In this study, the performance of three empirical methods for estimating reference evapotranspiration (ET0): Makkink (Mak) and Priestley-Taylor (PT) (radiation-based) and Hargreaves-Samani (HARG) (temperature-based) were assessed in semi-and regions. The values of ET0 derived using these three methods were compared to those estimated using the reference FAO Penman-Monteith (FAO-PM) method under semi-arid conditions of the Tensift basin (central of Morocco) and the Yaqui Valley (Northwest Mexico). The results showed that the HARG method is the best one to estimate ET0 over both semi-and test sites. Conversely, the performance of the other two empirical methods was poor except under humid conditions. However when the parameters alpha and C-m figurate in the PT and Mak equations are locally calibrated, the performance of these two methods greatly improved. Additionally, this study showed that, when measurements of meteorological parameters needed for estimating ET0 (which are not always available especially in developing countries) are lacking, the climatic data generated with numerical weather prediction models provide an alternative and effective solution to estimate ET0. In this regard, data generated using a weather forecast model (ALADIN) over the Tensift basin showed that the HARG model is the most accurate one for estimating the spatio-temporal variability of ET0.