ABSTRACT: Potential evapotranspiration (PET) is an important index of hydrologic budgets at different spatial scales and is a critical variable for understanding regional biological processes. It is often an important variable in estimating actual evapotranspiration (AET) in rainfall-runoff and ecosystem modeling. However, PET is defined in different ways in the literature and quantitative estimation of PET with existing mathematical formulas produces inconsistent results. The objectives of this study are to contrast six commonly used PET methods and quantify the long term annual PET across a physiographic gradient of 36 forested watersheds in the southeastern United States. Three temperature based (Thornthwaite, Hamon, and Hargreaves-Samani) and three radiation based (Turc, Makkink, and Priestley-Taylor) PET methods are compared. Long term water balances (precipitation, streamflow, and AET) for 36 forest dominated watersheds from 0.25 to 8213 km2 in size were estimated using associated hydrometeorological and land use databases. The study found that PET values calculated from the six methods were highly correlated (Pearson Correlation Coefficient 0.85 to 1.00). Multivariate statistical tests, however, showed that PET values from different methods were significantly different from each other. Greater differences were found among the temperature based PET methods than radiation based PET methods. In general, the Priestley-Taylor, Turc, and Hamon methods performed better than the other PET methods. Based on the criteria of availability of input data and correlations with AET values, the Priestley-Taylor, Turc, and Hamon methods are recommended for regional applications in the southeastern United States.