Assessing potential evapotranspiration methods in future drought projections across Canada

Abstract

Recently, concerns have arisen as to whether temperature-based proxy methods used to estimate potential evapotranspiration (PET) are reliable when examining future drought severity, especially in the context of a warmer climate. The objective of this study was to assess the effect of different PET approaches, focusing on proxies for radiation and humidity, on future Standardized Precipitation Evapotranspiration Index (SPEI) calculations across Canada. Using output from 22 CMIP6 global climate models (GCMs), seasonal and annual SPEI comparisons were carried out between the physically-based Penman-Monteith (PM) method and two approaches that incorporate temperature proxies to calculate radiation and/or humidity. These included the temperature-based Hargreaves (HG) approach and a PM method with derived humidity (PM-m). Results revealed that although the general patterns of SPEI projections across Canada were consistent among the methods, notable spatial and temporal differences were apparent. Specifically, both median and extreme SPEI projections based on the two temperature proxy methods revealed less annual and summer drying in much of central, eastern, and northern regions of Canada when compared to the physically based SPEI-PM. In extreme western regions (British Columbia, Yukon) these two methods, particularly HG, projected drier conditions. Differences of using temperature derived radiation and humidity were also most apparent in spring (and to a lesser degree, autumn), where the HG approach overestimated spring drying (and autumn wetting) over large regions of the country. Overall, differences tended to be more pronounced for the fully temperature-based HG approach during all periods considered. Results from this study strongly suggest that when possible, a physically-based approach be used when estimating PET to assess future drought projections. If a temperature proxy is used, the differences to a physically-based method should be understood and resultant implications be evaluated.