INTRODUCTION
Drivers of global environmental change such as habitat fragmentation, overexploitation, species invasions, climate change and pollution have the potential to modify plant-animal interactions (Laurance Reference LAURANCE2004, Turner Reference TURNER1996, Tylianakis et al. Reference TYLIANAKIS, DIDHAM, BASCOMPTE and WARDLE2008). In particular, habitat fragmentation can affect pollination processes because pollinator mobility may be restricted across fragments embedded in the matrix of heavily transformed landscape (Didham et al. Reference DIDHAM, GHAZOUL, STORK and DAVIS1996, Kearns et al. Reference KEARNS, INOUYE and WASER1998, Murcia Reference MURCIA, Schelhas and Greenberg1996). Pollination is the most studied interaction in the context of land-use change in many ecosystems (Aguilar et al. Reference AGUILAR, ASHWORTH, GALETTO and AIZEN2006, Jennersten Reference JENNERSTEN1988, Steffan-Dewenter & Tscharntke Reference STEFFAN-DEWENTER and TSCHARNTKE1999, Tylianakis et al. Reference TYLIANAKIS, DIDHAM, BASCOMPTE and WARDLE2008), but no definite conclusion has been reached as to whether habitat transformation has an overall negative impact on pollination. Nearly one third of the available studies show no effects or even positive effects of habitat fragmentation on pollination and plant reproductive success (Aguilar et al. Reference AGUILAR, ASHWORTH, GALETTO and AIZEN2006, Tylianakis et al. Reference TYLIANAKIS, DIDHAM, BASCOMPTE and WARDLE2008).
In tropical rain forests, over 90% of the tree species are pollinated by animals (mainly insects), and some plant-pollinator interactions have a high degree of specialization (Bawa Reference BAWA1990, Johnson & Steiner Reference JOHNSON and STEINER2000). For instance, palm pollination is often specialized (Henderson Reference HENDERSON1986) with cantharophily (Nitidulidae and Curculionidae) being the predominant syndrome, but melittophily (e.g. Melipona, Apis and Trigona) and myiophily (Calliphoridae, Syrphidae and Drosophilidae) are also common (Henderson Reference HENDERSON1986). In addition many species of palm are either monoecious or dioecious and rely on biotic vectors for their pollination (Henderson Reference HENDERSON1986, Knudsen et al. Reference KNUDSEN, TOLLSTEN and ERVIK2001). Because many monoecious species are dichogamous it is necessary to consider if the effects of fragmentation vary depending on the sexual phase of the inflorescences (male- and female-phase). If forest fragmentation were to decrease pollinator visitation (e.g. reduction in the abundance of pollinators or the ratio of pollinators/other floral visitors) to female-phase inflorescences, pollen may be a limiting factor and therefore a reduction in fruit set would be expected (Burd Reference BURD1994) but the only study addressing this issue in a palm species found no evidence in support of this hypothesis (Aguirre & Dirzo Reference AGUIRRE and DIRZO2008). Conversely, if pollination is affected by fragmentation mainly at the male-phase of inflorescences, pollen competition and cryptic sexual selection at the stigma are likely to be relaxed, jeopardizing the mean performance of the plants sired in forest fragments, or diminishing the genetic variation in populations in small fragments, even if fruit set is not pollen-limited as in the cantharophilus moecious herb Dieffenbachia seguine (Cuartas-Hernández et al. Reference CUARTAS-HERNANDEZ, NUÑEZ-FARFAN and SMOUSE2010).
The Neotropical palm Astrocaryum mexicanum Liebm. is a dichogamous monoecious species chiefly pollinated by beetles. This palm is the dominant understorey species at Los Tuxtlas rain forest, Mexico, and persists in forest fragments of all sizes (Arroyo-Rodríguez et al. Reference ARROYO-RODRÍGUEZ, AGUIRRE, BENÍTEZ-MALVIDO and MANDUJANO2007). Aguirre & Dirzo (Reference AGUIRRE and DIRZO2008) observed that the abundance of pollinators of A. mexicanum was low in small forest fragments (<35 ha) compared with that in large forest fragments (114 and 700 ha), but they found that there was no effect of forest fragment size on fruit set. Given that A. mexicanum is a dichogamous species it is necessary to consider if the effects of fragmentation vary depending on the sexual phase of the inflorescences (male- and female-phase), a topic not addressed by Aguirre & Dirzo (Reference AGUIRRE and DIRZO2008).
