Wolf depredation on livestock in central Greece

Acta Theriologica 54 (1): 11 22, 29. PL ISSN 1 751 Wolf depredation on livestock in central Greece Yorgos ILIOPOULOS*, Stefanos SGARDELIS, Vaios KOUTIS and Dimitrios SAVARIS Iliopoulos Y., Sgardelis S., Koutis V. and Savaris D. 29. Wolf depredation on livestock in central Greece. Acta Theriologica 54: 11 22. We studied wolf Canis lupus Linnaeus, 1758 livestock conflict in central Greece by investigating patterns of 267 verified wolf attacks on livestock for 21 months. Wolves attacked adult goats 43% and cattle 218% more than expected, whereas sheep 41% less than expected from their availability. Wolves killed less than four sheep or goats in 79%, and one cow or calf in 74% of depredation events, respectively. We recorded higher attack rates during wolf post-weaning season. Wolf attacks on strayed, or kept inside non predator- -proof enclosures, sheep and goats, were on average two to four times respectively more destructive than those when livestock was guarded by a shepherd. Sheepdog use reduced losses per attack. Optimal sheepdog number ranged from 3 to 9 animals depending on flock size. Losses per attack were positively related to the number of wolves involved. Total losses per farm were positively correlated with the size of livestock unit but percentage losses per capita increased with decreasing flock size. Management implications to mitigate livestock depredation are discussed. Callisto, Wildlife and Nature Conservation Society, Mitropoleos 123, Thessaloniki 54621, Greece (YI, VK, DS); Department of Zoology, School of Biology, Aristotle University of Thessaloniki 54124, Greece (YI); Department of Ecology, School of Biology, Aristotle University of Thessaloniki 54124, Greece (SS) Key words: wolf, livestock, depredation, husbandry methods, conservation, Greece Introduction Wolves Canis lupus Linnaeus, 1758 became extinct in late 193s from the isolated Peloponnesus peninsula in southern Greece, while current wolf distribution covers the majority of continental Greece (Hatzirvassanis 1991, Iliopoulos 1999a). Continued wolf presence, resulted in the maintenance of traditional livestock protection measures like shepherd surveillance, use of sheepdogs and enclosures, until present times. In central, northeastern Europe and North America wolves feed mainly on wild ungulates, (Pulliainen 1965, Fritts and Mech 1981, Fuller 1989, Okarma 1993, Jêdrzejewski et al. 2, Mattioli et al. 24, Gazzola et al. 25, Nowak et al. 25, Gula 28). In southern Europe, es- * Present address: Mitropoleos 123, Thessaloniki 54621, Greece, e-mail: yiliop@otenet.gr [11]

12 Y. Iliopoulos et al. pecially in areas with important livestock production, wolves may depend heavily on livestock as prey (Meriggi et al. 1991, Blanco et al. 1992, Fico et al. 1993, Papageorgiou et al. 1994, Meriggi et al. 1996, Vos 2, Pezzo et al. 23, Migli et al. 25). Depredation on livestock is a crucial factor for persecution incentives of wolves (Meriggi and Lovari 1996). Boitani (2) considers monitoring of livestock damages as of great importance for the species conservation and management in Europe. Figures on livestock damage are mainly based on official agencies data of farmer s compensation claims (Blanco et al. 199, 1992, Fico et al. 1993, Cozza et al. 1996, Ciucci and Boitani 1998, Gazzola et al. 28). The compensation scheme in Greece is uniform for the whole country and covers both depredation from wolves and free-ranging dogs (ELGA 23). Data provided by ELGA refer only to damages subject to compensation, usually the more severe incidents, excluding less severe but more numerous ones. Nevertheless they do not include information on attack circumstances and husbandry methods enforced by the affected farmers. Aims of this study were to: (1) describe actual, independently of their severity, figures on livestock losses caused by wolves (2) examine factors associated with the wolf depredation (3) suggest management priorities to mitigate the problem of wolf depredation. Study area Our study area comprises two regions in Central Greece. One region (12 km 2 ) is located in the prefecture of Trikala (39 44 N, 21 38 E) including 36 villages whereas another (1 km 2 ) includes 27 villages in the prefecture of Fthiotida and Magnesia (39 N, 22 3 E). Elevations range from 2 to 23 m a.s.l. Extensive evergreen oak Quercus coccifera, dominate areas up to 7 m a.s.l. Deciduous oak forests, Quercus spp. dominate the 7 12 m a.s.l. elevation zone. Fir forests, Abies borissii regis, Abies cephallonica, and black pine forests Pinus nigra dominate areas above 1 m a.s.l. In higher elevations and below the timberline (17 m a.s.l.) beech Fagus sylvatica, form mixed forests with Abies borissi regis stands. Average human population density, excluding cities, is 15 inhabitants per km 2 (Ministry of the Interior 1999). Availability of livestock expressed as the number of heads per species and per village, was provided by the Greek National Statistical Service and local veterinary offices of Trikala and Fthiotida. Seven thousand cattle, 119 sheep and 8 goats in total, graze inside the study area. Average summer density of sheep and goats is 9 head per km 2 and 3.5 per km 2 for cattle. Sheep and goat flocks graze only during daylight and are guarded by shepherds, with the aid of sheepdogs. Sheep and goat flocks consisted of one or both species (mixed flocks). Cattle herds graze unguarded during day and night, accompanied or not by sheepdogs. Concerning availability of wolf natural prey, wild boar Sus scrofa is widespread, while roe deer Capreolus capreolus occur in very low densities (Y. Iliopoulos, unpubl.). Balkan chamois Rupicapra rupicapra balcanica are rare and inhabit only alpine or sub-alpine remote areas (Papaioannou and Kati 27). Brown hares Lepus europeus exist at low densities compared to other European countries (Sfougaris et al. 1999). Winter wolf density in the study area has been estimated at 2.6 ±.1 (SE) wolves per 1 km 2 (Iliopoulos 1999b, 2) Material and methods Examination of damage claims We directly examined damage claims from February 1999 through October 2 every month. Since January 1999, a pilot compensation system was initiated in the framework of an EC LIFE project on wolf conservation in Central Greece. Each verified livestock loss caused by predation, was extra compensated if not covered by the national compensation system. A trained person (V. Koutis, D. Savaris and Y. Iliopoulos) visited the attack site within 24 hours. We examined attack sites for blood spilled over and close to carcasses, to exclude cases of post mortem consumption. We also used the following secondary criteria: presence of throat bites, presence and number of injured animals, signs of struggle, dragging of carcasses, predator consumption patterns, presence and location of subcutaneous hemorrhages in order to exclude fraud claims and identify predator involved (Kaczensky and Thomas 1994, Bousbouras 1997). We classified cause of livestock death into three categories: predation, when blood was found in the attack site and other secondary criteria were fulfilled indicating wolf predation, probable predation when several of the secondary criteria were fulfilled and other when evidence was poor or destroyed. In total, we examined 34 livestock damage claims. Claims for goat and sheep attacks were considered as one category, as in many cases attacks were on mixed flocks involving both species. We excluded cases classified as other cause of death and cases when dogs were suspected or observed to kill livestock from further analysis. Eight more cases were excluded as being attacks on dogs or donkeys. Figures on livestock losses are referred only to killed and seriously injured livestock found and examined in the attack site, thus they represent minimum levels of livestock losses per wolf attack.

Wolf depredation on livestock 13 Information on associated factors During field visits we asked each farmer to report: (1) if he was in the vicinity of the attack site and, accordingly, (2) date and hour of attack, (3) livestock flock size during attack, (4) number and reaction of sheepdogs present, (5) number of wolves observed, (6) weather conditions, and (7) attack circumstances. Attack circumstances were classified in three categories: (a) animals grazed with the rest of the flock guarded by a shepherd, (b) animals strayed away from the flock and out of the direct supervision of a shepherd, and (c) animals kept in non predator-proof enclosures in the absence of a shepherd. We classified weather conditions as (a) foggy and rainy with low visibility and (b) clear with good visibility. Farmers provided also information concerning husbandry methods, size and predominant birth season of their livestock flock. We assessed habitat in the attack site by describing three basic landscape features inside a radius of approximately 5 meters: (a) visibility according to main tree species and vegetation structure, (b) vegetation coverage and (c) topography (Appendix 1). Statistical analyses We used the 2 goodness-of-fit test to examine: (a) if cattle, sheep and goats suffered losses proportionally to their relative availability; (b) if the number of attacks on cattle were equally distributed among different age classes, and (c)ifrateofattackswasequallydistributedamongdifferent months or seasons of the year. In cases, where the 2 -test showed significant overall differences, we used the Bonferroni confidence interval method, eg Meriggi and Lovari (1996), to a) determine which of the three species suffered significantly different losses than expected and b) calculate deviation percentages from expected attack rates at different species and amongst seasons. We performed univariate non-parametric tests (Kruskall- -Wallis and Mann-Whitney) to compare the average number of sheep/goat killed per attack under different circumstances and seasons, and compare average total losses per farmer. We conducted multiple linear regression with stepwise variable selection for fitting a statistical model to predict the proportion of sheep and/or goats killed during each attack from the following possible predictor variables: number of dogs, season (month), vegetation visibility, forest cover, slope, weather, time of attack and flock type (sheep, goat, or mixed). Nominal variables were coded into a set of binary ones as dummy variables. The dependent variable was transformed according to Box-Cox family of transformations to select the appropriate scale. Continuous predictor variables were also subjected to various transformations for selecting the best set to fit the model. We selected the final model among those leaving normally distributed residuals. The selection criterion was the maximization of R 2 adjusted for the degrees of freedom. As reliable observations for the number of wolves involved during attacks were available only for a small subset of depredation events (n = 75), we fitted a second model for this data subset. We used the Spearman rank correlation test to examine relationship between the farm size and amount of losses. Results Number of attacks validated Two hundred sixty seven damage claims were assigned to predation and probable predation by wolves. Among them 35 attacks were on cattle and 224 on sheep and/or goats. We examined 95% of cattle, 88% of sheep and 76% of goats from those totally claimed by farmers as being killed by wolves. Wolf selection of livestock species and age classes Wolves attacked cattle 218% and goats 43% more often than expected, whereas sheep 41% less often than expected, based on their relative availability ( 2 = 94.86 df = 2, p<.1, Bonferroni confidence interval, p <.1, Table 1). From a total of 812 sheep and goats killed, 96% were adults (> 1 year old). In case of cattle, wolves preyed heavily on calves < 1 year old ( 2 = 56.58, df = 1, p<.1) that constituted 77% of all cattle Table 1. Wolf prey selection amongst the three livestock species (Bonferonni interval analysis). * significant at p <.1. Livestock species Expected predation rate (%) Differences from expected predation (%) Recorded attack events Recorded animals killed Cattle 3.4 + 218* + 45* Goats 38.4 + 43* + 52* Sheep 58.2 41* 37*

