Migration phenology plays a critical role in shaping bird life histories. While the spring migration phenology of birds has been widely studied, our understanding about the mechanisms underlying autumn migration is limited. Frost is an indicator of cold weather, food scarcity, and water unavailability, but how frost drives the autumn departure of migratory birds has not yet been quantified. In this study we propose the ‘frost wave hypothesis’, which posits that the autumn departure of waterfowl is driven by a successive wave of the onset of frost. Using bird satellite tracking data and generalized linear mixed models, we analyze how the departure probability of two waterfowl species during autumn migration is affected by frost, accumulated temperature, food, snow, ice, short-term weather conditions (i.e., wind, temperature and precipitation), remaining migration distances, relative stopover duration, and flight distances between stopover sites. We find that bird autumn departure probability sharply increases after the first frost spell when the accumulated temperature reaches 0 °C, facilitated by surface meridional wind and longer remaining migration distances. We underline the dominant effect of frost on autumn departure, as birds tend to leave even under head wind if the time lag since the onset of frost is large. Time constraints that trigger southward departure are likely to be stronger when migrating birds are still far from their wintering site. By riding the frost wave, birds manage to maximize stopover site utilization while escaping harsh environmental conditions. Our findings improve the understanding of annual avian migration and help quantify the impact of global climate change on migratory waterfowl.