Lianas as a food resource for brown howlers ( Alouatta guariba ) and southern muriquis ( Brachyteles arachnoides ) in a forest fragment

Lianas, woody vines, are abundant and diverse in tropical forests, but their relative contribution as a source of food for herbivores has been neglected. I compared feeding rates on lianas and trees of two sympatric primates, A. guariba and B. arachnoides, in Southeastern Brazil. Availability of liana foods was gathered in parallel with primate behavioral data collection. Liana represented 33.9% and 27.3% of food sources for A. guariba and B. arachnoides, respectively. Foods coming from trees, rather than from lianas, were significantly more consumed by B. arachnoides. However, both species took advantage of the continuously renewable and ephemeral food resources provided by liana. Availability of liana flowers correlated positively with A. guariba feeding proportions. The nutritional supply provided by lianas is apparently beneficial, or at least unharmful, but experiments comparing primate choices in forests with different liana abundances will help to shed light on their possible negative effect on communities


Introduction
Lianas are woody climbing vines (sensus Gentry, 1991) that rely on other plants for support.They en� compass 25% of species diversity in tropical systems (Schnitzer & Bongers, 2002).There is a growing body of evidence pointing to an increase in abundance of lianas in forests (Phillips et al., 2002;Wright et al., 2004;Swaine & Grace, 2007).Liana diversity and abundance has been shown to increase following disturbance in comparisons between forest edges and interiors (Oliveira-Filho et al., 1997;Laurance et al., 2001) and between secondary and old growth forests (DeWalt et al., 2000).By producing many rooting stems, lianas are able to rapidly colonize disturbed areas, thereby increasing their chances of survival (Schnitzer & Bongers, 2002).
It has been suggested that lianas are important to forest-dwelling animals because they provide food resources and pathways (Emmons & Gentry, 1983).However, investigations reporting on the relative contribution of lianas as a food source in comparison to other life forms are lacking.Because Neotropical primates have undergone a broad adaptive radiation towards arboreal lifestyle and folivorous / frugivorous habits (Cowlishaw & Dunbar, 2000), it would be expected that lianas would play a substantial role in their feeding preferences.The protein/fiber ratio and tannin content, for instance, are determinants of food choices of predominantly folivorous primates (Milton, 1979;Ganzhorn, 1992).As they rapidly colonize for� est gaps, lianas might be more effective providers of young leaves with favorable protein/fiber ratio and lower concentration of digestion-inhibitors when compared to trees.Also, since the production of foods offered by lianas fluctuates in a seasonal pattern (Putz & Windsor, 1987;Opler et al., 1991;Morellato & Leitão-Filho, 1996) and lianas reproduce during periods unfavorable to trees (Morellato & Leitão-Filho, 1996), it might be advantageous for primates to rely on food items provided by woody climbers.Whether higher rates of consumption of liana products lead to an increase in primate population size is an unresolved question because of the scarcity of data.One of the few studies on the effects of liana abundance, carried out in African forests, found evidence of a positive association between Colobus guereza densities and lianas in the larger size classes (Preece, 2006).
The brown howler (Alouatta guariba) and the south� ern muriqui (Brachyteles arachnoides) are primate species of the Atelidae family (Groves, 2001) living sympatrically in some Atlantic Forest fragments in Southeastern Brazil.The contribution of lianas to the overall diet of these two species is rarely mentioned in studies of their feeding habits.Thus, current available information is mainly anecdotal.After finding that 41% of leaves consumed by a group of A. guariba came from lianas, Chiarello (1994) suggested that habitat disturbance resulted in the proliferation of lianas, which was beneficial for howlers.Mendes (1989) reported that liana leaves represented 11.3% of the leaf diet of another howler group.B. arachnoides was observed feeding on leaves of two liana species (Torres de Assumpção, 1983).Woody vines made up 37-47% of feeding time of a B. arachnoides group studied by Milton (1984).Data on other congeneric species is also scanty: the liana Forsteronia glabrescens was the second most consumed species by A. caraya in Northern Argentina (Zunino, 1989).In a three-month study, Fonseca (1985) recorded 14% of liana leaves in the diet of B. hypoxanthus.Asensio et al. (2007) suggested that the high population den� (2007) suggested that the high population den� sity and increased rate of depletion of primary food sources are forcing groups of howlers in a Mexican forest fragment to forage on alternative resources, such as vines, lianas, shrubs and herbs.
Howlers and muriquis differ in terms of locomotion.While howlers walk in a predominantly quadrupedal, slow-motion fashion (Mendel, 1976), muriquis prac� tice brachiation, hanging by their long arms and tail to swing among crowns of adjacent trees.Muriquis are therefore able to cover a larger area in compari� son to howlers, and within the same period of time they can visit a higher number of widely dispersed food patches, such as flowering and fruiting trees.I investigated the feeding ecology of sympatric groups of A. guariba and B. arachnoides in a forest patch in Southeastern Brazil, where they occur at densities of 27 and 35 ind/km 2 , respectively (Martins, 2005).I compared the species in terms of feeding on resources available from trees and lianas.I also compared them for their intake of liana leaves, flowers and fruit on a monthly basis.I predicted that: 1) because muriquis move faster from one resource patch to another than howlers, they would feed less on lianas, and 2) food resources provided by lianas would be consumed by both species of primates in accordance with their availability.

