Title: Using wavelets to verify the correlation structure of meteorological forecast fields
Authors: Petra Friederichs - University of Bonn (Germany)
Sebastian Buschow - University of Bonn (Germany) [presenting]
Abstract: When predicting meteorological fields at high spatial resolutions, one important aspect of forecast quality is their spatial correlation structure. While current weather models and post-processing techniques may be unable to foresee the precise timing and location of small-scale event, a fair verification should nonetheless reward their progress in representing the overall spatial pattern. Since naive point-wise approaches fail to reward highly resolved forecasts in the presence of displacement errors, numerous so-called spatial verification techniques have emerged. The purpose is to deal with a recently developed methodology for the spatial verification of deterministic, as well as ensemble forecasts, based on discrete wavelet transformations: By projecting observed and predicted fields on a new set of basis functions with varying spatial scale, orientation and location, we can estimate a local wavelet spectrum at each grid point. These spectra intuitively summarize the distribution of spatial variability across scales and directions and can (under appropriate assumptions) directly be related to the spatial covariances themselves. We will briefly introduce the relevant theoretical background before demonstrating how the wavelet-approach can be used to analyse and compare precipitation structures in high-resolution ensemble forecasts over Germany.