Groundwater salinity modeling and mapping using machine learning approaches: a case study in Sidi Okba region, Algeria.

The groundwater salinization process complexity and the lack of data on its controlling factors are the main challenges for accurate predictions and mapping of aquifer salinity. For this purpose, effective machine learning (ML) methodologies are employed for effective modeling and mapping of groundwater salinity (GWS) in the Mio-Pliocene aquifer in the Sidi Okba region, Algeria, based on limited dataset of electrical conductivity (EC) measurements and readily available digital elevation model (DEM) derivatives. The dataset was randomly split into training (70%) and testing (30%) sets, and three wrapper selection methods, recursive feature elimination (RFE), forward feature selection (FFS), and backward feature selection (BFS) are applied to train the data. The resulting combinations are used as inputs for five ML models, namely random forest (RF), hybrid neuro-fuzzy inference system (HyFIS), K-nearest neighbors (KNN), cubist regression model (CRM), and support vector machine (SVM). The best-performing model is identified and applied to predict and map GWS across the entire study area. It is highlighted that the applied methods yield input variation combinations as critical factors that are often overlocked by many researchers, which substantially impacts the models' accuracy. Among different alternatives the RF model emerged as the most effective for predicting and mapping GWS in the study area, which led to the high performance in both the training (RMSE = 1.016, R = 0.854, and MAE = 0.759) and testing (RMSE = 1.069, R = 0.831, and MAE = 0.921) phases. The generated digital map highlighted the alarming situation regarding excessive GWS levels in the study area, particularly in zones of low elevations and far from the Foum Elgherza dam and Elbiraz wadi. Overall, this study represents a significant advancement over previous approaches, offering enhanced predictive performance for GWS with the minimum number of input variables.

Assessment of groundwater vulnerability using remote sensing, susceptibility index, and WetSpass model in an arid region (Biskra, SE Algeria).

Biskra region currently shows signs of stress and a high risk of groundwater contamination by various chemicals and pesticides. For this purpose, a modified integrated susceptibility index (SI) is coupled with remote sensing (RS) and WetSpass model to assess the sensitivity of the groundwater and the risk of pollution in the most exploited aquifer (Quaternary aquifer) in the study area. The results of the modified SI model show that a major part of the aquifer is at risk of contamination if the farmers do not implement good agricultural practices. Four sensitivity levels are considered, reflecting a vulnerability rating that ranges from low to very high. The very high category is observed in the agricultural areas with an estimated pollution index ranging from 84 to 90.57, while a large part of the aquifer shows a high vulnerability to contamination (64 < SI ≤ 84). This category is found in areas characterized by the dominance of bare soil. In urban areas, the vulnerability level decreases to low category (37 < SI ≤ 45). However, the area of forests is classified as moderate to vulnerability (45 < SI ≤ 64). The different statistical and GIS methods confirm the reliability of the obtained SI map. The combination of the SI method with WetSpass model and RS can give a reliable map to help and assist the authorities and decision-makers in groundwater resources planning and the implementation of monitoring programs and networks to control the quality of groundwater in arid environments.

Water quality index modeling using random forest and improved SMO algorithm for support vector machine in Saf-Saf river basin.

The water quality index is one of the prominent general indicators to assess and classify surface water quality, which plays a critical role in river water resources practices. This research constructs a hybrid artificial intelligence model namely sequential minimal optimization-support vector machine (SMO-SVM) along with random forest (RF) as a benchmark model for predicting water quality values at the Wadi Saf-Saf river basin in Algeria. The fifteen input water quality datasets such as biochemical oxygen demand (BOD), oxygen saturation (OS), the potential for hydrogen (pH), chemical oxygen demand (COD), chloride (Cl-), dissolved oxygen (DO), electrical conductivity (EC), total dissolved solids (TDS), nitrate-nitrogen (NO3-N), nitrite-nitrogen (NO2-N), phosphate (PO43-), ammonium (NH4+), temperature (T), turbidity (NTU), and suspended solids (SS) were employed for constructing the predictive models. Different input data combinations are evaluated in terms of predictive performance, using a set of statistical metrics and graphical representation. Results show that less than 40% of samples were observed to be poor quality water during the dry season in downstream northeastern part of the basin. The findings also show that the RF model mostly generates more precise water quality index predictions than the SMO-SVM model for both training and testing stages. Although thirteen input parameters attain the optimal predictive performance (R2 testing = 0.82, RMSE testing = 5.17), a couple of five input parameters, e.g., only pH, EC, TDS, T, and saturation, gives the second optimal predictive precision (R2 test = 0.81, RMSE testing = 5.55). The sensitivity analysis results indicate a greater sensitivity by the all input variables chosen except NO2- of the predictive outcomes to the earlier influencing water quality parameters. Overall, the RF model reveals an improvement on earlier tools for predicting water quality index, according to predictive performance and reducing in the number of input variables.