From seed dispersal service to reproductive collapse: density-dependent outcome of a palm-mammal interaction
Data files
Jun 08, 2023 version files 76.03 KB
-
HIGH_palms.dat
-
INTERMEDIATE_palms_seeds.dat
-
INTERMEDIATE_seedpredation.dat
-
INTERMEDIATE_trivariate.mcf
-
LOW_palms_seeds.dat
-
LOW_seedpredation.dat
-
LOW_trivariate.mcf
-
README.md
Abstract
Interspecific ecological interactions are inherently context-dependent. They may vary in both magnitude and sign depending on the biotic and abiotic conditions, depicting a mutualism-antagonism continuum. However, how population abundances and the activity of interacting species modulate these interactions remains underexplored. Here, we chose the interaction between the Mediterranean palm Chamaerops humilis and the feral goat Capra hircus in Mallorca (Balearic Islands, Spain). We selected three study plots with low, intermediate and high intensities of goat activity where we characterized palm distribution, seed rain, seed predation, and early palm recruitment during two consecutive years. Since goats can cause both costs (e.g. florivory) and benefits (e.g. seed dispersal) to C. humilis performance, we investigated the following three questions: (1) Does the spatial distribution of adult palms vary depending on the intensity of goat activity? (2) Does the intensity of goat activity influence seed rain and its potential spatial association with adult palms? (3) To what extent does the intensity of goat activity determine post-dispersal events such as seed predation and seedling emergence? We found that adult palms showed a more clumped and complex distribution (double-cluster process) in plots with low and intermediate goat activity compared to that with high goat activity (simple-cluster process). In the low goat activity plot, dispersed seeds were spatially aggregated around adult palms, showing twice insect-seed predation and nearly three times lower seed germination success than those in the intermediate goat activity plot. Palm seed dispersal and recruitment were almost nil in the high goat activity plot due to heavy consumption of palm inflorescences and developing fruits by goats. Our findings demonstrate how the net outcome of plant-animal interactions can change from mutualism to antagonism, from reproductive service to reproductive collapse, depending on the abundance and the activity of the interacting species.
Methods
To evaluate the hypothesis that the outcome of the palm-goat interaction depends on the intensity of goat activity, we first described the spatial distribution of C. humilis by geo-referencing in 2019 (with a Garmin GPSMAP) all adult palms within each study plot. Then, we searched for dispersed seeds in two consecutive years (2019 and 2020) from September to December, which covered the entire seed dispersal seasons of C. humilis. Every dispersed unit (defined as one dispersed seed or a group of dispersed seeds) was geo-referenced as an independent point and covered with a wire 1-cm mesh cage (10 x 10 x 9 cm) to prevent seed predation, seed removal and seedling trampling by vertebrates. (e.g. Fedriani and Wiegand 2014). GPS coordinates were then imported into a geographic information system (Quantum GIS Team 2023). Caged dispersed seeds were checked every two months up to January 2022 (i.e. two years after the end of the first sampling season) to record seed predation by scolytine beetles, seed germination, seedling emergence, and seedling survival. Then, we used recent extensions of spatial point pattern analyses (SPPA, hereafter; Wiegand and Moloney 2014, Velázquez et al. 2016, Ben-Said 2021) to explore the data. Specifically, to characterize the spatial distribution of adult palms and seed rain (question 1) we fitted univariate cluster point process models to the data, whereas to analyze the spatial association between adult palms and dispersed units (question 2) we employed bivariate pair correlation functions together with realizations of the spatial independence null model. Finally, to detect spatial patterns of seed predation and to test the Janzen-Connell hypothesis (question 3) we used mark connection functions and trivariate mark correlation functions together with the random labeling null model.
Usage notes
All spatial point pattern analyses were carried out with the free software Programita (http://programita.org, Wiegand and Moloney 2014). Detailed descriptions of the statistical procedures are provided in Supporting Information.