Abstract
An empirical statistical model is constructed to assess the forecast skill and the linear predictability of Atlantic Ocean sea surface temperature (SST) variability. Linear inverse modeling (LIM) is used to build a dynamically based statistical model using observed Atlantic SST anomalies between latitudes 20°S and 66°N from 1870 to 2009. LIM allows one to fit a multivariate red-noise model to the observed annually averaged SST anomalies and to test it. Forecast skill is assessed and is shown to be O(3–5 yr). After a few years, the skill is greatly reduced, especially in the subpolar region. In the stable dynamical system determined by LIM, skill of annual average SST anomalies arises from four damped eigenmodes. The four eigenmodes are shown to be relevant in particular for the optimal growth events of SST variance, with a pattern reminiscent of the low-frequency mode of variability, and in general for the predictability and variability of Atlantic SSTs on interannual time scales. LIM might serve as a useful benchmark for interannual and decadal forecasts of SST anomalies that are based on numerical models.
Laure Zanna
Professor of Mathematics & Atmosphere/Ocean Science [She/Her]
My research interests include Climate Dynamics, Physical Oceanography and Data Science.