Hurricanes cause dramatic changes to forests by opening the canopy and depositing debris onto the forest floor. How invasive rodent populations respond to hurricanes is not well understood, but shifts in rodent abundance and foraging may result from scarce fruit and seed resources that follow hurricanes. We conducted studies in a wet tropical forest in Puerto Rico to better understand how experimental (Canopy Trimming Experiment) and natural (Hurricane Maria) hurricane effects alter populations of invasive rodents (Rattus rattus [rats] and Mus musculus [mice]) and their foraging behaviors. To monitor rodent populations, we used tracking tunnels (inked and baited cards inside tunnels enabling identification of animal visitors’ footprints) within experimental hurricane plots (arborist trimmed in 2014) and reference plots (closed canopy forest). To assess shifts in rodent foraging, we compared seed removal of two tree species (Guarea guidonia and Prestoea acuminata) between vertebrate-excluded and free-access treatments in the same experimental and reference plots, and did so 3 months before and 9 months after Hurricane Maria (2017). Trail cameras were used to identify animals responsible for seed removal. Rat incidences generated from tracking tunnel surveys indicated that rat populations were not significantly affected by experimental or natural hurricanes. Before Hurricane Maria there were no mice in the forest interior, yet mice were present in forest plots closest to the road after the hurricane, and their forest invasion coincided with increased grass cover resulting from open forest canopy. Seed removal of Guarea and Prestoea across all plots was rat dominated (75%-100% rat-removed) and was significantly less after than before Hurricane Maria. However, following Hurricane Maria, the experimental hurricane treatment plots of 2014 had 3.6 times greater seed removal by invasive rats than did the reference plots, which may have resulted from rats selecting post-hurricane forest patches with greater understory cover for foraging. Invasive rodents are resistant to hurricane disturbance in this forest. Predictions of increased hurricane frequency from expected climate change should result in forest with more frequent periods of grassy understories and mouse presence, as well as with heightened rat foraging for fruit and seed in pre-existing areas of disturbance.

Each dataset was based on observations.

1) Tracking tunnels: To assess rodent incidence, which is a reflection of abundance and activity, we used tracking tunnels in each plot. Tracking tunnels are inked and baited cards placed in tunnels enabling foot prints of vertebrate visitors to be identified to genus (Shiels and Ramírez de Arellano 2018, Madden et al. 2019, Shiels and Ramírez de Arellano 2019). In each of the 12 CTE plots, two tracking tunnels that were 60 cm long with 10 × 10 cm openings were randomly placed approximately 10 m apart from each other; one tunnel was placed on the ground, and the other tunnel was secured by plastic straps on a branch or liana in the lower canopy at an average height of 135 cm (SE = 5 cm). The rodent tracks (rat or mouse) were identified on the tracking cards after 2 days. These are presence/absence data measured at each tunnel.

2) Seed removal trials: Trials were conducted to determine whether rats removed native seeds from the forest floor. Two common tree species, muskwood (Guarea guidonia, Meliaceae) and sierra palm (Prestoea acuminata, Arecaceae), were chosen based on their fruit and seed availability during the trial period (June-July 2017 and 2018). The seeds (6 per station for Guarea, 3 per station for Prestoea) were placed on the ground without any caging surrounding them and then they were visited to count the number of seeds in each treatment over time for 7 days, in the Summers of 2017 and 2018.

3) Canopy openness: Canopy openness was determined 9 months before and 9 months after Hurricane Maria by taking hemispherical pictures using a digital camera (Nikon 4500, Nikon Inc, Tokyo, Japan) with a fish-eye lens (Nikon FC-E8) positioned at 1 m height above ground using a tripod, at five locations (i.e., the four corners and the center) in each 20×20 m measurement area of the plot. The photos were taken on the following dates: 7-16 December 2016 (pre-hurricane), and 1-3 July 2018 (post-hurricane). Each picture was analyzed for percent canopy openness using Gap Light Analyzer software (version 2; Simon Fraser University, Cary Institute of Ecosystem Studies), setting thresholds at 82 for all 2016 pictures and 152 for all 2018 pictures; the higher threshold was needed in 2018 to prevent the greater light levels from obscuring the remaining vegetation. Percent canopy openness was based on the average of the five pictures per plot.

4) Percent grass cover: Grass cover (percent ground coverage < 1 m height) was visually estimated to the nearest 1% in September or early October of each year since the onset of the experiment within five permanent 1 × 3.5 m subplots within each of the 12 plots (see Shiels et al. 2010). We obtained an annual average of the five subplots per plot to use in our analysis (n = 3 for canopy trimmed plots; n = 9 for reference plots) for 2012 to 2019.