Correlation of the expansion coefficient of EOF-1 with a five-year mean near-surface temperature (top) and the spectrum of the expansion coefficient, K (bottom)
Figure 1. Correlation of the expansion coefficient of EOF-1 with a five-year mean near-surface temperature (top) and the spectrum of the expansion coefficient, K (bottom). The dashed line represents the significance level of spectral peaks at 99%.
Data from a 500-year preindustrial control run of climate model INMCM4 show distinct climate variability in the Arctic and North Atlantic with a period of 35–50 years. The variability can be seen as anomalies of upper ocean density that appear in the Arctic and propagate to the North Atlantic. The density gradient in a northeast–southwest direction alternates with the density gradient in a northwest–southeast direction. A positive density anomaly in the Arctic is associated with a positive salinity anomaly, a positive surface temperature anomaly and a reduction of sea ice in the Barents and Kara Seas. The nature of the variability is a vertical advection of density by thermal currents similar to that proposed in Dijkstra et al (2008 Phil. Trans. R. Soc. A 366). The cycle of model variability shows that after a negative anomaly of density in the northwest Atlantic, one should expect warming in the Arctic in 5–10 years. The ensemble of decadal predictions with climate model INMCM4 starting from 1995 shows that warming in the western Arctic and especially in the Barents Sea observed in 1996–2010 can be reproduced by eight of ten ensemble members. Arctic climate predictability in this case is associated with a proposed mechanism of a 35–50 year North Atlantic–Arctic oscillation.