ABSTRACT: Groundwater is used for drinking and irrigation purposes in many parts of the world. Irrigationpractices result in the deterioration of the quality of the groundwater over the time and this adversely affects the human health and plant growth. Monitoring of the vulnerable aquifers with cost-effectivemethods is important. In this study an artificial neural network model is proposed for the prediction ofsodium absorption ratio (SAR) in the unconfined aquifer, located in the east of Ergene basin. The samplestaken from 18 observation wells were analysed monthly for electrical conductivity, total dissolved solids,temperature, total hardness, chloride and pH. Levenberg–Marquardt (trainlm) was selected for backpropagation algorithm and 35 neurons were used in the model architecture. The model follows up theexperimental data very closely (R= 0,855). Application of the proposed model would make possible tomonitor the aquifers in a more cost-effective and easier way