Skip to main content

Data from: Induced phenological avoidance: a neglected defense mechanism against seed predation in plants

Cite this dataset

Sercu, Bram; Moeneclaey, Iris; Bonte, Dries; Baeten, Lander (2019). Data from: Induced phenological avoidance: a neglected defense mechanism against seed predation in plants [Dataset]. Dryad.


1.    Flowering phenology is an important life history trait affecting plant reproductive performance and is influenced by various abiotic and biotic factors. Pre-dispersal seed predation and pollination are expected to impose counteracting selection pressure on flowering phenology, with pre-dispersal seed predation expected to favor off-peak flowering and pollination to favor synchronous flowering. 
2.      Here we studied the effect of pre-dispersal seed predation by the beetle Byturus ochraceus, a specialist seed herbivore, on the flowering phenology of Geum urbanum. This forest understorey plant species is self-pollinating, so that the influence of seed predation can be studied independent from pollination. We measured in detail the timing and predation rate of individual flowers during two consecutive years in more than 60 individuals. We tested the hypotheses that pre-dispersal seed predation exerts selection for within-season compensatory flowering as well as for induced phenological avoidance in the following season.
3.      We found no indication for compensatory flowering within a growing season, but plants that experienced predation shifted their flowers to the end of the flowering season the subsequent year. This induced phenological avoidance points to a plastic response to pre-dispersal seed predation that may be adaptive. Importantly, the delay in flower production came at a cost, since flowers later in the season had a reduced seed output, presumably because of increasing light limitation following forest canopy closure. 
4.      Synthesis: Herbivory by specialist enemies can cause serious fitness decline in hosts. We here show that induced shifts in phenology can form an important defense strategy against pre-dispersal seed predation. The induced mismatches between herbivore and host phenology are anticipated to be adaptive when herbivory is predictable across successive flowering periods.


Study site and species

The research was conducted in the Aelmoeseneie forest, located south of Ghent (Gontrode, Belgium). This ca. 30 ha forested area entails a mix of ancient and post-agricultural forest and has been managed as high forest since 1950 (Vanhellemont & Verheyen, 2011). The soil consists of sand and sandy loam with alluvial ash forest on the more humid parts and acidophilous beech forest on the drier sandy parts. Within the forest, we selected ten groups of Geum urbanum plants in different locations spread over the forest. A group of individuals located close to each other, within a radius of 15 m, was considered a ‘plot’. In 2015, 59 plants were marked for detailed monitoring (N = 10 plots, 4-6 plants per plot). In 2016, the set of monitored plants was expanded with another 50 plants to obtain a larger sample size (N=12 plots, 4 – 15 plants per plot). In 2017 only the plants from 2016 were monitored. Differences in abiotic conditions, i.e. canopy openness, canopy phenology and chemical soil conditions were measured and controlled for but none of these variables were correlated with flowering phenology, reproductive output or presence of seed predators and could therefore be ignored in this study (unpublished data).

Plants were monitored once a week in 2015 from May until October. Timing of flower emergence was recorded as the day of the year and all open unmarked flowers were marked with a unique tag to enable identification of individual flowers. Total number of flowers per plant could be derived from these data at the end of the season. Based on the data of 2015, flowers were monitored on fewer occasions in 2016 (July 4, 12, 20, 27, August 12 and September 7; day of year 186, 194, 202, 209, 225, 251) and 2017 (July 13, September 19, October 18; day of year 194, 262, 291). We observed two flowering peaks per season during the three consecutive years of observations and during a preliminary study in 2014: high number of flowers in June (around day of year 150) and in August (around day of year 225) and a period with fewer flowers in between (around day of year 194) (SI 1 in the Supplementary Information). This period of fewer flowers coincides with the end of predator infection. Because of the bimodality of the flowering phenology of G. urbanum, off-peak flowering cannot be straightforwardly quantified. We therefore defined off-peak flowering relative to the activity period of the seed predators, that is, flowering away from the peak in seed predator activity. During the first flowering peak, most flowers emerged, and seed predation took place. We hence calculated, for each plant, the proportion of flowers that appeared in the second flowering peak relative to the total number of flowers produced on each plant as a measure of ‘off-peak’ flowering.

In 2015 and 2016, flowers were collected throughout the season when seeds were ripe, that is, for each plant we collected flowers on different occasions. In 2015, few flowers contained beetle larvae, which means that seed herbivory was still ongoing. In 2016, flowers were collected slightly later in order to ensure that seed herbivory was completed and no larvae were found in the flowers anymore. No seeds were collected in 2017. Only flowering phenology data were collected to test for lagged effects of seed predation on flowering phenology. The seeds from all flowers were checked for herbivory by Byturus ochraceus. The number of seeds and seed mass was determined separately for predated and unpredated seeds for each flower. From these data, we calculated the mean mass per seed for unpredated seeds. We used seed mass as a proxy for reproductive output instead of the number of seeds, since many flowers contained small seeds that were weakly developed. Moreover, we found a strong correlation between seed mass and germination probability (unpublished results). As a measure of reproductive output per flower, total seed mass per flower was calculated as the sum of the mass of predated and unpredated seeds. We included the weight of the predated seeds since predation rate could vary among seeds from completely eaten (with negligible weight) to minor damage, leaving most of the seed intact. Because we have no information on the germination probability of damaged seeds, the most conservative approach, that does not exaggerate the effect of predation on reproductive output, is to add the weight of predated seeds. As a measure of reproductive output per plant, total seed mass per plant was calculated as the sum of the seed mass of all flowers on that plant. For the analysis, we used a categorical variable for predation occurrence at flower and plant level with two categories: predated or unpredated (‘predation’ hereafter). A flower was considered predated if at least one seed showed signs of predation and a plant was considered predated if at least one flower was predated.


Research Foundation - Flanders, Award: W0.003.16N