GNN-SSC Sofrware: Computationally and Memory Efficient Graph Neural Network for Emulating Storm Surge Inundation State Over Extended Geospatial Domains

GNN-SSC performs node-wise binary classification of storm-surge inundation (dry vs. wet) over the computational domain. The model operates on a graph whose connectivity is defined either by response-based similarity (via cosine distance) or by geospatial proximity (via KNN in geographic space). The full pipeline is organized into four stages: data preparation, model training, baseline benchmarking, and evaluation. Main ideas: Preprocess raw gridded inundation data; select informative nodes (partially wet) and build a graph (edge list + weights). Train a GNN classifier on per-storm features to predict node-wise inundation. Compare with baselines; evaluate with MC/FP/FN metrics globally, per storm, and per node. Data Availability: All data supporting this study are openly available on:https://zenodo.org/records/17180282 Acknowledgement: This research was financially supported by the U.S. Army Corps of Engineers (USACE), Engineer Research and Development Center (ERDC), Coastal and Hydraulics Laboratory (CHL), as part of grant number W912HZ-25-C-0017. The first three authors gratefully acknowledge this support.