Single model initial-condition large ensembles (SMILEs) are a powerful tool for separating internal climate variability and forced climate change. SMILEs consist of individual climate models that have been run many times (20-100) under identical external forcing from slightly perturbed initial conditions. Here, I present work that leverages the power of these tools to investigate the El Niño Southern Oscillation (ENSO), using the new multi model large ensemble archive dataset (MMLEAv2). I first demonstrate that for highly variable metrics such as ENSO a model might evaluate poorly or favourably compared to the single realisation the observations represent, depending on the chosen ensemble member. Additionally I show that for robust detection of future changes to ENSO teleconnections model internal variability must be accounted for.

Second, I assess the temporal evolution of ENSO SST variability projections and find large model differences in ENSO SST variance itself, it’s pattern, evolution, and ENSO diversity. By assessing the evolution of ENSO variability and the mean state in combination, model differences are reduced, highlighting that the models tend to evolve in the same way on varying timescales. Third, I show how these differences in ENSO projections influence and confound the projections of other modes of variability in the Northern Hemisphere, and their precipitation impacts. Finally, I briefly show examples of other research using SMILEs to obtain a more accurate picture of ENSO. This work involves investigating predictability of ENSO using a model analogues technique and understanding rainfall risk related to ENSO events in Australia. Overall, I demonstrate the power of SMILEs for investigating highly variable phenomenon such as ENSO.

Lieu
Jussieu, salle du LOCEAN-IPSL, couloir 45-55, 4ème étage, pièce 417

Visio
https://visio.numerique.gouv.fr/ovk-eije-lnv