model2obs¶
Model–observation comparisons for regional MOM6 using CrocoLake, a cloud-optimized, Parquet-based observational database for fast access to observational data.
Get Started → model2obs on GitHub ↗What You Can Do¶
CrocoLake Access¶
Query a cloud-optimized Parquet database of QCed in-situ observations (Argo, GLODAP, Spray Gliders) without downloading large files.
Model-observation comparison¶
Use model2obs to interpolate MOM6 output onto the time and space of the observations, using DART's forward operators under the hood.
Visualization¶
Visualize CrocoLake data and model–observation differences on interactive maps and profile plots.
Heavy lifting¶
Work with larger-than-memory CrocoLake datasets, and use parallel execution capacity of model2obs to speed up interpolation on large-domain runs.
The Tutorial Series¶
This tutorial series includes three notebooks. The first two notebooks show how to interpolate MOM6 output onto CrocoLake observations’ space using model2obs: start small and serial to learn the machinery (first notebook), then scale up to a parallel, large-domain run (second notebook). The third notebook guides you through how to use CrocoLake-like datasets which are larger-than-memory.
If you don’t remember how to get started with the packages installation, see the Before you arrive instructions.
1. Small Domain, Serial¶
Interpolate a short CESM-MOM6 run around the Hawaiian islands onto CrocoLake observations. You leave knowing the basic model2obs workflow and how to plot the results on an interactive map.
2. Large Domain, Parallel¶
Run the same workflow but in parallel, on a larger NorthWest Atlantic domain. You leave able to compare large-domain runs against observations efficiently.
3. Exploring CrocoLake¶
Read and filter CrocoLake's Parquet files with pandas and dask, and map temperature measurements in the North West Atlantic. You leave able to handle larger-than-memory observational datasets.