Analysis of self-organizing maps and explainable artificial intelligence to identify hydrochemical factors that drive drinking water quality in Haor region

Identification of salinity sources in groundwater at Golgohar Mine using self-organizing maps (SOM) and correlation analysis: a hydrogeochemical and isotopic approach, south-central Iran

Understanding the hydrogeological regime and identifying the sources of salinity in mine pits are crucial for effective groundwater management and mining operations. This study investigates the hydrogeochemistry of the Golgohar Mines in south-central Iran, focusing on Mine Pit No. 2 and the Kheirabad Salt Pan, using a multi-proxy approach that includes Self-Organizing Maps (SOM), Hierarchical Clustering Analysis (HCA), hydrochemical diagrams, and stable isotopic analysis (δ2H, δ18O) and tritium. 20 sampling points were selected for hydrogeochemical analyses, 16 sampling points for stable isotopes of δ18O and δ2H, and 4 samples for the tritium radioisotope, with sampling conducted during both wet and dry seasons. Based on ionic ratio diagrams, halite dissolution is the predominant process contributing to salinity, with additional influences from reverse cation exchange and gypsum dissolution. The main type and facies of all groundwater samples in the region are Na-Cl and Ca-Cl. The alignment of self-organizing maps with HCA and the hydrochemical classification identified four distinct water clusters: Cluster I is associated with the Kheirabad Salt Pan, Cluster II originates from the hard rock aquifer, Cluster III results from the mixing of multiple sources, including the alluvial aquifer, hard rock aquifer, and brine from the Kheirabad Salt Pan, and Cluster IV is linked to the alluvial aquifer. A hydraulic and hydrogeochemical relationship was observed between the Kheirabad Salt Pan and Mine Pit No. 2, while the main source of salinity in the hard rock aquifer appears to be unrelated to the Kheirabad Salt Pan, potentially involving extensive mixing with other waters. The study highlights the effectiveness of SOM and HCA in elucidating hydrochemical processes in complex mining environments, offering valuable insights for groundwater management.

Spatial pattern of groundwater chemistry in a typical piedmont plain of Northern China driven by natural and anthropogenic forces

Groundwater is crucial for human society's development in piedmont plains, yet its hydrogeochemistry often exhibits complex spatial distributions due to the interplay of nature and human factors. Ninety-two phreatic groundwater samples were collected from a typical piedmont plain in northern China and analyzed using self-organizing map combined with hydrogeochemical simulation, diagrams, and the entropy-weighted water quality index. Groundwater samples were categorized into four clusters, demonstrating a gradual hydrogeochemical facies evolution from HCO3-Ca to Cl-Mg·Ca and Cl-Na, along with an increase in NO3- content in the order of clusters IV, II, III, and I. Natural processes, including silicates weathering and reverse cation-exchange, establish the natural fundamental framework of groundwater chemistry, which is furtherly sculptured by agricultural substances input. Groundwater quality was predominantly excellent or good, with entropy-weighted water quality index (EWQI) values below 100 at over 92% of the sampling sites. Groundwater quality is relatively poorer in the upstream areas near the mountains and along the Hutuo River, where the stratum permeability is high, but improves in the downstream areas where permeability is lower. Agricultural land use and spatial variation in aquifer permeability are responsible for the observed spatial variations in groundwater chemistry. Agricultural contaminants warrant attention for the protection of groundwater quality in piedmont plains that with long-term agricultural activities, especially in the upstream areas near the mountains. This research improves the understanding of the spatial distribution and variation of groundwater chemistry in piedmont plains, and provides scientific guidance for related groundwater development and management.

Innovative strategies for pollution assessment in Northern Bangladesh: Mapping pollution areas and tracing metal(loid)s sources in various soil types

This study assessed the risks of soil pollution by heavy metals in Chilmari Upazila, northern Bangladesh, using the static environmental resilience (Pi) model of soil. Geostatistical modeling and self-organizing maps (SOM) identified pollution areas and spatial patterns, while a positive matrix factorization (PMF) model revealed pollution sources. The results showed that the average concentrations of Cr, Pb and As were well above background levels. Agricultural and industrial soils were mainly contaminated with Cr, Pb and As according to the Nemerow Pollution Index (NPI), Ecological Risk (ER) and Pi Index. Over 70% of the sites were contaminated with Pb and Cr, while co-contamination was particularly high. A one-way ANOVA showed significant correlations between Pb, Cu and Zn levels and human activities. The PMF analysis revealed that industrial effluents, agrochemicals and lithogenic sources were the main contributors to soil contamination with 16%, 41% and 43%, respectively. The SOM analysis revealed three distinct spatial patterns (Pb-Zn, Cr-Cu-Ni and Co-Mn-As), which are consistent with the PMF results. These results emphasize the need for stringent measures to reduce industrial emissions and remediate soil contamination in order to improve soil quality and food security.

