Speaker: Henri Drake, PhD Candidate at the MIT/WHOI Joint Program

Henri will be presenting his talk "How skillful are climate models and have they improved over time?"

Abstract: As climate models have become increasingly comprehensive in their representations of the Earth system, they have also been increasingly used to project future climate impacts and inform environmental policy. Since observations of the future are presently unavailable and forced signals take decades to emerge from the noise of natural variability, modellers turn to other approaches to validate whether their models are fit for this purpose. Proposed alternative methods for validating climate models are: 1) empirical accuracy with respect to historical observations, 2) robustness across structural model differences, and 3) support by background knowledge.

In part 1, we revisit historical climate models and directly validate their projections of global warming against observations acquired subsequent to their publication dates. Projections of global warming were generally found to be consistent with observed trends, especially after correcting for differences between projected radiative forcing and the actualized radiative forcing. In part 2, we analyze trends in the empirical accuracy of the six generations of climate model ensembles used in the Intergovernmental Panel on Climate Change (IPCC)'s Assessment Reports. We find moderate improvement in climate model skill over thirty years of model development, from the handful of ground-breaking GCM experiments used in First Assessment Report (1990) to the state-of-the-art CMIP6 ensemble to be used in AR6 (2021).

The results from Part 1 are published in the attached Hausfather et al (2020) paper. Part 2 is work in progress and I welcome constructive criticism as we are currently putting together the manuscript. Those interested can read Reichler and Kim 2008, attached, for background on Part 2.