Correlation-based flow decomposition and statistical analysis of the eddy forcing

Agarwal, Niraj; Ryzhov, Eugene A.; Kondrashov, Dmitri; Berloff, Pavel

doi:10.1017/jfm.2021.604

Cited by 12 publications

(10 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is worth looking for the ways of imposing physical constraints, such as energy/mass conservation, into the emulators, for example, using an appropriate penalizing term in the loss function. Second, the results obtained here are directly relevant for emulation of various complex and multi-scale fields in the context of eddy parameterizations and test the alternative definitions of eddies investigated recently (Agarwal et al, 2021;Berloff et al, 2021). Finally, a possible sequel to this work is including more stochastic and deep-learning methods, or a mixture of both, for example, the Stochastic Neural Networks (Guillaumin & Zanna, 2021).…”

Section: Discussionmentioning

confidence: 79%

“…Additionally, we have diagnosed the stability and computational costs of the AGARWAL ET AL. Agarwal et al (2021), for future eddy emulators, but developing and applying such metrics is beyond the scope of this study.…”

Section: Discussionmentioning

confidence: 99%

“…Additionally, we have diagnosed the stability and computational costs of the methods using forecast horizon and training/prediction time complexities, respectively. It is also possible to define a few dynamically inspired performance metrics, such as 'product integral' discussed in Agarwal et al (2021), for future eddy emulators, but developing and applying such metrics is beyond the scope of this paper.…”

Section: Discussionmentioning

confidence: 99%

See 2 more Smart Citations

A Comparison of Data‐Driven Approaches to Build Low‐Dimensional Ocean Models

Agarwal

Kondrashov

Dueben

et al. 2021

J Adv Model Earth Syst

Self Cite

View full text Add to dashboard Cite

We present a comprehensive inter‐comparison of linear regression (LR), stochastic, and deep‐learning approaches for reduced‐order statistical emulation of ocean circulation. The reference data set is provided by an idealized, eddy‐resolving, double‐gyre ocean circulation model. Our goal is to conduct a systematic and comprehensive assessment and comparison of skill, cost, and complexity of statistical models from the three methodological classes. The model based on LR is considered as a baseline. Additionally, we investigate its additive white noise augmentation and a multi‐level stochastic approach, deep‐learning methods, hybrid frameworks (LR plus deep‐learning), and simple stochastic extensions of deep‐learning and hybrid methods. The assessment metrics considered are: root mean squared error, anomaly cross‐correlation, climatology, variance, frequency map, forecast horizon, and computational cost. We found that the multi‐level linear stochastic approach performs the best for both short‐ and long‐timescale forecasts. The deep‐learning hybrid models augmented by additive state‐dependent white noise came second, while their deterministic counterparts failed to reproduce the characteristic frequencies in climate‐range forecasts. Pure deep learning implementations performed worse than LR and its simple white noise augmentation. Skills of LR and its white noise extension were similar on short timescales, but the latter performed better on long timescales, while LR‐only outputs decay to zero for long simulations. Overall, our analysis promotes multi‐level LR stochastic models with memory effects, and hybrid models with linear dynamical core augmented by additive stochastic terms learned via deep learning, as a more practical, accurate, and cost‐effective option for ocean emulation than pure deep‐learning solutions.

show abstract

Section: Discussionmentioning

confidence: 79%

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

A Comparison of Data‐Driven Approaches to Build Low‐Dimensional Ocean Models

Agarwal

Kondrashov

Dueben

et al. 2021

J Adv Model Earth Syst

Self Cite

View full text Add to dashboard Cite

show abstract

“…This makes comparison between the two eddy types rather problematic. Here, we restrict our local filtering to the most simple (other alternatives may be correlation-based or spectral filtering approaches Agarwal et al 2021), spatial one based on running square-box averaging. Within this framework the only control parameter is the filter width.…”

Section: Appendix a Transport Tensor Properties: Full Fluxesmentioning

confidence: 99%

“…Agarwal et al. 2021) assumes that can be convolutionally filtered into the low-passed (over-barred) and residual eddy (primed) components: . Note that although in this case is nominally present on the coarse grid, it is still the object for parameterization, because its effects involving spatial derivatives are dynamically misrepresented.…”

Section: Introductionmentioning

confidence: 99%

On transport tensor of dynamically unresolved oceanic mesoscale eddies

Ryzhov

Berloff

2022

J. Fluid Mech.

Self Cite

View full text Add to dashboard Cite

Parameterizing mesoscale eddies in ocean circulation models remains an open problem due to the ambiguity with separating the eddies from large-scale flow, so that their interplay is consistent with the resolving skill of the employed non-eddy-resolving model. One way to address the issue is by using recently formulated dynamically filtered eddies. These eddies are obtained as the field errors of fitting some given reference ocean circulation into the employed coarse-grid ocean model. The main strengths are (i) no explicit spatio-temporal filter is needed for separating the large-scale and eddy flow components, (ii) the eddies are dynamically translated into the error-correcting forcing that perfectly augments the coarse-grid model towards reproducing the reference circulation. We uncovered physical properties of the eddies by interpreting involved nonlinear eddy/large-scale interactions via the classical flux-gradient relation. We described the eddies in terms of their full, space–time dependent transport tensor, which was made unique by constraining it to be the same for the potential vorticity, momentum and buoyancy fluxes. Both diffusive and advective parts of the transport tensor were found to be significant. The diffusive tensor component is characterised by polar eigenvalues and is further decomposed into isotropic and filamentation components. The latter component completely dominates, therefore, it should be taken into account by eddy parameterizations, which is not yet the case. We also showed that spatial inhomogeneities of the transport tensor components are important. Comparing these properties with those obtained for more common, locally filtered eddies revealed that they are distinctly different.

show abstract

Time-series analysis of differential pressure and flow velocity signals by means of weighted statistics

Chun

2022

Measurement

View full text Add to dashboard Cite

Correlation-based flow decomposition and statistical analysis of the eddy forcing

Abstract: Abstract

Cited by 12 publications

References 45 publications

A Comparison of Data‐Driven Approaches to Build Low‐Dimensional Ocean Models

A Comparison of Data‐Driven Approaches to Build Low‐Dimensional Ocean Models

On transport tensor of dynamically unresolved oceanic mesoscale eddies

Time-series analysis of differential pressure and flow velocity signals by means of weighted statistics

Contact Info

Product

Resources

About