Investigating light pollution trends at astronomical observatories worldwide

ABSTRACT The negative impact of light pollution on astronomical observatories threatens their research potential, necessitating continuous monitoring. This study examines artificial light at night (ALAN) trends at 2470 observatories worldwide from 2012 to 2024, with some hosting +4 m class telescopes. Using 15 regression methods, including machine learning techniques, along with Mann–Kendall and Sen’s Slope statistics, we observe significant ALAN trends. Results indicate 60 per cent of observatories show a positive ALAN trend, implying a loss of darkness. Among these, 1067 exhibited statistically significant trends, with 635 increasing in AL. In particular, the Mann–Kendall test highlighted changes in eight + 4 m class sites: Cerro Pachon, Paranal, Special Astrophysical, Mount Graham International, Roque de los Muchachos, Fred Lawrence Whipple, Eastern Anatolia, and Beijing Astronomical Observatories. Of the 15 regression models tested, the Bayesian Ridge, Decision Tree, and Random Forest regressors showed the best overall performance. Satellite imagery reveals that observatories such as Palomar, Eastern Anatolia, Beijing Astronomical, and Maple Ridge face increased ALAN risks due to urban encroachment. However, large observatories near natural darkness levels should be closely analysed, indicating successful preservation efforts. These findings underscore the need for protective measures and site identification less impacted by ALAN, advocating proactive legal protections.