Because male flowers of A. mexicanum offer rewards (pollen) to floral visitors, while female flowers do not, we hypothesized that assemblages of floral visitors (pollinators and non-pollinators) to this palm would be different in female- and male- phase inflorescences. Also, since A. mexicanum is a shade-tolerant, mature-forest species, we also predicted that floral visitors will shift towards a more heliophilous guild of visitors with fragmentation thus decreasing the proportional representation of true pollinators. We tested these expectations by censusing floral visitors to male- and female-phase inflorescences in forest fragments and continuous forest at Los Tuxtlas.
METHODS
Study site
The study site is located in the Los Tuxtlas region of the State of Veracruz, south-eastern Mexico. The predominant vegetation in the lowlands is tropical rain forest (Dirzo & Miranda Reference DIRZO and MIRANDA1991). The mosaic of vegetation types along the elevation range is considerably diverse at Los Tuxtlas, ranging from tropical cloud forest and mixed forests (conifer and broadleaved) at high elevations, where several species of boreal and tropical origins coexist, to typical tropical rain forest in the lowlands, with an overwhelming predominance of species of tropical origin in all the vertical strata of the forest (González-Soriano et al. Reference GONZÁLEZ-SORIANO, DIRZO, VOGT and México1997). In Los Tuxtlas forest fragmentation is severe. Mendoza et al. (Reference MENDOZA, FAY and DIRZO2005) documented that the landscape in the study area is composed of many small forest fragments and only a few large fragments: 1005 forest fragments (representing 90% of the total) with a median size of only 0.95 ha. The study was carried out in and around the Los Tuxtlas Research Station, administered by the National Autonomous University of Mexico. We selected six forest fragments (2, 4, 19.4, 34.6, 114.6 and 700 ha). The largest fragment (referred to as continuous forest) corresponds to the Los Tuxtlas Research Station, which extends westwards, towards the San Martin Volcano, encompassing a total of 1883 ha. The forest fragments were separated from the continuous forest c. 30 y ago, but have similar ecological and floristic compositions to those to the tropical rain forest in the research station. Further details of the sites can be found in Aguirre & Dirzo (Reference AGUIRRE and DIRZO2008).
Study species
Astrocaryum mexicanum Liebm. is the most abundant species in the understorey at the Los Tuxtlas region and its distribution is restricted to the lowlands (<700 m asl). At Los Tuxtlas, the average density of A. mexicanum > 1 m height is c. 1000 individuals ha−1 (Martínez-Ramos Reference MARTÍNEZ-RAMOS, González-Soriano, Dirzo and Vogt1997). Astrocaryum mexicanum reaches sexual maturity around 40 y of age, and its life span is well over 100 y. Most reproductive plants are in the range of 2–6 m in height (Piñero et al. Reference PIÑERO, MARTÍNEZ-RAMOS and SARUKHÁN1984), although some plants are up to 8 m (Búrquez et al. Reference BÚRQUEZ, SARUKHÁN and PEDROZA1987). A reproductive individual can have up to five inflorescences, but has only one mature inflorescence at any one time. Each inflorescence has an average of 4885 male flowers and 28 female flowers (Búrquez et al. Reference BÚRQUEZ, SARUKHÁN and PEDROZA1987). Inflorescences are dichogamous and protogynous. Female flowers are receptive in the morning, opening between 04h00–06h00 and remain functional for over 12 h. Male flowers open almost synchronously at night (after 20h00) (Búrquez et al. Reference BÚRQUEZ, SARUKHÁN and PEDROZA1987). Sexual functions within an inflorescence overlap through a short period but the pollen is largely immature while most female flowers are dysfunctional, thus minimizing self-fertilization (geitonogamy). Floral visitors are abundant and diverse, but the potential pollinators are four nitidulid beetles: Mystrops sp., M. mexicanus, Eumystrops centralis and Coleopterus aberrans (Aguirre & Dirzo Reference AGUIRRE and DIRZO2008, Búrquez et al. Reference BÚRQUEZ, SARUKHÁN and PEDROZA1987). Following Búrquez et al. (Reference BÚRQUEZ, SARUKHÁN and PEDROZA1987) we defined as pollinators those flower visitors that (1) were present in both phases of the inflorescences, (2) were observed transporting pollen on their bodies, and (3) were observed carrying pollen onto the stigmas on the female flowers.
