INTRODUCTION
Climate is known to affect the distribution of organisms and their capability to establish interactions (Hawkins et al. Reference HAWKINS, FIELD, CORNELL, CURRIE, GUÉGAN, KAUFMAN, KERR, MITTELBACH, OBERDORFF, O'BRIEN, PORTER and TURNER2003, Hegland et al. Reference HEGLAND, NIELSEN, LÁZARO, BJERKNES and TOTLAND2009). For instance, it is widely accepted that plant-pollinator assemblages differ along geographical (Cruden Reference CRUDEN1972, Kay & Schemske Reference KAY and SCHEMSKE2003, Kearns Reference KEARNS1992, Kessler & Krömer Reference KESSLER and KRÖMER2000, Krömer et al. Reference KRÖMER, KESSLER and HERZOG2006, Olesen & Jordano Reference OLESEN and JORDANO2002, Ollerton & Cranmer Reference OLLERTON and CRANMER2002, Ollerton et al. Reference OLLERTON, JOHNSON, HINGSTON, Waser and Ollerton2006, Primack Reference PRIMACK1983) and climatic gradients (Arroyo et al. Reference ARROYO, PRIMACK and ARMESTO1982, Dalsgaard et al. Reference DALSGAARD, MARTÍN GONZÁLEZ, OLESEN, OLLERTON, TIMMERMANN, ANDERSEN and TOSSAS2009, Devoto et al. Reference DEVOTO, MEDAN and MONTALDO2005, Medan et al. Reference MEDAN, MONTALDO, DEVOTO, MANTESE, VASELLATI, ROITMAN and BARTOLONI2002), particularly temperature and rainfall (Arroyo et al. Reference ARROYO, PRIMACK and ARMESTO1982, Dalsgaard et al. Reference DALSGAARD, MARTÍN GONZÁLEZ, OLESEN, OLLERTON, TIMMERMANN, ANDERSEN and TOSSAS2009, Devoto et al. Reference DEVOTO, MEDAN and MONTALDO2005, Hegland et al. Reference HEGLAND, NIELSEN, LÁZARO, BJERKNES and TOTLAND2009, Medan et al. Reference MEDAN, MONTALDO, DEVOTO, MANTESE, VASELLATI, ROITMAN and BARTOLONI2002). A widespread pattern often observed is a decrease in insect pollinator species richness at low temperatures and high humidity (Arroyo et al. Reference ARROYO, PRIMACK and ARMESTO1982, Cruden Reference CRUDEN1972, Devoto et al. Reference DEVOTO, MEDAN and MONTALDO2005, Elberling & Olesen Reference ELBERLING and OLESEN1999, Hodkinson Reference HODKINSON2005, Kearns Reference KEARNS1992, Olesen & Jordano Reference OLESEN and JORDANO2002, Tanaka & Tanaka Reference TANAKA and TANAKA1982). However, not all insect groups are affected equally (Arroyo et al. Reference ARROYO, PRIMACK and ARMESTO1982, Janzen Reference JANZEN1973, Janzen et al. Reference JANZEN, ATAROFF, FARIÑAS, REYES, RINCON, SOLER, SORIANO and VERA1976, Kearns Reference KEARNS1992, Warren et al. Reference WARREN, HARPER and BOOTH1988). For example, a widely cited global pattern is that flies replace bees in cooler and wetter habitats (Arroyo et al. Reference ARROYO, PRIMACK and ARMESTO1982, Devoto et al. Reference DEVOTO, MEDAN and MONTALDO2005, Elberling & Olesen Reference ELBERLING and OLESEN1999, Kearns Reference KEARNS1992, Medan et al. Reference MEDAN, MONTALDO, DEVOTO, MANTESE, VASELLATI, ROITMAN and BARTOLONI2002, Warren et al. Reference WARREN, HARPER and BOOTH1988).
