/SAR-Wave-Height

Predicting significant wave height from synthetic aperture radar (SAR) with deep learning on SAR-Altimeter colocation data.

Primary LanguageJupyter NotebookMIT LicenseMIT

SAR-Wave-Height

Predicting significant wave height from synthetic aperture radar (SAR) using the method described in Quach, et. al. 2020, Deep Learning for Predicting Significant Wave Height From Synthetic Aperture Radar. Also available here

Quick start:

  1. Process a netcdf file into a dataset for training or making predictions: scripts/create_dataset_from_nc.ipynb
  2. Train a model with uncertainty predictions (heteroskedastic regression): notebooks/train_model_heteroskedastic.ipynb 
  3. Load a model and make predictions: notebooks/predict.ipynb

Citation:

@article{quach2020deep,
  author={B. {Quach} and Y. {Glaser} and J. E. {Stopa} and A. A. {Mouche} and P. {Sadowski}},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={Deep Learning for Predicting Significant Wave Height From Synthetic Aperture Radar}, 
  year={2021},
  volume={59},
  number={3},
  pages={1859-1867},
  doi={10.1109/TGRS.2020.3003839}}