F-FNO basin-scale tsunami surrogate model (code, weights and training data sample)

Training, inference, and evaluation code for a Factorized Fourier Neural Operator (F-FNO) surrogate model of basin-scale tsunami propagation in the East Sea (Sea of Japan).
Companion archive for: Kim et al., "A Factorized Fourier Neural Operator Surrogate for Basin-Scale Tsunami Propagation", Geoscientific Model Development, 2026.

This archive contains two files:

1. ffno-tsunami-v1.0.0.zip (~421.6 MB) — Source code and model weights
  - train.py: full training code (architecture, loss, training loop)
  - inference.py: autoregressive rollout inference and figure generation
  - convert_comcot_to_nc.py: COMCOT raw output to NetCDF conversion
  - split_loader.py: function for loading train/val/test split lists in train.py
  - Pretrained model weights (.pt), two configurations
  - Scenario parameter table (864 logic-tree configurations)
  - COMCOT control file template and input generation script
  - Train/val/test split list files

2. ffno-tsunami-test-EM-data.zip (~44.12 GB) — Test-EM evaluation dataset
  - 54 NetCDF files for the most challenging test split (unseen epicenter + unseen magnitude, Ep 1 × Mw 8.0)
  - Sufficient to reproduce all Test-EM results reported in the paper

The full training dataset (~642 GB, 864 scenarios) can be regenerated from the provided scenario parameters using COMCOT v1.7 and is available from the authors upon request.