Compared with most insects, the activity of vertebrates is less constrained by rainfall and temperature, and vertebrates therefore gain importance as pollinators in wet and cold conditions (Cruden Reference CRUDEN1972, Dalsgaard et al. Reference DALSGAARD, MARTÍN GONZÁLEZ, OLESEN, OLLERTON, TIMMERMANN, ANDERSEN and TOSSAS2009, Kay & Schemske Reference KAY and SCHEMSKE2003, Kessler & Krömer Reference KESSLER and KRÖMER2000, Krömer et al. Reference KRÖMER, KESSLER and HERZOG2006, Stiles Reference STILES1978). For instance, in the New World, hummingbirds become especially important pollinators where insect species richness and activity are reduced due to low temperatures and high rainfall (Aizen Reference AIZEN2003), often encountered in mountains (Cruden Reference CRUDEN1972, Dalsgaard et al. Reference DALSGAARD, MARTÍN GONZÁLEZ, OLESEN, OLLERTON, TIMMERMANN, ANDERSEN and TOSSAS2009, Stiles Reference STILES1978). There are several reports of plant species that are visited exclusively by hummingbirds in the highlands while visited by insect and hummingbird species at lower elevations, e.g. Buddleja brasiliensis in south-eastern Brazil (Sazima et al. Reference SAZIMA, BUZATO and SAZIMA1996), Costus spp. in Central America (Kay & Schemske Reference KAY and SCHEMSKE2003), bromeliads in the Andes (Kessler & Krömer Reference KESSLER and KRÖMER2000, Krömer et al. Reference KRÖMER, KESSLER and HERZOG2006) and plant–hummingbird assemblages in the West Indies (Dalsgaard et al. Reference DALSGAARD, MARTÍN GONZÁLEZ, OLESEN, OLLERTON, TIMMERMANN, ANDERSEN and TOSSAS2009).
The aim of this study is to examine the effect of climate on pollinator richness and importance in pollination networks in the West Indies. The West Indies has a pollinator-limited flora (Spears Reference SPEARS1987) subject to major disturbances (Dalsgaard et al. Reference DALSGAARD, HILTON, GRAY, AYMER, BOATSWAIN, DALEY, FENTON, MARTIN, MARTIN, MURRAIN, ARENDT, GIBBONS and OLESEN2007), e.g. periodic hurricanes known to disrupt plant-pollinator associations (Rathcke Reference RATHCKE2000, Rivera-Marchand & Ackerman Reference RIVERA-MARCHAND and ACKERMAN2006). Although plant–hummingbird assemblages from the West Indies have been thoroughly studied (Dalsgaard et al. Reference DALSGAARD, MARTÍN GONZÁLEZ, OLESEN, TIMMERMANN, ANDERSEN and OLLERTON2008, Reference DALSGAARD, MARTÍN GONZÁLEZ, OLESEN, OLLERTON, TIMMERMANN, ANDERSEN and TOSSAS2009; Kodric-Brown et al. Reference KODRIC-BROWN, BROWN, BYERS and GORI1984, Lack Reference LACK1973), entire plant–pollinator communities have been neglected (although see Percival Reference PERCIVAL1974), and no study has addressed the effect of climate on interactions between plants and pollinators in these islands. We used a network approach to examine and compare plant–pollinator communities across gradients in rainfall and temperature, and hypothesized that: (1) bird species richness would be less affected by rainfall and temperature than insect richness; (2) the response of the different insect groups will be complex, mirroring previously observed patterns, with bee diversity higher in dry and warm sites, and dipteran diversity higher in cooler, wetter habitats. In order to asses the importance of each of the functional groups of pollinators, we used several metrics: (1) proportion of interactions established; (2) proportion of plants visited in each community; and (3) topological importance, measured as proportional betweenness centrality (BC). By using these complementary measures we are able to reveal any potential change in pollinator importance along the climatic gradient. Moreover, this allows us to identify whether rainfall or temperature has a stronger effect on the pollination assemblages in the West Indies.
