Storm Surge Forecasting with CNN-LSTM

This project trains a hybrid CNN + LSTM model to predict coastal storm surge (zeta) using ERA5 and CORA .

---

Files

• model.py — Defines the hybrid CNN + LSTM model
• dataloader.py — Loads ERA5 and CORA data from .nc files
• train.py — Trains the model and saves weights to cnn_lstm_model.pth
• evaluate.py — Evaluates model performance on the test set
• stack_era5.py— Preprocesses ERA5 data into ML ready format. Output is a .npy file that the dataloader can memory-map directly.
• cora_graph.py— Constructs a graph representation of the CORA dataset. Produces adjacency information that is fed into the graph neural network (GNN).

---