A Coherent Approach to Evaluating Precipitation Forecasts over Complex Terrain

Abstract

Precipitation forecasts provided by high-resolution NWP models have a degree of realism that is very appealing to most users of meteorological data. However, it is a challenge to demonstrate whether or not such forecasts contain more skillful information than their lower resolution counterparts. A verification procedure must be based on equally detailed observations that are also realistic in areas where ground observations are not available and remote sensing data can only increase the accuracy of the location of rain events at the cost of decreased accuracy in estimating the amount of rain that has actually reached the ground. Traditional verification methods based on station or grid point comparison yield poor results for high-resolution fields due to the double penalty error that is attributed to finite space and time displacement that such methods do not account for. A complete approach to evaluating precipitation forecasts over complex terrain is suggested. The method is based on realistic gridded precipitation observations generated by an interpolation method that uses long climate data series to determine the geographical characteristics that this parameter is best correlated with as well as remote sensing estimates as background information to cover the areas where observations are insufficient. Spatial verification methodologies are subsequently applied to a convective event that accentuate the relative skill of high-resolution COSMOGR forecasts in revealing characteristics in the precipitation patterns such as structure and intensity.