[00936] Recent advances in applications for large-scale data assimilation and inverse problems.
Session Time & Room : 4E (Aug.24, 17:40-19:20) @E504
Type : Proposal of Minisymposium
Abstract : The development of algorithms for data assimilation (DA) and inverse problems (IPs) has been traditionally driven by very specific applications in engineering, medicine and the geosciences. However, algorithms that were initially tailored for geophysical data assimilation have now been used for solving tomographic inverse problems in engineering. Similarly, developments on imaging methods have been used for DA. This exchange has not only led to further algorithmic developments but also deeper theoretical insights. In this minisymposium we bring together experts on DA and IPs that work at the interface with applied sciences, with the aim of fostering knowledge transfer across disciplines
00936 (1/1) : 4E @E504 [Chair: Svetlana Dubinkina]
[05372] Level-set parameterisations for Ensemble Kalman Inversion
Format : Online Talk on Zoom
Author(s) :
Marco Iglesias (University of Nottingham)
Abstract : We discuss a level-set approach to parameterise unknown interfaces and discontinuous properties with the Ensemble Kalman Inversion (EKI) framework for inverse problems. We demonstrate the applicability of this approach to solve various inverse problems where the unknown is a discontinuous property arising from the presence of an anomalous material/tissue. We will present numerical examples with applications to (i) non-destructive testing of composite materials, (ii) ground penetrating radar and (iii) magnetic resonance elastography.
[05415] Data assimilation for estimating nonlinear dynamics in earthquakes
Format : Online Talk on Zoom
Author(s) :
Femke Cathelijne Vossepoel (Delft University of Technology)
Hamed Ali Diab-Montero (Delft University of Technology)
Arundhuti Banerjee (Delft University of Technology)
Celine Marsman (Utrecht University)
Ylona van Dinther (Utrecht University)
Abstract : The highly nonlinear dynamics of earthquake sequences and the limited data availability make it very difficult, if not impossible, to forecast earthquakes. State- and parameter estimation with data assimilation can improve estimates and forecasts of earthquake sequences. We illustrate the challenges of data assimilation in earthquake simulation with a range of models using several ensemble data-assimilation methods, including the particle filter, the ensemble Kalman filter, the adaptive Gaussian mixture filter, and the particle flow filter.
[05358] Analysis of a localized ensemble Kalman-Bucy filter with sparse observations
Format : Talk at Waseda University
Author(s) :
Gottfried Hastermann (University of Potsdam)
Jana de Wiljes (University of Potsdam)
Abstract : With large scale availability of precise real time data, their incorporation into physically based predictive models, became increasingly important. This procedure of combining the prediction and observation is called data assimilation. One especially popular algorithm of the class of Bayesian sequential data assimilation methods is the ensemble Kalman filter which successfully extends the ideas of the Kalman filter to the non-linear situation. However, in case of spatio-temporal models one regularly relies on some version of localization, to avoid spurious oscillations.
In this work we develop a-priori error estimates for a time continuous variant of the ensemble Kalman filter, known as localized ensemble Kalman-Bucy filter. More specifically we aim for the scenario of sparse observations applied to models from fluid dynamics and space weather.
[05397] Edge-preserving inversion with 𝛼-stable priors
Format : Talk at Waseda University
Author(s) :
Jarkko Suuronen (LUT University)
Tomás Soto (LUT University)
Neil Chada (Heriot Watt University)
Lassi Roininen (LUT University)
Abstract : The 𝛼-stable distributions are a family of heavy-tailed and infinitely divisible distributions that are well-suited as prior distributions to edge-preserving inversion in the context of (discretization of) infinite-dimensional continuous-time statistical inverse problems. In this talk we present a new hybrid approximation method well-suited to the application of such priors.
