An observation-based progression modeling approach to spring and autumn deciduous tree phenology.

It is important to accurately determine the response of spring and autumn phenology to climate change in forest ecosystems, as phenological variations affect carbon balance, forest productivity, and biodiversity. We observed phenology intensively throughout spring and autumn in a temperate deciduous woodlot at Milwaukee, WI, USA, during 2007-2012. Twenty-four phenophase levels in spring and eight in autumn were recorded for 106 trees, including white ash, basswood, white oak, boxelder, red oak, and hophornbeam. Our phenological progression models revealed that accumulated degree-days and day length explained 87.9-93.4 % of the variation in spring canopy development and 75.8-89.1 % of the variation in autumn senescence. In addition, the timing of community-level spring and autumn phenophases and the length of the growing season from 1871 to 2012 were reconstructed with the models developed. All simulated spring phenophases significantly advanced at a rate from 0.24 to 0.48 days/decade (p ≤ 0.001) during the 1871-2012 period and from 1.58 to 2.00 days/decade (p < 0.02) during the 1970-2012 period; two simulated autumn phenophases were significantly delayed at a rate of 0.37 (mid-leaf coloration) and 0.50 (full-leaf coloration) days/decade (p < 0.01) during the 1970-2012 period. Consequently, the simulated growing season lengthened at a rate of 0.45 and 2.50 days/decade (p < =0.001), respectively, during the two periods. Our results further showed the variability of responses to climate between early and late spring phenophases, as well as between leaf coloration and leaf fall, and suggested accelerating simulated ecosystem responses to climate warming over the last four decades in comparison to the past 142 years.