METHODS
Study sites
Plots with dimensions of 200 m × 5 m were delimited in distinct habitats at two altitudes in the islands of Dominica, Grenada (Lesser Antilles) and Puerto Rico (Greater Antilles), i.e. a total of six plots (Table 1). In Dominica, we sampled in the coastal scrub forest south of Pointe Michel and at the montane thicket-elfin woodland on the ridge towards Boeri Lake in the UN World Heritage Site of Morne Trois Pitons National Park; in Grenada in a coastal dry scrub woodland in the north-east part of the Mt. Hartman Protected Area/Grenada Dove sanctuary and in a montane elfin woodland at the very top of Mt. Qua Qua in Grand Etang National Park; and in Puerto Rico in the coastal dry scrub forest of Guánica State Forest and UN Biosphere Reserve and at the montane thicket on the top of Monte de Guilarte State Forest. The lowland sites were situated in subtropical dry scrub forests, characterized by warm temperatures, mostly sunny days and low humidity. They are composed primarily of deciduous trees, shrubs and some cactus species. Some of the species, particularly in the Dominican lowlands, although native to the West Indies, are weedy species. The highland sites were situated in montane thicket-elfin woodlands, which are cooler, windier and with a higher rainfall, often being misty, especially in the early morning. These communities were composed mainly by evergreen short trees, herbs, epiphytes and palms. The relatively short vegetation at both lowland and highland habitats allowed us to sample the entire plant–pollinator assemblage.
Table 1. Description of the study sites from the West Indies. Total P = the number of plants observed in each site (i.e. total number of plant species in flower), P Visited = the number of plant species which received visits during our observations. A = Total number of pollinator species; Insect = Number of insect pollinator species; Insect Obs = Hours of insect observations; Av. Ins. Obs. = Average observation time (h) of each plant in each site for insects; Bird Obs. = Hours of bird observations; Av. Bird Obs. = Average observation time (h) of each plant in each site for birds. Rainfall data were collected by the Forestry and Wildlife Division of Dominica, by the Forestry Department, Grenada Ministry of Agriculture and the Departamento de Recursos Naturales y Ambientales of Puerto Rico. Mean annual temperature at each study site was calculated using the wet adiabatic lapse rate (0.65 °C per 100 m) as in Buckley & Roughgarden (Reference BUCKLEY and ROUGHGARDEN2006), the mean annual long-term sea-level temperature close to each of our lowland study sites (NOAA National Weather Service) and the elevation of each study site.
The mean annual temperature of each site was calculated using the 0.65 °C per 100 m wet adiabatic lapse rate as in Buckley & Roughgarden (Reference BUCKLEY and ROUGHGARDEN2006), taking the mean annual temperature at the closest weather station of each site on each island and calculating the resulting temperature at the elevation of each site (Table 1). The sites ranked in altitude from 9 m to 1154 m, mean annual temperature from 18 °C to 27 °C and mean annual rainfall from 799 mm to 7506 mm (Table 1). These differences in rainfall and temperature enabled us to study the diversity and topological importance of different pollinator groups along a climatic gradient.
Sampling methodology
We conducted our fieldwork between April and July 2005 (Dominica) and between March and June 2006 (Grenada and Puerto Rico). These months encompass the end of the dry and the beginning of the rainy season. We identified all plant species in flower within our study sites and observed their flowers for bird and insect visitation. Since most plant species did not flower for the entire fieldwork period, plants were sampled according to the length of their flowering.
