POLLINATION SYSTEMS AND FLORAL TRAITS IN CERRADO WOODY SPECIES OF THE UPPER TAQUARI REGION ( CENTRAL BRAZIL ) MARTINS

Plant species present flowers with varied morphological and functional features, which may be associated to pollination systems, including species pollinated by wind, beetles, moths, bees, small insects, birds, or bats. We calculated the frequencies of the pollination systems among woody species in five cerrado fragments in central-western Brazil and tested whether the pollination systems were indeed related to floral traits. We sampled 2,280 individuals, belonging to 121 species, ninety-nine of which were described in relation to all floral traits. Most species had diurnal anthesis, pale colors, and open flowers. The most frequent groups were those composed by the species pollinated by bees, small insects, and moths. A Principal Component Analysis of the species and floral traits showed that there was a grouping among species with some pollination systems, such as those pollinated mainly by beetles, moths, birds, and bats, for which inferences based on the floral traits are recommended in cerrado sites. For the species pollinated mainly by bees or small insects, inferences based on the floral traits are not recommended, due to the large dispersion of the species scores and overlapping between these two groups, which probably occurred due to the specificity absence in plant-pollinator relationships.


InTROdUCTIOn
Savannas are tropical and subtropical formations where the grass layer is almost continuous, interrupted only by shrubs and trees in varying proportions, and where the main growth patterns are closely associated with alternating wet and dry seasons (Bourlière & Hadley, 1983).The Cerrado domain formerly occupied 2 million km 2 of the Brazilian territory (Ratter et al., 1997), especially in the Central Plateau.As its name implies, in the Cerrado domain, the cerrado vegetation prevails.The cerrado vegetation is not uniform in physiognomy (Coutinho, 1990), ranging from grassland to tall woodland, but with most of its physiognomies within the range defined as tropical savanna (Bourlière & Hadley, 1983).The cerrado vascular flora has an herbaceous and a woody component, which are antagonistic because both are heliophylous (Coutinho, 1990).Scholes & Archer (1997) postulated that the climatic seasonal pattern of tropical savannas, with alternating warm dry and hot wet seasons, provides a potential axis of niche separation by phenology for the herbaceous and woody components.
Plant reproductive processes are determinants of the composition and structure of communities (Bawa, 1990;Oliveira & Gibbs, 2000).Among such processes, the plant-pollinator interactions form a dynamic, yet somewhat cohesive, ecological subunit of a community (Moldenke & Lincoln, 1979).Pollination biology at community level in the Neotropics has been studied in forest areas (Bawa et al., 1985;Kress & Beach, 1994) and also in the cerrado vegetation (Silberbauer-Gottsberger & Gottsberger, 1988;Barbosa, 1997;Oliveira & Gibbs, 2000).Pollination is an important ecological process for vegetation communities since it can directly affect the reproductive success of plants, being able to cause loss of species if affected (Wunderlee, 1997).Pollination usually involves abiotic (water or wind) or biotic (animal) vectors, including from non-specialist insects to animals strictly dependent on flowers for their survival, such as bees, birds, and bats (Proctor et al., 1996;Machado & Oliveira, 2000).
Plant species have flowers with many morphologic and functional features, which can be associated to pollination syndromes (Faegri & Pijl, 1979;Borges, 2000).These syndromes are composed of a set of attributes (e.g., color, odor, shape, reward, and anthesis), which are common or specific to plant species adapted to a certain type of pollinator (Faegri & Pijl, 1979;Waser et al., 1996).Pollination syndromes take into account these set of floral characters, which would allow determining the likeliest group of pollinators in certain species (Bawa et al., 1985).However, the concept of pollination syndromes must be applied with caution (Herrera, 1996), since it presupposes pollinator-plant specialization, which, in general, is rare (Waser et al., 1996).Pollination systems with a high level of specialization are exceptions; they often vary from less specialized to more generalized systems (Waser et al., 1996).Thus, Oliveira & Gibbs (2000) grouped species according to pollination systems or functional guilds.In this sense, pollination service from different types of animals may be seen as a resource that unifies different pollination systems (Oliveira & Gibbs, 2000).
