Evaluation of a Technique for Downscaling Climate-Model Output in Mountainous Terrain Using Local Topographic Lapse Rates

Abstract:

One of the challenges in using general circulation model (GCM) output is the need to downscale beyond the model’s coarse spatial grid in order to infer climate at any particular location. Traditionally, downscaling has been achieved either dynamically, through regional climate models (RCMs), or statistically, through empirical relationships between predictor variables in the GCM and observed variables. In mountainous terrain, elevation is one of the primary controls on temperature and precipitation at the local scale, which provides the potential for topographic variables to be used to adjust climate-model output. Here, local topographic lapse rates (LTLR) were estimated from gridded climate data for the Pacific Northwest, and those lapse rates were used to downscale RCM output. Skill scores were calculated for the LTLR-downscaled climate-model output relative to an existing set of model output downscaled using the well-established statistical downscaling technique of bias-corrected constructed analogs (BCCA). Spatial and temporal patterns in forecast skill and in bias of the LTLR downscaling method were also examined. The results indicate that the LTLR method performs well in the mountainous study region relative to the BCCA method. There is variability in the forecast skill, however, most notably the LTLR downscaling technique’s better performance in the eastern part of the study region for temperature and in the western part of the study region for precipitation. LTLR downscaling offers a promising method for downscaling climate-model output in regions in which elevation is a strong control on climate, particularly for studying impacts of past or future climate change.