The Role of Eddies in the North Atlantic Decadal Variability

The role of eddies in the North Atlantic decadal variability is investigated in this study by using two ocean reanalyses, including an eddy permitting (or eddy poor) reanalysis with horizontal resolution of 0.25 degree and 75 vertical levels and an eddy resolving (or eddy rich) reanalysis with horizontal resolution of 1/12 degree and 50 vertical levels. The prominent mid-1990s warming and post-2005 cooling trend as part of the North Atlantic decadal variability is well displayed in both reanalyses with no significant difference between them. The main driver of the mid-1990s warming and post-2005 cooling trend is the increase and reduction of the meridional ocean heat transport showing similar patterns in both reanalyses. The relative contribution of the heat transport anomalies from eddies to the total heat transport anomalies is slightly larger in eddy resolving than in eddy permitting ocean reanalysis. However, the total mean ocean meridional heat transport increases by 10% in eddy resolving reanalysis with respect to eddy permitting reanalysis and is mainly due to the associated increase of the mean states (temperature and velocity). Therefore, the increase of eddy population due to the increase of horizontal resolution, found by comparing the two datasets, does not affect the MHT anomalies significantly and, consequently, the North Atlantic decadal variability. It is found that the importance of the model horizontal resolution for the North Atlantic decadal variability depends on the interaction between the eddies (small scale) and the mean state (large scale) at decadal time scales. Although the fast increase of computational power will allow soon for eddy-resolving predictions, the need to use high resolution modeling tools for decadal predictions depends on the importance of initialization methods and the interaction between small scale and large-scale variabilities. This study has pivotal implications for the development of North Atlantic decadal prediction systems.