What does the wolf eat? Assessing the diet of the endangered Iberian wolf (Canis lupus signatus) in northeast Portugal

The Iberian wolf (Canis lupus signatus) is a top predator that inhabits the Iberian Peninsula. In Portugal, its numbers and distribution declined throughout the 20th century, due to human persecution, habitat degradation and prey decline, which have led to higher predation rates of livestock in the remaining packs. In Montesinho Natural Park (northeast Portugal), wild ungulate populations have been increasing in the last years, which may have led wolf to predate upon them. In order to assess Iberian wolf diet in this area, 85 wolf scats were collected from transects distributed throughout the study area in two periods between November 2017 and August 2019. Scat analysis indicated a high predation on wild ungulates, where the frequency of occurrence showed that roe deer was the most consumed prey (44%), followed by red deer (26%) and wild boar (24%). Domestic/wild cat (6%), domestic goat and stone marten (5%) were consumed in lower quantities. It was found a higher selection towards roe deer (D = 0.71) and this was the only prey item which was significantly dependent of the season of the year (χ2 = 16.95, df = 3, p < 0.001). This is the first study in Portugal where was recorded that wolves feed mainly on wild ungulates. We conclude that lower livestock predation may be correlated with higher wild ungulates densities in our study area, as well as suitable husbandry practices, leading to a shift on Iberian wolf diet from mainly livestock on previous studies to wild ungulates.


Introduction
The grey wolf (Canis lupus) is considered one of the world's most widespread mammal [1]. As an apex predator, this species can contribute to restore local biodiversity and trophic interactions, which ultimately leads to ecosystem recovery [1], [2]. In Europe, wolf original range was drastically reduced in the end of the 19 th century, mainly due to human persecution, habitat degradation and prey decline [3], being eradicated from most central and northern countries [4], [5], [6]. The rooted conflict between this predator and humans is mostly due to livestock predation [7]. This is mostly aggravated in areas where wild prey diversity and density is low a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 [6]. However, in the last decades, due to legal protection policies [4], natural recolonization [8] and wild ungulate increase [9], [10], wolf populations have recovered and are now expanding their ranges across some countries in Europe [1], [4], [5], [11].
In the Iberian Peninsula inhabits an endemic subspecies of the grey wolf, the Iberian wolf (Canis lupus signatus). However, in Portugal, contrasting to the European scenario, this species declined dramatically during the 20 th century, both in number and distribution, gradually disappearing from coastal, south and central regions of the country [3], [6], [12]. Facing this fast decline, since 1988 this species is protected by law and listed as "Endangered" in the Portuguese Red Data Book of Vertebrates [3], [13].
In Portugal, studies focused on wolf feeding ecology reveal higher preference for livestock, either on north and south of Douro River [6], [12], [16], [26], [27]. Torres et al [6] showed that livestock made up over 90% of this predator diet. However, increasing numbers and expansion of wild ungulates in the last decades across Portugal, either from natural re-colonization or reintroductions [10]

Ethics statement
Our research did not involve capture, handling or killing of animals, therefore did not require approval of animal care and use procedures. Permissions for field studies were given by Nature and Forestry Conservation Institute.

Study area
Our study was conducted in Montesinho Natural Park (MNP) (6˚30'-7˚12'W, 41˚43'-415 9'N), comprised as one of European Union's Natura 2000 Network sites (Fig 1). The total prospected area was 35,000 ha and is characterized by a mountainous landscape, with elevation ranging from 438 to 1,481 m. Our study area experiences a Mediterranean climate, with an annual average temperature ranging between 3˚C in the coldest month and 21˚C in the warmest, and precipitation between 600 and 1,500mm [

