Performance Evaluation of M5 and MLP models to TDS estimation (A case study: Lighvan Watershed)

TDS parameter is the main factor in water quality assessment and it has many effects on the material movement، chemical alteration and ionization. Also، TDS density has important role in optimum water allocation for drinking water، agricultural and industrial uses. However، the aim of this study is TDS prediction in a case study in Lighvan Watershed using M5 and MLP models. The study area is one of the sub-basin or watershed in Urmia Lake basin and monthly data collected from two hydrometric stations during 1972 to 2010. Generally، based on the criteria which evaluate the models، M5 and MLP had the good performance in the TDS prediction and also mostly they had the same results. In the Lighavn station، among 6 scenarios، the fifth and sixth scenarios showed the highest correlation coefficient between observed and estimated series based on the inputs combination in two models. Due to these scenarios in the most errors criteria were closed together، the fifth and sixth scenarios introduced as the best scenarios. In the Hervi station the results showed that 5th scenario was the best scenario as model inputs. In general، 5th scenario with HCO-3، Ca2+، Na+، Mg2+ parameters showed the best results for the TDS prediction.