Friday

Being situated in low Earth orbit, the NASA/ESA Hubble Space Telescope (HST) is susceptible to higher orbit artificial satellites and space debris crossing its field of view. We use Google's AutoML Vision object detection algorithm based on deep learning and trained on volunteers' classifications from the Hubble Asteroid Hunter citizen science project (www.asteroidhunter.org) to identify asteroids and satellite trails in HST archival images taken in the past 18 years.

We measure, for the first time, the fraction of images impacted by artificial satellites for different instruments, filters, and as a function of time. We find a fraction of 7.8% of the composite HST images and 2.5% of the individual exposures with a typical exposure time of 11 minutes being crossed by satellites. We show results from this project and discuss the prospects of using deep learning and citizen science to identify trail shapes in astronomical images. Many astronomers have recently raised concerns about the impact of satellite constellations on ground-based observations. We argue that satellite megaconstellations will also impact observations from low-Earth orbit, such as those with the iconic Hubble Space Telescope.