Breakdown of NAO reproducibility into internal versus externally-forced components: a two-tier pilot study

Abstract

Assessing the ability of atmospheric models to capture observed climate variations when driven by observed sea surface temperature (SST), sea ice concentration (SIC) and radiative forcings is a prerequisite for the feasibility of near term climate predictions. Here we achieve ensembles of global atmospheric simulations to assess and attribute the reproducibility of the boreal winter atmospheric circulation against the European Centre for Medium Range Forecasts (ECMWF) twentieth century reanalysis (ERA20C). Our control experiment is driven by the observed SST/SIC from the Atmospheric Model Intercomparison Project. It is compared to a similar ensemble performed with the ECMWF model as a first step toward ERA20C. Moreover, a two-tier methodology is used to disentangle externally-forced versus internal variations in the observed SST/SIC boundary conditions and run additional ensembles allowing us to attribute the observed atmospheric variability. The focus is mainly on the North Atlantic Oscillation (NAO) variability which is more reproducible in our model than in the ECMWF model. This result is partly due to the simulation of a positive NAO trend across the full 1920–2014 integration period. In line with former studies, this trend might be mediated by a circumglobal teleconnection mechanism triggered by increasing precipitation over the tropical Indian Ocean (TIO). Surprisingly, this response is mainly related to the internal SST variability and is not found in the ECMWF model driven by an alternative SST dataset showing a weaker TIO warming in the first half of the twentieth century. Our results may reconcile the twentieth century observations with the twenty-first century projections of the NAO. They should be however considered with caution given the limited size of our ensembles, the possible influence of other sources of NAO variability, and the uncertainties in the tropical SST trend and breakdown between internal versus externally-forced variability.