QG model

A quasi-geostrophic model (qg), written in Fortran by Dr. Shafer Smith, is implemented in NEDAS as a test model.

The model describes the evolution of the streamfunction \(\psi\) in a two-layer, doubly periodic domain under quasi-geostrophic dynamics (kmax = 127, 256 × 256 grid). Its realistic two-dimensional spatial structure makes it a standard benchmark for localization-based ensemble DA algorithms.

Topics covered in the tutorial notebook:

• Configuring ensemble size, observation network, localization radius, and inflation

• Running multi-cycle OSSE experiments with the Fortran QG model

• Reading and plotting RMSE versus DA cycle to verify filter convergence

• Visualizing streamfunction fields (truth, prior mean, posterior mean)

• Comparing batch (ETKF) and serial (EAKF) ensemble Kalman filter strategies

• Correcting position errors with the multiscale alignment updator (Horn-Schunck optical flow)

The notebook can be run in several environments:

• Docker (see below)

• Native Python — refer to the environment setup guide

docker pull myying/nedas-tutorials
docker run -it --rm -p 8888:8888 myying/nedas-tutorials

Then open the URL printed in the terminal and navigate to 3.multiscale_alignment_with_qgmodel.ipynb.

The full notebook is available on GitHub.