One Two
Sie sind hier: Startseite WEX-MOP Phase 2 Joining physical process understanding and statistical modeling

Joining physical process understanding and statistical modeling

Christ Weijenborg, Petra Friederichs and Andreas Hense, Meteorological Institute, Universität Bonn

In the second phase of the project we intend to combine the dynamical knowledge gained in phase one with staistical postprocessing. The COSMO-DE-EPS provides large amounts of data containing full three dimensional multivariate information over the 21 hour forecast period. However, the full information is a too large data set and thus cannot be used in the postprocessing. While purely statistical methods such as the (Bayesian) Lasso may be used to select a subset of informative predictors, physical reasoning is needed for the Lasso’s a priori selection process. In light of this, our aim is to perform postprocessing on an optimized subset of information that combines our physical process understanding with statistical predictor selection.