We used a semi-random sampling methodology for our observations of flower visitation. We divided our plots into 10 × 5-m segments and randomly chose a sampling order and segment for each plant species (Dominica) or a starting segment and observed all plant species not previously observed before moving on to the subsequent segment, etc. (Grenada and Puerto Rico). In order to record as many visitors as possible, we restricted the observations of each plant species to different individuals and at different times of the day, preferably. Observations were conducted in fair weather conditions between 06h00 and 18h00 (Dominica) and between 06h00 and 14h00 (Grenada and Puerto Rico). The visitation of insects (and other small-sized potential pollinators) was recorded during 10-min observation periods (a total of 275 h) at a distance of approximately 2 m from the flowers, while birds (and other large-sized potential pollinators) were monitored during 30-min periods (a total of 1215 h) from a distance of approximately 10 m (Tables 1 and 2). We recorded only floral visitors which touched the flower's reproductive organs, and thus may be regarded as potential pollinators.
Table 2. Absolute values of species richness and measure of pollinator importance for each pollinator group at each site. BC = betweenness centrality; Dipt. = dipterans; Lep. = lepidopterans; Col. = coleopterans.
Plant species were identified after Lack et al. (Reference LACK, WHITEFOORD, EVANS, JAMES and GREENOP1997) and with the help of several experts. Plant species which received no visits were excluded from the analysis (Table 1). Bird species were visually identified after Raffaele et al. (Reference RAFFAELE, WILEY, GARRIDO, KEITH and RAFFAELE1998). Bees were classified into families after Michener (Reference MICHENER2000), lepidopterans after Smith et al. (Reference SMITH, MILLER and MILLER1994) and the rest of the insect groups after Triplehorn & Johnson (Reference TRIPLEHORN and JOHNSON2005). Vouchers are kept at Aarhus University. A full list of the species included in the analysis is given in Appendix 1 and 2, and the networks will be deposited at the Interaction Web database (http://www.nceas.ucsb.edu/interactionweb) for open use.
Potential methodological biases
There were two potential methodological biases in this study. First, we were comparing communities belonging to islands of different sizes, which may affect species number and composition (MacArthur & Wilson Reference MACARTHUR and WILSON1967). However, total pollinator species richness is not related to island size (rs = −0.50, P > 0.05; two-tailed Spearman correlation analysis). Instead, there were strong, statistically significant relationships with rainfall and temperature (see Results).
A second potential bias was the difference in sampling effort between birds and insects. Birds were observed for a longer time than insect species (1215 h versus 275 h). This may inflate the importance of birds when compared with insect species. Nevertheless, a relative increase or decrease in visitation of birds or insect groups along the rainfall/temperature gradients are unaffected by differences in sampling effort between birds and insects. Hence, although caution should be taken when comparing the overall importance of birds versus insect species, the observed trends within groups along the climate gradient are genuine.
Data analyses
In our analyses, we focused on birds and insects since these groups accounted for more than 99% of the total species composition and interactions. The remaining 1% was comprised of lizards. Insects were further subdivided into bees, wasps, dipterans, lepidopterans and coleopterans. Other insect groups (hemipterans and thysanopterans) were represented by very few individuals and only in few sites and, due to their small number, were not analysed separately.
We calculated the proportional species richness of each pollinator group as the number of species belonging to that group out of the total number of species present in the community. In this way we can quantify how much each group accounts for the total biodiversity at each site. We measured different aspects of the importance of each pollinator group at each site as follows: (1) in order to have a measure of their activity, we calculated the proportion of interactions established by each pollinator group in each community; (2) to quantify their flower feeding niche width, we estimated the proportion of plant species which were visited by each pollinator group in each site; and (3) to determine their topological importance for the cohesion of the network we calculated their proportional betweenness centrality. Betweenness centrality (BC) of species i was calculated in Pajek (http://pajek.imfm.si/doku.php) as
![\begin{equation}
BC_{\rm i}=2\sum\limits_{{\rm j} \,{\textless}\, {\rm k};\ {\rm I} \,{\ne}\, {\rm j}} [ (g_{\rm jk} (i)/g_{\rm jk})/((n -1)(n -2))]
\end{equation}](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20151024053210521-0934:S0266467409990034_eqnU1.gif?pub-status=live){kind=small}
where n is the number of species in the community, gjk is the number of shortest paths linking any two species, and gjk(i) is the number of those shortest paths which pass through species i (de Nooy et al. Reference DE NOOY, MRVAR and BATAGELJ2005, Wasserman & Faust Reference WASSERMAN and FAUST1994). BC ranges from zero, when there are no shortest paths passing through the focal species, to one, when the focal species is the only connection between all other species in the community, i.e. its extinction would lead directly to complete network fragmentation. Therefore, high-BC pollinator groups are important for the cohesion of the community. BC and other centrality measures have recently been used to identify potential keystone species in ecological communities (Estrada Reference ESTRADA2007, Jordan et al. Reference JORDAN, LIU and DAVIS2006, Martín González et al., in press). To calculate the centrality of each pollinator group, we summed the BC scores of all species belonging to a given group and divided by the total BC sum of all species in the network.
