Climate Dynamics from Fusing Proxies and Models

Abstract:

Paleoclimate data assimilation is an emergent technique that provides an estimate of Earth's climate on a regular grid by fusing data from proxy records and climate model simulations. The resulting analysis is an optimal blend of the two original sources of information, and provides a novel opportunity to explore low frequency climate dynamics over long periods of time. Here we analyze initial data from the Last Millennium Climate Reanalysis Project (LMR) to explore climate dynamics on interannual to decadal timescales. In addition to identifying leading patterns of variability in the LMR, and their dependence on mean state, we also examine the dynamics implied by the LMR using a linear inverse modeling (LIM) approach. Specifically, we use a LIM to empirically estimate the low-order dynamics underlying the LMR reconstruction, to identify the evolution of leading modes of variability. We also use the LIM as a forecast model to estimate the predictive skill on decadal timescales over a large sample, and to identify the sensitivity of predictive skill to the state of the climate system.