MATERIAL AND METHODS
Study Area
We worked in two landscapes of Colombia, South America, as part of an
ongoing biodiversity conservation and monitoring initiative. The first
one (MML) is located within the Middle Magdalena River valley, and has
temperatures averaging between 23 °C to 29 °C and an average
precipitation of 2,900 mm/year. It overlaps the municipalities of Yondó
(Antioquia), Puerto Parra and Cimitarra (Santander), and extends over an
area of 253,600 has, from which 57% correspond to agriculture land use.
This landscape presented a total forest loss of 29, 631 ha between years
200 and 2014 according to data from Hansen et al., (2013). The second
landscape (OL), is situated within the Bita River basin, in the
municipality of Puerto Carreño (Vichada), in the Orinoquia region, and
has average temperatures between 23 °C and 36 °C and an average
precipitation range of 2,000 – 3,000 mm/year. It spans an area of
407,400 ha, from which 64% are natural savannas (grasslands), 15%
natural forests and 2.3% wetlands; while only 1% are agriculture land
uses, mainly pastures for livestock. This landscape lost around 74 ha of
forest between 2000-2014 according to data from Hansen et al., (2013).
Sampling
We divided the landscape into a 1x1km grid, and located sampling
stations within 11-27% of the cells, trying to cover as much habitat
heterogeneity as possible while constrained by access and permits to
enter private properties. Thus, the spatial sampling unit was a 1x1km
cell, and each sampling station consisted of a camera trap with movement
and infrared sensor (Reconix HC 500 Hiperfire) for detecting terrestrial
birds and medium-size and large mammals. We installed 68 sampling
stations in the MML between January-May of 2017 and 76 stations between
January-April in the OL, both periods corresponding to the dry season in
each landscape. Additionally, we conducted three independent 800 meter
transects within sampled cells to look for signs and prints of ungulates
and primates. Total sampling effort was 4,569 and 3,763 nights-camera in
the MML and OL respectively. Cameras were active an average of 67 nights
in the MML and 49 nights in the OL.
Covariables
One of our hypotheses is that factors that determine occupancy can vary
across landscapes. We assessed this by evaluating the effect of multiple
covariates on occupancy of studied species in each landscape. We
included covariates from two categories, those corresponding to natural
factors and those related directly or indirectly to anthropic factors.
The first category included the area of total forest, the area of each
type of forest according to a Corine Landcover classification (riparian
forest and high-ground forest in MML and flooded forest in OL), area of
natural grasslands (savanna), area of wetlands, and distance to water
bodies. These areas were estimated for each 1 x1 Km cell and in a
surrounding 1km buffer to explore different spatial scales of the
effect. The second group of covariates included distance to closest
human settlement, distance to roads, area of pastures (artificial land
cover), and the area of open pastures and secondary vegetation. We also
used land cover types to evaluate covariates affecting the probability
of detection for each species. All variables were estimated using ArcGIS
(ESRI 10.2.1).