PI: Sundar Niroula

ATS-GenAI is developing an AI-assisted framework to optimize when field samples are collected in resource-constrained environmental campaigns. The project fine-tunes a time series foundation model with physics-based hydrologic simulations to generate long-horizon streamflow forecasts then uses submodular optimization to select sampling days that maximize information content and hydrologic coverage. The approach outperforms heuristic field sampling designs and is broadly applicable to other scientific time series applications.