The effect of competitor presence on the foraging decisions of small mammals
Data files
Jun 25, 2024 version files 887.63 KB
-
FE_totalweights.csv
144.48 KB
-
Photo_Behaviours2.csv
732.98 KB
-
README.md
7.64 KB
-
sitevalues2.csv
2.53 KB
Abstract
Competitive interactions between species can have marked effects on the diets and foraging behaviours of the interactants. Dominant competitors may constrain the foraging decisions of subordinate competitors, reducing the individual fitness of subordinates, and potentially driving their populations to low levels. Following a sustained population decline of the bush rat (Rattus fuscipes) in the presence of the competitively dominant common brushtail possum (Trichosurus vulpecula) at Booderee National Park in south-eastern Australia, we investigated whether possums affected the foraging decisions of bush rats. Using a modified giving-up density experiment, we predicted that bush rats would: (a) increase visits to baited sites where possums had restricted access, and (b) restrict visits to baited sites where possums had free access. We used camera traps to investigate visitation patterns and foraging bout lengths at 40 baited sites with two treatments, one that allowed full access by both species (full access), and the other that attempted to prevent possum access (restricted access). We also measured additional covariate factors that may influence visitation. Bush rats visited both treatments less when there were more possum visits. We also found that bush rats spent less time eating bait at regularly visited sites, regardless of possums’ access level. Our results suggest a negative, potentially competitive interaction between the two species that is detrimental to bush rat foraging and is a potential factor contributing to bush rat's decline at Booderee National Park.
Description of the data and file structure
First Dataset: PhotoBehaviours2.csv
The data collected from camera trap images included the site visited, night visited, length of visit, species visiting, and description of behaviour observed during the visit.
The information is organised by site, with each line representing a visit event.
- The column named ‘bin type’ informs the access condition of the site.
- The column named ‘night’ informs which night of the experiment the visit occurred on.
- The columns named ‘Start Date’ and ‘End Date’ inform the dates on which the visit occurred.
- The columns named ‘Start Photo’ and ‘End Photo’ inform the first and last photo over which a visit took place, with the column named ‘No. of Photos’ describing the number of photos over which a visit took place.
- The column named ‘Start Time’ informs the time a visit started and the column ‘End Time’ informs the time a visit ended.
- The column ‘start’ is the combined ‘Start Date’ and ‘Start Time’, and the column ‘end’ is the combined ‘End Date’ and ‘End Time’.
- The column named ‘Species’ informs which species was identified visiting the site.
- The column named ‘Description’ is a description of the observed activities the species were performing during the visit.
- The column named ‘Activity’ is where the description has been categorised into one of three categories ((1) within site, (2) interacting with bin, (3) eating bait).
- The columns named ‘Last rat’ and ‘Last possum’ identify the time when a rat or possum was last observed at the site.
- ‘Time Since Rat’ and ‘Time Since Possum’ are the calculated time difference (in seconds) between the last rat or possum visit and this visit. ‘NA’ values refer to visitations that occur at the site before the other species were recorded at that site.
Second Dataset: FE_totalweights.csv
The data collected each day of the experiment.
The information is sorted by site, however each line represents a night of the experiment.
- The column named ‘Site’ is the site identifying code.
- The column named ‘Night’ informs the night of the experiment.
- The column named ‘Date’ informs what date the night was on.
- The columns ‘Weight’, ‘Change’, and ‘New’ described the weights of bait, the change from the previous night, and the amount added (if any) in grams.
- The columns named ‘Disturbance’ and ‘Spill’ describe the percentage of bait taken or spilled, based on visual inspection.
- The columns named ‘Photos’ and ‘Photo Change’ describe the number of photos taken overnight.
- The column named ‘Ground’ describes the observed condition of the ground for that night, which proxies for an estimate of rainfall.
- The column named ‘Notes’ includes any observed information about that day’s data collection.
