Calibrating a large-domain land/hydrology process model in the age of AI: The SUMMA CAMELS emulator experiments

Process-based (PB) hydrological modeling is a long-standing capability used for simulating and predicting complex water processes over large, hydro-climatically diverse domains, yet PB model parameter estimation (calibration) remains a persistent challenge for large-domain applications. New techniques and concepts arising in the artificial intelligence (AI) context for hydrology point to new opportunities to tackle this problem in complex PB models. This study introduces a new scalable calibration framework that jointly trains a machine learning emulator for model responses across a large-sample collection of watersheds while leveraging sequential optimization to iteratively refine hydrological model parameters. We evaluate this strategy through a series of experiments using the Structure for Unifying Multiple Modeling Alternatives (SUMMA) hydrological modeling framework coupled with the mizuRoute channel routing model for streamflow simulation. This "large-sample emulator"(LSE) approach integrates static catchment attributes, model parameters, and performance metrics, and yields a powerful new strategy for large-domain PB model parameter regionalization to unseen watersheds. The LSE approach is compared to using a more traditional individual basin calibration approach, in this case using a single-site emulator (SSE), trained separately for each basin. The jointly trained LSE framework achieves comparable or better performance to traditional individual basin calibration, while further enabling potential for probabilistic parameter regionalization to out-of-sample, unseen catchments. Motivated by the need to optimize complex hydrology models across continental-scale domains in support of applications in water security and prediction, this work leverages new insights from AI era hydrology research to help surmount old challenges in the calibration and regionalization of large-domain PB models.