When making foraging decisions, animals evaluate the quality and quantity of different food patches in the landscape, as well as the likelihood of being killed by a predator. As the perceived risk of predation increases, feeding animals resort to use anti-predator behaviours that confer safety at the expense of a reduction in the foraging efficiency and intake (Sih 1980, Fortin et al. 2005, Hernández & Laundré 2005). Such anti-predator responses usually involve reducing the time of foraging activity, shifting activity from risky to safe patches, and/or using higher levels of vigilance (Lima & Dill 1990, Kotler et al. 1994, Brown 1999). If the level of danger varies among patches throughout the landscape, the food resources left in each resource patch should reflect the foragers' perceptions of risk. The patches with higher food resources should be the ones that are perceived to confer higher risk of predation. In this way, one can quantify the animals' ‘landscape of fear’, which describes how the animals' foraging cost of predation varies spatially with respect to landscape features (Laundré et al. 2001, 2010). Such information can be very useful in understanding the distribution of species in the face of predators, as well as helpful for providing more accurate predictions concerning the effect of anthropogenic development and other changes in the environment. Hence, the identification of those environmental variables that are more likely to influence the behaviour of the target species may be valuable for meeting conservation goals.

Several studies have shown that ungulate habitat use has a strong association with abiotic factors that might influence the perceived risk of predation. For instance, when foraging, many animals reduce their use of habitat patches as the distance from a refuge increases (Berner et al. 1992, Molvar & Bowyer 1994, Kotler et al. 1999, Gates et al. 2005). Other species, though, prioritise visibility when selecting foraging habitats (e.g. mountain sheep Ovis canadensis; Risenhoover & Bailey 1985) because good visibility means that an animal has higher probability of detecting an approaching predator, and thus, more time to escape. Still, others prefer to forage on slopes with lower visibility than on plateaus since the safety of the slopes offsets the reduced visibility (e.g. soay sheep Ovis aries; Hopewell et al. 2005). Vegetation can be another important feature influencing habitat use (Fortin et al. 2005), and its effect may be determined by its obstructive or concealing and protective functions, depending on the anti-predator strategies of the species (Pays et al. 2009, Beauchamp 2010). The distance to a human-induced disturbance is also likely to affect habitat selection and foraging behaviour of certain wildlife species, with the magnitude of its effect varying greatly among species (Berger et al. 1983, Frid & Dill 2002).

Nubian ibex Capra nubiana is an endangered species protected by law in Israel (IUCN Red List, November 2005). These wild social goats forage during daylight hours, are more active during mornings and afternoons, and rest and ruminate during midday and at night (Levy & Bernadsky 1991). Differences in the vulnerability to predation risk and in forage requirements lead to sexual segregation in this species. Since females take care of the young and are more vulnerable than males, they are usually confined to safer and richer habitats (Gross et al. 1995a, Kohlmann et al. 1996). Potential predators of this species are leopards Panthera pardus, wolves Canis lupus, striped hyenas Hyaena hyaena and human beings Homo sapiens. The neonates can fall prey to golden eagles Aquila chrysaetos and eagle owls Bubo bubo (Levy & Bernadsky 1991). Nubian ibex are exceptional climbers so they are restricted to desert habitats with steep slopes and cliffs that provide escape terrain (Kotler et al. 1994, Gross et al. 1995b). As they venture farther from the cliff edges, their use of anti-predator behaviour increases (Kotler et al. 1994, Hochman & Kotler 2007).

The goal of our study was to quantify the Nubian ibex' landscape of fear. That is, we want to better understand how the ibex perceive features of their environment and how these features affect their use of the landscape with respect to the risk of predation. To this end, we used the giving-up density (GUD) technique to map the landscape (Brown 1988, Kotler et al. 1994, van der Merwe & Brown 2008, Druce et al. 2009). By measuring the variation in GUDs (i.e. the amount of food left behind after foraging exploitation) among patches, we drew contours of equal foraging costs of predation in the physical space occupied by the ibex.