Optimizing coastal groundwater quality predictions: A novel data mining framework with cross-validation, bootstrapping, and entropy analysis

Investigating the potential of novel data mining algorithms (DMAs) for modeling groundwater quality in coastal areas is an important requirement for groundwater resource management, especially in the coastal region of Bangladesh where groundwater is highly contaminated. In this work, the applicability of DMA, including Gaussian Process Regression (GPR), Bayesian Ridge Regression (BRR) and Artificial Neural Network (ANN), for predicting groundwater quality in coastal areas was investigated. The optuna-based optimized hyperparameter is proposed to improve the accuracy of the models, including optuna-GPR and optuna-BRR as benchmark models. Combined cross-validation (CV) and bootstrapping (B) methods were used to build six predictive models. The entropy-based coastal groundwater quality index (ECWQI) was converted into a normalized index (ECWQIn), which was divided into five classes from very poor to excellent. The self-organizing map (SOM), spatial autocorrelation and fuzzy logic model were used to identify spatial groundwater quality patterns based on 12 physicochemical variables collected from 67 groundwater wells. The SOM analysis identified four distinct spatial patterns, including EC-TDS-Cl-, MgpH, Ca2+K+NO₃-, and HCO₃-SO₄2-Na+F-. The results showed that both the ANN (CV) and ANN (B) models performed better than other optuna-based models during the test phase (RMSE = 0.041, MAE = 0.026, R2 = 0.971, RAE = 0.15 = 21 and CC = 0.986) and (RMSE = 0.041, MAE = 0.025, R2 = 0.969, RAE = 0.119 and CC = 0.975), respectively. SO42-, Cl- and F- played an important role in the prediction accuracy. F- and SO42- showed higher spatial autocorrelation, which affected groundwater quality degradation. In addition, the ANN (CV) and ANN (B) models showed a Gaussian distribution of model errors (small standard error, <1 %), indicating the stability of the model. These results indicate the efficiency of the ANN model in predicting groundwater quality in coastal areas, which would help regional water managers in real-time monitoring and management of sustainable groundwater resources.

Fluoride contamination in African groundwater: Predictive modeling using stacking ensemble techniques

Fluoride contamination of groundwater is a severe public health problem in Africa due to natural factors that include geological weathering of fluoride-bearing minerals and climatic conditions characterized by high evaporation rates that highly elevate fluoride levels. Anthropogenic activities further aggravate the problem and have affected millions of people in countries such as; South Africa, Tanzania, Nigeria, Ethiopia, Ghana, Kenya, Mauritania, Botswana, and Egypt. High fluoride levels of up to 10 mg/L have been encountered in parts of the East African Rift Valley, above the WHO's recommended limit of 1.5 mg/L, causing serious dental and skeletal fluorosis among the affected people. In this study, the distributions of F- in groundwater of Africa were forecast using an advanced stacking ensemble learning model based on 11 crucial groundwater physiochemical variables and 6270 accessible statistics of observed concentrations. The enhanced algorithm incorporates randomized trees, Tree-Bag, RF, DT, XGB, and ET Machine as base trainees, with a simple Naïve Bayes as the meta-analyzer. The model's AUC score of 0.86 accurately represented the uneven distributions of groundwater fluoride. The results showed that 20-35 % of the continent's eastern part and 10 % of its western region are at risk of having fluoride levels exceeding WHO limits, with an expected population of around 80 million. Regionally, fluoride contamination ranges from 0.1 to 3 mg/L in West Africa was range from 0.0 to 13.29 mg/L, 0.01-588 mg/L in East Africa, 0.04-65.9 mg/L in South Africa, and 0.1-10.5 mg/L in North and 0.01-1.9 mg/L in Central Africa. Na+ and HCO3- are Africa's leading primary causes of fluoride contamination, with Ca2+ and Cl- contributing to fluoride influence in some parts of the continent. This study helped identify health concerns linked to groundwater fluoride and offered guidance on assessing health risks in areas with sparse sample sizes.