- All of the above information was collected every other day by visiting each site. Cells with ‘na’ identify days when the site was set up, and therefore no data on weight measurements was taken. Cells with ‘nc’ identify that the site was not visited on that date, and therefore no information was collected.
- The next five columns include data collected from the Bureau of Meteorology.
- ‘Daily Rainfall’ was recorded in mililitres, and ‘Max Temperature’ and ‘Min Temperature’ were collected in Celsius.
- ‘Sunrise’ and ‘Sunset’ inform what time, in 24 hours, the sun rose and set.
- The final eight columns include information about moon phase and moonlight, collected from Tides4Fishing.com.au
- The column named ‘Moon Phase’ describes the categorical moon phase (Waxing Gibbous, Waning Gibbous, Waxing Cresent, Waning Cresent, New Moon, First Quater, Last Quater, Full Moon).
- The columns named ‘Moonrise (time)’ and ‘Moonset (time)’ describe the time the moon rose and set, in 24 hour time. Cells with ‘na’ refer to data points that were not available for that day on Tides4Fishing.com
- The columns named ‘Moonrise (degrees)’ and ‘Moonset (degrees)’ describe the position in the sky of the moon when it rose and set.
- The column named ‘Distance’ describes the distance of the moon from the earth in km.
- The column named ‘Visible moon’ describes the length of time the moon is visible.
- The column named ‘Illumination’ describes the amount of light from the moon in percentage.
Third Dataset: sitevalues2.csv
The data collected on the site-level variables.
The data was extracted from rasters and shape files, which were downloaded from GROCLIM (site productivity) (Xu and Hutchinson, 2011).
- The information is sorted by site, with each line representing a site. This is identified in the first column named ‘Site.no.’.
- The column ‘Bin.Type’ informs the access condition of the site.
- The column ‘Broad.Veg.Type’ informs the vegetation classification.
- The columns ‘Long’ and ‘Lat’ inform the longitude and latitude of each site using 0-metre easting and northing coordinates.
- The column ‘shape_values.FIRE_DATE’ includes data extracted from shape files regarding the date of the last fire from each site, sites with ‘NA’ refer to sites that were missing fire history information.
- The final column includes extracted data on the Topographic wetness index (TWI), which was extracted from a raster file, this falls under the column named ‘twi_twi’.
Sharing/Access information
Links to other publicly accessible locations of the data:
Bureau of Meteorology. (2022). Jervis Bay, NSW - March 2022—Daily Weather Observations. http://www.bom.gov.au/climate/dwo/202303/html/IDCJDW2067.202303.shtml
Tides4Fishing. (2023). Lunar calendar 2023 of lunar phases and eclipses. https://tides4fishing.com/lunar-calendar
Data was derived from the GROCLIM source.
Xu, T., Hutchinson, M., 2011. ANUCLIM VERSION 6.1 User’s Guide.
Code/Software
All code was from R. The package used to create and run the models was the glmmTMB package ver. 1.0.1. Model selection was done using the MuMIn package ver. 1.43.17. The exact R code has been included in a separate text file.
Models
There were 5 models created.
- The first modelled the response variable RF_count, which is the number of bush rat visits per night, and included a zero-inflation step.
- The second modelled the response variable RF_time, which is the time length of bush rat visits.
- The third modelled the response variable TV_count, which is the number of possum visits per night, and included a zero-inflation step.
- The fourth modelled the response variable TV_time, which is the time length of possum visits.
- The fifth modelled the response variable RF_eat_time, which is the time length of bush rat visits when they are eating bait.
Covariates
The covariates for site level variation were access type, broad vegetation type, no. of years since fire, and TWI. The covariates for night-level variation were illumination and daily rainfall.
Additional variables
The variables ‘Time Since Rat’ and ‘Time Since Possum’ were included as measures since the last visit by a rat for possum models, and the last visit by a possum for rat models. The site was included as a random effect.
