Multi-scale evaluation of dynamical modes of climate variability in CMIP6 models

Abstract:

The accuracy with which climate models capture climate variability across scales is of vital importance for historical attribution, subseasonal to seasonal prediction, and assessment of future trends. Here, we evaluate the skill of 17 state-of-the-art Coupled Model Intercomparison Project phase 6 (CMIP6) models in reproducing the variability of several oceanic and atmospheric variables over scales of 1 day to several decades. We then focus on two specific modes of special interest for sub-seasonal to interannual variability, the Madden-Julian Oscillation (MJO) and the El Nino Southern Oscillation (ENSO). We use a complex wavelet-based rotated spectral Principle Component Analysis (rsPCA), recently developed in our group, which is able to remove trends non-parametrically, has optimal time-frequency localization properties, and uses a regularization in the complex PC space to optimally extract dynamical modes with robust propagation dynamics. We show that while CMIP6 models tend to overestimate the variance associated with the ENSO, the MJO variance is substantially undermined in most of the models and we document differences in the phases and space-time dynamics of MJO in these models. Our results demonstrate the need to better simulate the coupled ocean-atmosphere dynamics in order to improve subseasonal to seasonal prediction. Moreover, studies using projected states of MJO for assessing future tropical and extratropical impacts should be examined with caution.