Water quality classification using self-organizing maps and cluster analysis: Case of Meknes-El Hajeb Springs, Morocco

The Ouislane sub-watershed is currently experiencing severe water shortages and is highly dependent on its water supply. The sub-watershed spans two communes: Meknes to the north and El Hajeb to the south. It serves as the primary water source for irrigation and drinking purposes for the local population. Consequently, it is crucial to assess the spatio-temporal variations of water quality to identify and address potential gaps; these focused on effective monitoring systems to detect contaminants, pollutants and health risks. This research project aims on the application of self-organizing map (SOM) techniques combined with cluster analysis to classify water quality in springs for drinking and irrigation purposes. The present study evaluates the water quality variations using physicochemical parameters of twelve water springs, collected during the wet and dry seasons of 2022. For this purpose, the water quality index (WQI), self-organizing map (SOM), hierarchical cluster analysis (HCA), and principal component analysis (PCA) are used as evaluation and classification methods. As a result, the SOM algorithm with a size of 5 × 5 units identified as the most suitable, based on the minimum quantization error (QE) and topographic error (TE), yielding a QE of 0.379 and a TE of 0.000. It grouped the water quality data into five distinct clusters, Cluster I represented 37.5% of the total samples, while cluster II represented 25%. Cluster III and IV each accounted for 8.33% of the samples, while 20.83% of the sampling water are classified in cluster V. Clusters I, II, and IV indicate good water suitable for drinking. However, cluster V had the highest WQI, suggesting very high contamination due to increased levels of the 10 studied physicochemical parameters. The water quality in this region (cluster V) is influenced by natural processes, such as precipitation intensity, weathering and vegetation cover, as well as anthropogenic factors like agriculture and urban concentration. PCA confirmed the clustering results obtained by SOM. However, SOM provides a more detailed classification and additional insights into the dominant variables influencing the classification processes. The results of this study suggest that SOM was an effective tool for gaining a better understanding of the patterns and processes driving water quality in the Ouislane sub-watershed and provides valuable avenues for further research to establish and monitor water quality for effective management of water resources.

Predicting groundwater phosphate levels in coastal multi-aquifers: A geostatistical and data-driven approach

The groundwater (GW) resource plays a central role in securing water supply in the coastal region of Bangladesh and therefore the future sustainability of this valuable resource is crucial for the area. However, there is limited research on the driving factors and prediction of phosphate concentration in groundwater. In this work, geostatistical modeling, self-organizing maps (SOM) and data-driven algorithms were combined to determine the driving factors and predict GW phosphate content in coastal multi-aquifers in southern Bangladesh. The SOM analysis identified three distinct spatial patterns: K+Na+pH, Ca2+Mg2+NO₃-, and HCO₃-SO₄2-PO43-F-. Four data-driven algorithms, including CatBoost, Gradient Boosting Machine (GBM), Long Short-Term Memory (LSTM), and Support Vector Regression (SVR) were used to predict phosphate concentration in GW using 380 samples and 15 prediction parameters. Forecasting accuracy was evaluated using RMSE, R2, RAE, CC, and MAE. Phosphate dissolution and saltwater intrusion, along with phosphorus fertilizers, increase PO43- content in GW. Using input parameters selected by multicollinearity and SOM, the CatBoost model showed exceptional performance in both training (RMSE = 0.002, MAE = 0.001, R2 = 0.999, RAE = 0.057, CC = 1.00) and testing (RMSE = 0.001, MAE = 0.002, R2 = 0.989, RAE = 0.057, CC = 0.998). Na+, K+, and Mg2+ significantly influenced prediction accuracy. The uncertainty study revealed a low standard error for the CatBoost model, indicating robustness and consistency. Semi-variogram models confirmed that the most influential attributes showed weak dependence, suggesting that agricultural runoff increases the heterogeneity of PO43- distribution in GW. These findings are crucial for developing conservation and strategic plans for sustainable utilization of coastal GW resources.

