Single measures of phenology do not accurately predict phenological shifts

Background/Question/Methods

Changes in the timing of species life cycle events (phenological shifts) disrupt ecosystems by altering the timing and duration of individual species and species overlap. Most phenology research considers only the first or mean occurrence of a phenological event, but since individuals vary in their timing, single metrics (e.g. mean) of phenology can be misleading. This research examined if shifts in the first phenological event were equal to shifts in the mean phenological event and how accurately these two metrics could predict a distribution curve for population level phenology. Timing of breeding for eleven frog species (gathered from frog calling data) over a fifteen year period was analyzed for eight Texas ponds.  From this data, we recovered species specific distributions of the timing of reproduction and made comparisons between species, location, and year.

Results/Conclusions

Our results indicate that the first and mean phenological events poorly predict population level phenology, and thus can be misleading indicators of phenological shifts. Within the eleven species monitored, almost all showed substantial variation in not only the first and mean phenological event, but also in the shape of the distribution of phenological events, both between ponds and between years. Furthermore, first and mean event did not shift equally, so the shape of the distribution could change substantially even if one or both metrics were unchanged. For example, some species showed shifts from a unimodal to bimodal distribution while maintaining the same first and mean metrics. Examining the entire distribution of phenological events enables better predictions about how phenological shifts affect population dynamics, interspecific interactions, and community composition. With climate change rapidly altering species phenologies, it is increasingly important to be able to accurately track phenological shifts and predict the net effects. We show that this requires measuring phenological distributions in lieu of single metrics.