AUC value for original DRASTIC model without any modification of weights and rates was 0.50. Modification of weights and rates resulted in better performance with AUC values of 0.64, 0.65, 0.75, and 0.81 for BA, SE, LMT, and WOE methods, respectively. This indicates that performance of WOE is the best in assessing groundwater vulnerability for DRASTIC model with 7 factors. The results also show more improvement in predictability of the WOE model by introducing 8 additional factors to the DRASTIC as AUC value increased to 0.91. The most effective contributing factor for ground water vulnerability in the study area is the net recharge. The least effective factors are the impact of vadose zone and hydraulic conductivity.

Introducing a new framework for mapping subsidence vulnerability indices (SVIs): ALPRIFT

Proof-of-concept (PoC) is presented for a new framework to serve as a proactive capability to mapping subsidence vulnerability of Shabestar plain of approximately 500km² overlaying an important aquifer supporting a region renowned for diversity of agricultural products. This aquifer is one of 12 in East and West Azerbaijan provinces, Northwest Iran, which surround the distressed Lake Urmia, with its water table declined approximately 4m in between 2004 and 2014. The decline of water table in aquifers undermines their soil texture and structure by exposure to pressures under their weight and thereby induce or trigger land subsidence. Inspired by the DRASTIC framework to map intrinsic aquifer vulnerability to anthropogenic pollution, the paper introduces the ALPRIFT framework for subsidence, which comprises the seven data layers of Aquifer media (A), Land use (L), Pumping of groundwater, Recharge (R), aquifer thickness Impact (I), Fault distance (F) and decline of water Table (T). The paper prescribes rates to account for local variations and weights for the relative importance of the data layers. The proof-of-concept for ALPRIFT is supported by the correlation of Subsidence Vulnerability Indices (SVIs) with measured subsidence values, which renders a value of 0.5 but improves significantly to 0.86 when using fuzzy logic. Similar improvements are suggested by the ROC/AUC performance metric.

Introducing the risk aggregation problem to aquifers exposed to impacts of anthropogenic and geogenic origins on a modular basis using 'risk cells'

Proof-of-concept is presented in this paper to a methodology formulated for indexing risks to groundwater aquifers exposed to impacts of diffuse contaminations from anthropogenic and geogenic origins. The methodology is for mapping/indexing, which refers to relative values but not their absolute values. The innovations include: (i) making use of the Origins-Source-Pathways-Receptors-Consequences (OSPRC) framework; and (ii) dividing a study area into modular Risk (OSPRC) Cells to capture their idiosyncrasies with different origins. Field measurements are often sparse and comprise pollutants and water table, which are often costly; whereas supplementary data are general-purpose data, which are widely available. Risk mapping for each OSPRC cell is processed by dividing a study area into pixels and for each pixel, the risk from both anthropogenic and geogenic origins are indexed by using algorithms related to: (i) Vulnerability Indices (VI), which identify the potential for risk exposures at each pixel; and (ii) velocity gradient, which expresses the potency to risk exposures across the risk cell. The paper uses DRASTIC for anthropogenic VI but introduces a new framework for geogenic VI. The methodology has a generic architecture and is flexible to modularise risks involving any idiosyncrasies in a generic way in any site exposed to environmental pollution risks. Its application to a real study area provides evidence for the proof-of-concept for the methodology by a set of results that are fit-for-purpose and provides an insight into the study area together with the identification of its hotspots.

Feature selection approaches for predictive modelling of groundwater nitrate pollution: An evaluation of filters, embedded and wrapper methods

Recognising the various sources of nitrate pollution and understanding system dynamics are fundamental to tackle groundwater quality problems. A comprehensive GIS database of twenty parameters regarding hydrogeological and hydrological features and driving forces were used as inputs for predictive models of nitrate pollution. Additionally, key variables extracted from remotely sensed Normalised Difference Vegetation Index time-series (NDVI) were included in database to provide indications of agroecosystem dynamics. Many approaches can be used to evaluate feature importance related to groundwater pollution caused by nitrates. Filters, wrappers and embedded methods are used to rank feature importance according to the probability of occurrence of nitrates above a threshold value in groundwater. Machine learning algorithms (MLA) such as Classification and Regression Trees (CART), Random Forest (RF) and Support Vector Machines (SVM) are used as wrappers considering four different sequential search approaches: the sequential backward selection (SBS), the sequential forward selection (SFS), the sequential forward floating selection (SFFS) and sequential backward floating selection (SBFS). Feature importance obtained from RF and CART was used as an embedded approach. RF with SFFS had the best performance (mmce=0.12 and AUC=0.92) and good interpretability, where three features related to groundwater polluted areas were selected: i) industries and facilities rating according to their production capacity and total nitrogen emissions to water within a 3km buffer, ii) livestock farms rating by manure production within a 5km buffer and, iii) cumulated NDVI for the post-maximum month, being used as a proxy of vegetation productivity and crop yield.