Tracing groundwater nitrate sources in an intensive agricultural region integrated of a self-organizing map and end-member mixing model tool

The traceability of groundwater nitrate pollution is crucial for controlling and managing polluted groundwater. This study integrates hydrochemistry, nitrate isotope (δ15N-NO3- and δ18O-NO3-), and self-organizing map (SOM) and end-member mixing (EMMTE) models to identify the sources and quantify the contributions of nitrate pollution to groundwater in an intensive agricultural region in the Sha River Basin in southwestern Henan Province. The results indicate that the NO3--N concentration in 74% (n = 39) of the groundwater samples exceeded the WHO standard of 10 mg/L. According to the results of EMMTE modeling, soil nitrogen (68.4%) was the main source of nitrate in Cluster-1, followed by manure and sewage (16.5%), chemical fertilizer (11.9%) and atmospheric deposition (3.3%). In Cluster-2, soil nitrogen (60.1%) was the main source of nitrate, with a significant increase in the contribution of manure and sewage (35.5%). The considerable contributions of soil nitrogen may be attributed to the high nitrogen fertilizer usage that accumulated in the soil in this traditional agricultural area. Moreover, it is apparent that most Cluster-2 sampling sites with high contributions of manure and sewage are located around residential land. Therefore, the arbitrary discharge and leaching of domestic sewage may be responsible for these results. Therefore, this study provides useful assistance for the continuous management and pollution control of groundwater in the Sha River Basin.

Appraising water resources for irrigation and spatial analysis based on fuzzy logic model in the tribal-prone areas of Bangladesh

The lack of quality water resources for irrigation is one of the main threats for sustainable farming. This pioneering study focused on finding the best area for farming by looking at irrigation water quality and analyzing its location using a fuzzy logic model on a Geographic Information System platform. In the tribal-prone areas of Khagrachhari Sadar Upazila, Bangladesh, 28 surface water and 39 groundwater samples were taken from shallow tube wells, rivers, canals, ponds, lakes, and waterfalls. The samples were then analyzed for irrigation water quality parameters like electrical conductivity (EC), total dissolved solids (TDS), sodium adsorption ratio (SAR), soluble sodium percentage (SSP), residual sodium bicarbonate (RSBC), magnesium hazard ratio (MHR), Kelley's ratio (KR), and permeability index (PI). Fuzzy Irrigation Water Quality Index (FIWQI) was employed to determine the irrigation suitability of water resources. Spatial maps for parameters like EC, KR, MH, Na%, PI, SAR, and RSBC were developed using fuzzy membership values for groundwater and surface water. The FIWQI results indicate that 100% of the groundwater and 75% of the surface water samples range in the categories of excellent to good for irrigation uses. A new irrigation suitability map constructed by overlaying all parameters showed that surface water (75%) and some groundwater (100%) in the northern and southwestern portions are fit for agriculture. The western and central parts are unfit for irrigation due to higher bicarbonate and magnesium contents. The Piper and Gibbs diagram also indicated that the water in the study area is magnesium-bicarbonate type and the primary mechanism of water chemistry is controlled by the weathering of rocks, respectively. This research pinpoints the irrigation spatial pattern for regional water resource practices, identifies novel suitable areas, and improves sustainable agricultural uses in tribal-prone areas.

Interpreting optimised data-driven solution with explainable artificial intelligence (XAI) for water quality assessment for better decision-making in pollution management

In Saudi Arabia, water pollution and drinking water scarcity pose a major challenge and jeopardise the achievement of sustainable development goals. The urgent need for rapid and accurate monitoring and assessment of water quality requires sophisticated, data-driven solutions for better decision-making in water management. This study aims to develop optimised data-driven models for comprehensive water quality assessment to enable informed decisions that are critical for sustainable water resources management. We used an entropy-weighted arithmetic technique to calculate the Water Quality Index (WQI), which integrates the World Health Organization (WHO) standards for various water quality parameters. Our methodology incorporated advanced machine learning (ML) models, including decision trees, random forests (RF) and correlation analyses to select features essential for identifying critical water quality parameters. We developed and optimised data-driven models such as gradient boosting machines (GBM), deep neural networks (DNN) and RF within the H2O API framework to ensure efficient data processing and handling. Interpretation of these models was achieved through a three-pronged explainable artificial intelligence (XAI) approach: model diagnosis with residual analysis, model parts with permutation-based feature importance and model profiling with partial dependence plots (PDP), accumulated local effects (ALE) plots and individual conditional expectation (ICE) plots. The quantitative results revealed insightful findings: fluoride and residual chlorine had the highest and lowest entropy weights, respectively, indicating their differential effects on water quality. Over 35% of the water samples were categorised as 'unsuitable' for consumption, highlighting the urgency of taking action to improve water quality. Amongst the optimised models, the Random Forest (model 79) and the Deep Neural Network (model 81) proved to be the most effective and showed robust predictive abilities with R2 values of 0.96 and 0.97 respectively for testing dataset. Model profiling as XAI highlighted the significant influence of key parameters such as nitrate, total hardness and pH on WQI predictions. These findings enable targeted water quality improvement measures that are in line with sustainable water management goals. Therefore, our study demonstrates the potential of advanced, data-driven methods to revolutionise water quality assessment in Saudi Arabia. By providing a more nuanced understanding of water quality dynamics and enabling effective decision-making, these models contribute significantly to the sustainable management of valuable water resources.