Since sampling effort was not equal for all communities, for all response variables we used proportional values in the analysis, as also done in similar studies (Devoto et al. Reference DEVOTO, MEDAN and MONTALDO2005). Alternatively, we could use absolute numbers and add sampling effort as a predictor variable into the analysis. However, due to the number of communities sampled (n = 6), it is preferable to keep the number of predictor variables as low as possible. Hence, our decision was to use proportions rather than absolute numbers. Prior to analysis, rainfall was log-transformed and all proportional variables arcsine square root-transformed. All variables were normally distributed (Kolmogorov–Smirnov test). We examined the relationships between the response variables against rainfall and temperature using both single and multiple regression analysis in SPSS 15.0 (SPSS Inc., Chicago, USA). Although multiple regression has the advantage of taking into account the effect of rainfall and temperature simultaneously, the lower number of degrees of freedom in the multiple regression analysis may cause otherwise significant relations to become non-significant. Therefore, significant results in both single and multiple regressions are considered. Multicollinearity was not an issue in the multiple regressions, since in all cases VIF <1.2 and Tolerance >0.8.
RESULTS
Networks ranged in size from 68 pollinator and 25 plant species in the Dominican lowland to 27 pollinator and 24 plant species in the Dominican highland. Pollinator species diversity, number of interactions established, number of plants visited and mean BC of each group also varied considerably among communities (Table 2).
For each pollinator group, we analysed how variation in pollinator richness and importance related to each climatic factor using different regression models (Table 3). Although birds did not have any significant relationship with the climatic variables, they did show some marked and near-significant tendencies, particularly with rainfall. Bird proportional species richness remained relatively constant along the gradients, but they tended to have a higher proportion of interactions on the wetter and cooler end of the gradient (Table 3). Birds also visited a higher proportion of plant species and had a higher BC towards the wet end of the gradient in the single regressions (Table 3).
Table 3. Results of the simple and multiple regressions. For each response variable (% species richness,% plant species visited,% interactions, and% BC), the effect of the predictor variables is based on a comparison of three regression models containing: (1) mean annual rainfall (MAR); (2) mean annual temperature (MAT); (3) multiple regression models containing MAR and MAT. For each model, the standardized coefficients and their significance level is indicate for each of the predictor variables, as well as the model fit based on Radj and P-value. ** P < 0.025, * P < 0.5. Dipt. = Diptera; Lep. = Lepidoptera; Col. = Coleoptera.
In contrast, for insects the proportional number of interactions significantly decreased with increasing rainfall when the single regression was performed (Table 3). However, none of the other variables was significant, although proportional species richness and proportional betweenness centrality (in single regressions) of insects were strongly affected by rainfall (Table 3).
Within insects, different functional groups showed significant and contrasting responses to climate, especially rainfall (Figure 1, Table 3). Bees were strongly and negatively affected by increasing rainfall: all variables observed (proportional species richness, proportional number of interactions, proportional number of plants visited and proportional BC) significantly decreased with rainfall both in the single and multiple regressions (Figure 1, Table 3). However, bees did not show any significant trend along the temperature gradient (Figure 1, Table 3).
