The study area, located in the center of Tierra del Fuego province (TF), Patagonia Argentina (54º 40’ S, 67º 35’ W, Figure 1 ), consists of three typical paddocks situated in landscapes that alternate undulated areas with valleys. One paddock was located in “Estancia María Cristina” (MC), and two in “Corazón de la Isla Reserve” (Reserve 1 and Reserve 2, R1 and R2, respectively, Figure 1 and Table 1 ). Their landscapes are composed of lenga beech forests with some very scattered ñire trees, surrounded by adjacent wet vegas in which herbs and grasses of the Cyperaceae, Juncaceae, and Poaceae families prevail. The regional climate is characterized by short, cool summers, and long, snowy, and frozen winters. Annual average wind speed outside the forests is 8 km h ^-1 , reaching up to 100 km h ^-1 during storms ( Martínez Pastur et al ., 2009 ). Average annual temperature is 4.5 °C; annual precipitation averages 467 mm and, in the form of either rain or snow, is evenly distributed along the year ( Pisano, 1977 ). In the three paddocks, the forest had been used for more than 50 years for timber production based on selective cut of the best lenga beech trees (high-grading). At present and after livestock exclusion (at least 10 years ago), natural populations of guanaco are the only ungulates grazing in these sites.

In each paddock and within their respective forest area, we selected a representative sub-area (~ 15 ha) to sample guanacos’ feces. Although guanacos may feed in the vegas or inside the forest, they showed a particular feces deposition behavior, concentrated in specific places within the Nothofagus spp. forests [named dung-piling, (Franklin, 1982)], and rarely in vegas ( Soler Esteban et al ., 2013 ). In each site, fresh fecal samples (recognized because they presented a typical greenish color as compared to older feces), were collected during spring (November 2009) and summer (February 2010). Feces collection was done from five active dung-pilings distant at least 50 m one to each other, mixed together and put in plastic bags totalizing 100 to 150 g each. Spring and summer feces were then considered as independent samples.

The botanical identification of plant tissues of reference plants collected in the three study sites was done by determining epidermal and non-epidermal plant fragments based on the micro-histological analysis method ( Williams, 1969 ) adapted for Patagonian plants by Latour & Pelliza Sbriller (1981) . At the same time of feces collection, we also sampled plants of all taxa present in either the vegas or in the forest understory. The analysis involved drying the plant material and feces at 60 ºC for 48 h, milled in an analytical mill (IKA A11) until reaching particle sizes of about 1 mm, and mounted on microscope slides for further identification. This microscopic identification allowed performing species-specific patterns of plant (items) for comparison with those found in feces. Microscopic observations of fecal samples showed different frequencies of species epidermis, so those results were expressed as percentage values of the items, at genera or species levels. Fecal samples were mounted on five microscope slides and 20 fields per slide were examined using 100 × magnifications. These analyses allowed grouping the results of these observations into six vegetation groups: trees, grasses, herbs, grass-like, shrubs, and lichens. Quantification of items composing each diet was done according to Holechek & Vavra (1981) , and Holechek & Gross (1982) . All analyses were conducted at the Botany laboratory, Universidad Nacional de la Patagonia “San Juan Bosco”, in Comodoro Rivadavia (Chubut, Argentina).

To determine biomass availability in the forest understory, we placed 3 transects per paddock. The transects, 320 m long each, were located in parallel and distant about 300 m one to the other, starting from the edge of each vega toward the interior of the forest. Along each transect, 1 m ² quadrats were placed at 10, 20, 40, 80, 160 and 320 m. The understory biomass inside each quadrat was cut at 2 cm aboveground, collected, separated into grasses, herbs, shrubs and trees, put in plastic bags, taken to the lab, oven-dried at 75 °C for 48 h, and weighed. Biomass was expressed as kg DM ha ^-1 , and was harvested in spring and summer simultaneously with feces collection.

Diet selection was analyzed considering aboveground biomass present in the forest understory during spring and summer, and used as an estimation of food availability. This was performed according to Ivlev’s selectivity index modified by Jacobs, 1974 ( Lechowicz, 1982 ), as:

The assumption of normality (Shapiro-Wilks) was investigated for the variables species and groups of species percentage in the diet. Mean differences of dietary composition between seasons was performed by using the Student T test (p<0.05) for paired samples, with n=3.

Linear mixed-effect models were used to analyze the effects of treatments: Site: MC, R1 and R2, and Season: spring and summer, on the biomass available. Treatments (Site, Season and Site *Season interaction) were considered as a fixed effect, and the variable transect was considered as a block (random effect). The amount of the model variance that was explained by differences among sites was calculated by dividing the random effect variance by the total variance.

