The State of the Art in Vortex Extraction

Günther, Tobias; Theisel, Holger

doi:10.1111/cgf.13319

Cited by 105 publications

(80 citation statements)

References 124 publications

(304 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, when averaging the deviation of vorticity from its local neighborhood along trajectories, the resulting Lagrangian-averaged vorticity deviation has a desirable property: Similar to the aforementioned LCS, it is invariant to rotations and translations of the frame of reference (Haller et al, 2016). Due to the relativity between observer and the observed feature, this formally guarantees that translating and rotating flow structures can be faithfully extracted (Günther & Theisel, 2018). Analyzing a vector field from the perspective of a local observer that moves with the vortices will enable the application of sophisticated steady-state vortex separation and extraction techniques (Rojo & Günther, 2019).…”

Section: 1029/2019jd032121mentioning

confidence: 99%

Evolution of an Atmospheric Kármán Vortex Street From High‐Resolution Satellite Winds: Guadalupe Island Case Study

et al. 2020

View full text Add to dashboard Cite

Vortex streets formed in the stratocumulus‐capped wake of mountainous islands are the atmospheric analogues of the classic Kármán vortex street observed in laboratory flows past bluff bodies. The quantitative analysis of these mesoscale unsteady atmospheric flows has been hampered by the lack of satellite wind retrievals of sufficiently high spatial and temporal resolution. Taking advantage of the cutting‐edge Advanced Baseline Imager, we derived kilometer‐scale cloud‐motion winds at 5‐min frequency for a vortex street in the lee of Guadalupe Island imaged by Geostationary Operational Environmental Satellite‐16. Combined with Moderate Resolution Imaging Spectroradiometer data, the geostationary imagery also provided accurate stereo cloud‐top heights. The time series of geostationary winds, supplemented with snapshots of ocean surface winds from the Advanced Scatterometer, allowed us to capture the wake oscillations and measure vortex shedding dynamics. The retrievals revealed a markedly asymmetric vortex decay, with cyclonic eddies having larger peak vorticities than anticyclonic eddies at the same downstream location. Drawing on the vast knowledge accumulated about laboratory bluff body flows, we argue that the asymmetric island wake arises from the combined effects of Earth's rotation and Guadalupe's nonaxisymmetric shape resembling an inclined flat plate at low angle of attack. However, numerical simulations will need to establish whether or not the selective destabilization of the shallow atmospheric anticyclonic eddies is caused by the same mechanisms that destabilize the deep columnar anticyclones of laboratory flows, such as three‐dimensional vertical perturbations due to centrifugal or elliptical instabilities.

show abstract

Section: 1029/2019jd032121mentioning

confidence: 99%

Evolution of an Atmospheric Kármán Vortex Street From High‐Resolution Satellite Winds: Guadalupe Island Case Study

et al. 2020

View full text Add to dashboard Cite

show abstract

“…The 2D wind direction is a promising atmospheric attribute, since it provides directional information on the possible pathways of clouds. When it comes to feature extraction from unsteady flows, the choice of the right reference frame is important [GT18]. To test, whether a neural network would learn to extract the reference frame internally on its own, we pre‐compute the optimal reference frame and compare the network output with the output obtained from the original vector field.…”

Section: Resultsmentioning

confidence: 99%

Visualization of Neural Network Predictions for Weather Forecasting

Roesch

Günther

2018

Computer Graphics Forum

Self Cite

View full text Add to dashboard Cite

Recurrent neural networks are prime candidates for learning evolutions in multi‐dimensional time series data. The performance of such a network is judged by the loss function, which is aggregated into a scalar value that decreases during training. Observing only this number hides the variation that occurs within the typically large training and testing data sets. Understanding these variations is of highest importance to adjust network hyper‐parameters, such as the number of neurons, number of layers or to adjust the training set to include more representative examples. In this paper, we design a comprehensive and interactive system that allows users to study the output of recurrent neural networks on both the complete training data and testing data. We follow a coarse‐to‐fine strategy, providing overviews of annual, monthly and daily patterns in the time series and directly support a comparison of different hyper‐parameter settings. We applied our method to a recurrent convolutional neural network that was trained and tested on 25 years of climate data to forecast meteorological attributes, such as temperature, pressure and wind velocity. We further visualize the quality of the forecasting models, when applied to various locations on the Earth and we examine the combination of several forecasting models.

show abstract

“…Over the past decades, dozens of vortex extraction algorithms have been proposed in the flow visualization and fluid mechanics literature. We refer to Günther and Theisel for a recent overview [GT18].…”

Section: Related Workmentioning

confidence: 99%

“…The robust extraction of vortices remained to this day one of the most difficult problems of unsteady vector field analysis [GT18]. Vortices themselves are studied in many different applications, such as for engine design [RP96, GLT*07], blood flow analysis [KGP*13, OJCJP16] or even in the atmosphere of other planets [HH16].…”

Section: Introductionmentioning

confidence: 99%

Robust Reference Frame Extraction from Unsteady 2D Vector Fields with Convolutional Neural Networks

Kim

Günther

2019

Computer Graphics Forum

Self Cite

View full text Add to dashboard Cite

Robust feature extraction is an integral part of scientific visualization. In unsteady vector field analysis, researchers recently directed their attention towards the computation of near‐steady reference frames for vortex extraction, which is a numerically challenging endeavor. In this paper, we utilize a convolutional neural network to combine two steps of the visualization pipeline in an end‐to‐end manner: the filtering and the feature extraction. We use neural networks for the extraction of a steady reference frame for a given unsteady 2D vector field. By conditioning the neural network to noisy inputs and resampling artifacts, we obtain numerically stabler results than existing optimization‐based approaches. Supervised deep learning typically requires a large amount of training data. Thus, our second contribution is the creation of a vector field benchmark data set, which is generally useful for any local deep learning‐based feature extraction. Based on Vatistas velocity profile, we formulate a parametric vector field mixture model that we parameterize based on numerically‐computed example vector fields in near‐steady reference frames. Given the parametric model, we can efficiently synthesize thousands of vector fields that serve as input to our deep learning architecture. The proposed network is evaluated on an unseen numerical fluid flow simulation.

show abstract

The State of the Art in Vortex Extraction

Cited by 105 publications

References 124 publications

Evolution of an Atmospheric Kármán Vortex Street From High‐Resolution Satellite Winds: Guadalupe Island Case Study

Evolution of an Atmospheric Kármán Vortex Street From High‐Resolution Satellite Winds: Guadalupe Island Case Study

Visualization of Neural Network Predictions for Weather Forecasting

Robust Reference Frame Extraction from Unsteady 2D Vector Fields with Convolutional Neural Networks

Contact Info

Product

Resources

About