14 Y. Iliopoulos et al. killed. From among 34 calves, 79% were 6 months old or younger. Temporal distribution of attacks Rate of wolf attacks (expressed as the number of attacks per day per month) on sheep and goats was not proportionally distributed throughout the year ( 2 = 41.6, df = 11, p <.1, n = 222) and peaked in early autumn whereas it was at minimum in winter (Fig. 1). Rate of attacks on cattle did not show any significant monthly differentiation ( 2 = 15.6, df = 11, p>.5, n = 35, Fig. 1). We also evaluated temporal distribution pattern of attacks, amongst the annual biological seasons of wolf packs, as proposed by Vila et al. (1995): (1) winter and early spring nomadic season (November April), (2) denning and wolf pup weaning (May July), and (3) post-weaning season (August October). Attack rate to sheep and goats during the wolf post-weaning season was by 48% greater than expected, whereas it was lower by 16% and 3% during the denningweaning season and the winter-early spring season, respectively ( 2 = 26.1, df = 2, p<.1 and Bonferroni confidence interval analysis, p<.1 in all cases, Fig. 2). Variation of livestock number killed per attack and associated factors Wolves killed less than 4 sheep or goats per attack in most cases (79%). Severe losses (> 15 animals) occurred only in 3% of all cases, (Fig. 3). When wolves killed cattle, 74% of all cases involved only one animal and on remaining occasions there were two individuals killed during each attack. All depredation on cattle occurred on pastures in the absence of a shepherd. Wolf attacks on strayed sheep and goats (11% of cases), or kept inside non predator-proof, enclosures (3% of cases), were on average, two and four times respectively more destructive, than.7.6 Attacks cattle sheep and goats Attacks per day per month.5.4.3.2.1 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Months Fig. 1. Monthly rate of wolf attacks: pooled data from both 1999 and 2.

Wolf depredation on livestock 15 Percentage 5 4 3 2 1-1 -2 Expected Observed Deviation -3 Winter-breed Den-weaning Post-weaning Fig. 2. Temporal distribution of expected and observed attack rates to sheep and goats, amongst the three wolf seasons. Black bars represent percentage deviation from expected values. those cases (86%) when livestock was clumped together and guarded by a shepherd (H = 22.56, df = 2, p <.1, Table 2). Most cases of strayed livestock killed (72%, n = 18) were goats. Wolves killed more sheep and goats per attack in mixed flocks containing both species, than in single-species flocks (H = 19.8, df = 2, p <.1, Table 2) and also during the post weaning season compared to other periods (H = 8.47, df = 2, p <.5, Table 2). We constructed two models to predict the proportion of sheep and goats killed in a livestock Percentage 8 7 6 5 4 3 2 1 1-4 5-7 8-15 Animals killed per attack >15 Fig. 3. Number of sheep and goats killed by wolves per attack. Percentage distribution of severity classes for 224 attack events. flock after a single attack event. In both models the independent variable was the logarithm of animal losses. In Model I, we analyzed 194 attack events and we did not include observed wolf number as a predictor. In Model II we used a subset of 75 attack events including observed wolf number as a predictor variable. The R 2 adjusted for the degrees of freedom was.453 for model I and.546 for model II. Model I predicts higher than average losses per attack for mixed (R 2 =.47) and sheep flocks (R 2 =.457) and when attacks occur in October (R 2 =.47). The number of dogs involved in protection of flocks affected the losses per attack following the formula: Z 1 =.37X 2.774ln(X+1) +.61ln(Y+1), where: X number of dogs and Y number of dogs per 1 animals. In Model II, losses increase as the number of observed wolves increases (R 2 =.44), while the effect of dog number is described by the following equation: Z 2 =.33X 2.622ln(X+1) +.121Y. Both models I and II indicate that when the number of guarding dogs increase, losses are gradually reduced until number of dogs reaches

16 Y. Iliopoulos et al. Table 2. Average number of goats and/or sheep killed per attack, according to factors related to severity of losses. Average loss ± CI (95%) Range, n Flock type Goat 3 ±.56 1 21 121 Sheep 3.43 ± 1.14 1 26 54 Mixed (sheep and goat) 4.49 ± 1.15 1 21 39 Attack circumstances Animals within the main flock guarded by shepherd 2.9 ±.41 1 21 181 Animals strayed away from the rest of the flock without guarding 6.66 ± 2.59 1 26 25 Animals in non predator-proof enclosure, not guarded 12.3 ± 6.7 2 36 7 Season Denning-weaning season 3.1 ±.31 1 26 72 Winter/wolf breeding season. 3.45 ±.97 1 17 53 Post weaning season 4.1 ±.95 1 36 97 a critical value, (we term it the optimal dog number). When number of dogs exceeds this optimal value, losses gradually start to increase, but dogs still maintain their protective role. When the number of dogs reaches very high numbers, model I and II predict losses even largerthanthecasewhennodogsatall,are used (Fig. 4). In model II, when number of wolves is used as a predictor variable of animal losses, effect of guarding dogs is more prominent: losses are lower for a certain number of dogs used, compared to those predicted from model I. The optimal dog number increases with livestock flock size from 3 4 in small flocks (n = 1) to 7 9 dogs in large flocks ranging from 5 to 1 animals (Fig. 5). Other factors like.2 Model 1 flock 3 Log relative losses -.2 -.4 -.6 -.8 Model 2 flock 3-1. 2 4 6 8 1 12 14 16 18 Number of dogs Fig. 4. Effect of guarding dogs on the relative animal losses (sheep and goats) killed by wolves per attack for flocks with a size of 3 animals.

