Predicting poaching hotspots in the largest remnant of the Atlantic Forest by combining passive acoustic monitoring and occupancy models

Abstract

Poaching can have major impacts on wild animal populations and is pervasive in tropical regions. The spatial distribution of this furtive activity is particularly difficult to estimate in large natural areas, and this hinders the development of effective anti-poaching strategies. We used passive acoustic recorders in combination with occupancy models to develop a predictive map of poaching presence in the Upper Paran´a Atlantic Forest of Argentina and Brazil. Poaching activity was measured by gunshots detected by the recorders that were active for 7 months (August 2018 to February 2019) on 90 sampling sites distributed in an area of 4637 km2. A total of 15,936 h of landscape sounds were recorded, detecting gunshots at 43 sites. Using occupancy models, we evaluated eight variables that might influence poaching occurrence and detectability. Poaching was higher in areas with higher accessibility, with a higher proportion of rural areas, and far from control posts of park rangers. The detectability of gunshots was lower during periods of heavy rainfall. We validated the occupancy models through field surveys conducted in the same period resulting in a predictive capacity of 82% of our best model. Our results show that this region is under very high poaching pressure, even within the protected area's boundaries and that urgent actions must be taken. The methods we used for estimating poaching pressure and the predictive maps developed could serve as a tool for developing and implementing anti-poaching strategies to reduce this pervasive threat.