It is important to consider the attractive and floral resources displayed by species to their visitors, since these aspects ensure the presence and the fidelity of pollinators to the flowers (Borges, 2000).Pollinators are able to distinguish floral differences and have preferences for some colors, forms, sizes, and odors (Levin & Anderson, 1970).Cerrado species ensure the attraction and permanence of pollinators in their environment, offering resources in morphologically different flowers (Borges, 2000).This attraction is reinforced in some species by the emission of odors, and in other ones, by morphologic characteristics, so that these features act as guides of resources (Borges, 2000).Bees are considered better-adapted animals to pollination (Faegri & Pijl, 1979) and are the main pollinators of tropical areas (Bawa, 1990;Ramirez & Brito, 1992).
In the cerrado vegetation, bees are also the main pollinators (Silberbauer- Gottsberger & Gottsberger, 1988;Barbosa, 1997;Borges, 2000;Oliveira & Gibbs, 2000).Nevertheless, other pollinators are also important for the cerrado flora, since most species rely on a wide spectrum of pollinators, defined more by their size and foraging requirements than by specific interaction (Oliveira & Gibbs, 2002).Many species have small, apparently generalist flowers, pollinated by a range of insects of different groups, such as flies, bees, and wasps (Oliveira & Gibbs, 2002).As in other tropical communities, plant-pollinator relationships in cerrado seem to involve guilds of pollinators associated with a given plant or group of plants (Bawa, 1990;Oliveira & Gibbs, 2000).Studies on the reproductive biology of cerrado plants have shown a great diversity of pollination systems, which are similar to those found in Neotropical forests (Oliveira & Gibbs, 2002), but is it possible to infer the main pollination system of cerrado woody species based on floral traits?We studied the woody component of some cerrado fragments in central Brazil, attempting to answer the following questions: What are the main pollination systems among these species?What are their floral traits?Do species with a given pollination system form distinct groups when all their floral traits are taken into account?Which floral traits are related to a given pollination system?

MATERIAl And METhOdS
We carried out this study in 2003 in Alcinópolis (Mato Grosso do Sul State), Alto Araguaia and Alto Taquari (Mato Grosso State), and Mineiros and Santa Rita do Araguaia (Goiás State), central-western Brazil, in the southwestern extremity of the Brazilian Central Plateau.The regional climate is Aw (Köppen, 1948), humid tropical with a wet summer and dry winter.Annual rainfall varies from 1,200 to 2,000 mm, concentrated from October to March, and the mean annual temperature is around 24.6 °C (Ramos-Neto & Pivello, 2000).This region was originally covered mainly by cerrado vegetation, from open (campo limpo, a grassland savanna) to closed (cerradão, a tall woodland) physiognomies, following Coutinho's (1990) classification.
We randomly placed 38 transects, perpendicular to the edge, in the cerrado sensu stricto physiognomy of the five fragments: eight in the first and fourth, four in the second, twelve in the third; and six in the fifth.In each transect, we placed 15 sampling points, 10 m apart one from the other, starting at 10 m from the fragment edge.We used the point-quarter method (Mueller-Dombois & Ellenberg, 1974) to sample the woody component.At each point, we sampled four woody plants with a stem diameter at soil level equal to or larger than 3 cm (SMA, 1997).Thus, in the 38 transects, we placed 570 points and sampled 2,280 individuals.We collected the sampled individuals and identified them by comparison with lodged material at the São Paulo Botanical Institute (SP), by comparison with the Emas National Park reference collection (Batalha & Martins, 2002), or by using a key based on vegetative characters (Batalha & Mantovani, 1999).