Wolf depredation on livestock 17 Optimum dog numbers 9 8 7 6 5 4 3 1 2 3 4 5 6 7 Flock size Model 2 Model 1 8 weather and habitat features (cover, visibility, and topography) did not have any significant effect to the level of livestock losses per attack. Level of total losses per livestock farm 9 1 Fig. 5. Optimum dog number for different flock sizes as indicated from Models I and II. Most farms (84%) experienced less than 2 wolf attacks and only 6% were chronically affected (> 3 attacks) during the 21-month period of the study (Fig. 6). Sixty one percent of all farmers lost less than 5% of their livestock during the study, or approximately less than 2.5% peryear(fig.6). Total losses averaged 5.8 ±.97% (CI = 95%) of sheep and goats (range = 1 38, n=131), or 2.3 ±.7% (CI = 95%) of cattle (range = 1 8, n=19). Total number of attacks and animal losses per sheep and goat farm were positively correlated with the size of the farm (r s =.269, p <.1and r s =.326, p <.1 respectively, n = 131). Proportional losses per livestock capita, however, were significantly negatively correlated with the size of the farm (r s =.555, p<.1, n=131). Positive, but not significant trend was found between cattle farm size and total number of attacks or overall number of cattle killed during the study (r s =.274 and r s =.24, respectively, p >.5, n = 19). As with sheep and goat farms, losses per cattle capita were significantly negatively correlated with the size of the farm (r s =.675, p <.1, n = 19). No significant differences were found between average total number of attacks and total animal losses per farm, among sheep farms, goat farms or mixed ones (Kruskall-Wallis test: H = 4.32 and H = 2.345, respectively, n = 114, df = 2, p >.5). Percentage of farms 8 7 6 5 4 3 2 (a) Sheep/goat farms Cattle farms Percentage of farms 8 6 4 2 (b) 1 1 2 3 >3.1-.5 >.5-1 >1-5 >5 Number of attacks Percentage of livestock capita Fig. 6. Total losses caused by wolf predation per livestock farm during the study. (a) Total number of attacks for sheep and goats (n = 131) and cattle (n = 19) units. (b) Percentage of livestock killed per flock for sheep and goat units (larger than 3 animals, n = 121) and cattle units (n = 19).

18 Y. Iliopoulos et al. Discussion Livestock species and age class selection Wolves killed more goats and less sheep than expected, although overall, the goat farms did not experience more attacks and losses compared to sheep farms. Analysis of wolf scats in part of the study area, showed also goats as a heavily preyed species (Migli et al. 25). Vos (2) in northern Portugal, found goats to be the main wolf prey, despite abundant sheep present. In central Italy instead, sheep was the preferred livestock prey, but in that case, availability of goats was very low (Ciucci and Boitani 1998). Huggard (1993) suggested that prey selectivity is depended on the degree of habitat overlap, encounter rates between predator and prey, vulnerability and predictability of prey occurrence. We do not expect sheep and goat predictability to differ, as both type flocks graze in steadily well defined pastures. Considering vulnerability, sheep should be equally and even more vulnerable than goats when attacked, as model I predicts. We suggest that wolves killed goats more often due to higher encounter rates and better attack opportunities with goat flocks. Goats utilize denser forested habitat of Quercus coccifera woodlands and young oak forests, like wolves do while moving during daytime, resting or denning (Ciucci et al. 1997). Gula (28) found that farms closer to wolf pack rendezvous sites experienced more attacks. Moreover, surveillance of goat flocks is more difficult as they are more scattered than sheep and forage in steeper slopes (Y. Iliopoulos, unpubl.). As goats tend to spread inside the forest, the encounter rate of wolves with isolated individuals may increase. Goats prevailed as the most preferred livestock species in a study area level, despite the fact that goat farms did not experience more attacks and losses compared to sheep farms. We suggest this to be a consequence of more goat units affected than expected, according to their numbers in the study area. Synchronized births in sheep and goat farms during winter months reduced losses of lambs and kids. In periods with higher predation risk (summer-autumn), this age class was almost absent from the pastures. Cattle predation was by 218% more frequent than expected and was actually a consequence of wolves predating mostly calves. Calves aged less than 6 months, grazed unattended during night hours, contrary to sheep and goats that normally stayed inside enclosures and were not available. We suggest that increased calf availability and accessibility to wolves was the main reason for high predation rates recorded. Bradley and Pletscher (25) in Montana and Idaho, found no relationship between cattle predation rates amongst farms and calving season or other husbandry techniques, but in that case, wolves fed mainly on elk. Habitat overlap of cattle with wolves and elk was the main predicting factor that determined loss levels. Wolf predation on cattle was a serious conservation