Potential Evapotranspiration as a Source of Uncertainty and Bias in Hydrologic Impact Analyses

Abstract:

The diversity of commonly used potential evapotranspiration (PET) models contributes uncertainty in the estimation of hydrologic response to anthropogenic climate change. The temperature sensitivity of six commonly used PET equations (Hamon, Oudin, Penman-Monteith, Priestley-Taylor, Samani-Hargreaves, and Thornthwaite) is readily shown to vary by almost an order of magnitude, with energy-unconstrained (i.e., temperature-based) methods showing the largest sensitivity. The change in annual multimodel (Coupled Model Intercomparison Project, Phase 5) PET under Representative Concentration Pathway 8.5 from 1981-2000 to 2081-2100 is typically 10-20% (20-40%) in the low (high) latitudes according to the physics-based Penman-Monteith (ASCE Standardized Reference Evapotranspiration) equation, but 20-40% (20-80%) according to the empirical, temperature-based Hamon equation. Radiation-based Priestley-Taylor changes are smaller than both of these, while empirical, temperature-based Thornthwaite changes are larger than both. These differences in PET change translate to large differences in change of water availability; when combined with a form of the Budyko water-balance relation, the PET methods predict a wide range of runoff changes. Furthermore, all PET methods result in bias that indicates drier conditions globally than those computed by the climate models themselves, and all PET methods overestimate the changes in actual evapotranspiration in non-water-stressed seasons/regions relative to the changes in the climate models. We conclude that use of PET methods that are inappropriate for climate-change applications is a source not only of uncertainty, but also of more drying than suggested by climate models, in hydrologic impact analyses. In view of the bias, it is advised that a no-PET-change analysis be used to define a wet upper bound on potential hydrologic impacts.