The Sun, the Snow, and Everything in Between

When Henrik first approached me about a project aiming to detect solar jets — short-lived plasma beams ejected from the coronal region of the sun—using artificial intelligence methods, my initial reaction was scepticism. What does AI have to do with such physical processes? At first glance, my scepticism seemed justified—the topics appear unrelated altogether!

I had the opportunity to take a deeper look alongside two fellow physics master students, Michel and Henrik (their recollections are also detailed on this page), at the physical-meteorological observatory in Davos (PMOD). Our task was to develop a proof of concept for detecting these small solar features. And a deeper look we got: during our two weeks in Davos, we encountered nearly every roadblock imaginable in utilizing astrophysical photometry data in machine learning pipelines.

This made our intermediate result all the more satisfying: a neural network trained solely on observational data to detect solar jets. The concept actually makes a lot of sense since the jets cannot be modeled analytically, being inherently irregular. This means they are detectable only by the human eye, which is exactly the approach neural networks seek to replicate.

Overall, this project has been an incredible learning opportunity, even for our own “neural networks” (our brains), as it allowed us to become more familiar with this emerging technology and the paradigms associated with it.

Our coding endeavors were complemented by a healthy dose of snow, which we got to enjoy on the weekend (and more cautiously during our walk to work). Additionally, we were fortunate enough to secure a flat for the three of us, making us roommates for the duration of the project. This was a great opportunity to spend some time together in a more relaxed atmosphere, including cooking, a movie night and even a home-workout. So, beyond the scientific work we had a fantastic time in Davos, leaving us motivated to finalize this project in the upcoming weeks—hopefully yielding conclusive results.

Interested in ETH Studios? Discover more here.