problem, despite the fact that only 19 farms were affected during the study. Cattle farmers were, in general, less tolerant to wolves, compared to sheep/goat farmers, as each adult cattle or calf killed represented an important economic asset. Wolf mortality related with depredation on cattle was important in the study area, as only a few affected cattle raisers were known to have shot or poisoned approximately 25% of the wolves totally killed during 1999 2 (Iliopoulos 1999b, 2) Seasonal changes in attack frequency Wolf attacks on livestock occurred all year round. There was, however, a marked seasonal pattern reported also from other parts of the wolf range (Fritts et al. 1992, Ciucci and Boitani 1998, Nowak et al. 25, Gazzola et al. 28). Attacks peak in summer and early autumn, when sheep and goat numbers and availability in pastures were relatively stable. More attacks than expected occurred during the post-weaning period (August October), when wolf pups require large food intake given their relatively high growth rate (Oftedal and Gittleman 1989). Although growth rate slows down with pup age, the net amount of food and energy intake is probably much higher in post weaning

Wolf depredation on livestock 19 than in the weaning period, imposing a need for increasing kill rates of livestock (Fritts et al. 23). Treves et al. (22) found that reproductive wolf packs, rather than lone wolves, caused most damages on livestock in Wisconsin. Attacks on cattle did not show a significant seasonal variability (see also Meriggi et al. 1991 and Fico et al. 1993). Small sample size may have obscured any seasonal trend. Role of husbandry methods and other factors Most of the attacks did not result in a severe loss of livestock and were comparable to average losses described by Ciucci and Boitani (1998) in Tuscany region where livestock production was also high. Serious damages happened in the absence of a shepherd, usually after a portion of the flock had strayed or was attacked inside non predator-proof enclosures. Shepherds can interrupt wolf attacks in the approach, attack, kill and eating phase (Linnell et al. 1996), although in our study area wolves were persistent killers of livestock even in the presence of humans during daytime. We suggest this is a result of low diversity and density of wild ungulates that render livestock as the only abundant prey for wolves. Moreover, wolves are well adapted to kill overabundant livestock having done so for centuries in the humanized environment of southern Europe. Harper et al. (25) found that learning was a prime factor that contributed to the increase of depredation events in Minnesota. Sheep dogs generally appeared to reduce losses. Wolves may avoid areas with livestock guarding dogs, as dogs may disrupt depredatory sequences by wolves, enforcing indirect or direct aggression (Coppinger and Coppinger 1995). When sheep dog number reached a critical level ( optimum dog number) their protective role was gradually reduced. Large number of sheepdogs may result in poor nutrition, lack of appropriate training and development of unsuitable behavioral traits, like killing of livestock. Training, amongst several other factors, affects sheep dog performance (Coppinger et al. 1983, Hansen and Smith 1999). Surplus killing of livestock inside enclosures was rare but very destructive. High density of livestock, lack of escape and ease of depredation inside enclosures stimulate the triggering of this predatory behavior (Linnell et al. 1996). Moreover, in four of these seven cases, guarding dogs were absent and were used to guard other portions of the flock in a different area. Thus, the effectiveness of enclosures may be only enforced or facilitated by the combined use of sheep dogs. Larger sheep and goat farms experienced more attacks and total animal losses, as also reported by Vos (2) in northern Portugal, Mech et al. (2) in Minessota and Brandley et al. (25) in Montana, concerning cattle farms. Larger flocks face higher encounter rates with wolves (Althoff and Gipson 1981), are more dispersed (Robel et al. 198) and serve as greater attractants for wolves, because they contain more individuals at lower condition (Brandley et al. 25). The losses in livestock due to depredation greatly increased with the number of wolves involved in the attack. Cooperative hunting by wolves may probably disrupt sheep dog coordination and effectiveness of shepherd surveillance. Many shepherds described a common hunting strategy of wolf packs, where one wolf distracts the dogs and others scatter the flock. Moreover, larger number of wolves involved also translates into higher energetic needs of the entire wolf pack. Severity of livestock losses at a study area level Levels of depredation and trophic dependence of wolves on livestock, observed in our study area (Migli et al. 25 and this study) were more severe compared to other areas. For instance, livestock losses in the province of Arezzo, Italy, were half the size of that recorded in our study, despite of larger area surveyed, comparable wolf density and high density of livestock (Gazzola et al. 28). However, in contrast to our area, a rich community of wild ungulates occurred in Arezzo. In Winsconsin, wolf predation, was negligible if compared to losses recorded in this study, with only one third of wolf packs were involved in killing livestock (Treves 22) while in our study area all wolf packs were actually involved. Studies on wolf

