gpd-error-propagation

To propagate uncertainties through a multi-step physics derivation, trace the derivation chain from inputs to the target quantity, identify uncertainty sources, compute sensitivities, and generate a formal error budget tracking correlated inputs across dependency trees.

You calculate sensitivities for correlated inputs using analytical or numerical methods, quantifying how intermediate results and approximations interact across dependency trees to produce a validated error budget for your target quantity.

Yes, you can track uncertainty sources across state-tracked project workflows by classifying inputs, numerical factors, and correlations, then updating the project state with propagated uncertainties for full reproducibility.

A phase-based workflow is supported for generating an error budget, applying to state-tracked projects where intermediate results and correlated inputs must be quantified across derivation dependency trees to validate final target quantities.

Limitations include the complexity of accurately classifying and quantifying approximation and model uncertainties across deep dependency trees, requiring careful state management to ensure intermediate results remain validated throughout the derivation chain.