A biodiversity hotspot losing its top predator: The challenge of jaguar conservation in the Atlantic Forest of South America

The jaguar is the top predator of the Atlantic Forest (AF), which is a highly threatened biodiversity hotspot that occurs in Brazil, Paraguay and Argentina. By combining data sets from 14 research groups across the region, we determine the population status of the jaguar and propose a spatial prioritization for conservation actions. About 85% of the jaguar’s habitat in the AF has been lost and only 7% remains in good condition. Jaguars persist in around 2.8% of the region, and live in very low densities in most of the areas. The population of jaguars in the AF is probably lower than 300 individuals scattered in small sub-populations. We identified seven Jaguar Conservation Units (JCUs) and seven potential JCUs, and only three of these areas may have ≥50 individuals. A connectivity analysis shows that most of the JCUs are isolated. Habitat loss and fragmentation were the major causes for jaguar decline, but human induced mortality is the main threat for the remaining population. We classified areas according to their contribution to jaguar conservation and we recommend management actions for each of them. The methodology in this study could be used for conservation planning of other carnivore species.

Abbreviations: (     Methods S1. Information about the sources and type of data of jaguar presence used to develop and test the habitat suitability model. The sources of data of jaguar presence that we used to develop the habitat suitability model analysis and to estimate the area of jaguar occupancy in our study came from different sources. Some of the data were collected during a participatory initiative to update the After the conclusion of the jaguar habitat suitability analysis, we collected additional data from the field (n= 107) to use as an independent dataset to further test the model's predictive ability for jaguar occurrence in the AF. These recent data correspond to camera traps pictures, tracks and feces obtained in systematic and unsystematic surveys developed in different regions of the Atlantic Forest.
Methods S2. Description of the methodology used to estimate jaguar density with spatially explicit capture-recapture (SECR) models.
Jaguar density was estimated using spatially explicit capture-recapture models (Efford, 2004;Royle et al., 2009). The assumptions of these models are that 1) the study animals have circular home ranges that are constant during the survey, 2) the activity centers are randomly distributed within the study area, 3) the detection probability decrease with distance from the home range center following a detection function, and 4) the population is closed during the survey (Efford, 2004;Royle and Young, 2008).
We defined a priori the extent of the surveyed area as the polygon that includes all the camera traps plus an external buffer of 25 km (see Gopalaswamy et al., 2012). Inside the defined surveyed area we located the potential activity centers uniformly distributed at a distance of 1 km among them. The activity centers that overlay with the area of jaguar occupancy (AJO) were considered as the potential home ranges for the analysis. To run SPACECAP we selected the following model definitions: trap response absent, spatial capture-recapture, half normal detection function, and Bernoulli´s encounter model. We set the MCMC simulations to run 100,000 iterations, discard (burn in) the first 10,000 iterations, assume a thinning value of five and a data augmentation value equal to eight times the number of jaguar individuals recorded in every survey. After every analysis we checked the values of the Geweke test, sample size, model fit, density plot for Psi and N, and the detection function plot to confirm that the results were adequate. If some of these parameters indicated that something was wrong, we ran the models again changing the MCMC simulation parameters until obtaining a model with adequate parameters (for details about all SPACECAP procedures see Gopalaswamy et al., 2012).
SECR models assume a closed population, and to reduce the risk of violation of this assumption, we restricted the density estimation analysis of every survey to a period of between 96 and 120 days, except for the Morro do Diabo and Mbaracayu surveys. For these two surveys because they had a low number of camera traps, we moved the camera traps through different locations during the survey periods to evenly cover these protected areas (Karanth & Nichols, 2002)