Visualising the results
We first visualised the results using forest plots, created using the ‘plot_model()’ function from the sjPlot package. We then model the conditional effects of the chosen plots for relevant variables using the ‘plot_model()’ function from the sjPlot package. All visualisation was done using the ggplot package.
Ethics Statement
This study was conducted in strict accordance with the recommendations in the Australian Code for the Care and Use of Animals for Scientific Purposes. The protocol was approved by the Animal Experimentation Ethics Committee at the Australian National University (Protocol Number: A2021_52).
Study Location
Our experiment was conducted in March and April 2022 in Booderee National Park (BNP), a 6,000 ha protected area on the south coast of New South Wales, Australia (Fig 1). The Park is owned by the Wreck Bay Indigenous Community and is jointly managed by them and Parks Australia. The Park has a temperate climate, with an average annual rainfall of 1,213 mm (Bureau of Meteorology 2022). However, due to the La Niña weather system occurring at the time of data collection, the Park experienced higher-than-average rainfall, including partial flooding, which occurred in the weeks before the experiment. The temperatures at the time of the experiment ranged from 13.5° to 26.6°C (Bureau of Meteorology 2022). The Park has a heterogeneous environment, with vegetation types ranging from forests and woodlands to sedgelands and heathlands (Taws 1998). The Park has experienced several wildfires in the last decades, with the most recent large fire occurring in 2017 (Lindenmayer et al. 2023).
Experimental Setup
We established our experiment at 40 sites across BNP, selecting these by referring to trapping records from the immediately previous five years (Lindenmayer et al. 2008; Lindenmayer et al. 2016). The 40 sites were established in places selected from a long-term monitoring program established in 2002, and had annual mammal trapping records, with trapping occurring at each site every other year (Lindenmayer et al. 2008; Kanishka et al. 2023). The criteria for site selection were records of both bush rats and common brushtail possums being trapped at the same site within the last five years, with preference given to the most recent trapping sessions or sites where both species had been trapped most regularly. We selected only those sites within woodland and forest vegetation types.
We placed 60-litre black garbage bins upside down at 40 sites, with an entrance at the base that was modified to create two site conditions: full-access (entrance: 10 x 10 cm), where both species could easily gain entry, and restricted-access (entrance: 5 x 5 cm), where only bush rats could gain entry. To attract animals to enter the bins, we used 100 g of rodent pellets as bait, placed on a ceramic dish in the centre of the bin. We placed remote cameras facing the entrance to each bin on the bottom of a star picket (1 – 2 m above the ground depending on the slope of the ground) 2 metres away from the bin.
Cameras recorded animal activity for 28 days and we collected data on the amount of bait taken at each site every other day (depending on weather conditions). To confirm visitation to our sites, we measured the amount of bait taken by both weight and visual assessment. We replenished the baits at least once a week, or if more than 20 g was taken, or if there was evidence of the bait going mouldy.
Camera Data Collection
We collected data on bin visitation from the remote cameras. We used two brands of cameras: Boly ScoutGuard Trail Cameras (Boly Media Communications Inc., California, USA) and Bushnell Core DS No Glow Trail Cameras (Bushnell Outdoor Products, Kansas, USA). We set the cameras to take photos only at night when both species were active and to take three photos in succession upon detecting movement, with a minimum of a 10-second gap between sets of photos.
We recorded information for periods when an animal was visible on camera, which we refer to as ‘visits’. A visit began when the animal was visible on camera and ended when they were seen exiting the site or there were more than five minutes between photos. During visits, we recorded the species identity, time of arrival and exits, and a brief description of the activities the animal was performing during this time. We categorised this description into one of three broad activities: the species was (1) within the site (i.e., visible on camera, but not interacting with the bait or bin), (2) interacting with the bin (i.e., sniffing/touching it, trying to move it, climbing on top, or entering/exiting the bin), or (3) eating the bait.
