Thank you for all your efforts in helping this round of data to a completion! Watch this space while we work on getting more data to continue the project!

Results

Paper published in the Planetary Science Journal

Our first paper is published on PSJ here. We also discuss these results in the AAS Journal Author Series

In this paper, we used your classifications—vortex size, color, and location—to better understand Jupiter’s atmosphere. One key thing we looked at was something called the Rossby deformation radius. It basically tells us how big a vortex can get before it gets distorted by the planet’s rotation. This value matters because it helps us understand how stable or turbulent the atmosphere is. We found that while this radius is mostly consistent across Jupiter, there are some regions where it changes a lot—these areas might be hotspots for storms and turbulence.

You can look at this data yourself:

• Vortex properties consensus: https://zenodo.org/records/11659728
• Images and Consensus Labels: https://huggingface.co/datasets/ramanakumars/JovianVortexHunter

You can also explore the consensus and vortex properties in this data dashboard that we are building: JVHExplorer. Note that this is currently in beta so you might experience some slow performance.

Paper published in Citizen Science Theory and Practice

Our second paper was published in the Citizen Science Theory and Practice special issue on "The Future of Artificial Intelligence and Citizen Science". You can read the paper here

During the project, many of you came across features that didn’t clearly fall into “vortex” or “turbulence” categories. We trained a machine learning model to help understand these confusing cases. Interestingly, these ambiguous features might be in-between stages in vortex development, giving us clues about how these storms evolve over time. Our machine model showed us that it was possible to find these confusing in-between vortices (and other vortices that showed anomalous signatures) with a simple architecture, which can potentially be applied to other citizen science projects.

You can explore the machine learning that we used here: https://github.com/ramanakumars/cvae. It was trained on the data in the HuggingFace repository listed above.