AOPP Seminar: Examining Fast Hydrological Responses to Anthropogenic Forcing Agents Using Machine-Learning Climate Emulators 5

The response of the climate system to increased greenhouse gases is governed by a combination of fast responses and slow feedbacks. Slow feedbacks are typically activated in response to changes in ocean temperatures on decadal timescales and manifest as changes in climatic state with no recent historical analogues. Fast responses are activated in response to rapid atmospheric physical processes on weekly timescales, and they can be of comparable magnitude to the slow feedbacks. Fast responses are also operative in response to natural variability, so their physical mechanisms are present in historical earth system model (ESM) simulations and meteorological reanalyses used to train many recent successful machine-learning (ML) weather and climate emulators. Our findings establish that ML emulators trained only on stationary climate simulations can nonetheless quantify the fast precipitation response to reduced and elevated concentrations of greenhouse gases and aerosols with no recent historical precedent. We show that the responses from historically trained emulators agree with theoretical estimates and with ESMs. In conclusion, we demonstrate that ML emulators trained solely on natural climate variability, without exposure to prospective future scenarios, can predict key components of climate change. This opens the possibility to use ML emulators to study a variety of fast processes in global climate.