New study shows predictive accuracy of risk map for wolf threats to livestock

June 30, 2017

A new study published Friday (June 30, 2017) in the journal PLOS One shows that a risk map for wolf predation on livestock remained predictive for five years or more after its construction, with predictive accuracy exceeding 91 percent.

Risk maps serve as predictive tools to identify high-risk and low-risk locations for various environmental hazards, such as predation on livestock. The spatial models distinguish sites by the probabilities that an environmental hazard will occur there. The findings of this study carry potential implications for preventive action, suggesting that wildlife management interventions be refocused to high-risk areas. 

“Predicting where livestock may be at risk from predators can promote preventive methods by owners and managers,” explains Adrian Treves, an associate professor of environmental studies at UW-Madison who led the study. 

Treves founded and directs the Carnivore Coexistence Lab in the Nelson Institute for Environmental Studies at UW, examining patterns of conflict with wildlife predators where crop and livestock production overlap carnivore habitat, and human responses to these conflicts. Having authored more than 100 scientific papers on predator-prey ecology and conservation, he has previously studied the effectiveness of predator control methods across North America and Europe, changes in attitudes toward wolves before and after an inaugural public hunting and trapping season in Wisconsin, and how to balance human needs with predator conservation

In this latest study, Treves and Mark Rabenhorst, an affiliate of the Carnivore Coexistence Lab who works in the private sector, tested the long-term validity of a published risk map built from 133 locations in Wisconsin where gray wolves attacked livestock from 1999-2006.

Risk map for wolf predation on livestock in Wisconsin
A: The analysis area (black) showing published risk probabilities in six categories [3]. B: Wisconsin and the analysis area (green) with risk colors dichotomized into high-risk (red), lower-risk (orange), and very low risk (green). Symbols depict locations of verified incidents of wolf attack on livestock (magenta triangles), perceived threats to livestock (yellow squares), and sites where the Wisconsin DNR killed one or more wolves to prevent livestock loss (blue circles). Note that symbols are larger than the pixels colored by risk category, so they may obscure underlying risk levels. Source: PLOS One

Using data collected after the model’s construction related to verified additional incidents of wolf attacks on livestock statewide from 2007–2011, the researchers verified that the risk map continued to accurately predict locations that might risk wolf attacks on livestock, as late as 2011.

Specifically, the risk map correctly predicted that 91 percent of the new sites of wolf attacks would face some risk of an incident. And the risk model situated almost half of all incidents in the highest-risk category of locations, thereby focusing attention on less than 7 percent of the statewide analysis area. 

“Predictive power lasting five years or more substantiates the claim that risk maps are both valid and verified tools for anticipating spatial hazards,” the authors write in PLOS One.

Given the predictive power of risk maps, the researchers recommend that such models be used to improve efficiency and selectively target interventions — such as the killing of wolves to prevent livestock loss in areas demonstrated to be high risk.