Climate Variability vs. Forced Change: Insights from Large Ensemble Climate Model Simulations and Climate Emulators

Convener: Aneesh SundaresanECSECS | Co-conveners: Alexia KarwatECSECS, Debashis PaulECSECS, Pengfei Lin, Yukiko Imada

The historical changes in the global climate systems are mainly attributed to the joint or individual influence of the internal variability and external forcing. By utilizing observational dataset or single realization of the climate models, it is difficult to differentiate the true forced component from the noisy internal variability of the climate system. Both initialized and uninitialized large ensemble climate simulations provide parallel climate realizations: they capture the range of possible trajectories of the climate under both internal variability and external forcing. Their primary value lies in disentangling the anthropogenic signals from internal variability and enabling more robust detection and attribution. Large ensembles support diverse applications, including the estimation of time of emergence, risk assessment of extreme events and compound events, projection of the climate modes of natural variability and its teleconnections, quantifying model uncertainty, testing the robustness of future projections, evaluating tipping points and climate hysteresis, and studying Earth system feedback, such as those in the carbon cycle. Thus, the extensive availability of ensemble data makes them well-suited for deep learning applications and the development of climate emulators.
The climate emulator development has rapidly advanced in recent years with the innovative statistical and machine learning approaches. The computationally efficient climate emulators are a good tool for modelling the forced response and internal variability of the part of the climate system. Although the output of typical emulators has a limited number of variables; its use for evaluating scenario uncertainty, geoengineering applications, impact assessment, policy making, and high-resolution regional projections is quite prominent in the recent times.
This session welcomes a broad range of contributions focused on the analysis of large ensemble datasets and climate emulators spanning all components of the climate system. It specifically covers topics: (a) the detection and attribution of climate change, (b) new statistical and machine learning methodologies to identify the forced change and internal variability, (c) assessment of model uncertainty and climate projection, (d) geoengineering studies, (e) development and applications of the climate emulators, and (f) tipping point and climate hysteresis.