SCALE ISSUES IN QPF VERIFICATION

Efi Foufoula-Georgiou

St. Anthony Falls Laboratory and Department of Civil Engineering

University of Minnesota

Mississippi River at 3^rd Avenue SE

Minneapolis, MN 55414

Phone: 612-627-4595; E-mail: efi@tc.umn.edu

Observations typically available for QPF verification are at scales different from the grid scale of a numerical forecast model. The discrepancy between observation scales (i.e., the spacing between samples for rain gauges or the integration volume of a sample for radar estimates) and modeling scales (i.e., the grid size of a model) makes QPF verification challenging. This is especially so since precipitation variability highly depends on the scale at which the process is considered (referred to as multi-scale variability). This talk will discuss problems that arise when typical methodologies, based on spatial interpolation or spatial averaging, are used to transform point observations to the scale of the model output (e.g., point-to-area conversion) or vice-versa (area-to-point conversion). It will also present preliminary results from a newly developed framework capable of accounting for the multi-scale variability of precipitation such that observations and model outputs at different scales can be compared while explicitly accounting for their inherent scale-dependent variability and observational uncertainty. This framework is especially appealing when observations at more than one scale (e.g., rain gauges, radars and satellites) are available for model verification.