Long-term monthly evapotranspiration estimates from Brutsaert and Stricker's Advection-Aridity model were compared with independent estimates of evapotranspiration derived from long-term water balances for 139 undisturbed basins across the conterminous United States. On an average annual basis for the period 1962--1988 the original model, which uses a Penman wind function, underestimated evapotranspiration by 7.9% of precipitation compared with the water balance estimates. Model accuracy increased with basin humidity. An improved formulation of the model is presented in which the wind function and the Priestley-Taylor coefficient are modified. The wind function was reparameterized on a seasonal, regional basis to replicate independent proxy potential evapotranspiration surfaces. This led to significant differences from the original Penman wind function. The reparameterized wind function, together with a recalibrated Priestley-Taylor coefficient in the wet environment evapotranspiration formulation, reduced the underestimation of annual average evapotranspiration to only 1.15% of precipitation on an independent set of validation basins. The results offered here lend further support for Bouchet's hypothesis as it applies to large-scale, long-term evapotranspiration. ¿ 2001 American Geophysical Union