Data Assimilation (Inversion)

This subpackage provides implementations to solve data assimilation (DA) problems (inverse problems) for both time-independent and time-dependent simulation models.

The package distinguishes between two categories of DA methods:

• Filters (time-independent): methods that use a single observation to update the model state or parameter (e.g., Kalman filter, 3D-Var, EnKF).

• Smoothers (time-dependent): methods that use multiple observations at different times to update the state trajectory (e.g., 4D-Var, Kalman smoother).

All DA methods inherit from InverseProblem, which manages registration of the simulation model, observation operator, error models, prior, and observational data.

Abstract Base Classes

pyoed.assimilation.InverseProblem	Base class for implementations of Inversion/Inference/DA (Data Assimilation) methods/approaches.
pyoed.assimilation.Filter	Base class for all filtering (time-independent) DA implementations.
pyoed.assimilation.Smoother	Base class for all smoothing (time-dependent) DA implementations.

Class Hierarchy

Data-assimilation abstract base class hierarchy

• DA Core (Base Classes)

• Filtering DA Methods
  • DA Filtering Algorithms
  • DA Filtering Base Classes
• Smoothing DA Methods
  • DA Smoothing Algorithms
  • DA Smoothing Base Classes

• Goal-Oriented DA
• DA Utilities
• Extras (ROM, etc.)