Several methods allow the determination of evapotranspiration (ET), either by direct measurement or by estimation from weather data. The most used estimation method is the FAO56 (FAO-PM), based on the concept of reference evapotranspiration (ETo) and crop coefficient (Kc). The eddy covariance technique (EC) developed for measurement of convective fluxes between land surface and the atmosphere is commonly used to estimate ET. However, these methods were established under standard conditions in flat terrain, and their use in hilly areas is questionable. In this work, the variability of ETo and ET measured by EC and energy balance (EB) in a hilly area of northern Tunisia is studied for different relief configurations. The experiment was conducted using a meteorological (M), and three EC-EB measurement stations in three wheat fields. Two stations were installed on opposite slopes of a ridge with apposing aspects and moderate slopes (A, B), and one station was installed in a flat site (C). Results of monitoring during the mid-season of wheat growth showed similar ETo levels in all sites for hourly and daily time steps with relative RMSE in the range 0.03-0.08 compared to M. Average ET values in sloping fields (A, B) were, respectively, 15% and 10% lower than in (C). However, hourly values of ET/ETo obtained from EC measurements were smaller than FAO-Kc, contrarily to those based on the EB method.