The Atlantic Multidecadal Oscillation in models and observations

Abstract

We use a simplified model of the North Atlantic ocean to study the Atlantic Multidecadal Oscillation (AMO), which is a low-frequency variation found in sea surface temperatures (SST) over the North Atlantic ocean. A mechanism for the AMO has previously been described; here we study the excitation of the variability when atmospheric noise is included in the system. It is found that spatial and temporal patterns in the atmospheric noise (such as the North Atlantic Oscillation (NAO) in the real atmosphere) can be effective at exciting the AMO. One important phenomenon associated with the minimal model mechanism is the westward propagation of temperature anomalies in the surface ocean. This propagation is observed in temperature measurements taken below the surface of the ocean in the North Atlantic. No firm conclusions about the time scale can be drawn due to the limited length of the observational record but it appears that the variability occurs on a time scale closer to 20-30 years than to 50-70 years, which is the period more usually associated with the AMO. The variability of temperature causes variability in sea surface height (SSH). Tide gauges on the North American and European sides of the North Atlantic both show coherent variability along the coast. The link between SSH and the AMO is made using results a coupled climate model (the GFDL CM2.1 GCM) which shows AMO-like variability in the North Atlantic on 20-30 year time scales. The longer time series of tide gauge observations that is available around the North Atlantic allows us to distinguish the 20-30 year period variability from the 50-70 year period. We found that the 20-30 year period is present in other data sets such as the Central England Temperature record and two records of net snow accumulation from ice cores in Greenland. The 20-30 year period can be found, along with the 50-70 year period, upon closer analysis of the SST record. The two time scales are also found in the control run of the GFDL CM2.1 GCM. The time scale of westward propagating anomalies in observations suggests that the shorter period is caused by the mechanism from the simple model but the observational record is too short to make any similar statements about the longer period variability. In the GFDL CM2.1 model there is variability of salinity in the Arctic ocean from which anomalies may propagate into the North Atlantic. Salinity variability in the Arctic was further investigated using a minimal model of multidecadal variability in the Arctic where the normal modes of the salinity driven flow were investigated. Under zero forcing the modes are all damped and stationary, as forcing is increased a merger between two stationary modes occurs, resulting in an oscillatory mode with a multidecadal period. The spatial pattern of this mode, bearing in mind the deformational effect of realistic bathymetry, resembles the salinity variability found in the GFDL CM2.1 model