Reproduce “Redesigning the ensemble Kalman filter with a dedicated model of epistemic uncertainty” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Chatchuea Kimchaiwong, Jeremie Houssineau, Adam M. Johansen

A novel possibilistic Ensemble Kalman Filter (p-EnKF) was developed to explicitly model epistemic uncertainty in data assimilation, demonstrating lower estimation errors and better calibrated uncertainty with small ensemble sizes across linear and nonlinear systems. This method provides a principled approach for implementing localization and inflation without ad-hoc tuning.

Key insights

  • The deterministic tracking of the mode particle in possibility theory allows the p-EnKF to maintain highly accurate mean estimates without re-estimation, contributing to its stable performance with small ensembles.
  • The Gaussian fitting procedure for possibility functions provides an intrinsic and principled method for implementing localization and inflation through conditional independence constraints, automatically adjusting for precision loss.
  • Possibility theory's properties, such as the preservation of Gaussianity under linear transformations and the property that the expected value of a transformed variable is the transformation of the expected value, enable a direct adaptation of Kalman Filter-like updates.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.