The protocol for sampling floral visitors was reported in a previous study (Aguirre & Dirzo Reference AGUIRRE and DIRZO2008). In sum, all reproductive plants of A. mexicanum were mapped in permanent plots. Within each plot we randomly selected 15 adult plants from which we collected all the floral visitors from 10 inflorescences in the female phase and in five inflorescences in the male phase. The sampling was conducted between March and May 1999 in a haphazardous fashion. All inflorescences were collected between 07h00 and 08h00, when the activity among the floral visitors to A. mexicanum is most intense (Aguirre & Dirzo Reference AGUIRRE and DIRZO2008, Búrquez et al. Reference BÚRQUEZ, SARUKHÁN and PEDROZA1987). In the laboratory, the floral visitors were identified to species or morphospecies level, separated and counted.
Statistical analyses
To explore the differences in the structure of the assemblages of floral visitors in the female- and male-phase inflorescences and the differences related to forest fragmentation, we used non-metric multidimensional scaling on log10-transformed abundances of floral visitors. An importance value index (IV) was calculated for each species of floral visitor. This index was used to include, simultaneously, the relative abundance and frequency of each flower visitors. The relative frequencies were calculated as the number of inflorescences (male or female) in which each species of floral visitor was observed divided by the sum of all the frequencies across all floral visitors. The relative abundance was calculated by dividing the abundance of each species by the overall abundance of floral visitors. IV is the sum of the two relative values, and because we sampled roughly twice the number of female inflorescences compared with male ones, calculations of IV were done separately for male and female inflorescences.
To explore the effects of forest fragment size on evenness of the community of floral visitors we used Pearson correlation analysis. To explore the effects of forest fragment size on species richness, we used mixed-effects models where sexual phase was the fixed factor and log-transformed fragment size was used as covariate, whereas fragment identity and individual palm within each fragment appeared in the random component of the model. In this way we avoided potential pseudo-replication. The same model structure was used to explore the effects of forest fragment size on pollinator abundance and the ratio of pollinators to other floral visitors. All statistical analyses were performed with R-2.5.1 (www.R-project.org).
RESULTS
Assemblages of floral visitors
We collected 228 786 arthropod floral visitors, corresponding to 57 species (or morphospecies), from 86 inflorescences of A. mexicanum (30 in the male phase and 56 in the female phase) across six forest fragments. The major contingent of visitors belonged to two groups: (1) Coleoptera, which included the pollinators (Eumystrops centralis, Mystrops mexicanus, Coleopterus aberrans and Mystrops sp.) and two non-pollinators (Enochrus sp. and Ciclocephala fasciolata), and (2) Hymenoptera (Plebeia frontalis, Plebeia pulchra, Polybia occidentalis, Apis mellifera, Trigona fulviventris, Pheidole sp. 1, Camponotus sericeiventris, Pachycondylla ferruginea, Pachycondylla villosa, Zacriptocerus spinosus and Camponotus sp.). Beyond these two orders, other prominent visitors comprise species of Orthoptera and Diptera (Appendix 1). Abundance of floral visitors increased significantly with fragment size (F = 9.5, df = 1, 4, P = 0.04) although this relationship was not linear, particularly in female-phase inflorescences. The 10 most abundant species of floral visitor were predominately found in inflorescences from the three large fragments: their relative abundances ranged from 61% to 93% and included the three reported pollinators of A. mexicanum, whose relative abundances in the three large fragments ranged between 72% and 92%. In contrast, 60% of the less common species of floral visitor, excluding singletons and doubletons, were more abundant in the three small fragments and half of them had at least 70% of their abundance represented in inflorescence from the three small fragments.
