Boston University Boston, Massachusetts, United States
Background/Question/Methods
Phenology is a primary ecological indicator of climate change and has substantial ecosystem and climatic impacts. Leaf senescence is important because it prepares deciduous broadleaf forests for dormancy, but it is less studied and understood than spring phenology. Traditionally, senescence studies and models assume that senescence occurs due to a trigger when a certain threshold of cooling degree days or decreased photoperiod is reached. We investigated if a model could be created without a trigger that predicts the start of senescence inflection by modeling daily greenness as the sum of the breakdown remainder from the previous day and newly synthesis that is linearly related to daily temperatures and day lengths. We used the model to predict greenness at 24 PhenoCam sites throughout the autumn and then compared the amount of inflection in the predictions for each site-year to the inflections in the PhenoCam data. In order to further investigate the lack of a need for a trigger, we attempted to calibrate the model to each site using only pre-senescence data. We also used the model calibrations for each of the sites to predict greenness at 69 other PhenoCam sites and compared these predictions to the validation site’s climatological average estimates.
Results/Conclusions
Our model was able to produce reasonable estimates of greenness for withheld data even without including any senescence greenness data and potentially predicted regreening events. We also found that our model predicted significant variation in the amount of maximum inflection between site-years in the calibration sites when not including post start of senescence phenology data in the calibration. Additionally, our model consistently and significantly better predicted greeness throughout the autumn better than climatology in withheld years. Furthermore, our site-specific model predicted a start of senescence inflection in 49% of the site-years without including any post start of senescence calibration data. Lastly, our model could regularly predict greenness at other sites better than their climatologies and could predict significant variation in the timing of the start of senescence even when it was not calibrated to any senescence data. This mechanistic model is novel because it shows that a trigger is not necessary for senescence to start and that summer only data can be used to estimate breakdown and synthesis that will predict senescence at the calibration and validation sites.