Study site
The study was carried out at the Fazenda Barreiro Rico (22 o 41' S, 48 o 06' W), a cattle ranch located in the eastern range of the central Plateau, in the state of São Paulo, Southeastern Brazil.Low topographic relief and poor, sandy soils characterize the central Plateau region, which is bordered by the humid Atlantic Forest in the coastal eastern range and the Cerrado domain to the west.I collected data in a 1,450 ha semideciduous forest fragment.Seventy-six species of trees and 21 species of vines were recorded at the site (Assumpção et al., 1982).There is a profu� sion of canopy gaps and considerable abundance of lianas, as 32.8% of 400 phenologically censused trees bore lianas (unpubl.data).This probably re� sulted from many decades of selective logging in this patch.Large, emergent trees still remain in old, regenerated logged areas.The predominant climate is mesothermic.Mean annual rainfall is 1,284.5 mm (data from a local climatological station), with rains falling mostly between January and March.A moder� ate to severe drought (< 70 mm of monthly rainfall) occurs from April to September.

Study groups
Two groups of A. guariba and B. arachnoides (one species each) were habituated to the presence of observers before data collection.These groups were randomly selected in respect to the proportion of trees infested by lianas in their ranging areas.Like other groups observed in the population sur� vey, the ranging area of study groups comprised regions with higher and lower visibility on account of lianas.The group of A. guariba comprised six members: one adult male, one sub-adult male, two adult females, a juvenile male, and an infant male.The non-cohesive, ever-changing social units of B. arachnoides averaged 3.25 ± 1.65 individuals.While the members of the A. guariba group were easily recognized by their natural markings, only 13 adult females and dependent young B. arachnoides could be reliably identified.The muriqui group probably contained 25 to 30 members, a group size estimated by adding the total number of adult females, their offspring, and the largest clump (11) recorded for males on an incomplete day of observation.

Sampling of feeding behavior
From June 2001 to May 2002, observations were carried out on four to five consecutive days per month (from dawn to dusk) for A. guariba and B. arachnoides groups, totaling 40 and 38 full days, respectively.During this year-long period, I col� lected 2,038 feeding records for A. guariba and 2,122 for B. arachnoides.To collect feeding data, I used an instantaneous scan sampling technique: scans lasted one minute followed by an interval of five minutes.The behavior of all the animals that came into view was assigned to one of the following categories: moving, resting, eating, or interacting socially.Whenever individuals were eat� ing, I recorded whether the food source was leaves, flowers, or fruits.I also recorded whether the food source had been obtained from a liana or a tree.Although individual recognition of B. arachnoides was constrained by the absence of natural markings, I took care about not recording the same animal twice per observation session.
Phenology of resources I selected 131 bunches of lianas that were hanging on tree branches.I considered a strict definition of liana, i.e., climbers that germinate on the forest floor and produce true wood (see Gerwing et al., 2006).Due to the difficulty of collecting and identifying lianas, each bunch was considered as a single sampling unit, regardless of how many and which species it was composed of.If more than one bunch was being supported by the tree, one of them was randomly se� lected.The bunch was sampled regardless of whether the lianas climbed the tree directly or came from the crown of an adjacent tree.These bunches of lianas were carefully observed at three-weekly intervals in parallel with sampling of the behavior.Abundance of resources (leaves, flowers and fruits) was evaluated using a scale of degree of availability that ranged from 0 to 4 (0 = total absence, 4 = full crown).To avoid the subjectivity associated with the classifica� tion, I did not distinguish between mature and young leaves, nor between ripe and unripe fruits.An index of availability (IA) was calculated monthly for each food resource using the formula: Sum of degrees of availability IA = x 100 Total degree of availability Percentages of monthly IA for all three food items were calculated.Voucher specimens were collected whenever possible and deposited at the Herbarium of the University of São Paulo.Identification, whenever possible, was carried out to the species level.

