Abstract Detail

Biodiversity Informatics & Herbarium Digitization

Hearn, David [1].

A general theory of phenological timing reveals counterintuitive interpretations of historical patterns of phenological change.

Digitized herbarium specimens provide a wealth of phenological data, and multiple studies have used these data to examine the relationships between changes in climate, biological environment, and timing of phenological events. Analyses of phenological data have, thus far, taken place in the absence of a general theory of phenological timing and have relied on more traditional statistical analysis that assume normality. Inferences based on normality have subsequently suggested widespread shifts of the onset of phenophases to earlier dates in the year in more recent years, as would be predicted from a climate warming scenario. However, conclusions about historical shifts in phenological timing can be incorrect using this traditional framework when the specific components that contribute to the sampled distribution of phenophase observation times are not considered. To address this issue, I developed a general theory of phenological timing that decomposes a phenophase into two separate components: the onset time and the duration of the phenophase. When phenophase data are analyzed using this theory-driven framework, a nonintuitive result is possible: shifts to later phenophase onset times in the population may be associated with earlier observed times, on average. This outcome can occur when the onset shifts to a later time in the year, on average, whereas the duration is greatly shortened, on average. I apply Bayesian MCMC techniques to infer population-level parameter values of the onset and duration for 13 species of NE North American spring ephemerals by analyzing sampling-bias-corrected phenophase data extracted from over 10,000 digitized herbarium specimens. The Bayesian inferences indicated, on average, that five species had shifts to earlier onset and shorter durations of active growth, six had shifts to earlier onset and longer durations, and two species had later onsets and shorter durations. The two species with inferred later onset dates fell into the above-mentioned, nonintuitive category in which their average sampled dates appeared to occur earlier in the year in more recent years compared to sampled dates from the late 1800's and early 1900's. In all species, Bayesian measures of uncertainty of parameter values were largely overlapping when phenophases from earlier years were compared to those of more recent years; this overlap indicates that much of the variation in phenological timing can be attributed to uncertainty and / or variation in the data. In conclusion, the newly developed, general theory of phenological timing provides a novel framework with which to analyze variation in phenophase timing and avoid incorrect inferences based on misapplication of more traditional statistical approaches.

1 - Towson University, Biology, 8000 York Road, Towson, MD, 21252, United States

Bayesian inference
phenological shift
biological timing
Herbarium Digitization
spring ephemeral.

Presentation Type: Oral Paper
Number: BI&HD II007
Abstract ID:394
Candidate for Awards:None

Copyright © 2000-2022, Botanical Society of America. All rights reserved