Atlantic Multidecadal Variability: Mechanisms, Impact, and Predictability: A Study Using Observations and IPCC AR4 Model Simulations

Atlantic multidecadal sea surface temperature variability (AMV) is a prominent phenomenon that is thought to arise from the natural or internal interaction of the atmosphere and ocean (in contrast with the response to anthropogenic forcing). It is also associated with a wide array of significant global impacts. Models of different complexity strongly support the assertion that AMV is related to the variability of the Atlantic Meridional Overturning Circulation (AMOC) and can be thought of as the surface expression of the latter and the communicator of deep ocean variability to the atmosphere. There is also indication that the related AMV/AMOC variability is potentially predictable. Prediction of AMV should be a crucial element of any attempt to predict the evolution of climate in the coming decades even as the major element of change in this period is the effect of anthropogenic greenhouse gas (GHG) emissions. Several modeling centers have already begun to take such action with models of the class used in the IPCC Fourth Assessment. In preparation for a broad community attempt at addressing nearterm climate change prediction we propose a diagnostic analysis of output from a set of IPCC coupled models to systematically catalog the AMV exhibited in these models, its climatic impacts over land, its link to AMOC and atmospheric variability, and its predictability. Within each representative model we will also use three classes of model output: pre-industrial control runs, ensemble integrations with known 20th century external forcing (GHG, aerosols, solar, and volcanoes), and ensemble integration with projected 21st century GHG forcing. The analysis will deploy an array of diagnostic tools, such as optimal methods for detecting and separating between the externally forced signal and internal variability, multivariate techniques to study the spatial and temporal properties of AMV and its instantaneous and time-lagged associations to subsurface and atmospheric variability. To study modeled AMV predictability and to provide alternative insight to its dynamics we propose to use linear inverse modeling (LIM) methodology that fits model output in multivariate fields to explore modes that display non-normal growth. Identifying such modes allows for efficient assessments of error growth and predictability. Application of the analysis on a range of fully coupled modes will directly contribute to the NOAA goal to “understand and describe climate variability and change to enhance society’s ability to plan and respond.”