Open AccessAccounting for initial condition uncertainties in COSMO‐DE‐EPS
Author(s) -
Peralta C.,
Ben Bouallègue Z.,
Theis S. E.,
Gebhardt C.,
Buchhold M.
Publication year - 2012
Publication title -
journal of geophysical research: atmospheres
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.67
H-Index - 298
eISSN - 2156-2202
pISSN - 0148-0227
DOI - 10.1029/2011jd016581
Subject(s) - ensemble forecasting , probabilistic logic , ensemble average , statistical ensemble , ensemble learning , statistical physics , numerical weather prediction , environmental science , meteorology , computer science , mathematics , canonical ensemble , physics , statistics , monte carlo method , climatology , artificial intelligence , geology
This study describes the experimental setup of an atmospheric ensemble prediction system (EPS) based on a high resolution numerical weather prediction model developed at Deutscher Wetterdienst (DWD). The focus is set on uncertainties of initial conditions, and how to combine them with perturbations of the boundary conditions and model physics variations, discussed in a previous investigation. Two ensemble setups are constructed: one with model physics diversity and variations of boundaries (BP ensemble) and one which also includes perturbations of initial conditions (IBP ensemble). Experimental ensemble forecasts of precipitation with a lead time of 24 hours, for a period of 29 days in the summer of 2009, are generated. Deterministic verification shows that the inclusion of the initial perturbations does not decrease the quality of the forecast, with individual members of the IBP and BP ensembles being statistically similar. The probabilistic scores of the IBP ensemble are better than those of the BP ensemble during the first 12 hours of the forecast, and afterwards both have similar performances. Similarly, the IBP ensemble provides more spread during the first 12 hours of the forecast, decreasing to levels comparable to the BP ensemble afterwards.