Particle filters for high-dimensional geoscience applications: A review.

Area of Science:

• Geosciences
• Atmospheric Science
• Oceanic Science

Background:

• Particle filters offer potential for fully nonlinear data assimilation.
• Application in high-dimensional geoscience systems was limited by inefficiency.
• Recent advancements address these limitations, improving applicability.

Purpose of the Study:

• To present a state-of-the-art discussion on developing particle filters for high-dimensional nonlinear geoscience state-estimation.
• To emphasize atmospheric and oceanic applications.
• To provide a guide with new ideas, derivations, and unifications.

Main Methods:

• Developments in proposal densities.
• Application of optimal transportation concepts.
• Implementation of localization and adaptive resampling strategies.
• Hybridization with ensemble Kalman filters and variational methods.

Main Results:

• Significant progress in particle filter efficiency for high-dimensional systems.
• Demonstrated competitiveness with current numerical weather prediction methods.
• Successful unification of various particle filter developments.

Conclusions:

• Particle filters are rapidly overcoming previous limitations in high-dimensional geoscience.
• They are becoming a viable and competitive tool for state estimation.
• Particle filters are poised for mainstream adoption in fields like numerical weather prediction.