Plant phenology is a sensitive indicator of climate change^1,2,3,4 and plays an important role in regulating carbon uptake by plants^5,6,7. Previous studies have focused on spring leaf-out by daytime temperature and the onset of snow-melt time^8,9, but the drivers controlling leaf senescence date (LSD) in autumn remain largely unknown^10,11,12. Using long-term ground phenological records (14,536 time series since the 1900s) and satellite greenness observations dating back to the 1980s, we show that rising pre-season maximum daytime (T_day) and minimum night-time (T_night) temperatures had contrasting effects on the timing of autumn LSD in the Northern Hemisphere (> 20° N). If higher T_day leads to an earlier or later LSD, an increase in T_night systematically drives LSD to occur oppositely. Contrasting impacts of daytime and night-time warming on drought stress may be the underlying mechanism. Our LSD model considering these opposite effects improved autumn phenology modelling and predicted an overall earlier autumn LSD by the end of this century compared with traditional projections. These results challenge the notion of prolonged growth under higher autumn temperatures, suggesting instead that leaf senescence in the Northern Hemisphere will begin earlier than currently expected, causing a positive climate feedback.

Spring warming substantially advances leaf unfolding and flowering time for perennials. Winter warming, however, decreases chilling accumulation (CA), which increases the heat requirement (HR) and acts to delay spring phenology. Whether or not this negative CA-HR relationship is correctly interpreted in ecosystem models remains unknown. Using leaf unfolding and flowering data for 30 perennials in Europe, here we show that more than half (7 of 12) of current chilling models are invalid since they show a positive CA-HR relationship. The possible reason is that they overlook the effect of freezing temperature on dormancy release. Overestimation of the advance in spring phenology by the end of this century by these invalid chilling models could be as large as 7.6 and 20.0 days under RCPs 4.5 and 8.5, respectively. Our results highlight the need for a better representation of chilling for the correct understanding of spring phenological responses to future climate change.

Significance: Decline in winds over past decades were observed over high northern latitudes (>50°), yet its influence on the date of autumn leaf senescence (DFS) remains unknown. Using ground observations, flux measurements, and remote sensing imagery, here we show that decline in winds significantly extended DFS over high latitudes at a magnitude comparable with the temperature and precipitation effects. We found that decline in winds reduces evapotranspiration, causes fewer damaging effects, and also results in decreased cooling effect. Our results therefore are of great significance for carbon cycle modeling because an improved algorithm based on these findings projected overall widespread earlier DFS by the end of this century, contributing potentially to a positive feedback to climate.