Scat collection and laboratory analysis
Experienced and field-trained personnel collected wolf scats between November 2017 and August 2018 and between June and August 2019, throughout our study area (Fig 1). Paths, dirt roads, firebreaks, forest trails and crossroads were prospected using a vehicle (< 10km/h) or by foot. Morphology, size, scent, colour, contents and spatial position were used to identify wolf scats. Scats collected along the trails were stored in plastic bags, labelled and registered using a Global Positioning System (GPS) [6]. The collected scats were submitted to genetic analysis to confirm the species Canis lupus signatus, avoiding misclassification of domestic dog (Canis lupus familiaris) and red fox (Vulpes vulpes). DNA extraction was performed using QIAamp1 DNA Stool Mini Kit (QIAGEN Hilden, Germany) following manufacturer instructions. A fragment of 350bp from the control region (mitochondrial region) [37], was amplified using the universal primers Thr-L 15926 5'-CAATTCCCCGGTCTTGTAAACC-3 0 and DL-H 16340 5 0 -CCTGAAGTAGGAACCAGATG-3 0 [36]. PCR mix reaction was performed with 2.5 μL of BSA, 0.85 μL of MgCl2, 0.5 μL of dNTPs, 0.3 μL of each primer (Thr-L and DL-H), 0.2 μL of Taq and 12.88 μL of double-distilled H2O, and then 5μL extracted DNA was added to the mix. Reaction mixtures were initially denatured at 94˚C for 3 min, followed by 42 amplification cycles (94˚C for 1 min; annealing for 2 min at 50˚and extension for 1,5 min at 72˚C) and a final extension step at 72˚C for 10min (adapted from [38], [39]). Samples were visualized by electrophoresis on 1.4% agarose gel. Mitochondrial fragments were purified using ExoSap-IT1 (USB Corporation) and sent to sequence in both directions using sequencers ABIPRISM1 3730-XL DNA Analyser from Applied BiosystemsTM. Sequences were then manually aligned using MEGA version 6.0 [40] and compared with previously published Canis lupus signatus sequences [37]. Since some scats were older, we were only able to extract DNA for wolf confirmation from 50% of them. Of that 50%, only 2.2% were excluded from belonging to red fox.
Regarding hair identification procedures, we followed Teenrik et al [41], De Marinis & Asprea [42] and Valente et al [43] protocols in order to prepare the hair slides for further identification. Wolf scats were firstly washed with water and examined macroscopically in order to differentiate hair from bones, feathers, vegetable and mineral material, insects and garbage. After drying, hair slides were prepared, and the consumed prey items were identified through microscopic examination of their cuticular pattern, medulla and cross-section [6].

Wolf diet analysis
To evaluate wolf diet based on hair identification, we used frequency of occurrence (FO) as a measure to quantify prey items. FO is considered the most common method used in diet analysis; however, it may overestimate the frequency of difference size preys, taking into account their ratio surface/volume (preys either with bigger body mass or small ones) [6], [44], [45]. For each prey item, FO was categorized according to Ruprecht [46]: basic food (�20%); regular food (5-20%); supplementary food (1-5%) and sporadic food (�1%). A χ 2 -test (significance level ρ�0.05) was performed to assess differences in the FO of each prey item. The analyses were performed using the R software [47].
FO was then converted into the percentage of biomass of prey consumed, using the linear regression of Floyd et al [41], modified by Weaver [48]: Where y = biomass consumed per scat x = average weight (kg) of each prey identified in the scats. Each y was then multiplied by the actual number of scats containing each prey type in order to estimate the total amount of biomass for each prey class.
Levins' index of niche breadth [55] was calculated according to the following formula: Where B = Levins' measure of niche breadth p j = proportion of prey items from food category j Levins' measure was then standardized on a scale of 0 (specialist predator-strong specialization in one group of prey) to 1 (generalist predator-opportunistic preying on all groups of preys), according to Hurlbert's formula [56]: Where B A = Levins' standardized Food Niche Breadth B = Levins' Food Niche Breadth n = number of prey items found in the diet Shannon's Diversity Index (H') was calculated in order to obtained diet evenness [57], [58]: Where p j = proportion of prey species i in the diet Diet evenness was standardized on a scale of 0 (uneven) to 1 (complete evenness), following Shannon equitability diversity (Eh) formula: Where Eh = Shannon equitability diversity H' = Shannon's Diversity Index H'max = ln(S) where S represents the total number of prey items The last index calculated was Ivlev's electivity index (D) [59], modified by Jacobs [60], to measure wolf prey preference from -1 (total avoidance of a species) through 0 (no selection) to +1 (maximum positive selection): , using the FO of the prey items found in their study for further comparison with our indices. Ivlev's electivity index (D) was not possible to estimate because we did not have access to prey density data for both livestock/wild species in the previous study.
Seasonal variation of the main prey species found in wolf diet was calculated using a χ 2 -test, defining a significance level of ρ �0.05. We used the χ 2 -test to understand if (1) the FO is different for each season (Autumn, Winter, Spring and Summer); and (2) the FO of each individual prey item (red deer, roe deer, wild boar, domestic/wild cat, stone marten, domestic goat and small mammals) are seasonally dependent.