2 Y. Iliopoulos et al. predation in multi-prey ecosystems has shown that wolves still prefer wild ungulates despite the presence of livestock (Meriggi and Lovari 1996, Capitani et al. 24, Gazzola et al. 25, Nowak et al. 25, Gula 28). Management implications Efforts to restore ungulate communities in wolf-livestock conflict areas in order to reduce wolf attacks to livestock in Europe have been encouraged as a very important conservation measure (Meriggi and Lovari 1996, Linnell et al. 1996). The enforcement of such a measure in Greece should be carefully examined in areas used by livestock as domestic ungulate densities are locally still high. Predation on domestic animals may remain relatively high, if livestock is locally abundant, even in areas with high densities of wild prey populations, if protective methods are not effectively enforced (Patalano and Lovari 1993, Gazzola et al. 28). Enhancement of natural prey availability in selected suitable areas with relatively lesser livestock densities may be more feasible and effective and should be combined with intensification of husbandry methods. Surveillance of livestock is probably the most effective way of protecting livestock against wolf attacks. Although this husbandry method is still widely used in Greece and southern Europe, it becomes less and less attractive for younger farmers especially in areas recently re-occupied by wolves in the periphery of wolf distribution (Y. Iliopoulos, unpubl.). This leads to numerous socially unacceptable severe attack events, jeopardizing any other management tool to mitigate wolf-livestock conflicts. Maintenance of an optimum number of well trained guarding dogs adapted to flock size, instead of keeping a large uncontrolled number of dogs, may result in more effective protection of livestock and avoidance of unnecessary costs. Use of traditional guarding dog breeds should be encouraged, due to their long history of use, even for centuries (Linnell et al. 1996). In Greece, such a breed is the Greek sheepdog Ellinikos poimenikos, however it is not commonly used by farmers, as pure descendants of the breed are very rare (Drivas 1996). Intensification of husbandry methods should focus particularly during the wolves post-weaning period (August October) when the rate and severity of attacks is higher and at goat farms that have a higher probability to be attacked by wolves. Duetotheirlargebodysizeandlowervulnerability to wolf depredation cattle have been proposed as an alternative for sheep and goat farming to reduce net number of carnivore attacks on livestock (Linnell et al. 1996). We do not recommend the use of cattle instead of sheep and goats to reduce overall livestock losses, unless serious improvements in husbandry, birth synchronization and herding techniques have also been enforced. Confinement of calves younger than 6 months of age during night could significantly reduce losses to this age class and increase the effectiveness of sheepdogs but requires substantial effort by farmers. Although large farms experienced the highest losses due to depredation, the small farms should be prioritized in conservation and management projects, because the economic loss per capita is proportionally higher there, which greatly influences the social acceptability of wolves. Use of predator-proof enclosures combined with use of guarding dogs, should be more widely implemented. Aknowledgements: This study was part of the LIFE project Lycos, NAT97-GR4249, considering wolf conservation in Greece. Study was funded from the EC-DG Env directorate (5%) and Arcturos NGO (5%), former affiliation of Y.I, V. K. and D. S. authors. We greatly thank, Ir. Hatzimichael, P. Menounos, P. Pavlides, S. Tzortzakis, N. Kanellopoulos, G. Giannatos, D. Vassilakis, for contribution on fieldwork, I. Aravidis for helping with GIS work and Dr. Y. Mertzanis for linguistic help on the manuscript. We also greatly thank Dr J. D. C. Linnell for his helpful comments, professors S. Lovari, A. Meriggi, two anonymous referees and especially Dr K. Schmidt for thoroughly reviewing an earlier draft of this manuscript. References Althoff D. P. and Gipson P. S. 1981. Coyote family spatial relationships with reference to poultry losses. The Journal of Wildlife Management 45: 641 649. Blanco J. C., Cuesta L. and Reig S. (eds) 199. El lobo (Canis lupus) en Espana. Situation, Problematica y

