Skip to main content
Dryad logo

Detection histories of mesocarnivores in agricultural areas of Southern Chile

Citation

Gálvez, Nicolás et al. (2021), Detection histories of mesocarnivores in agricultural areas of Southern Chile, Dryad, Dataset, https://doi.org/10.5061/dryad.5qfttdz3h

Abstract

We obtained mesocarnivore detection/non-detection data from 180 sampling units (4 km2 each) located in the agricultural landscapes of southern Chile from January-April of 2019. We used single-species occupancy models to investigate the associations of forest fragmentation, forest loss, and private land ownership subdivision (as a measure of human use intensification) with the occurrence of four mesocarnivores (güiña, grey fox, culpeo fox, and Molina’s hog-nosed skunk), and extended this framework to two-species occupancy models to assess patterns of mesocarnivore co-occurrence with domestic carnivores. We also assessed whether co-occurrence of native and domestic carnivores led to shifts in species’ temporal activity.

We hypothesized that the occurrence of native mesocarnivores was largely mediated by human use intensification variables only, the occurrence of domestic mesocarnivores only, or a combination of both. Our results largely supported the human use intensification hypothesis, with some influence of domestic mesocarnivores. Mesocarnivore occurrence shifted from a native to a domestic species composition as private land ownership subdivision increased, and native mesocarnivores shifted their behaviour temporally when co-occurring with domestics. In addition, the presence of domestic dogs was associated with an absence of native mesocarnivores, possibly driving a defaunation process in agricultural areas.

Methods

We obtained mesocarnivore detection/non-detection data via a camera trap survey from January to April of 2019 (i.e. summer months). Potential sampling units (hereafter SUs) were defined via a grid of 4 km2 cells across the study region. The size of the SUs was informed by the mean estimated home range size of güiña from collared individuals in the study area, a size comparable to home ranges from the other native mesocarnivores under study (Schüttler et al., 2017; Johnson & Franklin, 1994; Salvatori et al., 1999; Kasper, Soares, & Freitas, 2012). Detectability was modelled based on the assumption that a 5-day survey block is a separate independent sampling occasion, a survey period used previously for mesocarnivores in southern Chile (Moreira-Arce et al., 2016).

Minimum survey effort requirements (i.e. number of SUs and sampling occasions) were determined following Guillera-Arroita, Ridout, and Morgan (2010), using species-specific parameter values from Gálvez et al. (2013) and a target statistical precision in occupancy estimation of SE < 0.075. Our target was a minimum of 120 SUs and 10 sampling occasions. A total of 180 SUs were selected at random from the grid of 2,243 cells (Fig. 2). Each camera was active 48.6 days on average (SD = 7.77), with 4 to 14 sampling occasions at each SU (average = 10.14 occasions, SD = 1.55).

The extent of forest loss and fragmentation were evaluated using ecologically meaningful metrics reported in the literature as being relevant to mesocarnivores and our hypotheses (see below), using remotely‐sensed landcover data (Table 1). The metrics were measured at two scales: landscape-scale (400 ha circular buffer around each camera) and fine-scale (10 ha buffer) in order to examine the landscape-level and core home range-level effects, respectively, of habitat loss and fragmentation on species’ occupancy. FRAGSTATS 4.1 was used to create these buffers (McGarigal, Cushman, Neel, & Ene, 2002). These core area variables were also used to model detection (Guillera-Arroita, 2017). Importantly, land subdivision was assessed as the number of land properties within each 400-ha or 10-ha buffer.

Depending on the species, we used 5 to 7 covariates for modelling detection probability, including quadratic effects on continuous covariates that could have a non-monotonic effect on species detection (e.g. forest cover; Table 1). See Supporting Information Appendix S1 for methods used to confront the multicollinearity of our covariates.

Funding

FONDECYT-Ministry of Science of Chile, Award: 11170850

FONDECYT-Ministry of Science of Chile, Award: 11170850