Plant phenology is the timing of recurring life history events such as leaf-out, flowering, seed production, and senescence, and is a key indicator of climate change. Plant phenology also affects nutrient and energy cycles through the timing of photosynthesis and is critical to animal species that rely on seasonal plant resources.
However individuals of the same plant species in the same geographic area often vary markedly in the timing of key phenology events. This variation may be explained by factors that are intrinsic to individuals (e.g. tree height), local communities (e.g. position in canopy), as well as immediate abiotic environmental conditions (e.g. elevation). Understanding this variation is important for building general predictive plant phenological models and connecting ground-based phenology measurements of individuals with remote sensing phenology data.
I used a dataset produced by the National Ecological Observatory Network (NEON) from 2014 to 2016 to explore the drivers of local variation in plant phenology of 2610 individual plants comprising 83 plant species at 35 sites across the United States. I employed hierarchical Bayesian modeling to (1) characterize the uncertainty in the timing of phenology events due to sampling regime; (2) characterize the variability in phenology events by species, site, and growth form; (3) analyze the effects of local-scale factors, such plant height, canopy position, and slope, on phenology events.
Results/Conclusions
Uncertainty in phenology events due to sampling regime ranged from 5.4 days (start of new growth) to 7.7 days (end of leaf shed). Uncertainty in end of new growth (7.5 days), first color (6.4 days), first flower (6.7 days), and last flower (6.9 days) fell in between. However, there was variation among site-species (standard deviation of 1.9 to 3.1 days) and variation among individuals within site-species (1.9 to 2.4 days). Uncertainty in spring phenology events (new growth and flowering start and end) showed more variation among site-species than among individuals, whereas for fall phenology events (first color and end of leaf shed) variation among site-species was less than among-individual variation.
These preliminary results suggest that the standardized ground-based phenology methods used by NEON (based on those of the National Phenology Network) produce data with an uncertainty of about a week. Analyses that use the datasets based on these methods should consider this uncertainty when evaluating the drivers and effects of phenology and propagate it into larger models.