La prévision saisonnière et les modèles numériques de la circulation générale

Click to increase image sizeClick to decrease image size

Recent experiments (**PROVOST, ELMASIFA, SMIP, POTENTIALS... ) have shown that it is possible to obtain significant information about certain climatic characteristics, on 4-month time-scales, from numerical simulations of atmosphere models, whether or not they are coupled with ocean models. The nature of phenomena studied and the sensitivity of equation systems with regard to small perturbations, make the deterministic approach ineffective. In addition, estimates of such predictions show clearly the probabilistic nature of information which is helpful to seasonal forecast users who have to manage uncertainty like a gambler do, but with a responsibility. In addition to the fundamental scientific interest linked to numerical research, a major point concerns the usefulness, i.e. the value, for the user. This concept need not necessarily reveal the same conclusions as those found when using pure verification scores (anomaly correlation, etc.) GCM and seasonal forecast.