Hydrogeochemical evolution processes, groundwater quality, and non-carcinogenic risk assessment of nitrate-enriched groundwater to human health in different seasons in the Hawler (Erbil) and Bnaslawa Urbans, Iraq

The main objectives of this research are to assess groundwater, a primary source of drinking water in the urban areas of Hawler (Erbil) and Bnaslawa in northern Iraq, and the non-carcinogenic human health risks of nitrate contamination associated with drinking water quality. For this purpose, twenty-seven groundwater samples were collected from wells to assess the hydrogeochemical characteristics and groundwater quality for both natural and anthropogenic purposes during the wet (May 2020) and dry (September 2020) seasons. During the wet and dry seasons, NO3- in groundwater ranged from 14.00 to 61.00 mg/L and 12.00 to 60.00 mg/L, with an average value of 35.70 and 29.00 mg/L, respectively. Approximately 25.92% of the samples exceeded the permissible limit of the WHO (2011) drinking water standard. The ratios of NO3-/Na+ vs. Cl-/Na+ and SO42-/Na+ vs. NO3-/Na+ indicate the effect of agricultural activities and wastewater leaking from cesspools or septic tanks on the quality of groundwater during the wet and dry seasons. The entropy weighted water quality index method ranked 62.5% and 75% of the urban groundwater as not recommended for drinking, and the remaining samples are moderately suitable in both wet and dry seasons. The non-carcinogenic human health risk assessment displayed that during the wet and dry seasons, 29.6% and 25.9% of adults, 48% and 30% of children, and 48.1% and 29.6% of infants were exposed to increased concentrations of nitrate in groundwater. Due to high nitrate in drinking water, non-carcinogenic human health risk levels vary as infant > child > adults. The main findings obtained from this study can assist policymakers in better understanding the hydrogeochemical properties of groundwater in terms of drinking water safety, thereby facilitating the management of water resources to take the necessary measures.

Using unsupervised machine learning models to drive groundwater chemistry and associated health risks in Indo-Bangla Sundarban region

Groundwater is an essential resource in the Sundarban regions of India and Bangladesh, but its quality is deteriorating due to anthropogenic impacts. However, the integrated factors affecting groundwater chemistry, source distribution, and health risk are poorly understood along the Indo-Bangla coastal border. The goal of this study is to assess groundwater chemistry, associated driving factors, source contributions, and potential non-carcinogenic health risks (PN-CHR) using unsupervised machine learning models such as a self-organizing map (SOM), positive matrix factorization (PMF), ion ratios, and Monte Carlo simulation. For the Sundarban part of Bangladesh, the SOM clustering approach yielded six clusters, while it yielded five for the Indian Sundarbans. The SOM results showed high correlations among Ca2+, Mg2+, and K+, indicating a common origin. In the Bangladesh Sundarbans, mixed water predominated in all clusters except for cluster 3, whereas in the Indian Sundarbans, Cl--Na+ and mixed water dominated in clusters 1 and 2, and both water types dominated the remaining clusters. Coupling of SOM, PMF, and ionic ratios identified rock weathering as a driving factor for groundwater chemistry. Clusters 1 and 3 were found to be influenced by mineral dissolution and geogenic inputs (overall contribution of 47.7%), while agricultural and industrial effluents dominated clusters 4 and 5 (contribution of 52.7%) in the Bangladesh Sundarbans. Industrial effluents and agricultural activities were associated with clusters 3, 4, and 5 (contributions of 29.5% and 25.4%, respectively) and geogenic sources (contributions of 23 and 22.1% in clusters 1 and 2) in Indian Sundarbans. The probabilistic health risk assessment showed that NO3- poses a higher PN-CHR risk to human health than F- and As, and that potential risk to children is more evident in the Bangladesh Sundarban area than in the Indian Sundarbans. Local authorities must take urgent action to control NO3- emissions in the Indo-Bangla Sundarbans region.