Climate model diversity in the North Atlantic and its impact on prediction skill on interannual-to-decadal timescales

Description du projet

De meilleures prédictions du modèle climatique pour l’Atlantique Nord

Prédire avec précision le climat dans l’Atlantique Nord sur une échelle de temps interannuelle à décennale dépend de modèles simulant correctement les processus atmosphériques et océaniques, y compris l’oscillation nord-atlantique (NAO) et la circulation méridienne de retournement de l’Atlantique (AMOC). Le projet DivPredSkill, financé par l’UE, appliquera l’idée novatrice qui consiste à relier la prévisibilité aux biais du modèle via une meilleure compréhension des interactions entre le NAO et l’AMOC et en reliant les résultats de longues simulations non initialisées aux prédictions initialisées, fournissant de nouvelles informations pour la modélisation du climat et la communauté des prévisions. Plus précisément, le chercheur boursier en charge du projet identifiera les différences de modèle dans la relation NAO-AMOC simulée, évaluera le rôle des biais d’état moyen dans les différences de modèle et fera le lien entre ces résultats et les compétences en matière de prévision dans l’Atlantique Nord.

Objectif

Europe is vulnerable to climate extremes and climate change, making predictions valuable to political and economic stakeholders. The European winter climate depends a lot on sea level pressure differences between Iceland and the Azores, described by the North Atlantic Oscillation (NAO). Thus, many modelling centres working on predictions, focus on improving NAO predictions. Recent studies find the NAO could be highly predictable, much higher than suggested by individual models - calling for a better process understanding to improve models and predictions. The major source of our ability to make predictions for years and beyond are slow ocean processes, e.g. linked to the Atlantic Meridional Overturning Circulation (AMOC). Both, AMOC and NAO are considered key for North Atlantic decadal variability and predictability. They interact, especially via the NAO driving AMOC, but also vice versa. Consequently, prediction skill is reduced by uncertainties in the NAO-AMOC relationship, but also by severe mean state biases common among models. As of yet, not much effort has been made to understand skill difference across models - motivating the proposed research: The overarching research objective is to understand the causes for inter-model differences in interannual-to-decadal prediction skill - with a focus on the role of the model-dependent NAO-AMOC relationship and mean state biases. Specific objectives are to: (i) identify model differences in the NAO-AMOC interaction and their effect on differences of the potential to predict the North Atlantic; (ii) assess the role of mean state biases for model differences; (iii) link these findings with North Atlantic forecast skill. The proposed research comprises: the novel idea to relate predictability to model biases via NAO-AMOC interaction; and the innovative approach to link findings from long, uninitialised simulations to initialised predictions, enabling new insights valuable to the climate modelling and prediction community.