DART Data Assimilation¶
Ensemble data assimilation for regional MOM6 in CESM using DART: the Data Assimilation Research Testbed from NSF NCAR. Constrain your model with observations to produce improved ocean state estimates.
Get Started →CESM-DART Tutorial¶
Build a regional CESM-DART data assimilation experiment for Hawaii, from preparing observations through ensemble cycling and diagnostics.
This tutorial series includes three notebooks, designed to be worked through in order. Each notebook builds on the work from the previous one.
Want to work only with real observations? Skip the synthetic observations notebook and jump from Tutorial 1 directly to cycling DART-CESM.
1. Working with Real Observations¶
Turn real Argo profiles from CrocoLake into DART obs_seq files with dartobsgen. You leave with a directory of observations ready to assimilate.
2. Synthetic Observations¶
Sample a model state with DART's perfect_model_obs at the Tutorial 1 locations plus random ones you design. You leave with a synthetic observing network for an OSSE.
3. Cycling DART–CESM¶
Build a multi-instance regional MOM6 ensemble in CESM, assimilate your observations every 24 hours, and diagnose the results. You leave with a running DA experiment.
What You Can Do¶
Observation Preparation¶
Convert CrocoLake observations (Argo, GLODAP) into DART's obs_sequence format. Or create synthetic observations for Observation System Simulation Experiments
Ensemble Filter¶
Configure DART's Ensemble Filters, adaptive inflation and localization for regional ocean domains.
Assimilation Cycling¶
Run forecast–assimilate–update cycles via the DART–CESM interface, fully integrated with the standard CESM job-submission workflow.
Diagnostics¶
Assess assimilation quality with observation-space RMSE/spread diagnostics and state-space increment maps.