This proposal addresses area (1) of the FOA, “Interaction of Climate Change and Low Frequency Modes of Natural Climate Variability”. Our overarching objective is to detect, describe and understand the changes in low frequency variability between model simulations of the preindustrial climate and simulations of a doubled CO2 climate. The deliverables will be a set of papers providing a dynamical characterization of interannual, decadal, and multidecadal variability in coupled models with attention to the changes in this low frequency variability between pre-industrial concentrations of greenhouse gases and a doubling of atmospheric concentrations of CO2. The mode of analysis, singular vector decomposition, is designed to advance our physical, mechanistic understanding. The simulation of the past millennium will help to characterize the deficiencies of model simulations of such variability. This study will include external natural variability due to solar and volcanic aerosol variations as well as variability internal to the climate system. An important byproduct will be a documented, easy to use archive of the data from the model runs. Part of the expression of “climate change” is expected to be a change in the amplitude and frequency of occurrence of “natural” interannual to multidecadal climate variability. In addition changes in the mean climate state should engender structural changes in the modes of natural variability. For each of the 3 models investigated (GFDL, CAM, CFS), we will compare the low frequency variability in a long (1000 years or more) run with pre-industrial (Pi) conditions to a similarly long run with doubled CO2 (2C). These simulations will be the basis for assessing changes due to variability internal to the climate system. Analysis of the differences will be based on both multivariate empirical orthogonal functions – modes that are selected to account for the most variance and hence describe the peaks of the climate system’s preferred modes of variability – and singular vectors, modes that are selected by model dynamics, and capture how the system evolves to the modes with the most variance. Thus we rely on objective methods that:

1. Do not impose such specific expectations of spatial or temporal structure on the analysis
2. Default to the assumption that all modes are global, making no a priori restriction to single ocean basins.

This leaves the task of interpreting the modes we find in terms of the familiar descriptions of climate variability in terms of El Niño, the North Atlantic Oscillation, etc., but since we still lack accepted theories for the origins and mechanisms of the NAO, PDO, AMO, perhaps an alternate approach is needed to provide new insights. A second set of runs will consider the external natural variability forced by variations in solar radiation and volcanic aerosol. There is a great deal of evidence indicating that such climate variations in the past have substantially to impacted societies and ecosystems. We will account for this external forcing, by rerunning the Pi and 2C runs for 1000 years with a representation of the solar and volcanic forcing of the past 1000 years. This PiX simulation is a simulation of the past millennium and may be compared with the growing inventory of paleoproxy data, enabling additional tests of the model’s skill in simulating climate. The 2CX simulation will provide a better estimate of the range of variability that might occur a century from now than will the usual IPCC simulations of only greenhouse gas forced changes and so will be useful to planners.