Figure 1. For the insect groups, the relationships between rainfall and the studied measures: pollinator species richness (a); pollinator group's number of interactions (b); pollinator group's number of plant species visited (c); and pollinator group's topological importance as betweenness centrality (BC) (d). Notice that raw values were used for the figures whereas transformed values were used for the analyses. Only significant relationships are shown, i.e. P < 0.05.
The responses of wasps were different from those of bees. Wasp proportional species richness increased with rainfall, although their proportional BC decreased, when single regressions were performed (Figure 1, Table 3). The other variables did not show any significant trend (Figure 1, Table 3).
Dipterans had significantly lower proportional species richness in warmer sites, both in single and multiple regressions (Table 3). They were in fact the only insect group affected significantly by temperature. Moreover, their BC increased significantly with rainfall both with single and multiple regressions (Figure 1, Table 3).
Lepidopterans were the only insect group which did not show any significant trend with rainfall or temperature.
Finally, for coleopterans the proportion of visited plants and proportional BC increased with rainfall, although the former only in the single regressions (Figure 1, Table 3). They were unaffected by temperature.
DISCUSSION
In this study, we have examined the response of the main pollinator groups to climatic gradients in six plant-pollinator networks on three islands in the West Indies. Besides reporting proportional species richness of each pollinator group, we also used several complementary parameters describing their importance as pollinators, including both simple proportions of interactions or plant species visited, and a more sophisticated network measure, i.e. betweenness centrality. All these different measures provide valuable information about how pollinators change in their patterns of interaction along the climatic gradient, giving us a stronger analytical background to support our conclusions than could be obtained by simple species counts alone.
As expected, rainfall and temperature significantly affected the composition and importance of pollinators, but in different ways, depending upon the functional group considered. Overall, rainfall was the more critical for most of the groups, being the main driver of pollinator change along the climatic gradient. A reason for this might be that the lowest temperature at our sites (c. 18 °C) is relatively high, whereas the highest rainfall (c. 7500 mm y−1) is extremely wet and might therefore be a constraining factor for some insect pollinators. Yet, humidity was also found to be the main driver of pollinator turnover in a similar study in Patagonia, South America (Devoto et al. in press). Moreover, humidity has, at the macro-ecological scale, also been shown to be the most important factor determining invertebrate richness in warm climates (Hawkins et al. Reference HAWKINS, FIELD, CORNELL, CURRIE, GUÉGAN, KAUFMAN, KERR, MITTELBACH, OBERDORFF, O'BRIEN, PORTER and TURNER2003). Still, not all pollinator groups responded equally along the gradient.
In concordance with studies from Central and South America, birds tended to become, although not significantly, more important as rainfall increases. This is probably because birds have a high energy demand and therefore feed on flowers even during rain, whereas insect pollinators are inactive under these conditions (Cruden Reference CRUDEN1972, Dalsgaard et al. Reference DALSGAARD, MARTÍN GONZÁLEZ, OLESEN, OLLERTON, TIMMERMANN, ANDERSEN and TOSSAS2009). The importance of birds as pollinators in wet areas is further supported by a higher proportion of plants adapted to ornithophily in humid areas (Aizen Reference AIZEN2003, Dalsgaard et al. Reference DALSGAARD, MARTÍN GONZÁLEZ, OLESEN, OLLERTON, TIMMERMANN, ANDERSEN and TOSSAS2009). This tendency of birds was coupled with an overall significant decrease in the importance of insects with increasing rainfall.
