Statistical downscaling of atmospheric variables

How much information about local-scale weather can be gained from large-scale atmospheric models? This work investigates downscaling methods for temporally high-resolution, normally distributed target variables. The performed procedures, based on model output statistics, are particularly appropriate when only short-term observations (i.e., few-years) are available. A case study is shown, with air temperature and humidity records from automatic weather stations in the Cordillera Blanca (Peru) representing the target variables, and reanalysis data from different institutions the predictors. The goal is to extend information available from the existing short-term observations into the past, in order to provide a better understanding about high-resolution atmospheric variability in the Cordillera Blanca. In principle, the downscaling techniques are portable to different sites or variables, as long as a few years of observational data are available for training. Yet we recommend their use primarily for skill assessment of large-scale predictors, and the subsequent application of results in impact studies only if high skill is demonstrated.

Autorentext

Marlis Hofer was born in 1982 in Brixen (Italy). After a diploma study of meteorology in Innsbruck (Austria), she continued with a PhD and a Post-Doc in the same field. She participated to numerous meteorological training courses in Europe and America, and was responsible of glacier field work campains in the Cordillera Blanca (Peru).