Speaker: Greg Hakim (U-Washington)
Climate variability involves low-frequency signals that are typically recovered for the past from statistical analysis of proxy data, and projected into the future from integration of numerical models from arbitrarily chosen initial states. Combining models and observations to constrain estimates of both past and future low-frequency climate signals is an emerging discipline, and the subject of this talk. First, I describe an efficient method for estimating low-frequency climate signals from noisy proxy measurements. Second, I propose a modeling framework, motivated by Shannon's Noisy Channel Coding Theorem and the dimensional reduction of balance models, where information is carried by a small number of distinguished variables.