Providing forecast of water balance components such as precipitation, evapotranspiration, deep percolation and runoff is important for water management and irrigation scheduling. Reference evapotranspiration (ETo) prediction will greatly enhance our capability to manage high-frequency irrigation systems and shallow-rooted crops. Reference evapotranspiration can be calculated on daily or hourly basis using analytical models (Penman-Monteith, Penman, etc.) and meteorological forecasts from numerical weather prediction models. One can also use time series analysis of ETo and meteorological variables related to evapotranspiration process. For example, autoregressive integrated moving average (ARIMA) models and artificial neural networks (ANN) can be applied in time series modeling and forecasting. The main aims of this study were to analyze and compare the performance of the above-mentioned techniques in short-term prediction of hourly and daily ETo. Reference evapotranspiration rates were calculated using the hourly Penman-Monteith equation, weather data provided by the Agrometeorological Service of Sardinia, Italy (SAR), and weather forecasts from a limited area model (BOLAM2000). Both ARIMA and ANN models were developed using four years of hourly meteorological data from three meteorological stations of SAR. Models were validated using a two-year data set from the same locations. The accuracy of models was evaluated comparing the forecasts with ETo values calculated using observed weather data from SAR weather stations. The use of meteorological variables from numerical weather forecast gave better results than those obtained from ARIMA and ANN models. The Limited Area Model gave root mean squared difference values of the forecasted ETo smaller than 0.15 mm on a hourly basis and near 1.0 mm on a daily basis. However, the analysis showed a large scatter of calculated versus predicted ETo values, in particular for hourly values. The evaluation of the effect of weather forecast variables on forecast ETo accuracy showed that solar irradiance is the main parameter affecting ETo forecast.