We classified species into families following Judd et al. (1999) and determined their floral traits based on the literature (e.g., Barbosa, 1997;Borges, 2000;Oliveira & Gibbs, 2000;Wanderley et al., 2002;Wanderley et al., 2003).We classified flowers according to their morphologic features (color and shape), functional features (time of anthesis, presence of odor, and presence of resource guides), and reward offered to pollinators (nectar and pollen).We considered the classes "open", "closed" and "tubular" for flower shape and "white", "cream", "green", "lilac", "red", "orange" and "yellow" for flower color.We classified those species for which we determined all floral traits into pollination systems based on the literature (e.g., Silberbauer-Gottsberger & Gottsberger, 1988;Barbosa, 1997;Borges, 2000;Oliveira & Gibbs, 2000).Occasionally, when we could not find data for a given species, we classified its pollination system based on data for related species of the same genus.We recognized the following pollination systems: wind, beetles, moths, bees, small insects, hummingbirds, and bats, calculating the frequency of each system among the sampled individuals.The 'small insects' class included a varied assemblage of relatively small insects, such as small bees, beetles, flies, wasps, and butterflies (Bawa et al., 1985).
We constructed a matrix of the floral traits in relation to the species, excluding those species for which we did not find information for all floral traits.For the two nominal variables -flower shape and flower color, we created dummy variables, which had value 1 when a given trait was present and value 0 when it was absent (Jongman et al., 1995); for example, the dummy variable for tubular flowers had value 1 when the flower was tubular and value 0 when the flower was closed or open.To test for relationships between pollination systems and floral traits, we used a Principal Components Analyses (PCA) (Jongman et al., 1995) with the Multivariate Statistical Package software (Kovach, 1999).

RESUlTS
In the 2,280 sampled individuals, we found 121 species, and for 99 of which we obtained information for all floral traits (Table 1).These 99 species comprised 2,085 individuals and 30 families.The most representative families were Myrtaceae (18 species), Fabaceae (17), and Apocynaceae, Malpighiaceae, and Melastomataceae, each one with five species.Most species had flowers with diurnal anthesis (79%) and pale colors (89%), that is, white, cream, yellow, or green.In relation to the shape, 68% of the species had open flowers; 24%, tubular flowers; and 8%, closed flowers.Pollen was a resource present in 65% of the species; and nectar, in 54%.Of the 99 species, 55.6% were mainly pollinated by bees; 20.2%, by small insects; 13.1%, by moths; 5.0%, by bats; 3.0%, by beetles; 2.0%, by hummingbirds; and 1.0%, by wind.
In the PCA, the eigenvalues of the first two axes were 0.73 and 0.37, explaining 46.8% of the variation in the data.In the ordination diagram (Fig. 1), the floral traits with longer vectors were shape, reward, anthesis, and, to a lesser extent, odor.Species pollinated by moths formed a distinct group, related to nocturnal anthesis and tubular flowers.These two characters, together with the presence of odor, were also related to the species pollinated by beetles and bats.Species pollinated by birds presented positive scores in the first axis and negatives scores in the second one, being related to tubular flowers, diurnal anthesis, presence of nectar, and showy colors.The only species we found pollinated by wind presented negative scores in both axes, with green, diurnal, and open flower.Species pollinated by bees and small insects were dispersed throughout both axes, one considerably overlapping with the other.

dISCUSSIOn
All plant species have particular morphological and physiological features that can attract some groups of floral visitors to the detriment of others (Bosch et al., 1997).Many morphological and functional aspects of floral biology are important for the establishment of plant-animal interaction (Faegri & Pijl, 1979;Waser, 1983).Shape and odor are floral traits that have been included in community studies, because they have important implications not only for the plant-pollinator relationship, but also for the reproductive success of the plant (Endress, 1994).Flower color was particularly good at separating bird-and hymenopteran-flowers for Pentstemon species (Wilson et al., 2004).For us, however, color had a minor importance in separating pollination systems, since the two largest pollinator classes, bees and small insects, visit flowers of several colors.The exception was white, which had a relatively long vector, and was associated with moths, beetles, and bats in the second axis.We observed that the floral traits with longer vectors and, thus, best related to the pollination systems were flower shape, reward, anthesis, and odor.
