RESEARCH PAPER
Assessing wild canid depredation risk using a new three steps method: the case of Grosseto province (Tuscany, Italy)
 
More details
Hide details
1
Dipartimento di Biologia e Biotecnologie “Charles Darwin”, Università di Roma La Sapienza, Viale Università 32, 00185 Roma
Online publish date: 2017-03-06
Publish date: 2017-03-06
 
Hystrix It. J. Mamm. 2017;28(1):21–27
KEYWORDS
ABSTRACT
The recovery of large carnivores in human dominated landscapes can cause controversy and concern for livestock producers, especially where wild predator populations and farmland overlap. This is the case in the Grosseto province, located in the southern part of Tuscany, Italy. Anticipating where predator attacks are likely to occur can help focus mitigation efforts. We suggest a three-step method to predict wild canid depredation risk using presence only data on wild canid detections and confirmed depredation events in the study area. We obtained the probability of occurrence for canids and depredation events based on ecological variables and then performed an ensemble model following an ad-hoc procedure. We compared models’ outputs obtained from two different approaches to species distribution modeling: Maximum Entropy (Maxent) and Bayesian for Presence-only Data (BPOD) testing their ability to predict the occurrence of events. The ecological niche factor analysis (ENFA) was used to assess the importance of each environmental variable in the description of the presence points. Forested areas were identified as the most important attribute predicting wild canid occurrence. Livestock predation was most likely to occur close to farms where sheep densities were higher and more accessible. Higher depredation risk zones were characterized by proximity to forested areas and the presence of landscape features that allowed wild canids to reach pastures with minimum effort such as the network of smaller watercourses. Only 15% of the total sheep farms fall within higher risk areas, indicating that depredation was facilitated by environmental conditions (e.g. closeness to the woods) rather than the availability of prey. Overall BPOD performed better than Maxent in terms of sensitivity, suggesting that BPOD could be a promising approach to predict probability of occurrence using presence only data.
eISSN:1825-5272
ISSN:0394-1914