Predicting the Atlantic multidecadal variability from initialized simulations

Abstract

In contrast to dominant interannual time-scale variability in other ocean basins, the leading observed mode variability in the Atlantic is characterized as a basinwide seesaw-like sea surface temperature variability between the North and South Atlantic on a multidecadal time scale (approximately 60–80 years), known as the Atlantic multidecadal variability (AMV). AMV has been identified as a key driver for climate shifts that occurred in the mid-1960s and late 1990s. Here we attempt to predict the summer AMV by analyzing decadal prediction experiments from two climate models. Results show that these climate models with proper initialization do a better job than uninitialized historical runs, and are capable of predicting the observed AMV time evolution. Our models predict that the AMV will be in a neutral to slightly negative phase, leading to a warm–dry trend over western Europe and North Africa and a cold–wet trend (cold relative to the warming trend) over southeastern China and Indochina in the next few years.