Project Grant 2435756

This $193,500 federal Project Grant award from the National Science Foundation's (NSF) Computer and Information Science and Engineering (CISE) program (CFDA 47.070) supports the development and validation of an AI framework to automate and accelerate the generation of interpretable environmental science hypotheses using a broad array of image data collected from different sensing modalities. The goal is to address the challenge of applying environmental research insights from one location or time to other contexts, which is critical for understanding and mitigating the effects of environmental changes, especially in vulnerable communities. The framework integrates new techniques into satellite imagery models to leverage sparse, multi-modal data and couples this with self-improving language models to provide interpretable feedback and hypotheses to researchers. This approach aims to go beyond current state-of-the-art computer vision methods. The project will be validated on the task of detecting submerged aquatic vegetation, a technically challenging environmental monitoring application, with the intent to demonstrate the broader applicability of the model to other terrestrial environmental science challenges. The award was co-funded by the CISE and Biological Sciences directorates and runs from June 2025 to November 2026.