Key points are not available for this paper at this time.
Earth system models for deep time have typically been unable to represent geological timespans in 3D because climate and ocean circulation plays a key role in global biogeochemistry and generating a 3D physical climate simulation is extremely computationally expensive. This means that Earth System Modelling for periods of over 1 Myr has been exclusively carried out in nondimensional box models, which leads to oversimplification of spatially heterogeneous processes like continental weathering and marine carbon burial. This simplification may be a key reason why so many climate questions over deep time remain unresolved. The SCION (Spatial Continuous IntegratiON) project aims to produce a 3D and self-consistent climate and biogeochemical system that can be run over billion-year timeframes. To do this, it employs a physical climate emulator which is developed using a Deep Learning method trained on hundreds of General Circulation Model runs over different paleogeographies and CO2 levels. The SCION development project SIM-EARTH also includes a new process-based reconstruction of paleotopography using the GPlates kinematic plate model, development of a long-term dynamic global vegetation module and ocean biogeochemical module, and databasing projects to establish 3D datasets for marine and terrestrial palaeontology and geochemistry that can be compared to model outputs at the local scale to test hypotheses. We hope that new model frameworks like this can help us better understand the evolution of Earths surface conditions over time, assess the contribution of the biosphere to global environmental change, and help determine what fundamental characteristics are required for a planet to be habitable for complex life.
Mills et al. (Mon,) studied this question.