Seminar

Abstract

Climate projections are made computationally demanding because the thermal inertia of the ocean means that changes in radiative forcing require several thousand years to be fully manifested as climate change. That timescale makes state-of-the-art coupled climate models computationally prohibitive for use in many impacts studies and complicates understanding of fundamental climate physics from either models or data. As part of the University of Chicago's Center for Robust Decision-making on Climate and Energy Policy (RDCEP), we have used large libraries of climate model output in studies that aim at core climate science questions. Research efforts include developing a statistical approach for rapid "emulation" of the the output of large climate models for arbitrary climate forcings; using spectral methods to characterize and emulate changes in climate variability; identifying the cause of long-term nonlinearity in ocean heat uptake with warming; and understanding the energetic constraints that cause rainfall in climate models to increase in equilibrated high-CO2 climates, but to decrease following an abrupt addition of CO2. Results suggest both simplifying and complicating features of climate model behavior. Statistical emulation appears a feasible technique for many aspects of climate projections, but different climate models show changes in variability that differ even by sign. Trajectories of heat uptake and precipitation changes, on the other hand, appear robust between models and rooted in simple physics. Nonlinearity in ocean heat uptake appears a simple consequence of the differential warming rates across the Earth, and precipitation changes are readily reproducible from surface energy-budget constraints even in simple one-column models.