Limited-area short-range ensemble predictions targeted for heavy rain in Europe

摘要

Inherent uncertainties in short-range quantitative precipitation forecasts(QPF) from the high-resolution, limited-area numerical weather predictionmodel DMI-HIRLAM (LAM) are addressed using two different approaches tocreating a small ensemble of LAM simulations, with focus on prediction ofextreme rainfall events over European river basins. The first ensemble typeis designed to represent uncertainty in the atmospheric state of the initialcondition and at the lateral LAM boundaries. The global ensemble predictionsystem (EPS) from ECMWF serves as host model to the LAM and provides thestate perturbations, from which a small set of significant members isselected. The significance is estimated on the basis of accumulatedprecipitation over a target area of interest, which contains the riverbasin(s) under consideration. The selected members provide the initial andboundary data for the ensemble integration in the LAM. A second ensembleapproach tries to address a portion of the model-inherent uncertaintyresponsible for errors in the forecasted precipitation field by utilisingdifferent parameterisation schemes for condensation and convection in theLAM. Three periods around historical heavy rain events that caused orcontributed to disastrous river flooding in Europe are used to study theperformance of the LAM ensemble designs. The three cases exhibit differentdynamic and synoptic characteristics and provide an indication of theensemble qualities in different weather situations. Precipitation analysesfrom the Deutsche Wetterdienst (DWD) are used as the verifying reference anda comparison of daily rainfall amounts is referred to the respective riverbasins of the historical cases.