Climate Model Emulation

Climate model emulation uses machine learning to create faster, cheaper surrogate models of complex Earth System Models (ESMs), enabling more extensive climate scenario analysis. Current research focuses on improving the accuracy and equity of these emulators, exploring various neural network architectures (including diffusion models, randomly wired networks, and those incorporating physical constraints) and regression models (like Gaussian Processes) to achieve high-fidelity emulation at various temporal resolutions (daily to monthly). This work is significant because it allows for computationally efficient exploration of a wider range of climate scenarios and improved accessibility of climate data for researchers and policymakers, ultimately aiding in climate change mitigation and adaptation strategies.