Groundwater quality assessment for irrigation in coastal region (Güzelyurt), Northern Cyprus and importance of empirical model for predicting groundwater quality (electric conductivity)

Abstract

Variations in precipitation patterns, the exact moment of aquifer recharge, and methods of groundwater extraction and usage all contribute to variations in the quality of groundwater used for irrigation between regions. Therefore, the present study aims to evaluate the seasonal groundwater suitability for irrigation purposes in the Güzelyurt region, Northern Cyprus for the first time using indexical techniques. The results demonstrated that \({Na}^{+}\) values, which ranged from 6 to 15 mg/l, were within allowable limits. Besides, the concentrations of \({Ca}^{2+}\) and \({Cl}^{-}\) that ranged from 2 to 7 mg/l and 4 to 12 mg/l, respectively, were appropriate for irrigation. However, \({Mg}^{2+}\) concentrations between 4 and 9 mg/l exceed FAO requirements. The \({N{O}_{3}}^{-}\) levels of 20–80 mg/l raised concerns about pollution and salinity. Furthermore, \({HC{O}_{3}}^{-}\) and \({S{O}_{4}}^{2-}\) concentrations fell between 3 and 8 mg/l and 2 and 5 mg/l, respectively, within safe limits. Additionally, the results showed that most of the samples are in the “suitable” or “excellent” category, which means that the water quality is generally appropriate for irrigation, according to the IWQI. However, there are observable declines in water quality after the monsoon season, especially in sodicity and sodium levels, which can negatively impact soil quality and crop production. This highlights the significance it is to maintaining irrigation systems and ensuring agricultural yield over time by effectively controlling water quality. Moreover, hydrogeological features, irrigation return water, maritime invasion, and aquifer communication can all be connected to the region’s groundwater salinity as identified by Electric Conductivity (EC). Therefore, a novel method based on Multi-Layer Perceptron Neural Network (MLP), K-Nearest Neighbor Algorithm (KNN), Support Vector Regression (SVR), and Non-Linear Neural Network Ensemble (NL-NNE) models optimized by Whale Optimization Algorithm (WOA) is proposed in this work for determining seasonally EC as a function of groundwater quality, groundwater depth, and weather parameters. The results demonstrated that the NL-NNE may increase the average performance of a single model during the verification phase. This showed that NNE’s potential ability to solve nonlinear processes supported its resilience and reliability in modeling EC.