Joint Seminar: QuickClim-v2: Climate emulator of the global moist three-dimensional atmosphere

Our previous emulator, QuickClim (

https://doi.org/10.1038/s43247-023-01011-0 ), was a physically-inspired machine learning approach learnt purely from CMIP data, with a unique representation for each climate model in the archive. Here we take an alternate approach, where the climate emulator is instead physically-constrained and learnt from both CMIP and reanalysis. A reduced order model of the global atmosphere is developed by solving the moist hydrostatic equations of motion projected onto a multi-variate set of three-dimensional mode shapes learnt from the underlying data. This approach transforms the atmospheric primitive equations, which is a system of partial differential equations dependent upon time and space, into a system of ordinary differential equations dependent upon only time and mode index. This massively reduces the dimensionality of the problem. The emulator coefficients are calculated using an optimisation approach, with remaining model errors accounted for via a data-driven stochastic parameterisation. Temporal integrations of the climate emulator over a recent 40-year period demonstrate excellent agreement with the underlying data. Statistical properties for each mode, of the climate change trend, and of the El Nino Southern Oscillation are all shown to be reproduced at a fraction of the computational cost. The climate emulator has also been demonstrated to produce plausible future projections of high and low carbon concentration pathways.

Bio: Dr Kitsios completed a PhD with the University of Melbourne and the Université de Poitiers (France) on fluid dynamical stability and model reduction of aerospace flows. He then undertook post-doctoral research with the CSIRO and then Monash University, on the numerical simulation and stochastic parameterisation of atmospheric, oceanic and boundary layer turbulence. He then held an industrial research position at a hedge fund developing trading algorithms based on macroeconomic themes and market conditions. Since re-joining CSIRO, he has been undertaking research on data assimilation methods for improved climate state / parameter estimation and forecasting. His most recent research involves the application of machine learning for climate emulation, and quantifying the influence of climate on agriculture, financial markets, health indicators and social unrest. He is currently a co-chair of the Machine Learning for Climate and Weather Working Group of the Australian Climate Community Earth System Simulator National Research Infrastructure consortium, an associate editor for the Theoretical and Computational Fluid Dynamics journal, and an elected committee member of the Australasian Fluid Mechanics Society.

Datum

28.01.2026

Uhrzeit

13:30 h

Ort

Bundesstr. 53, room 022/023

Seminar Room 022/023, Ground Floor, Bundesstrasse 53, 20146 Hamburg, Hamburg

Chair

Wolfgang Müller

label_back