Attribution of Predictive Skill Along the Atlantic Water Pathway

The variable heat transport of the Gulf Stream’s extension toward the Arctic Ocean influences western European climate, Arctic sea ice conditions, and northern fisheries. It would accordingly be most beneficial if one could skilfully predict the state of the ocean several years into the future. In this study we assess to what extent six different dynamical prediction systems can predict the winter sea surface temperature (SST) in the subpolar North Atlantic and the Nordic Seas in the time period 1960-2010. We focus in particular on the region where warm water flows poleward; the Atlantic water pathway. To understand why there is predictive skill or lack thereof, we confront the models with a mechanism identified from observations – an oceanic teleconnection from low to high latitudes – on each lead time from one year to ten years ahead. This observed mechanism shows that warm and cold anomalies propagate along the Atlantic water pathway with a certain time lag, and that this occurred repeatedly over the last 60 years. We find that most models represent this mechanism to some extent, but mostly on longer lead times. The reason for this might be that predictions on longer lead times allow the time for correctly initialised SST anomalies in the subtropics to be advected northwards. The propagation mechanism appears to be a source for skill, but with large model differences in location and timing. However, long-term trends also contribute to skill on longer lead times and we aim to disentangle how much of the skill that can be attributed to the propagation mechanism. Furthermore, the mean state of surface ocean velocities and surface heat fluxes at each lead time are investigated to better understand why there is predictive skill or lack of skill.