Data analyses
Mean annual and monthly percentages of food re� sources in the diet of the primates were calculated using only the feeding records gathered on complete days of observation.I calculated daily proportions of food resources that the primates collected from each source (trees or lianas) by dividing the number of records for that source by the total daily number of feeding records.The same procedure was carried out to calculate the daily proportions of leaves, flowers, and fruits that primates collected from lianas.Feed� ing proportions for a given month were obtained by dividing the sum of proportions by the number of the full days of observation in that month.
Differences between the species relative to the daily proportions of feeding on resources provided by trees and lianas was tested by two-way ANOVA.The data set was previously tested for homogeneity of variances by means of Levene's test (Zar, 1999).Interactions between monthly proportions of feeding on leaves, flowers, and fruits gathered from lianas and availability indexes were tested by Spearman rank correlation test.The software STATISTICA v.5.0 was used to carry out the tests, and the significance level was set at 0.05.

Results
Resources from liana represented 33.9% (691/2,038) of total feeding records for howlers and 27.3% (579/2,122) for muriquis over the 12 month study period.A. guariba and B. arachnoides fed on at least 12 and 10 species of lianas, respectively.Taxonomic determination was possible for only nine species (table 1), mostly because plant specimens grew in inaccessible places high in the tree canopies.
The significant difference between sources (trees and lianas) in the diet of the primates (F = 9.95; P = 0.00) along with the significant interaction (F = 9.73; P = 0.02) between species and sources indicated that foods coming from trees are consid� erably more consumed by B. arachnoides when compared to foods coming from lianas (fig.1).
Difference between the species in terms of relative contribution of the two sources was not significant (F = 0.0; P = 1.0).Availability of liana leaves followed a regular pat� tern (fig.2A), as illustrated by the proximity of the   Availability of flowers and fruits followed seasonal patterns (figs.2C, 2E).Flower production peaked at the beginning of the wet season (September), and availability values were high throughout the first months of the season (October and Novem� ber).There was a small, second peak in May.
High abundance (8-9%) of liana fruits occurred in November-December, dropping sharply after that (fig.2E).Fruit production remained below 4% during the rest of the year.The percentage of total feeding records for a given month varied widely among food items consumed by each species (fig.2).While leaves were the staple food item throughout the year both for howlers and muriquis, flowers and fruits were consumed much less or not consumed at all.On a monthly basis, the percentage of liana leaves in the howlers' diet varied widely in comparison to that in the muriquis' diet.This food item composed a minimum of 7% of the diet of A. guariba, but peaked at 49% in August (fig.2A).Unlike howlers, the maximum leaf consumption of muriquis remained at intermediate levels, such as 28% (fig.2E).Liana leaves were consumed less by B. arachnoides from January to March, following the overall tendency of reduced leaf consumption, probably as a response to higher fruit availability during the wet season.the first peak of availability in September, but not during the second flowering in May (fig.2D).Feed� ing on liana fruits peaked in December (55%) for A. guariba and peaked twice (June and December, both 29%) for B. arachnoides.(figs.2E, 2F).The primate species shared a seasonal pattern of fruit consumption: June-July and December.In these periods both primate species mainly consumed the fruits of two liana species: Pereskia aculeata in the former period and Diclidanthera sp. in the latter period.Both of these plant species were abundant and easily recognized by the unique morphology and color of their fruit.The yellow-colored fruits of Pereskia aculeata bore spines on the epicarp, but these did not seem to be a deterrent to the animals.Fruits of Diclidanthera sp. were juicy, violet drupes.
The monthly patterns of flower and fruit feeding by the two primate species revealed opposite phases: higher relative feeding rates on flowers were recorded during periods in which fruits were eaten less or not eaten at all and vice-versa.Only the availability of liana flowers presented a positive significant correla� tion with consumption by howlers.