The mean differences of D index between seasons was performed by using the Student T test (p<0.05) for paired samples, with n=3.

This test was applied by using the Satterwai correction in the case the variances were not homogeneous (p<0.05). All statistical analyses were performed by using the INFOSTAT Statistical Software ( Di Rienzo et al ., 2015 ).

The percentages of the different groups of plants composing guanacos’ diet during spring and summer, in the three study sites, is shown in Table 2 . In the two seasons studied and considering all sites together, guanacos’ diet presented high percentage of tree lenga beech leaves and twigs (38.7%), followed by grass-like (27.6%) and grass species (27.3%), and showing a low presence of herbs (3%), Lichens (2.1 %) and shrubs (1.2%). The most important items found in the diet, with frequencies higher than 10%, were Nothofagus sp. (mainly lenga beech leaves and shoots), Uncinia lechleriana , Poa sp, Carex sp. and Festuca spp. In spring, 55% of the diet was composed of species only found in the forest understory ( Nothofagus sp., Osmorhiza chilensis , Berberis sp., Gaultheria sp. and lichens), while 30% corresponded to grass-like species only found in vegas. The rest of the diet (15%) corresponded to other grass species found in either the forest understory or in the nearby vega. During the summer, the percentage of Nothofagus spp. in the diet significantly decreased in all three sites (29%, p: 0.03) and increased the grass component (40%, p: 0.003). The grass-like species showed a mean percentage of 25% in the diet in the two seasons analyzed. Of these grass-like species, U. lechleriana marginally decreased (p: 0.05) its frequency during the summer ( Table 2 ).

Total biomass showed significant differences among sites, being higher in R2 in both seasons analyzed and lower during spring in R1 and in summer in MC ( Table 3 ). The higher values found in R2 could be related to the higher lenga seedling and sapling biomass availability as compared to the other sites. The lowest lenga seedling and sapling biomass availability in R1 and MC were registered during the summer season ( Table 3 ). Biomass of all other analyzed groups of species did not show significant differences. However, biomass of grasses presented a maximum value (~160 kg DM ha ^-1 ) in R2 and a minimum (~50 kg DM ha-1) in MC, both during the spring. All other herb species showed a peak of biomass (~115 Kg DM ha ^-1 ) in R2 during the summer and a minimum (~18 kg DM ha ^-1 ) in R1 also during the summer. Shrub biomass was in general scarce (< 1 kg DM ha ^-1 ) in all three sites. In neither case, the seasons considered nor the interaction between season and site did show significant differences ( Table 3 ).

The Ivlev’s index showed that lenga regeneration (seedlings and saplings) were selected or consumed in relation to its availability in all three sites and during both seasons ( Figure 3 ). In spite of their low availability, shrubs were also highly selected. Herbs were rejected in all three sites during both seasons. Grasses, instead, were rejected during spring and consumed according to their availability, or slightly selected, during summer, being significant the consumption difference between seasons (spring and summer) (T: 3.29, p: 0.03, Figure 3 ).

In the particular environments that compose the landscapes of Tierra del Fuego, in which Nothofagus forests are intermixed with vegas, guanacos’ populations may show either migratory or sedentary behavior ( Raedeke, 1982 ; Schiavini et al ., 2009 ; Moraga et al ., 2015 ). Some studies showed that guanacos migrate to lowlands and coastal areas of Tierra del Fuego during cold winter periods, returning to forest areas of higher altitudes in spring and summer ( Raedeke, 1980 ; Bonino & Fernández, 1994 ; Montes et al l., 2000 ), or mainly during the summer ( Moraga et al ., 2015 ). Schiavini et al . (2009), however, found that there are sedentary guanacos populations that remain in the forests and surrounding vegas of Tierra del Fuego all year around. While guanacos in the three studied sites may have shown different migratory behavior, we restricted our study to determine their diet during the relatively warm period comprising from mid spring to mid-summer, when the populations wandered around Nothofagus forests and vegas. During these periods, high amounts of leaves and twigs of seedlings and saplings of Nothofagus spp., and also of grass species (mainly of Poa sp. and Festuca sp.), were consumed by guanacos; grass-like species such as Carex sp. and U. lechleriana completed its diet. The consumption pattern was similar in the three analyzed sites. These results are in accordance with some studies carried out in other areas of Tierra del Fuego (either in Chile or Argentina) ( McNaughton, 1983 ; Bonino &Pelliza Sbriller, 1991 ; Muñoz González, 2008 ; Soler Esteban et al ., 2013 ), and emphasize the importance that Nothofagus spp. leaves and twigs have in guanacos’ diet during the spring in this island ( Soler Esteban et al ., 2011 ). Our results, however, highly contrast with guanacos’ diet determined in other environments outside Tierra del Fuego. In continental Patagonia, for instance, it has been demonstrated that guanaco diet overlap with that of sheep during the summer season, when both species feed on grasses and grass-like species found in meadows ( Baldi et al ., 2004 ). During the winter months, instead, guanaco shifts its preference to shrubs ( Senecio spp., Mulinum spp., Berberis spp.), which comprise about 55 to 80% of its diet during this cold period ( Baldi et al ., 2006 ). In the north-central region of Argentina (in the provinces of Mendoza and Córdoba), several studies revealed that guanacos mainly consume grass species, shifting their preferences to some portions of woody species only when grasses are very scarce ( Puig et al ., 1997 ; Barri et al ., 2014 ). It is interesting to note that this north-central region includes a variety of environments in which the trees Polylepis australis and Maytenus boaria are abundant. While leaves of these species are highly preferred by domestic grazers (cows, goats, and sheep) ( Renison et al ., 2015 ), guanacos prefer to graze on short grasses, avoiding foraging on these and other tree species ( Flores et al ., 2013 ).

