NSF Award
2425735
Uncertainty quantification—determining how much a prediction can be trusted—is a fundamental challenge in geoscience and is central to cost-effective decision-making, mitigation of extreme weather hazards, and adaptation to a changing climate. Similarly, inequity quantification—measuring how well a predictive model serves different populations—is critical to ensuring that historically marginalized communities, disproportionately vulnerable to extreme weather and climate change, are adequately served by weather and climate models. The growing use of artificial intelligence (AI) models in geoscience has made uncertainty and inequity quantification more important, and difficult, than ever. This project supports a partnership between geoscientists and computer scientists to co-develop novel AI-based approaches to quantify and explain uncertainty and inequity in geoscience. By considering social inequities in geoscience, this project will help shape the future of geoscience through a more interdisciplinary understanding of the Earth system. This project will also build capacity for cross-discipline collaboration and education between computer science and geoscience students, helping meet workforce demands for scientists with experience in both AI and geoscience.<br/> <br/>Three algorithmic innovations will be developed as part of this project. The first innovation develops a computationally efficient framework capable of producing easily interpretable estimates of aleatoric and epistemic uncertainty in geoscience. The second innovation develops generalizable metrics that quantify inequities in Earth system model performance. The third innovation develops a novel AI-ready multimodal (text and image) geoscience dataset that will be used to fine-tune a large multimodal model, capable of describing geoscience imagery and associated uncertainties and inequities. Collectively, these innovations will enable the contextualization of several sources of uncertainty and inequity in geoscience.<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
