On the predictability of North Atlantic ocean state

This study investigates the decadal predictability of the ocean climatic state in the North Atlantic in an ocean-forced context. To assess this oceanic predictability, we compute Linear Optimal Perturbations (LOPs) in a realistic Ocean General Circulation Model in a global configuration (NEMO-ORCA2) and estimate the maximum impact of small disturbances on ocean dynamics. Our calculations of LOPs involve a maximization procedure based on Lagrangian multipliers in a non-autonomous context. As the metrics of the ocean state we use four different measures: the Meridional Volume Transport (MVT), the Meridional Heat Transport (MHT), the mean Sea Surface Temperature (SST), and the Oceanic Heat Content (OHC), all in the North Atlantic. We show that the four metrics are dramatically different in regard to predictability. Whereas the SST and OHC can be modified only by relatively large-scale anomalies, the MVT and MHT are strongly affected by small-scale anomalies as well (acting along the basin eastern and western basin boundaries and changing the East-West density difference across the Atlantic). This suggests that MVT and MHT are much less predictable than SST and OHC. It is only when MVT and MHT are averaged over climatically relevant timescales (e.g. 30 years) that the four metrics have comparable predictability. This result stresses the need for long-term measurements of the Atlantic Meridional Overturning Circulation intensity and its associated heat transport in order to have climatically relevant data. Our study also suggests that initial errors of a few milli-Kelvins can lead on a decadal timescale to an error of 0.1 K in North Atlantic mean sea surface temperature estimates. This transient error growth is maximal after about 17 years and can be interpreted as a decadal predictability barrier.