Defining the internal component of Atlantic multidecadal variability in a changing climate

The canonical index of "Atlantic Multidecadal Variability" (AMV) is the low-pass filtered timeseries of sea surface temperature anomalies (SSTA) averaged over the North Atlantic. This index and its associated SSTA spatial pattern confound externally forced climate change and internally generated climate variability. The internal component of AMV is commonly isolated by either subtracting the global-mean SSTA or removing the pattern of SSTA associated with the global-mean. This study evaluates the skill of each method with regard to the spatial pattern of internal AMV, using nine coupled model Large Ensembles over the period 1940-2100 as a testbed in which the true internal AMV is known a priori. The first method aliases the structure of forced climate change onto internal AMV, while the second method is generally robust to climate change. The models simulate realistic patterns of internal AMV, although such an assessment is hampered by the brevity of the observational record.