Sequential Markov chain Monte Carlo for Lagrangian data assimilation with applications to unknown data locations
Author(s)
Ruzayqat, Hamza
Beskos, Alexandros
Crisan, Dan
Jasra, Ajay
Kantas, Nikolas
Type
Journal Article
Abstract
We consider a class of high-dimensional spatial filtering problems, where thespatial locations of observations are unknown and driven by the partiallyobserved hidden signal. This problem is exceptionally challenging, as not onlyis it high-dimensional, but the model for the signal yields longer-range timedependences through the observation locations. Motivated by this model, werevisit a lesser-known and provably convergent computational methodologyfrom Berzuini et al. (1997, Journal of the American Statistical Association, 92,1403–1412); Centanniand Minozzo (2006, Journal of the American StatisticalAssociation, 101, 1582–1597); Martin et al. (2013, Annals of the Institute of Sta-tistical Mathematics, 65, 413–437) that uses sequential Markov Chain MonteCarlo (MCMC) chains. We extend this methodology for data filtering prob-lems with unknown observation locations. We benchmark our algorithms onlinear Gaussian state-space models against competing ensemble methods anddemonstrate a significant improvement in both execution speed and accuracy.Finally, we implement a realistic case study on a high-dimensional rotatingshallow-water model (of about 104 –105 dimensions) with real and synthetic data.The data are provided by the National Oceanic and Atmospheric Administra-tion (NOAA) and contain observations from ocean drifters in a domain of theAtlantic Ocean restricted to the longitude and latitude intervals [−51◦, −41◦],[17◦, 27◦], respectively.
Date Issued
2024-04-01
Date Acceptance
2024-02-24
Citation
Quarterly Journal of the Royal Meteorological Society, 2024, 150 (761), pp.2418-2439
ISSN
0035-9009
Publisher
Wiley
Start Page
2418
End Page
2439
Journal / Book Title
Quarterly Journal of the Royal Meteorological Society
Volume
150
Issue
761
Copyright Statement
© 2024 The Authors. Quarterly Journal of the Royal Meteorological Society published by John Wiley & Sons Ltd on behalf of the Royal Meteorological Society. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
Identifier
10.1002/qj.4716
Subjects
high-dimensional filtering
Markov chain Monte Carlo
MCMC
Meteorology & Atmospheric Sciences
PARTICLE FILTER
Physical Sciences
Science & Technology
spatial filtering
Publication Status
Published
Date Publish Online
2024-05-07