Document Type : Original Article
Authors
1 Assistant Professor, Department of Civil Engineering, Materials and Energy Research Center, Dez.C., Islamic Azad University, Dezful, Iran.
2 Associate Professor, Department of Civil Engineering, Islamic Azad University, Khorramabad branch, Khorramabad, Iran
3 PhD in Water Sciences and Engineering, PhD in Water Sciences and Engineering, Soil Conservation and Watershed Management Research Department, Lorestan Agriculture and Natural Resources Research and Education Center, AREEO, Khorramabad, Iran
Abstract
Accurate estimation of groundwater salinity in the coastal areas of the Caspian Sea is of great importance due to challenges such as saltwater intrusion and spatiotemporal complexity. In this study, hybrid models based on support vector regression (SVR) combined with advanced optimization algorithms, including wavelet transform, algorithm of innovative gunner (AIG), and particle swarm optimization (PSO), were used to predict groundwater salinity in the coastal region of the Caspian Sea. The data included water quality parameters such as bicarbonate (HCO₃), sodium (Na), total hardness (TH), total dissolved solids (TDS), magnesium (Mg), potassium (K), pH, and calcium (Ca) as inputs, and electrical conductivity (EC) as the output parameter. The data were collected from monitoring wells in the region over a period from 2003 to 2023, using eight combined scenarios. To evaluate the performance of the models, correlation coefficient (R), root mean square error (RMSE), mean absolute error (MAE), and Nash-Sutcliffe coefficient (NSE) were used. The results showed that combined scenarios in the studied models improved their performance. The evaluation metrics also indicated that the support vector regression-wavelet (SVR-Wavelet) model achieved a correlation coefficient of 0.985, RMSE of 0.206 dS/m, MAE of 0.105 dS/m, and NSE of 0.990 in the validation phase. Sensitivity analysis revealed that TDS and total hardness (TH) had the most significant impact on the prediction accuracy of the models. This study confirms the effectiveness of hybrid approaches, particularly the combination of SVR with the cultural algorithm, for sustainable management of groundwater resources in coastal areas.
Keywords
Main Subjects
)