Wolf depredation on livestock 21 apuntes sobre su ecologia. Ministerio de agricultura pesca y alimentation. Icona, Madrid: 1 118. Blanco J. C., Reig S. and Cuesta L. 1992. Distribution, status and conservation problems of the wolf Canis lupus in Spain. Biological Conservation 6: 73 8. Boitani L. 2. Action plan for the conservation of wolves in Europe (Canis lupus). Nature and Environment, No 113. Council of Europe Publishing, Strasbourg: 1 84. Bousbouras D. 1997. Guide to the identification of wounds inflicted on livestock by predators. Arcturos, Thessaloniki: 1 24. Bradley E. H. and Pletscher D. H. 25. Assessing factors related to wolf depredation of cattle in fenced pastures in Montana and Idaho. Wildlife Society Bulletin 33: 1256 1265. Capitani C., Bertelli I., Varuzza P., Scandura M. and Apollonio A. 24. A comparative analysis of wolf (Canis lupus) diet in three different Italian ecosystems. Mammalian Biology 69: 1 1. Ciucci P. and Boitani L. 1998. Wolf and dog depredation on livestock in central Italy. Wildlife Society Bulletin 26: 54 514. Ciucci P., Boitani L., Francisci F. and Andreoli G. 1997. Home range, activity and movements of a wolf pack in central Italy. Journal of Zoology, London 243: 83 819. Coppinger R. and Coppinger L. 1995. Interactions between livestock guarding dogs and wolves. [In: Ecology and conservation of wolves in a changing world. L. N. Carbyn, S. H. Fritts and D. R. Seip, eds]. Canadian Circumpolar Institute, Alberta, Edmonton: 523 526. Coppinger R., Lorenz J., Glendinning J. and Pinardi P. 1983. Attentiveness of guarding dogs for reducing predation on domestic sheep. The Journal of Range Management 38: 89 18. Cozza K., Fico R. and Battistini L. 1996. The damage-conservation interface illustrated by predation on domestic livestock in central Italy. Biological Conservation 78: 329 336. Drivas H. 1996. [Greek sheepdog. A course of race reconstruction]. Cynologia 25: 28 35. [In Greek] ELGA 23. [Animal insurance regulation. Hellenic Farmers Insurance Organization]. Athens: 1 24. [In Greek] Fico R., Morosetti G. and Giovannini A. 1993. The impact of predators on livestock in the Abruzzo region in Italy. Revue Scientifique et Technique, Office International Epizooties 12: 39 5. Fritts S. H. and Mech L. D. 1981. Dynamics, movements and feeding ecology of a newly protected wolf population in northern Minnesota. Wildlife Monographs 8: 1 79. Fritts S. H., Paul W. J., Mech L. D. and Scott D. P. 1992. Trends and management of wolf-livestock conflicts in Minessota. United States Department of the Interior. Fish and Wildlife Service, Resource publication 181, Washington, DC: 1 27. FrittsS.H.,StephensonR.O.,HayesR.D.andBoitaniL. 23. Wolves and humans. [In: Wolves: behavior, ecology and conservation. D. L. Mech and L. Boitani, eds]. The University of Chicago Press, Chicago and London: 289 316. Fuller T. K. 1989. Population dynamics of wolves in North- -Central Minnesota. Wildlife Monographs 8: 1 41. Gazzola A., Bertelli A., Avanzinelli E., Tolosano A., Bertotto P. and Apollonio M. 25. Predation by wolves (Canis lupus) on wild and domestic ungulates of the western Alps, Italy. Journal of Zoology, London 266: 25 213. Gazzola A., Capitani C., Mattioli L. and Appolonio M. 28. Livestock damage and wolf presence. Journal of Zoology, London 274: 261 269. Gula R. 28. Wolf depredation on domestic animals in the Polish Carpathian Mountains. The Journal of Wildlife Management 72: 283 289. Hansen L. and Smith M. E. 1999. Livestock-guarding dogs in Norway: part II. Different working regimes. Journal of Range Management 52: 312 316. Harper E. K., Paul W. J. and Mech L. D. 25. Causes of wolf depredation increase in Minnesota from 1979 1998. Wildlife Society Bulletin 33: 888 896. Hatzirvassanis V. 1991. Observations sut l etat des populations du loup en Gr ce. Biologia Gallo-hellenica 18: 13 2. Huggard D. J. 1993. Prey selectivity of wolves in Banff National park. I. Prey species. Canadian Journal of Zoology 71: 13 139. Iliopoulos Y. 1999a. [Distribution and population estimates of the wolf in Greece. Project LIFE Lycos NAT97- GR4249: Conservation of the wolf (Canis lupus L.) and its habitats in Greece]. Arcturos, Thessaloniki: 1 53. [In Greek] Iliopoulos Y. 1999b. [Distribution, population estimates and mortality of the wolf in Central Greece. Annual report 1999. Project LIFE Lycos NAT97-GR4249: Conservation of the wolf (Canis lupus L.) and its habitats in Greece]. Arcturos, Thessaloniki: 1 19. [In Greek] Iliopoulos Y. 2. [Distribution, population estimates and mortality of the wolf in Central Greece. Annual report 2. Project LIFE Lycos NAT97-GR4249: Conservation of the wolf (Canis lupus L.) and its habitats in Greece]. Arcturos, Thessaloniki: 1 13. [In Greek] Jêdrzejewski W., Jêdrzejewska B., Okarma H., Schmidt and Zub K. 2. Prey selection by wolves in Bia³owie a primeval forest, Poland. Journal of Mammalogy 81: 197 212. Kaczensky P. and Thomas H. 1994. Wer war es? Dokumentation und identifikation von Raubtierrissen. Zentralstelle Österr. Landesjagdverbände, Wien: 1 4. LinnellJ.D.C.,SmithM.E.,OddenJ.,SwensonJ.E.and Kaczensky P. 1996. Carnivores and sheep farming in Norway. 4. Strategies for the reduction of carnivore- -livestock-conflicts: a review. NINA Oppdragsmelding 443: 1 118. Mattioli L., Capitani C., Avanzinelli E., Bertelli I., Gazzola A. and Apollonio M. 24. Predation by wolves (Canis lupus) on roe deer (Capreolus capreolus) in north-eastern Apennine, Italy. Journal of Zoology, London 26: 249 258. Mech L. D., Harper E. K., Meier T. J. and Paul W. J. 2. Assessing factors that may predispose Minnesota farms to wolf depredation on cattle. Wildlife Society Bulletin 28: 623 629.