Material and methods

Results

The tolerance values indicated multicollinearity problems in the model. In order to overcome these problems and to reduce the number of predictors, we dropped some of the intercorrelated variables from the model. Among these variables were ‘elevation’ and ‘pebble’, which had tolerance levels < 0.1. ‘Visibility’ was also involved in high intercorrelation, but since we have special interest in this variable, due to results from an experiment which we have recently conducted and due to the fact that it also evaluates visibility, we kept it in the model. Since ‘drainage’ and ‘big rocks’ were highly correlated with ‘visibility’, we dropped them instead.

Patch use by Nubian ibex varied significantly with environmental features along the landscape. The percentage of vegetation cover, the distance to the main cliff and the distance to nearest slope significantly influenced the GUDs of Nubian ibex. The polynomial regression analysis suggested no non-linear problems, therefore, we only report the results of the linear multiple regression (Table 1). Vegetation cover had a negative influence on the GUDs (see Table 1 and Fig. 2A). The distance to the main cliff and to the nearest slope both affected the GUDs positively (see Table 1 and Fig. 2B-C). The rest of the environmental variables were not significant.

We created a map with contour lines of equal GUDs. The lines represent areas that differ in the predation risk perceived by the ibex and can be viewed as the Nubian ibex' landscape of fear. Then, we determined the contours of the GUD on the map of the features of the landscape (see Fig. 1). The features of the landscape were drawn based on a satellite photo. Areas of lower GUDs correspond to areas of greater safety. These correspond to areas near the cliff and more heavily vegetated areas.

Discussion

Different environmental features may serve as cues of predation risks to foraging animals. If so, animals should alter their foraging and anti-predator behaviour in response to these cues as they move within a landscape. In our study, we demonstrated that the foraging behaviour of Nubian ibex varies according to different characteristics of their environment. The behaviour of the ibex reveals that they perceive greater predation risk as they venture farther from the cliff areas. This was reflected in the increasing GUDs with the distance from both the main cliff and the nearest cliff or slope. The observation of Nubian ibex leaving higher GUDs as the distance to the main cliff increases has been shown by Hochman & Kotler (2007). However, our work provides a more complete view of the most important environmental features working together to shape the Nubian ibex behavioural responses to the risk of predation, revealing the animals' preferences for escape routes and cover while foraging.

In 19 out of 24 artificial feeding patches, the nearest cliff was represented by a small and shallow slope, while in only five occasions, the nearest cliff coincided to be the main cliff. The small slopes have a smaller effect than the main cliff and at a smaller scale. This suggests that less steep and shallower slopes provide less protection for the ibex. Still, when venturing farther from the main cliff, an ibex may gain some protection by being close to a slope. The last observation is consistent with how an ibex responds to threats such as dogs or humans. The moment that the threat is detected, the ibex stop their current activity and becomes vigilant. If the agent of threat comes closer, the ibex retreat towards the nearest slope edge, and if it becomes even more threatening, the ibex retreat to the main cliff and may even descend the slope in extreme cases (C. Iribarren, pers. obs.).

The fact that a reduction of foraging effort by ibex occurs as the distance from the refuge increases, suggests that this species behaves like a central place forager. A central place is a location such as a nest, a refuge, a burrow or any other site to which an animal returns between foraging bouts. In general, the cost of foraging increases as the forager travels to farther areas from the central place (Thorson et al. 1998, Druce et al. 2006). As a consequence, patches that are closer to the central place are more valuable. This leads the foragers to harvest those patches more thoroughly. The pattern of increasing perceived predation risk with the distance from the refuge has been shown previously in several other species where the strategy of avoiding predators consists on fleeing to a refuge (see review by Stankowich & Blumstein 2005).

GUDs were lower when thick (40%) vegetation cover was present, i.e. Nubian ibex increased their foraging efforts in the presence of vegetation. This suggests that the costs of predation when the ibex forage far from the cliff might be assuaged by an increase in the vegetation cover. Such effect was especially evident at station A5 (see Fig. 1), located in the row farthest from the cliff (60 m). This station had considerably higher vegetation cover than the other four stations located at approximately 60 m from the main cliff. Nubian ibex always left lower GUDs at this station compared to the other four.