Results
Between November 2017 and August 2019, a total of 85 wolf scats were collected throughout the study area. Seven different prey items (wild boar, red deer, roe deer, domestic goat, domestic/wild cat, stone marten and small mammals) were identified (see supplementary information). Fig 2 shows diet composition expressed in FO of prey remains in scats. In total, 87% of all analysed scats presented only one prey item, whereas in 13% were found two prey items. According to the FO, roe deer was the most consumed prey (44%), followed by red deer (26%) and wild boar (24%), categorized as basic food items on wolf's diet. Domestic/wild cat (6%), domestic goat and stone marten (5%) were also found in wolf's diet, although in lower percentage as regular food items, while small mammals were the least consumed prey (4%), being categorized as supplementary food item [46]. The χ 2 -test showed significant differences in the FO of each prey item in the total analysed samples (χ 2 = 76.38, df = 6, p<0.001).
When considering prey consumed biomass, red deer was the most consumed prey (27.3%), followed by roe deer (23.3%) and then wild boar (19.5%) (Fig 3). For both FO and consumed biomass, wild ungulates were the most consumed prey of wolf's diet (83%), and both domestic goat and all the other wild preys found on the scats represent only a small fraction of its diet. Niche breadth index for this study, estimated by the standardized Levin's index was closest to zero (B = 0.35), indicating a tendency towards a more specialist feeding habit. Shannon equitability diversity indicated a higher evenness, suggesting that each identified prey item is almost equally consumed by the Iberian wolf (Eh = 0.83). On the other hand, Levin's index for Paixão de Magalhães and Petrucci-Fonseca [26] study was closest to 1 (B = 0.69), indicating a tendency towards a more generalist feeding habit, while Shannon equitability diversity indicated a higher evenness, practically equal to the one found in the present study (Eh = 0.85).
Ivlev's electivity index calculated for this study (Fig 5) showed a higher selection towards roe deer (D = 0.71), and almost no selection towards red deer (D = 0.05). Domestic goat was negatively selected, taking into account their availability in the study area, meaning that it is consumed less than expected by their availability (D = -0.21). Considering sheep, cattle and domestic pig availability and densities in the study area and given that these species were never identified in the analysed samples, Ivlev's index showed that wolf never select any of these species (D = -1.00).
Seasonal variation, expressed in FO, demonstrated that wild ungulates represent the main prey items on wolf's diet during all seasons, with values ranging from 68% in Summer and 89% in Spring (Fig 6). All seven prey items found in the collected scats were only consumed in Spring, while Autumn was the season with less prey diversity. Roe deer was the most consumed prey in both Spring and Autumn (51% and 38%, respectively), being equally consumed as the red deer in Summer (26%) and slightly less consumed than the red deer (27%) in Winter (23%). χ 2 -test analysis revealed that (1) the FO is significantly different for Winter (χ 2 = 12.91, df = 5, p = 0.02) and Spring (χ 2 = 54.51, df = 6, p<0.001) but not for Autumn and Summer; and (2) the only prey dependent of the seasons was the roe deer (χ 2 = 16.95, df = 3, p<0.001).

