Small selective ungulate species differ in habitat use within transformed landscapes of the Overberg, South Africa

Description

The Renosterveld vegetation in the Overberg, Western Cape (South Africa) is listed as critically endangered with more than 80% of Renosterveld transformed for agricultural purposes. Remaining natural vegetation consists mostly of small patches that are situated far apart. The landscape between these patches is filled with agricultural land used for grain crops, vineyards, indigenous flower crops and livestock farming. The remaining patches of natural vegetation is being conserved providing the wildlife in the area with corridors and small protected areas. The objective of this study was to determine how the habitat use of five small specialist browser species are affected by anthropogenic land use. These species were bushbuck (Tragelaphus sylvaticus), Cape grysbok (Rhaphicerus melanotis), common duiker (Sylvicapra grimmia), grey rhebok (Pelea capreolus) and steenbok (Rhaphicerus campestris). To determine which site-specific characteristics were the strongest drivers of habitat use during each season we ran occupancy models using camera trap data and covariate data collected around camera trap sites to run single species, single season occupancy models. The results suggest that some species appear dependent on the anthropogenic landscape, while still relying on the natural vegetation for cover and resources. In contrast, other species however appear to be more dependent on the natural landscape and occasionally use the anthropogenic landscape to their advantage.

Steps to reproduce

We established a plot with a 10 m radius around each camera trap to determine the characteristics that make up the habitat. In each plot, we determined tree/shrub height, distance to water, distance to crops and land use. The land use was classified either as “De Hoop (Protected Area, PA)”, or “Overberg (Agricultural, OAA)”. We used camera trap data in an occupancy framework to determine the probability of habitat use for each species during the two seasons (Mackenzie et al., 2002). To do this we determined which of the covariates affected the occupancy of each species at a camera trap site and how it differed between seasons. We used the detection histories of four species for each season to do a single species single season occupancy analysis for each species. Detection history: we separated the data set into two seasons. No-crop season - the time of harvest to the time of planting (November 2018 – May 2019). Crop season - the time of planting until harvest (May 2019 – October/November 2019). We filtered the camera trap data and kept one capture per species per hour per camera. Camera trap effort: using exploratory analysis using the first and last capture dates. We filtered the camera trap data to fit the “crop season” dates and the “no-crop” season dates and removed the inactive camera traps. We then ran a camera operation matrix using the ’cameraOperation‘ function from the ’camtrapR‘ package in R. Detection history: we used the function ’detectionHistory’. The occasion length was set as 15 days to avoid excessive non-detections, resulting in 13 occasions for the “crop season” and 12 occasions for the “no-crop season”. We used the ’unmarked‘ package to run the occupancy analysis. Firstly, we determined the naïve occupancy. We then created a data summary of the observations per species at each site and created a plot to illustrate it. Steenbok was removed from the results as a low detection resulted in models not converging. We evaluated each variable separately. Models were compared to the null model, using a model selection approach. We used constant occupancy as well as distance to the nearest crop, distance to the nearest water, average tree/shrub height and land use to explain occupancy. We used occupancy as a substitute for habitat use. We ran all potential models at the same time following Mackenzie et al. (2002). To determine model fitting we manually ran a Null model, sub models for detection probability (p) and sub models for occupancy (psi). We determined the estimated occupancy for the distance to the nearest crop, distance to the nearest water, average tree/shrub height and land use and we plotted the results using the ’ggplot‘ function from the ’ggplot2‘ package. 

Institutions

Categories

Funding

Licence