Weather types prediction at medium-range from ensemble forecasts
Résumé
Medium-range weather forecasts can be of
high economic value in many fields: agriculture, renewable
energy production, maintenance operations planning.
Such forecasts can be based on ensembles derived
from weather models, and the postprocessing of such
ensembles is an active research problem in the statistical
weather community. In this work, we try to face the
problem of long forecasting horizons, and focus on the
multivariate case where different meteorological variables
interact. The prediction problem is simplified and defined
as the prediction of a weather type, which is a categorical
variable defined by the interaction of the meteorological
variables. We use machine learning techniques to predict
this weather type from the multivariate ensemble
forecasts. The algorithms are applied to a 5 to 10 days
weather forecasting in the north-west of France, based on
wind and precipitation data from the ECMWF ensemble
system.
Domaines
Statistiques [math.ST]
Origine : Fichiers éditeurs autorisés sur une archive ouverte