Macroscale prediction of autumn leaf coloration throughout the continental United States
Previous studies have shown that warming temperatures can affect the phenology of cold deciduous forests, delaying the timing of leaf coloration. However, these works have principally been based on observations from a small number of sites. Consequently it has been challenging to infer continental-scale variations in the phenology of individual deciduous forest species and the extent to which there may be underlying climate drivers for these variations. To address that problem, this study evaluated and predicted the large-scale variations of leaf colouring by using macroscale observations and models.
We developed leaf colouring models using select observations (1) from Harvard Forest only and (2) from both Harvard Forest and a new, ground-based, Alaskan dataset from the USA National Phenology Network (USA-NPN). Both model types were evaluated using reserved observations from the continental-scale USA-NPN that were not used in model calibration. Validated models were then used to assess the spatial scaling and interspecies variation in the timing of leaf coloration. The sensitivity of the models to projected climate change was also evaluated.
Using a model calibrated only with data from Harvard Forest, significant biases were found in predictions of leaf colouring date for species with broad habitat ranges in the temperate to boreal regions. When calibration data from both Harvard Forest and Alaska were used, model performance improved throughout the whole continent. It was also found that species with similar shade tolerance could be described by similar models. Finally, the models indicated that climate change over the next century will affect leaf coloration in Alaska less than in the Harvard Forest region.
For a given species, continental-scale variations in the timing of autumn leaf coloration can be predicted using a model driven by photoperiod and daily temperature. The temperature sensitivity of the leaf colouring date is nonlinear, such that warmer regions have a larger temperature sensitivity than cooler regions. Species-specific measurements from multiple environments are essential for model parameterization.