TorchDA

Use deep learning in data assimilation workflow. Paper link

Tutorials (4 notebooks)

notebook	open in colab / kaggle
EnKF Testing Demo on Lorenz 63 Model
LSTM on Lorenz 63 Model
Testing Demo on Variational Methods
Testing Demo of Variational Methods on Shallow Water Model

Package Dependencies

• Python 3.10 or later
• PyTorch (Recommend 2.0 or later)

Installation

• Build a conda environment, run:

  conda env create -f environment.yml

  then activate it with:

  conda activate TorchDA

• Install PyTorch with pip, run:

  pip3 install torch torchvision --index-url https://download.pytorch.org/whl/cu118
  pip3 install -r requirements.txt

  then at the root folder of this repo, run:

  pip3 install .

• Install torchda from source, run:

  pip3 install git+https://github.com/acse-jm122/torchda.git