The problem of identifying arsenic anomalies in the basin of Sahand dam through risk-based 'soft modelling'

An investigation is undertaken to identify arsenic anomalies at the complex of Sahand dam, East Azerbaijan, northwest Iran. The complex acts as a system, in which the impounding reservoir catalyses system components related to Origin-Source-Pathways-Receptor-Consequence (OSPRC) viewed as a risk system. This 'conceptual framework' overlays a 'perceptual model' of the physical system, in which arsenic with geogenic origins diffused into the formations through extensive fractures swept through the region during the Miocene era. Impacts of arsenic anomalies were local until the provision of the impounding reservoir in the last 10years, which transformed it into active system-wide risk exposures. The paper uses existing technique of: statistical, graphical, multivariate analysis, geological survey and isotopic study, but these often seem ad hoc and without common knowledgebase. Risk analysis approaches are sought to treat existing fragmentation in practices of identifying and mitigating arsenic anomalies. The paper contributes towards next generation best practice through: (i) transferring and extending knowledge on the OSPRC framework; (ii) introducing 'OSPRC cells' to capture unique idiosyncrasies at each cell; and (iii) suggesting a 'soft modelling' procedure based on assembling knowledgebase of existing techniques with partially converging and partially diverging information levels, where knowledgebase invokes model equations with increasing resolutions. The data samples from the study area for the period of 2002-12 supports the study and indicates the following 'risk cells' for the study area: (i) local arsenic risk exposures at south of the reservoir, (ii) system-wide arsenic risks at its north; and (iii) system-wide arsenic risk exposures within the reservoir even after dilution.

A linear and non-linear polynomial neural network modeling of dissolved oxygen content in surface water: Inter- and extrapolation performance with inputs' significance analysis

Accurate prediction of water quality parameters (WQPs) is an important task in the management of water resources. Artificial neural networks (ANNs) are frequently applied for dissolved oxygen (DO) prediction, but often only their interpolation performance is checked. The aims of this research, beside interpolation, were the determination of extrapolation performance of ANN model, which was developed for the prediction of DO content in the Danube River, and the assessment of relationship between the significance of inputs and prediction error in the presence of values which were of out of the range of training. The applied ANN is a polynomial neural network (PNN) which performs embedded selection of most important inputs during learning, and provides a model in the form of linear and non-linear polynomial functions, which can then be used for a detailed analysis of the significance of inputs. Available dataset that contained 1912 monitoring records for 17 water quality parameters was split into a "regular" subset that contains normally distributed and low variability data, and an "extreme" subset that contains monitoring records with outlier values. The results revealed that the non-linear PNN model has good interpolation performance (R²=0.82), but it was not robust in extrapolation (R²=0.63). The analysis of extrapolation results has shown that the prediction errors are correlated with the significance of inputs. Namely, the out-of-training range values of the inputs with low importance do not affect significantly the PNN model performance, but their influence can be biased by the presence of multi-outlier monitoring records. Subsequently, linear PNN models were successfully applied to study the effect of water quality parameters on DO content. It was observed that DO level is mostly affected by temperature, pH, biological oxygen demand (BOD) and phosphorus concentration, while in extreme conditions the importance of alkalinity and bicarbonates rises over pH and BOD.

Mapping vulnerability of multiple aquifers using multiple models and fuzzy logic to objectively derive model structures

Driven by contamination risks, mapping Vulnerability Indices (VI) of multiple aquifers (both unconfined and confined) is investigated by integrating the basic DRASTIC framework with multiple models overarched by Artificial Neural Networks (ANN). The DRASTIC framework is a proactive tool to assess VI values using the data from the hydrosphere, lithosphere and anthroposphere. However, a research case arises for the application of multiple models on the ground of poor determination coefficients between the VI values and non-point anthropogenic contaminants. The paper formulates SCFL models, which are derived from the multiple model philosophy of Supervised Committee (SC) machines and Fuzzy Logic (FL) and hence SCFL as their integration. The Fuzzy Logic-based (FL) models include: Sugeno Fuzzy Logic (SFL), Mamdani Fuzzy Logic (MFL), Larsen Fuzzy Logic (LFL) models. The basic DRASTIC framework uses prescribed rating and weighting values based on expert judgment but the four FL-based models (SFL, MFL, LFL and SCFL) derive their values as per internal strategy within these models. The paper reports that FL and multiple models improve considerably on the correlation between the modeled vulnerability indices and observed nitrate-N values and as such it provides evidence that the SCFL multiple models can be an alternative to the basic framework even for multiple aquifers. The study area with multiple aquifers is in Varzeqan plain, East Azerbaijan, northwest Iran.

Assessment of groundwater vulnerability using supervised committee to combine fuzzy logic models

Vulnerability indices of an aquifer assessed by different fuzzy logic (FL) models often give rise to differing values with no theoretical or empirical basis to establish a validated baseline or to develop a comparison basis between the modeling results and baselines, if any. Therefore, this research presents a supervised committee fuzzy logic (SCFL) method, which uses artificial neural networks to overarch and combine a selection of FL models. The indices are expressed by the widely used DRASTIC framework, which include geological, hydrological, and hydrogeological parameters often subject to uncertainty. DRASTIC indices represent collectively intrinsic (or natural) vulnerability and give a sense of contaminants, such as nitrate-N, percolating to aquifers from the surface. The study area is an aquifer in Ardabil plain, the province of Ardabil, northwest Iran. Improvements on vulnerability indices are achieved by FL techniques, which comprise Sugeno fuzzy logic (SFL), Mamdani fuzzy logic (MFL), and Larsen fuzzy logic (LFL). As the correlation between estimated DRASTIC vulnerability index values and nitrate-N values is as low as 0.4, it is improved significantly by FL models (SFL, MFL, and LFL), which perform in similar ways but have differences. Their synergy is exploited by SCFL and uses the FL modeling results "conditioned" by nitrate-N values to raise their correlation to higher than 0.9.