METEORv1.0.1: a novel framework for emulating multi-timescale regional climate responses

Abstract. Resolved spatial information for climate change projections is critical to any robust assessment of climate impacts and adaptation options. However, the range of spatially resolved future scenario assessments available is limited, due to the significant computational and human demands of Earth System Model (ESM) pipelines. In order to explore a wider variety of societal outcomes and to enable coupling of climate impacts into societal modelling frameworks, rapid spatial emulation of ESM responses to climate change is therefore desirable. Many existing pattern scaling methods assume spatial climate signals which scale linearly with global temperature change, where the pattern of response is independent of the nature and timing of emissions. However, this assumption may introduce biases in emulated climates, especially under net negative emissions and overshoot scenarios. To address these biases, we propose a novel emulation system, METEOR, which represents multi-timescale spatial climate responses to multiple climate forcers. The mapping of emissions to forcing is provided by the CICERO Simple Climate Model, combined with a calibration system that can be used to train model-specific pattern response engines using only core training simulations from CMIP. Here, we demonstrate that our fitted spatial emulation system is capable of rapidly and accurately predicting gridded annual mean temperature and precipitation responses to out-of-sample scenarios.