Bayesian comparison of stochastic models of dispersion

DOI10.1017/JFM.2022.472zbMATH Open1505.76048arXiv2201.01581OpenAlexW4283325476MaRDI QIDQ5085103FDOQ5085103

Authors: Martin T. Brolly, A. L. Teckentrup, J. R. Maddison, J. Vanneste 

Publication date: 27 June 2022

Published in: Journal of Fluid Mechanics (Search for Journal in Brave)

Abstract: Stochastic models of varying complexity have been proposed to describe the dispersion of particles in turbulent flows, from simple Brownian motion to complex temporally and spatially correlated models. A method is needed to compare competing models, accounting for the difficulty in estimating the additional parameters that more complex models typically introduce. We employ a data-driven method, Bayesian model comparison (BMC), which assigns probabilities to competing models based on their ability to explain observed data. We focus on the comparison between the Brownian and Langevin dynamics for particles in two-dimensional isotropic turbulence, with data that consists of sequences of particle positions obtained from simulated Lagrangian trajectories. We show that, while on sufficiently large timescales the models are indistinguishable, there is a range of timescales on which the Langevin model outperforms the Brownian model. While our set-up is highly idealised, the methodology developed is applicable to more complex flows and models of particle dynamics.

Full work available at URL: https://arxiv.org/abs/2201.01581

Recommendations

• Analysis of relative dispersion of two particles by Lagrangian stochastic models and DNS methods
• Use of a modified Langevin equation to describe turbulent dispersion of fluid particles in a channel flow
• A Lagrangian stochastic model for nonpassive particle diffusion in turbulent flows
• Relative dispersion in isotropic turbulence. Part 2. A new stochastic model with Reynolds-number dependence
• Parameter estimation in multi particle Lagrangian stochastic models

zbMATH Keywords

machine learningturbulent mixingdata-driven methodBrownian particle dynamicsLangevin particle dynamicstwo-dimensional isotropic turbulence

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05) Turbulent transport, mixing (76F25) Statistical turbulence modeling (76F55) Basic methods in fluid mechanics (76M99)

Cites Work

• Algorithm 778: L-BFGS-B
• Asymptotic Statistics
• Title not available (Why is that?)
• The Bayesian Choice
• Strictly Proper Scoring Rules, Prediction, and Estimation
• Elements of Information Theory
• A Simplex Method for Function Minimization
• Title not available (Why is that?)
• Bayes Factors
• New introduction to multiple time series analysis.
• Expected information as ecpected utility
• Probability Theory
• Atmospheric and Oceanic Fluid Dynamics
• Stochastic processes and applications. Diffusion processes, the Fokker-Planck and Langevin equations
• Criteria for the selection of stochastic models of particle trajectories in turbulent flows
• Estimating eddy diffusivities from noisy Lagrangian observations
• Parameterization of dispersion in two-dimensional turbulence
• Title not available (Why is that?)
• Title not available (Why is that?)
• Coherent vortices and kinetic energy ribbons in asymptotic, quasi two-dimensional f-plane turbulence

Cited In (4)

• Evaluating predictors of dispersion: a comparison of dominance analysis and Bayesian model averaging
• Conditional relative acceleration statistics and relative dispersion modelling
• Title not available (Why is that?)
• Improved inference in dispersion models

This page was built for publication: Bayesian comparison of stochastic models of dispersion

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5085103)