Despite the abundant offer of herbaceous species during spring and summer in the areas surveyed in Tierra del Fuego, our results showed some preference of guanacos to browse on Nothofagus spp. leaves and twigs to satisfy part of their nutritional requirements. This particular feeding behavior may be related to the efficiency of gastric digestion that guanacos have ( Marín et al ., 2006 ). In fact, guanacos and other camelids regurgitate and re-chew the forage they eat, and are much more efficient than ruminants in extracting protein and energy from poor quality forages ( San Martin & Bryant, 1989 ; Pinto Jiménez et al ., 2010 ). On the other hand, the corporal mass plays an important role in the foraging strategy ( Gordon, 2003 ). In this regard, it has been proposed that an inverse relationship exists between body size and selectivity, being the guanaco a herbivore with an intermediate selectivity (species from 50 to 200 kg of body weight ( Baldi et al ., 2004 ).

The analysis of diet selectivity of guanacos, taking into consideration exclusively species grown in the forest understory (excluding grass-like species), showed that they prefer to consume woody vegetation (leaves and twigs of lenga seedlings and saplings) and grasses only during the summer, avoiding selecting grasses during the spring, and herbs during either spring or summer. Other studies carried out in Tierra del Fuego emphasized the high frequency of lenga leaves and twigs found in the diet of guanaco, although neglected the importance of shrub species in that diet ( Martínez Pastur et al ., 2010 ; Soler Esteban et al ., 2011 ; Arias et al ., 2015 ). In our study, however, although we recorded little abundance of shrubs in the forest understory and that these species were infrequent in the diet, the Ivlev's index indicated that shrubs are highly preferred should they are available.

On the other hand, although the availability of herbs was relatively high related to the total available biomass inside the forest (and much higher if we consider the adjacent vegas), they were rejected by the guanacos, coinciding with the results presented by other studies ( Soler Esteban et al ., 2011 ; Muñoz & Simonetti, 2013 ). Grasses, instead, although rejected during spring, they were slightly selected during the summer and consumed according to their availability. Contrasting with our results, Puig et al . (2011) indicated that in steppe zones of Central Argentina, grasses and graminoids species were preferred by free ranging guanacos, and that woody species were rejected. Furthermore, these authors reported that in these steppe zones, wetland meadows constitute a focal sector for guanacos foraging. Based on the classification proposed by Hofmann (1989) the guanaco could be then defined as an intermediate herbivore, or opportunistic (mixed) feeder. This is so because it forages on a highly diverse range of food sources ( González et al ., 2006a ), having developed physiological and morphological adaptations for including in its diet a high proportion of woody species ( Searle & Shipley, 2008 ). In continental Patagonia, the negative impacts of understory grazing by domestic and wild ungulates has been considered as one of the most important disturbances that affect Nothofagus forests regeneration ( Veblen et al ., 1992 ; Relva & Veblen, 1998 ; Raffaele et al ., 2011 ; Barrios-Garcia et al ., 2012 ). In this region, guanacos avoid grazing on Nothofagus forests. In Tierra del Fuego, instead, guanacos select feeding areas within Nothofagus forest in spring and summer, despite these are seasons of great abundance of other forages outside the forest. The significant consumption of lenga could adversely affect the development of regeneration of the species, and this aspect should be carefully considered in planning and implementing management plans for sustainable regeneration and growth of Nothofagus forests. In order to cast light on this issue, it is necessary that future research should be focused in determining vegetation dynamics comparing grazed and ungrazed lenga forests in Tierra del Fuego.