On average (±SE), there were 3376 ± 409 floral visitors to female-phase inflorescences and 1324 ± 303 to male-phase inflorescences. Overall, the assemblages of floral visitors to inflorescences in the male and female phases were structurally different, as revealed by a non-parametric multidimensional scaling analysis based on the abundance of the floral visitors (Figure 1a). The robustness of this analysis was satisfactory, as suggested by the value of 8.3 for the stress parameter.
Figure 1. Assemblage of floral visitors to male-phase (black) and female-phase (white) inflorescences of Astrocaryum mexicanum. Non-metric multidimensional scaling based on abundance in log10 scale of floral visitors in six forest fragments: circles (2 ha), squares (4 ha), diamonds (19.4 ha), triangle (34.6 ha), inverted triangle (114.6 ha) and stars (700 ha) (a). Importance values of floral visitors, including the main pollinators: 57 = Eumystrops centralis, 56 = Mystrops mexicanus and 54 = Coleopterus aberrans. O = other floral visitors (b). For the complete list and species codes see Appendix 1.
The distribution of importance values of pollinators (E. centralis, M. mexicanus, Mystrops sp. and C. aberrans) was similar in female- and male-phase inflorescences (Figure 1b) while the IVs of the other floral visitors were biased to one phase.
Impact of forest fragmentation on evenness, species richness and diversity of floral visitors
The index of evenness of the community of floral visitors across fragments ranged from 0.37 to 0.51, and this index correlated negatively with the natural logarithm of forest fragment size (r = −0.82, t = 2.95, df = 4, P = 0.04). The mixed-effects model showed that the overall species richness of floral visitors to A. mexicanum was significantly higher (F = 10.2, df = 1, 4, P = 0.04) for inflorescences in the male phase than for female-phase inflorescences. However, the species richness of the floral visitors was unaffected by forest fragment size, either as a main effect (F = 0.03, df = 1, 4, P = 0.86) or in interaction with the sexual phase of the inflorescence (F = 0.56, df = 1, 4, P = 0.49). Consistent with this finding, Shannon's diversity index of the floral visitors was also significantly higher (F = 12.8, df = 1, 4, P < 0.023) for inflorescences in the male phase (1.60 ± 0.07) than that for female-phase inflorescences (1.37 ± 0.03). Shannon's diversity index decreased with forest fragment size (Figure 2), and this effect was significantly more pronounced in male phase inflorescences compared with female phase ones (F = 7.96, df = 1, 4, P = 0.04).
Figure 2. Forest fragment size effects on Shannon's diversity index of floral visitors in inflorescences of Astrocaryum mexicanum in male (black circles) and female (white circles) phases. Circles represent individual palms within each of the six forest fragments and lines are the fitted mixed effects model's predictions for each sexual phase.
Impact on pollinators of Astrocaryum mexicanum
Considering only those taxa that we defined as pollinators, we observed that their abundance was positively associated with forest fragment size (F = 8.69, df = 1, 4, P = 0.04, Figure 3a), but there was no significant interaction between the sexual phase of the inflorescences and the size of the forest fragments (F = 0.71, df = 1, 4, P = 0.44). The abundance of pollinators changed by over 10-fold when the smallest forest fragment (2 ha) was compared with the largest forest fragment (700 ha) in both sexual phases of the inflorescences. Nevertheless, across all the sizes of forest fragments, the abundance of pollinators in the female-phase inflorescences was higher than that in the male-phase inflorescences (F = 9.49, df = 1, 4, P = 0.03).
