Modeling of groundwater quality index by using artificial intelligence algorithms in northern Khartoum State, Sudan

Abstract

In the present study, multilayer perceptron (MLP) neural network and support vector regression (SVR) models were developed to assess the suitability of groundwater for drinking purposes in the northern Khartoum area, Sudan. The groundwater quality was evaluated by predicting the groundwater quality index (GWQI). GWQI is a statistical model that uses sub-indices and accumulation functions to reduce the dimensionality of groundwater quality data. In the first stage, GWQI was calculated using 11 physiochemical parameters collected from 20 groundwater wells. These parameters include pH, EC, TDS, TH, Cl−, SO4−2, NO3−, Ca+2, Mg+2, Na+, and HCO3−. The primary investigation confirmed that all parameters except for EC and NO3− are beyond the standard limits of the World Health Organization (WHO). The measured GWQI ranged from 21 to 396. As a result, groundwater samples were classified into three classes. The majority of the samples, roughly 75%, projected into the excellent water category; 20% were considered good water and 5% were classified as unsuitable. GWQI models are powerful tools in groundwater quality assessment; however, the computation is lengthy, time-consuming, and often associated with calculation errors. To overcome these limitations, this study applied artificial intelligence (AI) techniques to develop a reliable model for the prediction of GWQI by employing MLP neural network and SVR models. In this stage, the input data were the detected physiochemical parameters, and the output was the computed GWQI. The dataset was divided into two groups with a ratio of 80% to 20% for models training and validation. The predicted (AI) and actual (calculated GWQI) models were compared using four statistical criteria, namely, mean square error (MSE), root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination (R2). Based on the obtained values of the performance measures, the results revealed the robustness and efficiency of MLP and SVR models in modeling GWQI. Consequently, groundwater quality in the north Khartoum area is evaluated as suitable for human consumption except for BH 18, where highly mineralized water is observed. The developed approach is advantageous in groundwater quality evaluation and is recommended to be incorporated in groundwater quality modeling.

Machine Intelligence in Africa: a survey

“…The work in [52] investigated groundwater quality in northern Khartoum State, Sudan, utilizing MI algorithms. The authors employed multilayer perceptron neural network and support vector regression to assess groundwater suitability for drinking.…”

Multi-criterion approach for assessment of hydrochemical characteristics and human health risk of groundwater in Khartoum State, Sudan

“…The most crucial step in creating the GWQI model is allocating weights for each parameter 8 . The correlation coefficient between the physiochemical parameters is used in this research to determine the appropriate weights for each parameter.…”

“…The term groundwater quality index (GWQI) refers to a numerical value that can be calculated to characterize and classify groundwater using physical, chemical, and biological parameters dissolved in groundwater 8 . The classification of water quality using GWQI is more efficient than conventional approaches in which the detected parameters are only compared to a given water quality standard 9 .…”

Management of Agricultural Groundwater in Sudan: The Use of Artificial Intelligence Algorithms in Khartoum State

“…As a result, groundwater levels range from 294 m in the southern parts to 366.6 m in the central part (Figure 2). Consequently, groundwater flows mainly from the western to the eastern and from northern to southern parts of the region [17].…”