Evapotranspiration is a significant variable of a watershed hydrological and water quality processes for surface and subsurface modelling in a basin scale. Most of the calibration and validation hydrological parameters rely on streamflow discharge. Inaccuracy calculation of the PET could disturb all the simulation processes. Hence, this study aimed to evaluate the SWAT capability on the estimation of potential evapotranspiration in Kelantan characterised as forest-dominated topography in upstream catchment. The model calibration applied for the data set from 1985 to 2000 by utilising the SUFI-2 algorithm and validated for the period from 2001 to 2016 with three different PET methods: P-M, P-T, and HG. The model indicated that there was no significant difference in the value of NSE and R2. The calibration simulations give very good result in term of NSE and R2 with HG achieved 0.85, 0.86, P-T achieved 0.84, 0.86 and P-M achieved 0.83 and 0.85 respectively. However, HG tends to give lower simulated flow than P-T and P-M in terms of PBIAS were 2.5%, −5.9% and −1.9% respectively. While during validation, all three PETs indicated simulated higher than the observed with PBIAS were −23.3%, −29.3% and −28.5%. The P-M method tends to estimate lower PET than HG and P-T with yearly simulated was accounted for approximately 66%, 70%, and 71% respectively of annual precipitation. Overall, all the available options PETs methods in SWAT, give good performance model after calibration and validation.