Estimating the onset of spring from a complex phenology database: trade-offs across geographic scales

Phenology is an important indicator of ecological response to climate change. Yet, phenological responses are highly variable among species and biogeographic regions. Recent monitoring initiatives have generated large phenological datasets comprised of observations from both professionals and volunteers. Because the observation frequency is often variable, there is uncertainty associated with estimating the timing of phenological activity. "Status monitoring" is an approach that focuses on recording observations throughout the full development of life cycle stages rather than only first dates in order to quantify uncertainty in generating phenological metrics, such as onset dates or duration. However, methods for using status data and calculating phenological metrics are not standardized. To understand how data selection criteria affect onset estimates of springtime leaf-out, we used status-based monitoring data curated by the USA National Phenology Network for 11 deciduous tree species in the eastern USA between 2009 and 2013. We asked, (1) How are estimates of the date of leaf-out onset, at the site and regional levels, influenced by different data selection criteria and methods for calculating onset, and (2) at the regional level, how does the timing of leaf-out relate to springtime minimum temperatures across latitudes and species? Results indicate that, to answer research questions at site to landscape levels, data users may need to apply more restrictive data selection criteria to increase confidence in calculating phenological metrics. However, when answering questions at the regional level, such as when investigating spatiotemporal patterns across a latitudinal gradient, there is low risk of acquiring erroneous results by maximizing sample size when using status-derived phenological data.