MATERIALS AND METHODS
Study site — The study was conducted in two adjacent areas with distinct protection categories (Figure 1). The Rio Doce State Park (RD) is an IUCN protected area category II (National Park) and the largest remnant of Atlantic Forest of State of Minas Gerais, southeastern Brazil, with ca. 36,000 ha of stational semi-decidual forest (SOS Mata Atlântica 2019). The RD has a large lacustrine system composed of 42 natural lakes and it is limited in the eastern part by the Rio Doce river. In 2010 the RD importance was internationally recognized when it became a Ramsar site by UNESCOs Convention on Wetlands. Despite being one of the largest remnants of Atlantic Forest, RD is isolated from other forest remnants and is surrounded by several types of human-related habitats, especially a mosaic formed by eucalyptus plantations, but also native forest fragments.
The second study area (EP) is a private property of ca. 23,000 ha composed mainly by eucalyptus plantations, but also by fragments of Atlantic Forest and natural lakes. This area is an IUCN protected area category VI (Protected area with sustainable use of natural resources) and during the evaluation time, eucalyptus management and logging were regularly done. Hunting is prohibited either in RD or in EP, but fishing is allowed in EP lakes, where fishermen and illegal hunters are common. The EP area is located at the southern border of the RD buffer zone, and it is commonly frequented by mammals. The region has a tropical climate and during this study had an annual precipitation of 1,035 mm and the average temperature was 25ºC (INMET 2019).
Data collection — Sample period was from April 2014 to January 2015, where camera traps (Bushnell®) where deployed on man-made trails and game trail in RD and EP, totalizing 30 camera sites (15 on each area). A minimum distance between camera sites was 1km to minimize a lack of independence between sampling sites. Camera traps were deployed 30cm above ground level, allowing the detection of medium- to large-sized mammals, and operated 24-h.
Activity patterns analysis — We defined an activity sample as all photographs of the same species detected at a camera site within an 1-h period, thus minimizing the nonindependence of consecutive photographs. The hour of each activity sample recorded by the camera traps was transformed into a solar time based on sunrise and sunset times of our study area. This is important to accurately define the activity pattern of the species and also to allow comparisons with other studies (Foster et al . 2013). Sunrise and sunset times were obtained from the software Moonrise v.3.5 (Romero-Muños et al . 2010; Foster et al . 2013), and we used the following formula described by Woolf (1968) for solar conversion:
\begin{equation} \mathrm{LCT\ =\ }\mathrm{t}_{\mathrm{s}}\ \mathrm{-\ }\frac{\mathrm{\text{EOT}}}{\mathrm{60}}\mathrm{\ +\ LC\ +\ D}\nonumber \\ \end{equation}
Where LCT is the standard clock time, ts the solar time, EOT the equation of time, LC the longitudinal correction, and D the daylight saving time (see Woolf (1968) for further details).
Then, we used the Rao’s spacing test (Rao 1976) to verify whether the species activity pattern was uniformly distributed (i.e., cathemeral) or associated with a specific time period (i.e., diurnal, nocturnal or crepuscular). We categorized the activity pattern of each species into diurnal (>60% of records between 1h after the sunrise and 1h before the sunset), nocturnal (>60% of records between 1h after the sunset and 1h before the sunrise), crepuscular (>50% of records occurring 1h before and after sunrise and sunset) and cathemeral (peaks of activity through the diurnal and nocturnal period). To compare the 24-h cycles of each species between RD and EP we used the Mardia-Watson-Wheeler test (W ). When theW test revealed no significant differences (P> 0.05) in the 24-h cycles of a given species between the studied areas, we combined species data from both areas for the subsequent analyses. The analyses were performed using the package “circular” v.0.4-93 (Lund and Agostinelli 2017) in R Software v.3.6.3 (R Development Core Team 2019).
Activity overlap analysis — To evaluate the temporal activity overlap between predators, as well as between predators and their potential prey, we calculated the coefficient of overlap (Δ; Ridout and Linkie 2009) that varies from 0 (no overlap) to 1 (complete overlap). We used the Δ1 estimator when the number of independent records of at least one species in the pairwise comparisons was <75 photographs. Otherwise, we used the Δ4estimator. We calculated the 95% confidence intervals for\(\hat{\Delta}\) from 10,000 bootstrap samples (Ridout and Linkie, 2009). To complement the coefficient of overlap, we compared the 24-h cycles between species using the W test. To calculate the coefficient of overlap and the W test statistics we used the package “overlap” v.0.3.2 (Linkie and Ridout 2011) and the package “circular” respectively, both available in the R Software.
Potential preys were based on studies of feeding habits for each predator species (Appendix 1). We considered as potential preys only those preys found at least once in any study. Rarely, some prey species that are much larger than the predator were described as a diet item, but as it was related to a scavenging behavior, we did not consider it directly as a potential prey. We did not find in any study that the giant-armadillo (Priodontes maximus ) could be a prey item for jaguars, but because we believe that this predator can prey upon it, we included the giant-armadillo as potential prey for jaguars.
We used the study of Oliveira and Pereira (2014) to either verify the relationships of dominance and subordination among predators or the possibilities of IGP/IK among them. The analysis of temporal activity overlap was performed only if either IGP or IK was noticed between the given predators in this study. Also, according to this study, jaguars are the top predators with no natural predators. The puma has the jaguar as a potential predator, and ocelots have jaguars and pumas as potential predators. These three felids are potential predators for crab-eating foxes, tayras, and coatis, and there were no records of IGP or IK between these latter species.