Published September 4, 2025 | Version v0
Software Open

A Generalised Data-Driven Shoreline Model at the Regional Scale

Creators

  • 1. EDMO icon University of New South Wales , Water Research Laboratory (WRL), School of Civil and Environmental Engineering

Contributors

  • 1. ARC Training Centre in Data Analytics for Resources and Environments (DARE)
  • 2. EDMO icon University of Sydney
  • 3. EDMO icon University of New South Wales , Water Research Laboratory (WRL), School of Civil and Environmental Engineering

Description

This repository accompanies the manuscript “A Generalised Data-Driven Shoreline Model Trained on the South-East Australian Coastline”. It provides code, configuration files, and example data to reproduce the workflow for a regional-scale shoreline forecasting model.

The model adapts the Temporal Fusion Transformer (TFT) architecture to integrate dynamic wave forcing with static site descriptors, enabling forecasts that generalise across more than 300 beaches and 2000km without site-specific calibration. Included materials demonstrate model training, weight loading, evaluation, and ablation experiments.

To keep the archive lightweight, only a small sample transect dataset is included. The full shoreline and wave datasets used for training and evaluation are available separately on Zenodo. A pre-trained set of model weights (model_weights.pth) and an example Jupyter notebook (shorelineModel.ipynb) are provided.

Links

Manuscript dataset: https://doi.org/10.5281/zenodo.16749486

GitHub repository: https://github.com/KitNOTpat/A_Generalised_ShorelineModel

Interactive map (GitHub Pages): https://kitnotpat.github.io/A_Generalised_ShorelineModel/

 

Files

A_Generalised_ShorelineModel.zip

Files (19.9 MB)

Name Size Download all
md5:426a9363e8ef4e1bdd04de40dcf50560
19.9 MB Preview Download