22 Y. Iliopoulos et al. Meriggi A., Brangi A., Matteucci C. and Sacchi O. 1996. The feeding habits of wolves in relation to large prey availability in northern Italy. Ecography 19: 287 295. Meriggi A. and Lovari S. 1996. A review of wolf predation in southern Europe: does the wolf prefer wild prey to livestock? Journal of Applied Ecology 33: 1561 1571. Meriggi A., Rosa P., Brangi A. and Matteucci C. 1991. Habitat use and diet of the wolf in northern Italy. Acta Theriologica 36: 141 151. Migli D., Youlatos D. and Iliopoulos Y. 25. Winter food habits of wolves in central Greece. Journal of Biological Research 4: 217 22. Ministry of the Interior 1999. Population census of Greece- -Kapodistrias project. http:www.ypes.gr/kapodistrias/index.html. Nowak S., Mys³ajek W. and Jêdrzejewska B. 25. Patterns of wolf Canis lupus predation on wild and domestic ungulates in the Western Carpathian Mountains (S Poland). Acta Theriologica 5: 263 276. Oftedal O. T. and Gittleman J. L. 1989. Patterns of energy output during reproduction in carnivores. [In: Carnivore behavior,ecology,andevolution.j.l.gittleman,ed]. Cornell University Press, Ithaca, New York: 355 378. Okarma H. 1993. Status and management of the wolf in Poland. Biological Conservation 66: 153 158. Papaioannou H. I. and Kati V. I. 27. Current status of the Balkan chamois (Rupicapra rupicapra balcanica) in Greece: Implications for conservation. Belgian Journal of Zoology, London 137: 33 39. Papageorgiou N., Vlachos C., Sfougaris A. and Tsachalidis E. 1994. Status and diet of the wolf in Greece. Acta Theriologica 39: 411 416. Patalano M. and Lovari S. 1993. Food habits and trophic niche overlap of the wolf Canis lupus L., 1758, and the red fox Vulpes vulpes (L., 1758) in a Mediterranean mountain area. Revue d Ecologie (La Terre et La Vie) 48: 23 38. Pezzo F., Parigi L. and Fico R. 23. Food habits on wolves in central Italy based on stomach and intestine analyses. Acta Theriologica 48: 265 27. Pulliainen E. 1965. Studies on the wolf (Canis lupus) in Finland. Annales Zoologici Fennici 2: 215 259. Robel R., Dayton A. D., Henderson F. R., Meduna R. L. and Spaeth C. F. 198. Relationships between husbandry methods and sheep losses to canine predators. The Journal of Wildlife Management 45: 894 911. Sfougaris A., Papageorgiou N., Giannakopoulos A., Goumas H., Papaevangelou E. and Anni A. 1999. Distribution, populations and habitat of the European hare (Lepus europeus Pallas, 1778) in central and western Greece. Proceedings of the 24th Congress of the International Union of Game Biologists: Agriculture-Forestry-Game: Intergrading wildlife in land management. Thessaloniki: 423 43. Treves A., Jurewicz R. R., Naughton-Treves L., Rose R. A., Willging R. C. and Wydeven A. P. 22. Wolf depredation on domestic animals in Wisconsin, 1976 2. Wildlife Society Bulletin 3: 231 241. Vila C., Urios V. and Castroviejo J. 1995. Observations on the daily activity patterns in the Iberian wolf. [In: Ecology and conservation of wolves in a changing world. L. N.Carbyn,S.H.FrittsandD.R.Seip,eds].Canadian Circumpolar Institute, Alberta: 335 34. Vos J. 2. Food habits and livestock depredation of two Iberian wolf packs (Canis lupus signatus) in the north of Portugal. Journal of Zoology, London 251: 457 462. Received 18 March 28, accepted 18 September 28. Associate editor was Krzysztof Schmidt. Appendix 1. Landscape features and their given numerical values per class used in analysis of factors associated with wolf depredation on livestock in Greece. Landscape feature Description Rank classes Visibility according to main vegetation species and form V >V1 >V2 Open pasture or alpine areas High visibility V Beech and oak forests in winter Medium visibility forest V1 Beech and oak forests in summer Low visibility forest V2 Fir forest Low visibility forest V2 Black pine forest Low visibility forest V2 Quercus coccifera shrublands Medium visibility forest V1 Quercus coccifera woodlands Low visibility forest V2 Vegetation cover Partially forested 4% forest coverage C1 Forested 41 6% forest coverage C2 Dense forested >6% forest coverage C3 Topography Hilly terrain-forest plateau T1 Rugged terrain steep slopes T2 Gorge-stream T 3