Figure 3. Forest fragment size effects on the abundance (a) and the ratio (b) of pollinators:non-pollinators visiting inflorescences to male (black circles) and female (white circles) phases. Circles represent individual palms within each of the six forest fragments and lines are the fitted mixed effects model's predictions for each sexual phase.
When we considered the ratio of the abundance of pollinators to that of other floral visitors (Figure 3b), we again observed an overall positive effect of forest fragment size (F = 12.8, df = 1, 4, P = 0.023), but the magnitude of the effect was higher in male-phase inflorescences than in female-phase inflorescences (F = 16.1, df = 1, 4, P < 0.001) and no significant effects of sex and the covariate (fragment size) as main effects were detected (F < 0.87, df = 1, 4, P > 0.354).
DISCUSSION
Our results are consistent with other studies with palms showing a great diversity of floral visitors (Ervik & Bernal Reference ERVIK and BERNAL1996, Meléndez-Ramírez et al. Reference MELÉNDEZ-RAMÍREZ, PARRA-TABLA, KEVAN, RAMÍREZ-MORILLO, HARRIES, FERNÁNDEZ-BARRERA and ZIZUMBO-VILLAREAL2004, Siefke & Bernal Reference SIEFKE and BERNAL2004), and although variation across sexual phases regarding assemblages of floral visitors of palms have been documented with inconsistent results, no other study seems to have taken into account the effect of forest fragmentation on floral visitors of male- and female-phase inflorescences. We found higher abundance of floral visitors on male-phase inflorescences, which is consistent with reports of other palm species, including Orbignya phalerata (Anderson et al. Reference ANDERSON, OVERAL and HENDERSON1988), Euterpe precatoria (Küchmeister et al. Reference KUCHMEISTER, SILBERBAUER-GOTTSBERGER and GOTTSBERGER1997), Prestoea shultzeana (Ervik & Feil Reference ERVIK and FEIL1997), Wettinia quinaria, Attalea allenii (Nuñez et al. Reference NÚÑEZ, BERNAL and KNUDSEN2005), Oenocarpus bataua. (Nuñez-Avellaneda & Rojas-Robles Reference NÚÑEZ-AVELLANEDA and ROJAS-ROBLES2008). In contrast, in other species (Phytelephas seemannii, Aphandra natalia) (Bernal & Ervik Reference BERNAL and ERVIK1996, Ervik et al. Reference ERVIK, TOLLSTEN and KNUDSEN1999) higher abundance of floral visitors has been reported in female-phase inflorescences compared with male ones. The available evidence, including the present study, indicates an inconsistent pattern in the abundance of the non-pollinating floral visitors to inflorescences in the male and female phases of monoecious palms. However, detailed analysis of the likely subset of non-pollinating visitors has been poorly investigated. This is especially important in studies of the floral visitors to palms because their inflorescences are known to attract a large number of species of arthropods, including pollinators and many other visitors searching for food (flowers, pollen, and other floral visitors). Also, palm inflorescences can serve as mating arenas and oviposition sites (Búrquez et al. Reference BÚRQUEZ, SARUKHÁN and PEDROZA1987, Consiglio & Bourne Reference CONSIGLIO and BOURNE2001, Henderson Reference HENDERSON1986, Siefke & Bernal Reference SIEFKE and BERNAL2004).
Considering all floral visitors, the average species richness was higher in male-phase inflorescences than in female-phase inflorescences. In other words, male inflorescences harboured a greater number of what seemed to be accessory floral visitors than did female-phase inflorescences. This finding is in agreement with the higher importance value observed in over 50 of the less important floral visitors, many of which were only found in male-phase inflorescences.
