How Do Biases in General Circulation Models Affect Projections of Aridity and Drought?

Abstract:

Unless corrected, biases in General Circulation Models (GCMs) can affect hydroclimatological applications and projections. Compared to a raw GCM ensemble (direct GCM output), bias-corrected GCM inputs correct for systematic errors and can produce high-resolution projections that are useful for impact analyses. By examining the difference between raw and bias-corrected GCMs for the continental United States, this work highlights how GCM biases can affect projections of aridity (defined as precipitation (P)/potential evapotranspiration (PET)) and drought (using the Palmer Drought Severity Index (PDSI)). At the annual time scale for spatial averages over the continental United States, the raw GCM ensemble median has a historical positive precipitation bias (+24%) and negative PET bias (-7%) compared to the bias-corrected output. While both GCM ensembles (raw and bias-corrected) result in drier conditions in the future, the bias-corrected GCMs produce enhanced aridity (number of months with PET>P) in the late 21^st century (2070-2099) compared to the historical climate (1950-1979). For the western United States, the bias-corrected GCM ensemble estimates much less humid and sub-humid conditions (based on P/PET categorical values) than the raw GCM ensemble. However, using June, July, and August PDSI, the bias-corrected GCM ensemble projects less acute decreases for the southwest United States compared to the raw GCM ensemble (1 to 2 PDSI units higher) as a result of larger decreases in projected precipitation in the raw GCM ensemble. A number of examples and ecological implications of this work for the western United States will be presented.