Two possible explanations may help explain the habitat preference of Nubian ibex for patches with higher vegetation. One explanation is that patches with higher vegetation are more attractive to the ibex simply because they offer greater feeding opportunities. This agrees with Searle et al. (2008), who demonstrated that after altering the natural environment of brown bandicoots Isoodon macrourus, the animals spent more time foraging in feeding stations within areas of the landscape that had greater resource density, independent of the predation risk. In this example, bandicoots harvested food only from feeding trays. However, in our case, the extra food was found outside of trays and would more likely have diverted ibex from feeding trays.

A second explanation may be related to a reduction in the perceived risk of predation. The vegetation present in the area was not tall enough to restrict the ibex' sight lines, but in most cases it was sufficiently tall to wholly or partially conceal an ibex. Moreover, vegetation was distributed in a manner that did not represent a barrier to the ibex, ensuring a safe escape route if necessary. Therefore, vegetation could increase the ibex' concealment without impeding their vigilance. In this way, the vegetation in our study area may play a role in reducing risk of predation, leading the ibex to forage more thoroughly from areas where vegetation was abundant. Note that GUDs were lowest at 40% cover, indicating that really thick vegetation might obstruct vigilance or escape opportunities. If so, very dense vegetation should have the opposite effect compared to the moderate (40%) vegetation, hence increasing the ibex' perception of predation risk.

Sight lines per se did not affect the ibex' patch use significantly. This is in contrast to a recent experiment in which we obstructed sight lines surrounding food patches for the same population of ibex (C. Iribarren & B.P. Kotler, unpubl. data). In response to reduced sight lines, ibex increased their GUDs and increased their vigilance. What may explain these differences? A possibility is that the criterion used to measure the visibility in our current work was inappropriate. We assumed that all sightline directions were of equal importance. However, it is possible that some visual directions are more relevant for reducing risk of predation than others, especially those sight lines directed to escape routes or to areas where the risk increases (e.g. with higher human disturbance or opposite to the cliff area). Similarly, Arenz & Leger (1997) suggested that the position of visual obstructions within the visual field is important to thirteen lined ground squirrels Spermophilus tridecemlineatus, even more than the amount of occlusion per se. In addition, the perceived risk of predation may also depend on the distance to the obstruction. However, in our study, we assumed that all visual obstructions within 10 m were equally important.

An alternative explanation for the lack of correlation between visibility and risk of predation is that, in general, the ibex forage in groups where individuals are separated from each other by a few metres. This means that the presence of sentinels located in areas that offer better visibility (e.g. at higher elevations) may reduce the predation risk of the whole group. Numerous works have shown that ungulates rely on the vigilance of others, and that it leads to a reduction of vigilance of group members in several cases (Alados 1985, Roberts 1996, Hunter & Skinner 1998). If this is the case with Nubian ibex, it is possible that visibility has a stronger effect when individuals forage solitarily or in very small groups. However, we note that ibex do not appear to alter their vigilance behaviour according to group size (Gross et al. 1995b, Hochman & Kotler 2007).

Our study provides information about the important role that habitat structure plays in the ecology of Nubian ibex, even when considering a small area (relative to their home range). Environmental features influenced the ibex' perceptions of predation risk. They ranked the patches based on these perceptions and varied their foraging efforts among the patches accordingly. This means that any manipulations of the natural environment of the Negev Desert could strongly influence the ibex' fear of predation leading to changes in their patch use and foraging efficiency. As previously noted, Nubian ibex are endangered (IUCN Red List, November 2005) and a reduction in their foraging efficiency or in the easy access to escape terrain could affect the future of the local population. Midreshet Ben-Gurion and environs (the location where this research was carried out) has become a popular recreational destination and it has been in continuous development during the last several years. Therefore, proper conservation planning and management is needed. Based on this work, special attention should be paid to the most prominent environmental features that affect the Nubian ibex landscape of fear, which are the distance to the cliff and edge slopes that offers escape terrain, and the vegetation cover that provides concealment to these animals.