When is the onset of a phenophase? Calculating phenological metrics from status monitoring data in the National Phenology Database
Detecting large scale phenological signals of climate change requires expansive datasets. Nature's Notebook, developed by the USA National Phenology Network (USA-NPN) is a national on-line program that facilitates scientists and citizens in collecting ground-based phenology data through standardized protocols. However, data can be limited across spatial, temporal, and ecological resolutions, creating challenges for estimating phenological metrics such as onset of phenophases (e.g. leaf-out). To address this challenge, we have developed criteria for calculating phenological metrics at differing spatial and temporal resolutions using status monitoring data within the National Phenology Database (NPDb). Standardizing the calculation of fundamental metrics is essential for performing subsequent statistical analyses to determine trends in spatiotemporal phenological responses within and among ecological communities. Here we present three case studies utilizing five increasingly conservative criteria for extracting individual phenophase recrods from the NPDb, the first positive record, first positive preceded by a negative record, the midpoint between the negative and positive, and the first positive and midpoint date only for observations with < 8 days between the negative and positive. We examined differences among metric estimates (e.g. mean onset date), their variance, and sample size at local, state, and bioregional spatial scales for species and functional types including bud break and flowering of temperature-sensitive deciduous trees and shrubs in the northeastern US and mid-Atlantic and flowering and fruiting of shrubs in water-limited ecosystems of California.
Although the most conservative criteria had the lowest mean standard deviation across metric estimates, it also had the lowest sample size, excluding up to 87% of data used in the least conservative method. Metric estimates within spatial scales varied from <1 to 33 days. Differences between criteria are minimized by frequent sampling, at least once per week. The balance between sample size and variance depends on the spatiotemporal scale and species or community of interest, particularly the inherent variability of ecosystems (e.g. repeated flowering events throughout the year). We make recommendations on criteria usage for data users of the NPDb based on the scientific question or management goals of interest and desired applications of the metric produced (e.g. general phenology calendar, annual interpolated maps).