Discussion
The mean annual percentage of lianas in the diet of A. guariba (27%) is similar to that of B. arachnoides (33%).However, on a daily basis, the latter rely much more on resources from trees than from lianas.The greater quantity of trees in the muriquis' diet matches expectations as they can reach widely disperse areas by brachiation.It is not yet known whether muriquis would replace resources from trees with those from lianas if the latter became more abundant as the result of habitat disturbance.Lianas may outcompete trees through the combined effects of shading and nega� tive impact on sapling growth (Putz, 1984;Schnitzer et al., 2005).If muriquis are unable to make such a replacement due to constraints of dietary requirements changes in their feeding rates would be expected in the long-term.It appears that A. guariba would consume a greater amount of lianas the greater their abundance, but this would likely not be the case of B. arachnoides unless the tree resources they rely on have their growth, fecundity and/or recruitment rates negatively affected by lianas.On a monthly basis, liana resources were present in the diets of both A. guariba and B. arachnoides.It is known that atelines are flexible in terms of food choice (Chapman, 1987) and high dietary flexibility of both ateline species was evident in this study, illustrated by the consumption of resources from plants with different growth habits.Food choices also showed a temporal pattern of flexibility.The monthly fluctuation of flower and fruit availability favored the shift from one resource to another in consecutive months.Fleshy fruits supply non-structural, easily digested carbohydrates, the major components of the pulp (Waterman, 1984).Flowers eaten by primates provide protein and minerals (Castellanos & Chanin, 1996;Silver et al., 2000).By shifting between these foods, primates can fulfill their requirements at any time of the year.
With the exception of flower consumption by A. guariba, no significant correlations were found between the availability of food resources and the monthly proportion of feeding records for either A. guariba or B. arachnoides.This finding suggests that attributes other than abundance might govern consumption.The howlers' recognized strategy of maximizing protein intake (Milton, 1979) might have favored their reliance on a patchily distributed re� source such as flowers.Flowering trees are perhaps distributed further apart from each other, on average, than lianas, and this would limit the consumption of their flowers by A. guariba, a species that moves in a quadrupedal, slow-motion fashion.The same limitation would be expected for fruit, since they do not fulfill the howlers' high protein requirements as well as flowers (and leaves) do.In the case of B. arachnoides, however, brachiation would not prevent them from opportunistically exploiting liana flowers.Differential nutritional demands and abilities to deal with secondary compounds possibly underlie the distinct findings for howlers and muriquis.However, there may be other explanations.Researchers have failed to find correlations between consumption by primates and availability of food resources (e.g.Chapman, 1987;Maisels et al., 1994;Peres, 1994;Stoner, 1996;Heiduck, 1997;Kaplin et al., 1998).It seems that whenever the pool of surveyed plants harbors few individuals of the species consumed, the temporal variability in the abundance of these species' resources remains hidden within the overall pattern.Indeed, this may be the case for liana spe� cies in the present study.Here, a bunch composed of perhaps two or more species, rather than a single species as usually carried out in tree phenological assessments, was selected as a sample unit.Thus, it is hard to tell whether liana species preferred by howlers and muriquis were poorly represented in the phenological sample, except for those bearing flowers and fruits with specific traits as mentioned above.
It is clear that lianas play an important role in the feeding patterns of howlers and muriquis at the present study site.This may be the result of two environmental factors, rainfall and forest disturbance, either alone or in combination.A high abundance of liana has been reported in areas of Ghana, Africa, where the rainfall is low (Swaine & Grace, 2007) and also in fragmented forests or those woodlots where timber harvesting has prevailed for years (DeWalt et al., 2000).The study site receives less than 1,300 mm of rainfall annually.Moreover, se� lective logging has been carried out for decades at Barreiro Rico, although rates are currently lower.Howlers and muriquis take advantage of continu� ously renewable (leaves) and ephemeral (flowers and fruits) liana resources.This plant group may be contributing to the growth of populations, because both species present moderate population density at the study site.Whether food items from liana have been continuously increasing in the diet of A. guariba and B. arachnoides at the study site as a response to liana abundance is difficult to know, given the lack of long-term monitoring of their feeding choices.
Although the food supplied by lianas is apparently beneficial, or at least unharmful in the setting of the mesic forest and the disturbed nature of the study site, experiments comparing primate choices in forests with different rates of liana infestation would help to shed light on the possible negative effects of such consumption in these communities.

Fig. 1 .
Fig. 1.Mean annual proportions of feeding records on tree and liana by Alouatta guariba and Brachyteles arachnoides.
The highest percentages of liana flowers in the diet of A. guariba and B. arachnoides were 23% and 17%, recorded in February and September, respectively (figs.2C, 2D).The explosive consumption shown by howlers in February, in discordance with availability, would led to a negative relationship had February not been an exception to the rule.On one day in this month, the group twice visited an unidentified species of flowering liana and devoured almost all the flowers.Even so, the percentage of consump� tion varied significantly in accordance with avail� ability (fig.2C). A. guariba showed an opportunistic behavior during the two blooms of liana flowers, in the early wet and in the dry season.In contrast, B. arachnoides fed substantially on liana flowers during

Fig. 2 .
Fig. 2. Monthly percentage of feeding records by Alouatta guariba and Brachyteles arachnoides and percentage of availability index of leaves, flowers, and fruits of lianas.Spearman correlations are between monthly proportions of feeding records on liana foods and availability indexes of lianas: * Significant at 0.05.