Small mammals reduce distance-dependence and increase seed predation risk in tropical rainforest fragments
Data files
Jun 12, 2022 version files 234.80 KB
-
camera_trap_data.csv
-
focal_trees_latlong.csv
-
README.txt
-
seed_plots_raw.csv
-
seed_survival_data.csv
-
seed_traits.csv
Abstract
Methods
Data collection:
We set up a field experiment to compare risk of seed predation by mammals on four native rainforest tree species — Acronychia pedunculata, Cullenia exarillata, Ormosia travancorica, and Syzygium rubicundum, between fragmented forests in Valparai Plateau and contiguous forests of the Anamalai Tiger Reserve, India . At focal trees of each species in contiguous forest and fragments, we set up a field experiment with seed plots consisting of 10 seeds placed ‘near’ and ‘far’ from focal trees. Further, to examine seed predation in the absence of mammal predators, we set up seed plots within mammal-exclosures, both near and far from focal trees beside the open plots. We monitored the seed plots weekly for 6 – 8 weeks. Each week, we recorded the number of intact seeds remaining in the plot and seeds with visible signs of predation by insects (identified by a boring hole), or mammals (identified the remains of chewed seeds with tooth marks, or remains of tags without seeds in the plots). Additionally, to monitor mammal visitation, we set up infra-red motion detection cameras at 16 locations each in contiguous forests and fragments near trees of Cullenia exarillata, Ormosia travancorica, and Syzygium rubicundum. Cameras were placed 30-60 cm above the ground, and 2 m away from the seed plots with the field of view focused towards the seed plots.
Data Analysis:
We used Cox proportional hazards regression in a mixed-effects framework (Therneau, 2015) to model seed survival (time until predation or the end of monitoring) as a function of the habitat status (fragment/contiguous), plot location with respect to the focal tree (near/far), and their interaction. For seeds that were intact at the end or that went missing during the study, we modelled seed survival for the duration that they were monitored. These seeds were ‘censored’, (i.e. marked alive at the end of monitoring), as is a standard practice in survival analysis. To account for spatial proximity of seeds within a plot, and between pairs of plots at focal trees, we included plot ID nested within focal tree ID as a random effect. We conducted the analysis for seeds within and outside mammal exclosures and for each species separetly. Additionally, we also plotted the average proportion of seeds predated over time until 6 weeks for graphical representation of seed survival over time. From the camera trap footage we calculated the mean visitation rate and proportion of seeds removed or eaten by each seed predator species and group (large and small mammals).
Usage notes
All statistical analysis were carried out using the software R (Core Team, 2020). We used packages ‘ggplot2’ for graphical representation (Wickham, Chang, & Wickham, 2016) and ‘coxme’ to implement the mixed-effects Cox proportional hazards model (Therneau, 2015).