On the other hand, we found that forest fragment size was negatively correlated with the diversity (Shannon's index) of floral visitors, the abundance of pollinators and the ratio between pollinators and other floral visitors, but not with the total numbers of floral visitors. Also, these findings are in agreement with the negative correlation between forest fragment size and evenness of the community of floral visitors. Furthermore, the effect of forest fragment size on the ratio pollinators to non-pollinators was relatively small in female-phase inflorescences, varying from 4:1 in the smallest fragment to 20:1 in the largest fragment. In contrast, the ratio varied from 2:3 to 23:2 in the smallest and largest forest fragments, respectively, in the male-phase inflorescences. The effect of forest fragment size on the ratio of pollinators to non-pollinators in male-phase inflorescences resulted from a combination of a steady reduction in the abundance of pollinators as forest fragment size decreased, and the accumulation of non-pollinating visitors in small forest fragments. Male flowers of A. mexicanum are attractive to a large set of floral visitors, foraging for pollen without performing any pollination service for the palm (e.g. Apis mellifera). Because the prevalence of this kind of floral visitors was higher in small forest fragments than in large fragments, it is possible that the main threat of forest fragmentation to the pollination system of A. mexicanum is the increased competition (by exploitation) between non-pollinators and pollinators in male-phase inflorescences. This finding is consistent with the reduction in diversity observed as forest fragment size increased, particularly in male-phase inflorescences.
In the smallest fragments, we observed on average 150 pollinators in male-phase inflorescences and 500 pollinators in female-phase inflorescence. Each inflorescence of A. mexicanum has, on average, more than 170 times more male flowers (4885) than female flowers (28), and this ratio should hold at the population level because inflorescences are dichogamous (Búrquez et al. Reference BÚRQUEZ, SARUKHÁN and PEDROZA1987). Therefore, even if exploitation competition between pollinators and accessory floral visitors is high in small forest fragments, fruit set would be unlikely to be limited by pollen. This argument is supported by the results of Aguirre & Dirzo (Reference AGUIRRE and DIRZO2008), who found no effect of forest fragment size on the fruit set of A. mexicanum at the same study site. Nevertheless, even if pollen flow is not affected by forest fragmentation beyond a critical threshold, and thus does not limit fruit set, reduced pollen loads may relax pollen tube competition. In consequence, reduced selection pressure may operate on the siring of seeds in small forest fragments (Janse & Verhaegh Reference JANSE and VERHAEGH1993, Richardson & Stephenson Reference RICHARDSON and STEPHENSON1992). This scenario may be more applicable to abundant plant species with many floral visitors such as A. mexicanum, but not necessarily to less abundant species with relatively few floral visitors. Clearly, these aspects warrant further investigation. Also, this study raises an important question as to whether the observed drop in pollinator abundance among small forest fragments occurs in other plant species, particularly those which naturally have far fewer floral visitors than A. mexicanum and would therefore be at greater risk of reduced pollen flow. Two other studies at Los Tuxtlas with perennial herbs showed different effects of fragmentation on pollination. Cuartas-Hernández & Nuñez-Farfán (Reference CUARTAS-HERNANDEZ and NUÑEZ-FARFAN2006) found that fragmentation negatively affects the fruit set and gene flow of Dieffenbachia seguine, suggesting a reduction of the populations of their main beetle pollinators (Cyclocephala sp. and Erioscelis sp.). In contrast, Suárez-Montes et al. (Reference SUÁREZ-MONTES, FORNONI and NUÑEZ-FARFAN2010) found no effects of forest fragmentation on the genetic variability and gene flow of the hummingbird-pollinated Heliconia uxpanapensis (Heliconiaceae). As pointed above, there is not enough evidence to draw an overall conclusion of the effects of habitat fragmentation on pollination dynamics. Therefore, studies of the effects of habitat fragmentation on pollination are warranted.
ACKNOWLEDGEMENTS
Field work was supported with funds from UNAM assigned to RD. INECOL supported AA with a postdoctoral fellowship. RG was supported by CONACYT (sabbatical fellowship 79846). J. Ghazoul and two anonymous reviewers offered constructive comments and suggestions on an early draft. The Biological Station of Los Tuxtlas-UNAM provided all necessary facilities to conduct this work.
Appendix 1. List of species/morphospecies of floral visitors to male and female phase inflorescences of Astrocaryum mexicanum at Los Tuxtlas Veracruz, Mexico.
