V. Balaji - From Simulation to Emulation: Climate Modeling at the Crossroads

Traditional general circulation models, or GCMs – i.e. 3D dynamical models with vertical “columns” containing unresolved terms represented in equations with tunable parameters – have been a mainstay of climate research for several decades, and some of the pioneering studies have recently been recognized by a Nobel prize in Physics. Yet, there is considerable debate around their continuing role in the future. Frequently mentioned as limitations of GCMs are: the non-local nature of certain unresolved processes in atmosphere and ocean, which do not fit readily into a GCM’s “column” abstraction for unresolved scales; the “structural” uncertainty across models with different representations of unresolved scales; and the fact that the models are tuned to reproduce certain aspects of the observed Earth. It is now often contended that nothing short of resolving finer-scale motions, coupled with assimilation of present-day observations to control model biases, will in fact address these shortcomings, and that a future generation of models – including what are sometimes referred to as “digital twins” – will address these issues through substantially higher resolution. At the same time, models used in service of decision-making and policy, including the recently concluded IPCC AR6, rely on emulators allowing of rapid exploration of multiple future scenarios through the use of statistical techniques connecting forcings to outcomes, and for the correction of biases. Furthermore, it is noted that models pegged to present-day climate do not do a good job of representing the climate fluctuations of the past, including past warm climates that may hold lessons for the climate emergency.

In this talk, Balaji argues that the sense that GCMs may be obsolescent comes from these conflicting demands: very high resolution for some key processes, which restrict the ability to explore and quantify uncertainties and study low-frequency variability; the need to explore many counterfactual scenarios for constructing climate policy, which cannot be guided simply by present-day observations; the long simulation times needed to understand prior episodes of abrupt climate change and global warming. He presents potential approaches founded on a renewed emphasis to mathematical methods of calibration and dimensionality reduction, using the GCM’s structure to connect twins to emulators to long-running models of the Earth’s past and future.

Bio:
V. Balaji joined Schmidt Futures as Distinguished Fellow in June 2022 to bring his thought leadership towards new programs in climate and computational sciences. Balaji previously held a 25 year career at Princeton University, during which time he served as head of the Modeling Systems Division at NOAA’s Geophysical Fluid Dynamics Laboratory, one of the world’s pioneers in climate modeling, from 2003-2021. In 2017 he was among the first recipients of President Macron’s Make Our Planet Great Again award and through that has an academic appointment at the Institut Pierre-Simon Laplace in Sorbonne Universités. He has served on National Academies panels related to climate, computing, and geophysical data, and has been on scientific advisory boards to several US agencies and climate research institutions across the world (currently to the UK Met Office). His interests are in the application of advanced technologies and methods to hard science problems, and broadening access to these tools in underserved communities and the developing world.