Europlanet Science Congress 2022
Palacio de Congresos de Granada, Spain
18 – 23 September 2022
Europlanet Science Congress 2022
Palacio de Congresos de Granada, Spain
18 September – 23 September 2022
EPSC Abstracts
Vol. 16, EPSC2022-1061, 2022
Europlanet Science Congress 2022
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Vortex Statistics in the Saturn DYNAMICO GCM: Manual and Automated Detection

Padraig Donnelly1, Aymeric Spiga1, Sandrine Guerlet1, Matt James2, and Deborah Bardet2
Padraig Donnelly et al.
  • 1Laboratoire de Météorologie Dynamique, Sorbonne University, Paris, France
  • 2Planetary Science Group, University of Leicester, Leicester, UK


The Saturn DYNAMICO Global Climate Model (GCM) is a high-resolution, multi-annual numerical simulation of Saturn's atmospheric dynamics [1], combining a radiative-convective equilibrium model [2] and a hydrodynamical solver on an icosahedral grid. The model reproduces well the observed behaviour of jets and eddy-momentum transfer to the mean flow. Vortices arise naturally in the model over time but until now they have not been given direct consideration. Here we investigate the long-term statistical distribution and organization of vortices using (1) a manual visual inspection method and (2) automated techniques that utilise machine learning and analytical calculations as a means of validating the first approach.

Manual Detection

This vortex detection method is similar to previous observational studies of Jupiter and Saturn [3, 4, 5, 6] and shows how the spatial and temporal distributions of the modelled vortices compares to those observed on Saturn [4, 5, 6], as well as studying the formation conditions and long-term temporal evolution of vortex distributions. With seven simulated model years at ½-degree spatial resolution, we constrain well the size and location of vortices, the horizontal wind field components and the magnitude and sign of horizontal vorticity, enabling direct comparison of the manual and automated methodologies.

Automated Detection

A convolutional neural network is used to reproduce the manual visual detection method across the entire timeseries using the same assumptions and using the results of the manual study as a training set. We also study the Angular Momentum Eddy Detection Algorithm (AMEDA, [7]) designed for the analysis of terrestrial oceanic eddies. The machine learning study is ongoing and the results of the AMEDA algorithm are largely consistent with the manual approach, meaning that this algorithm can be used in future studies of Jupiter and Saturn DYNAMICO GCM outputs.

Figure 1: Vortex count over the entire mature model timeline for the manual (black) and AMEDA (red) approaches. Manual technique measures at each seasonal peak, AMEDA measures at each timestep to create a seasonal average for comparison. AMEDA analysis to be extended to earlier years.