To quantify factors in addition to species presence that could affect foraging by bush rats, we collected information on topographic wetness, years since fire, and the broad vegetation type at each site, as well as estimated illumination from moonlight and rainfall each night. A topographic wetness index (TWI) was calculated across the park for all sites from raster grids using GROCLIM (site productivity) (Xu and Hutchinson 2011) and extracted using R (R Core Team 2021). We categorised the broad vegetation type based on semi-annual vegetation surveys (Macgregor et al. 2020), and calculated the number of years since the last fire at each site based on historical and on-ground records. We estimated illumination from moonlight, based on moon phase, from fishing/tide records for the coast of BNP (Tides4Fishing 2023), and extracted the daily rainfall data from the nearby Point Perpendicular weather station (Bureau of Meteorology 2022).
Statistical Analysis
To examine the effect of common brushtail possums on bush rat foraging activity, we constructed generalised linear mixed models (GLMMs) using the glmmTMB package ver. 1.0.1 (Brooks et al. 2017) in R (R Core Team 2021). We used five different response variables (number of bush rat and possum visits within a night, length of time of bush rat and possum visits, and length of visits when bush rats ate bait) (Table 1). We included TWI, years since fire, broad vegetation type, illumination, and rainfall as covariates and included sites as a random intercept effect to account for the correlation between repeated measures at the same sites. There was no animal activity at some sites on many nights, yielding zero-count data that could bias our models (Welsh et al. 1996). Therefore, we used zero-inflated models, which address excess zeros by calculating a probability of absence (Welsh et al. 1996) when modelling the number of visits. We also used the response variables as explanatory variables in the other models, to test associations between foraging activities between species. The access condition for the sites (either full- or restricted-access) was also used as an explanatory variable. To compare model effects, we scaled all continuous variables to have a mean of zero and a standard deviation of one.
To select the most important variables for each model, we conducted Akaike’s Information Criterion for small sample sizes (AICc) model selection (Burnham and Anderson 2002) on all subsets of the five models described in Table 1 using the dredge function in the MuMIn package ver. 1.43.17 (Bartoń 2023). We chose the simplest model within two ∆AICc scores of the top-ranked model (Burnham and Anderson 2002; Bartoń 2023).
Question One: Differences in visitation of the two species at restricted-access and full-access sites
We evaluated differences in the number and length of visits of bush rats and common brushtail possums between the two site conditions (full- and restricted-access to food). We used zero inflated negative binomial error distribution for the zero-inflated models (number of visits, models 1 and 3), and a gaussian error distribution for the other models (time length of visits, models 2 and 4). To do this, we looked at the response of both species in the first four models (number of visits and time length of visits for each species as response variables, models 1-4, Table 1) to the access condition and the covariates (TWI, broad vegetation type, years since fire, illumination, rainfall).
Question Two: The effect of possums on bush rat visitation
We evaluated the response of bush rats from two of the models (bush rat number of visits and time length of visits as response variables, models 1-2) to the number and time length of visits by possums between the two site conditions. For this, we used the models with bush rat numbers and bush rat visit time lengths as the response variables. We included two interactions: between the number of possum visits and site-access condition, and between the lengths of time of possum visits and site-access condition.
Question Three: The effect of possums on bush rats eating bait
We evaluated the time bush rats spend at sites when their main activity was eating bait, and how this varied in response to possum visitation. To do this, we constructed a GLMM where the response variable was the length of time of bush rat visits when eating bait (model 5, Table 1). We used a Gaussian error distribution for this model. The explanatory variables were the number and time lengths of possum visits, with an interaction with the site-access condition for both variables.
Question Four: The effect of time since visitation
We evaluated the visitation of bush rats and possums based on the time since the last visit by the other species. To do so, we used the four models used in the first question but focused on a new explanatory variable (models 1-4, Table 1). The variable was the time since the last visit by a possum for the bush rat models and the time since the last visit by a bush rat for the possum models. We included an interaction with the access condition.