The reference evapotranspiration (ET o ) is an important variable for hydrological studies, crop water requirements estimations, climatic zonification and water resources management. The FAO recommends the Penman-Monteith (P-M) and/or the Hargreaves models as the worldwide useful for the ET o calculation. The objective of this work was to test the performance of these models in one place of the Colombian Coffee Belt, and identify limitations and proposes modifications. The ET o calculation were compared with daily lysimeter measurements. The principal disadvantages of the P-M model were: the lack of calibrated coefficient for the long wave radiation estimation (R nl ) which affected seriously the net radiation estimation and finally the ET o , highly sensitivity at the wind speed changes, that make it inappropriate for locations without this data. The Hargreaves model, as FAO proposed, overestimate the ET o , which made necessary a modification. The ET o estimation for this location was most sensible to atmospheric vapour and air temperature than the available energy in the atmosphere (R n -G).