Discussion
In the Iberian Peninsula, the Iberian wolf shows a large spectrum of diet (livestock [6], [ and in our study, where it is observed a clear shift on wolf's feeding habits over the last decades, from mainly livestock (e.g. sheep (28.8%), domestic goat (19.2%)) to wild ungulates. At that time, low densities of roe deer were reported, and the red deer population was just starting to settle due to the natural expansion from the border   Spanish population [62]. Other studies in Portugal showed that livestock make up most of the wolf diet, even at different rates. For example, Vos [16] found a prevalence of 97.5% for domestic goat in south of Douro river population, but in Peneda-Gerês (northern Portugal), goat made up 58.7% of wolf diet. Roque et al [27] and Á lvares [12] showed that in the northwest population, livestock was the most consumed prey (84.7% and 81.7%, respectively), while wild ungulates were far less consumed (7.2% and 10.8%, respectively). The same trend was reported by Torres et al [6] for wolf packs located at south of Douro river. (>94% of livestock for the three packs evaluated). So, based on previous studies, wolves' diet was mostly focused on livestock and, to the best of our knowledge, this is the first study in Portugal where wolves feed mainly on wild ungulates.  [63]. Cuesta et al [33] found in three out of five analysed areas, which cover most of the Iberian wolf's Spanish distribution that wolves feed mostly upon wild ungulates. Similar results were found by Barja [34] were roe deer was considered the most important prey species in Iberian wolf diet in north-western Spain (see also Lagos & Bárcena [61]).
Seasonal variation showed significant differences for Winter and Spring season, and roe deer was the only prey significantly dependent of the seasons. These results can be explained due to spring births of the roe deer litter, which are an easy prey for the Iberian wolf [64]. Likewise, Paixão de Magalhães and Petrucci-Fonseca [26] reported higher FO of roe deer in spring due new-born birth. High availability and predation of juvenile roe deer, particularly in Spring and Summer, was also described by Barja [34], which is critical during the cub's period, where a higher demand of energy is required.
Several studies in central and northern Europe have already reported a preference of wolf towards wild ungulates, due to higher density and diversity of wild preys [7], [22], contrasting with some southern Europe countries [6], [17]. Our results show that the larger percentage of consumed biomass was from red deer (27.3%), which is consistent with other European studies, although they found higher percentages: 42% [18], 43% [19], 59% [63], and 59.9% [65]. Likewise, other studies performed in northern Europe reported higher predation rates of roe deer (53.3%, [22]) and wild boar (35.6%, [21]), and that happened even in north Italy (wild boar, 58.9%, [45] and red deer, 28% [66]). According to Huggard [67], prey selectivity is dependent on habitat overlap, vulnerability and probability of prey occurrence and encounter rates between predator and prey. Across Europe, studies showed that roe deer was positively selected, and both red deer and wild boar were negatively selected [19], [22], which is consistent with our results. Wild boar is generally avoided by wolves in Europe [68], however can be positively selected if there is lower abundance of both red and roe deer [18], [69]. Although wolves main prey items are ungulates (either wild or domestic), they are considered generalist carnivores, having a significant diet breadth and feeding from different species throughout their range [70], [71]. Our results show that wolves tend to have a more specialist diet towards wild ungulates, however Shannon's diversity index indicates evenness, meaning that each prey item found in our study have a tendency to be equally consumed by wolves. Wolf selection of a given prey may be dependent not only on the relative abundance of that species, but also on alternative prey availability [66]. Livestock species are the most consumed item by the Portuguese wolf packs, which can be explained by the low abundance and diversity of wild ungulates, either at south of Douro River packs [6], [16], and at northwest packs [12], [16], [27]. In Paixão de Magalhães and Petrucci-Fonseca [26] study, low abundance and diversity of wild ungulates, lead wolves to acquire a more generalist feeding habitat, and Shannon's diversity index also indicates evenness on prey consumption. Since 1978 [26], besides the increasing wild prey availability in MNP, improving husbandry practices play a major role in explaining the current lower rates of livestock predation found in our study area, since shepherds and guarding dogs are always seen during daylight, escorting cattle and small ruminants, which is considered one of the most effective ways of protecting livestock from wolf predation [17]. Although semi-confined farming systems are dominant in our study area, guarding dogs are essential, especially at night, in order to reduce wolf attacks and costs of compensation schemes [71], but also human-wildlife conflicts [72]. Our study shows a significant domestic goat avoidance, which may be related with improved livestock protection measures, which consequently reduces the predation. Pimenta et al [73] predicted wolf predation probability for our study area, where sheep was the livestock species with higher predation risk, while goat was the second and donkeys the third. However, Pimenta et al [73] results are correlated with livestock densities estimated for our study area, where sheep was the species with higher density (9.211 ind/100 ha) (DGAV, 2017, personal communication). Our results did not support this hypothesis.
Stone marten and domestic/wild cats were found with the same frequency of occurrence of domestic goat on wolf's diet. Carnivore's consumption by wolves can be correlated with interspecific interactions, which may be associated with spatial and trophic competition but also as an important additional food source [12], [74]. It is important to stress that scat analysis, through hair identification, only reveals what wolves ate and that does not necessarily have to correspond to what they killed, since scavenging events can occur. Yet, as the wolf is the only top predator in the study area, there are no reasons to believe that wild ungulate presence in the wolf diet does not reflect wolf predation.

Conclusion
Our results showed that Iberian wolf diet in northeast Portugal is mainly composed by wild ungulates, contrasting with studies performed in other Portuguese packs. These results can be explained by both a decrease in the number of head of cattle since the 80's and 90's and an increase of wild ungulates densities in our study area in the last decades, in comparison with other areas in Portugal. The presence of wild preys can be envisioned as the most welcome step towards the mitigation of human-predator conflicts, contributing to the conservation of the Iberian wolf in Portugal.