One of the most widespread consequences of climate change is the disruption of trees’ phenological cycles. The extent to which tree phenology varies with local climate is largely genetically determined, and while a combination of temperature and photoperiodic cues are typically found to trigger bud burst (BB) in spring, it has proven harder to identify the main cues driving leaf senescence (LS) in autumn. We used 925 individual field-observations of BB and LS from six Fagus sylvatica provenances, covering the range of environmental conditions found across the species distribution, to: (i) estimate the dates of BB and LS of these provenances; (ii) assess the main drivers of LS; and (iii) predict the likely variation in the growing season length (GSL; defined by BB and LS timing) across populations under current and future climate scenarios. To this end, we first calibrated linear mixed-effects models for LS as a function of temperature, insolation and BB date. Secondly, we calculated the GSL for each provenance as the number of days between BB and LS. We found that: i) there were larger differences among provenances in the date of LS than in the date of BB; ii) the temperature through September, October and November was the main determinant of LS in beech, although covariation of temperature with daily insolation and precipitation-related variables suggests that all three variables may affect LS timing; and iii) GSL was predicted to increase in northern beech provenances and to shrink in populations from the core and the southern range under climate change. Consequently, the large differences in GSL across beech range in the present climate are likely to decrease under future climates where rising temperatures will alter the relationship between BB and LS, with northern populations increasing productivity by extending their growing season to take advantage of warmer conditions.