Measurements made in the past few decades undeniably indicate change in the climate. The most visible sign of global climate change is air temperature, while less visible indicators include changes in river water temperatures. Changes in river temperature can significantly affect the environment, primarily the biosphere. The physical, biological and chemical characteristics of the river are directly affected by water temperature, although estimation of this relationship presents a complex problem. Although river temperature is influenced by hydrological and meteorological factors, the purpose of this study is to model daily water temperature using only one known parameter, mean air temperature. The relationship between the daily mean air and daily water temperature of the River Drava in Croatia is analysed using linear regression, stochastic modelling or nonlinear regression and multilayer perceptron (MLP) feed-forward neural networks. The results indicate that the MLP models are much better models which can be used for the estimation and prediction of daily mean river temperature.