AbstractThe predictability of the minimum sea ice extent (SIE) in the Arctic in September, especially for large anomaly years, is of strong current interest, given the rapid decline in sea ice amount. Our results show that June reflected solar radiation (RSR) is closely related to the underlying sea ice condition in that month and can be used to achieve September SIE predictions with good accuracy. The correlation coefficient between detrended June RSR and September SIE reaches 0.91 based on 16 year satellite observations, and the relatively high forecast skill using Modern‐Era Retrospective analysis for Research and Applications Version 2 reanalysis data is similar to or better than other complex prediction models. The results confirm the particular importance of the early summer sea ice state and help to explain the abrupt declines of September SIE in the 21st century (2007 and 2012).