Authors: Lirong Yin – University of Arizona
Title: A Python version of the restructured NOAA-OWP Module (NOM) for swapping alternative models and AI applications
Presentation Type: Lightening talk
Abstract: Our group has translated the full NOAA-OWP Module (NOM) into Python (PyTorch) and restructured NOM for swapping/testing alternative modules (e.g., soil and snow modules), especially for machine learning and AI-based systems. We verified that the Python/PyTorch versions produce the same states/fluxes as does the original Fortran version. We also tested its compatibility with ML applications and showed that it can be used with frameworks such as differentiable modeling. This work creates a practical bridge between a physical land surface model and modern AI methods, supporting future development of hybrid and physics-informed ML modeling approaches.