In the present work, two modelling and forecasting techniques were evaluated on the basis of: a) their efficiency to forecast and b) their ability to utilize auxiliary environmental information: ARIMA and Transfer Function (TF) models. For this purpose, almost three years (2005–2007) daily measurements of Water Temperature (Tw) data in four different depths (1 m, 20 m, 40 m and 70 m) of Thesaurus dam-lake in River Nestos, were used to obtain the best models for these time series. For all seven models, the Mean Square Error (MSE) was calculated and used to evaluate the accuracy of the forecast and to compare the forecasting ability of each approach.