The development of meso- scale numerical weather prediction (NWP) models requires a substantial investment in research, development and computational resources. Traditional objective verification of deterministic model output fails to demonstrate the added value of high-resolution forecasts made by such models. It is generally accepted from subjective verification that these models nevertheless have a predictive potential for small-scale weather phenomena and extreme weather events. This has prompted an extensive body of research into new verification techniques and scores aimed at developing mesoscale performance measures that objectively demonstrate the return on investment in meso- NWP.
In this article it is argued that the evaluation of the information in mesoscale forecasts should be essentially connected to the method that is used to extract this information from the direct model output. This could be an evaluation by a forecaster, but, given the probabilistic nature of small-scale weather, is more likely a form of statistical postprocessing. Using Model Output Statistics (MOS) and traditional verification scores the potential of this approach is demonstrated both on an educational abstraction and a real world example. The MOS approach for this article incorporates concepts from fuzzy verification. This MOS approach objectively weighs different forecast quality measures and as such it is an essential extension of fuzzy methods.