Suitability of Global Circulation Model Downscaled BCCA Daily Precipitation for Local Hydrologic Applications

Abstract

Monthly precipitation hindcasts/projections for various climate change scenarios have been available for over a decade. More recently, daily precipitation hindcasts/projections using downscaled bias-corrected constructed analog (BCCA) have garnered wide usage for climate change investigations. Dry bias and drizzle effect reported in previous studies are among known limitations of downscaled BCCA precipitation hindcasts/projections which affect end-user applications. End users also require a reliable representation of precipitation sequence as well as physical consistency of precipitation series for climate change impact assessment applications. This study focuses on the evaluation of the daily sequence of downscaled BCCA precipitation hindcasts. Herein, direct use of BCCA daily precipitation hindcasts are further assessed to determine their suitability to study daily soil moisture dynamics at the field-scale for central Oklahoma climatic conditions. Three daily precipitation data sets were considered: (1) the 1961–1999 BCCA precipitation hindcasts for a 12 km grid in central Oklahoma, (2) the 1961–1999 spatially interpolated daily precipitation data used in the BCCA downscaling procedure, and (3) the 1961–1999 observed daily precipitation observations at the Weatherford COOP Weather Station located within the 12 km BCCA grid. The BCCA daily precipitation hindcasts showed a larger number of rainy days, lower rainfall amounts per rainy day, and longer sequences of consecutive rainy-day clusters than found in observations. These differences were large enough to suggest that BCCA daily precipitation hindcasts may not reflect the characteristics of actual precipitation observations at a point location, i.e., weather station. Additional analyses at Walnut Creek, Iowa and Moorhead, Mississippi confirmed the results are not location specific but numerically induced as a result of the downscaling process. The underlying cause for the noted differences was traced back to the differences in spatial scales of the BCCA outputs and observed daily precipitation at a station. Thus, caution is advised to end users using BCCA daily rainfall hindcasts/projections directly in local and field-scale water investigations, particularly for applications requiring reliable representation of precipitation sequence. Alternatively, a statistical downscaling method based on stochastic weather generation that includes wet-day dry-day transition probabilities would provide the desired temporal disaggregation and sequencing of daily rainfall events.