Disentangling risk in a multi-predator landscape : roe deer respond to differing patterns of risk to lynx, wolves and humans through shifts in their habitat selection

Predation risk is known to evoke behavioural responses in prey animals, and prey are often faced with a trade-off between lowering their risk to predation and acquiring resources. This situation becomes more complex in a multi-predator landscape, especially if those predators employ different hunting strategies, and induce different spatial patterns of risk. In this study, the spatial patterns of predation risk that roe deer face from humans, as well as their natural predators, lynx and wolves, were identified. Using the natural experiment provided by the return of large predators and the coinciding decline in hunting mortality, the behavioural responses of roe deer to shifting patterns of predation risk were examined. Using predation and hunting mortality locations, combined with used locations from the same 149 roe deer, predation risk to each predator was related to habitat and infrastructure attributes. Mostly in line with predictions, agricultural lands were found to present the highest risk to human hunting, while old forest provided a safe habitat from lynx predation, and no strong pattern was found for wolf predation. Habitat selection in relation to the most important risk factors was then compared between the period before and after lynx recolonized the area, using location data from 231 roe deer individuals. All analyses were conducted at the within home-range scale. I found that in general, agricultural lands were selected for, although less intensively during the hunting season both before and after lynx recolonisation. Moreover, there was a tendency for weaker differences between hunting and non-hunting seasons in the period after the return of lynx, potentially reflecting the lower hunting intensity in that period, and the added risk to lynx in that habitat throughout the year. An increased use of old forest was found after the recolonisation of lynx, potentially due to the relative safety from lynx attack that this habitat provides. Hence, this study demonstrates that different predators can generate different spatial patterns of predation risk, and roe deer seem to respond to these differential patterns through spatial and temporal habitat selection shifts.POPULAR SCIENCE SUMMARY Many of us have heard about what happened in Yellowstone national park in the USA when wolves were reintroduced there after over 50 years of absence. Their main prey, elk started spending less time in confined stream and river valleys where they were at most risk to wolf attack. Supposedly as a result of this, tree species in those areas started to bounce back from heavy browsing, and wolves were accredited with the recovery of these river-side ecosystems. The story has become a classic example of how predators can control natural ecosystems both directly and indirectly through “trophic cascades”, although recent studies have questioned whether wolves are solely responsible. In south-central Sweden, wolves had been absent for over 100 years, due to persecution by humans, but have naturally recolonized the area during the last decade. Likewise, Lynx had been absent for around 30 years, but started to recolonize the southern half of Sweden during the mid-1990s. This situation posed the question of whether the return of these top predators could lead to an ecological cascade similar to Yellowstone’s. At the Grimsö Wildlife Research Area, Roe deer have been monitored using VHF and GPS collars since the 1970s. This provided the opportunity to study the changing patterns of predation risk that these deer have been exposed to with the return of their natural predators. By comparing the death site characteristics of roe deer with locations that they had used during their life-times, I was able to identify which habitats presented the most risk from each predator, as well as from human hunting. I found that older forest stands (> 60 years old) provided a relatively safe habitat to roe deer from lynx predation, while wolf predation didn’t show any strong patterns, and as expected, agricultural lands presented the most risk to hunting by humans. Using this information, I set out to determine whether roe deer have responded to these changing spatial patterns of risk by selecting for safer habitats. To do this I compared location data from roe deer before and after the return of lynx and wolves. I found that during the hunting season, roe deer use agricultural lands considerably less. Furthermore, after the return of lynx, they have used old forest more than they did before. Hence, it seems that roe deer respond to the different spatial patterns of predation risk induced by their top predators, by avoiding the most risky habitats, and using safer habitats more often. The consequences of these indirect predator effects to the ecosystem itself however, are unknown. Conversely, other studies have found that the direct effects of lethal predation by lynx are likely responsible for the strong reduction in roe deer population growth rate and density after the return of this predator. It may be that in these types of human dominated ecosystems which are common throughout Europe, the potential for indirect predator effects to cause phenomenon such as trophic cascades, is limited. Nevertheless, this study may be one of the first to show that roe deer in Europe are responding to the return of their natural predators through shifts in their habitat selection