FourDVar-Lorenz96

This is scripts for 4D-Var using a neural network surrogate model obtained by machine learning for Lorenz 96 model.

Contents

• Misc

  • README.md: this file
  • LICENSE: license file

• For Experiments

  • lorenz96.py: Lorenz 96 model
  • net.py: neural network model
  • train_1step.py: for one-step learing
  • train_10step.py: for ten-step learing
  • 4dvar.py: for 4D-Var experiment without observation error
  • 4dvar_err.py: for 4D-Var experiment with observation error
  • 4dvar_phy.py: for 4D-Var experiment with manually constructed adjoint model

• For Analysis and Visualization

  • run_err.py
  • learning_curve_1step.py
  • graph_loss.py
  • graph_4dvar.py
  • accuracy_4dvar.py
  • lorenz96.rb
  • learning.rb
  • cost.rb
  • 4dvar.rb
  • 4dvar_rmse.rb