Forecasting of Groundwater Total Dissolved Solids in Khan Younis City

Abstract

Wells of groundwater is the main resource of water supplies in Gaza Strip. The elevated water consumption, the rainfall shortage and the intrusion of seawater in some aquifers increases the creation of total dissolved solids (TDS) in water. The early forecasting of TDS levels in wells provides more timely information for better management of groundwater resources. In this paper we applied various techniques on timeseries data to forecast the TDS levels of wells in Khan Younis city as a case study. For this purpose, annual timeseries data were collected for seven wells from main two sources to predict TDS amounts for five years in advance. Five different forecasting models which are autoregressive integrated moving average (ARIMA), k-nearest neighbors (KNN), linear regression (LR), random forests (RF) and artificial neural networks (ANN) were applied individually on each well data. According to the mean absolute percentage error (MAPE) metric, the best results of the applied models on six wells (Ahrash, Ayia, El Amal, Riada City, UN Khan Younis and Islamic Relief) is ARIMA (MAPE ratios is between 3% and 9.8%). While the RF model has the best MAPE ratio (5.5%) in the seventh well (Abu Khaled). The results demonstrated that in the next five-year period of each well according to its horizon, the TDS ratio will increase for Ayia, Riadia City, UN khan Younis, Ahrash and Abu Khalid in comparison with the previous five-year period by 22.1%, 5.6%, 4.4%, 2.7% and 1.6% respectively. In contrast, the TDS levels in Islamic Relief and El Amal wells will decrease by 7.2% and 2.1% respectively. The results of this study contribute in better management, planning and risk minimization of water quality.