The AI & Localized Policy project is aimed at using advanced AI for simultaneous bias-correction and downscaling of climate model projections for localized policy relevance.
This project explores the capacity of cutting-edge Artificial Intelligence (AI) to simultaneously downscale and bias-correct general circulation climate models (GCMs) and climate projections. The general task at hand is to develop AI models that take low-resolution climate maps (produced by GCMs) as input and produce high-resolution counterparts that capture the observed climatic structure (e.g., marginal distribution properties, spatiotemporal dependence, etc) and honor mass conservation. When successful, this project will lead to obtaining downscaled and bias-corrected climate projections that are more relevant and actionable at scales that matter for societal decision-making, especially for local governance and community preparedness.
The team's Postdoctoral Research Associate, Shivam Singh, is a Climate Fellow working collaboratively with his mentors from the environmental science and data science disciplines at UVA, to leverage their combined expertise to pioneer the application of state-of-the-art AI techniques, including knowledge-guided Style Generative Adversarial Networks (StyleGAN) and diffusion models, to simultaneously correct biases and downscale climate projections.