Generally, the relationships between floral traits and pollination systems we found were similar to those described in other studies (e.g., Faegri & Pijl, 1979;Silberbauer-Gottsberger & Gottsberger, 1988;Bawa, 1990;Barbosa, 1997;Oliveira & Gibbs, 2000).Flowers pollinated mainly by beetles were yellow, closed, nocturnal, and with odor; flowers pollinated mainly by birds were orange, tubular, diurnal, and with nectar; flowers pollinated mainly by bats were white, nocturnal, and with odor; and flowers pollinated mainly by moths were pale, tubular, nocturnal, and with odor and nectar.Most of the flowers pollinated mainly by bees were open and yellow, white, cream, or lilac.Such colors, as well as the sweet and pleasant odor emitted by the flowers during the anthesis, are attributes related to the attraction of bees, since they are sensitive to olfactory and visual stimuli (Faegri & Pijl, 1979;Kevan & Baker, 1983).

Floral traits and pollination systems of woody species in cerrado fragments located in Upper Taquari region (central-western Brazil). Family names were abbreviated according to Weber (1982). Pol -pollination systems; CO -color; A -anthesis
Flowers pollinated mainly by small insects were small, usually open, and green or yellow.On the one hand, these traits imply floral rewards that are accessible to many insects; on the other hand, a certain insect from such a group can visit flowers of several species, since there is a lot of species with these traits (Bawa & Opler, 1975;Bawa, 1980).Flowers pollinated by this group are generally small (less than 1 cm in length), pale green or pale yellow, and lack morphological specialization with the result that floral rewards are accessible to a wide variety of insects (Bawa et al., 1985).We found a large overlapping between the two most frequent groups, bees and small insects, which probably occurred due to the absence of specificity, which is the most frequent condition in different vegetation types (Bawa, 1980), including the cerrado (Borges, 2000).In tropical forests, many species have morphologically simple flowers, allowing the access of different categories of visitors, such as bees, butterflies, moths, flies, and wasps (Bawa, 1990).Even species with more complex flowers are visited by several species that act as pollinators (Bawa, 1990).
In the cerrado, bird-and wind-pollinated flowers are rare (Silberbauer- Gottsberger & Gottsberger, 1988;Oliveira, 1991) and were the groups with lowest frequencies in the area we studied as well.Although hummingbirds are important pollinators of herbs in open cerrado areas (Silberbauer-Gottsberger & Gottsberger, 1988;Barbosa, 1997), they seem to be only secondary or opportunistic visitors of cerrado woody species (Oliveira & Gibbs, 2000).The only species pollinated by wind that we sampled had inconspicuous flowers, without odor and rewards, corroborating Faegri & Pijl (1979), who related these traits to abiotic pollination.Wind pollination, which has been associated with seasonally dry areas (Bullock, 1994), is rare among cerrado woody species and occurs commonly only among grasses and sedges (Silberbauer- Gottsberger & Gottsberger, 1988;Oliveira & Gibbs, 2002).We did not find butterfly-pollinated flowers, which are poorly represented among cerrado woody species (Oliveira & Gibbs, 2000;2002).
Information from studies on pollination in cerrado sites (Silberbauer-Gottsberger & Gottsberger, 1988;Oliveira, 1991;Oliveira & Gibbs, 2000;2002) pointed out the existence of several pollination systems, suggesting that plant species depend on the interaction with several groups of animals, and vice-versa.Our ordination analysis of floral traits and plant species showed that there was a grouping among species with some pollination systems (beetles, moths, birds, and bats), for which inferences based on the floral characters are recommended for cerrado sites.In these cases, one could infer the pollination system based on the floral traits.On the contrary, regarding the species pollinated mainly by bees or small insects, inferences based on the floral traits are not recommended, due to the large dispersion of the species scores and overlapping between these two classes, which occurred, probably, due to the specificity absence in the plant-pollinator relationships.