Within insects, the two major insect pollinator groups (bees and dipterans) showed opposite responses to climate, and thus partially replaced each other. The higher importance of dipterans at colder and wetter sites is not necessarily because of any selective preference from the flies, but because they are less affected than other insect groups by these factors (Kearns Reference KEARNS1992). The reason may be their low energetic requirements, ability to sun-bask and use of heliotropic flowers (Arroyo et al. Reference ARROYO, PRIMACK and ARMESTO1982, Kearns Reference KEARNS1992, Kevan & Baker Reference KEVAN and BAKER1983). In contrast all variables observed for bees were greatly affected by increasing rainfall. Bees are very energy-demanding insects which require constant foraging for their nest building and offspring provisioning (Kearns Reference KEARNS1992, Warren et al. Reference WARREN, HARPER and BOOTH1988). High rainfall and low temperatures are unfavourable to their foraging (Cruden Reference CRUDEN1972, Kearns Reference KEARNS1992, Michener Reference MICHENER2000, Roubik Reference ROUBIK1989) and their nests are susceptible to fungal and bacterial diseases in high humidity conditions, which may be the reasons for their peak in diversity in dry, warm climatic zones (Michener Reference MICHENER2000, Ollerton et al. Reference OLLERTON, JOHNSON, HINGSTON, Waser and Ollerton2006, Roubik Reference ROUBIK1989). Hence, the importance of bees at the wet end of our gradient decreases drastically, mirroring the global trend in bee species richness.
Whereas the trends shown by bees and dipterans can be explained by direct responses to climate, the effects of rainfall and temperature on the minor pollinator groups is more obscure. Wasp diversity increased with rainfall, mainly due to a higher diversity of pompilids. However, other non-climatic factors such as availability of hosts, strongly determine the distribution of these species. Interestingly, although more diverse in species at wetter sites, the interactions that they established were peripheral to the community, i.e. they visited relatively specialized plants. Lepidopterans did not show any significant trend. Lacking a general response to climate, non-thermoregulatory factors seem to have a greater effect on lepidopterans (Warren et al. Reference WARREN, HARPER and BOOTH1988). Finally, regardless of the expected lower species richness of coleopterans with increasing rainfall (Janzen Reference JANZEN1973, Warren et al. Reference WARREN, HARPER and BOOTH1988), coleopterans gained importance at this end of the gradient. However, since this group was formed by only a few individuals these results should be taken cautiously.
In conclusion, the effects of climate in pollinator species distribution and interaction patterns in the West Indies are complex. As in other geographic areas, most pollinator groups are affected by climate in terms of species richness and importance. The general trends found in our work are consistent with those found in similar studies, both from tropical (Cruden Reference CRUDEN1972, Kay & Schemske Reference KAY and SCHEMSKE2003) and temperate mainland assemblages (Arroyo et al. Reference ARROYO, PRIMACK and ARMESTO1982, Devoto et al. Reference DEVOTO, MEDAN and MONTALDO2005, Kessler & Krömer Reference KESSLER and KRÖMER2000, Krömer et al. Reference KRÖMER, KESSLER and HERZOG2006, Medan et al. Reference MEDAN, MONTALDO, DEVOTO, MANTESE, VASELLATI, ROITMAN and BARTOLONI2002, Sazima et al. Reference SAZIMA, BUZATO and SAZIMA1996, Warren et al. Reference WARREN, HARPER and BOOTH1988). Most of these patterns can be directly explained by the pollinators' physiological capabilities to respond to rainfall and temperature. However, climate can also indirectly affect pollinators by controlling, for example, habitat heterogeneity, plant flowering patterns, floral productivity and phenotype, or nesting conditions (Devoto et al. in press). Understanding how rainfall and temperature affect pollinator species composition and importance is crucial for the preservation of pollination processes under future climatic change scenarios (Devoto et al. in press, Hegland et al. Reference HEGLAND, NIELSEN, LÁZARO, BJERKNES and TOTLAND2009, Ings et al. Reference INGS, MONTOYA, BASCOMPTE, BLUTHGEN, BROWN, DORMANN, EDWARDS, FIGUEROA, JACOB, JONES, LAURIDSEN, LEDGER, LEWIS, OLESEN, VAN VEEN, WARREN and WOODWARD2009). In this respect we advocate the use of complementary network measures to document any potential change in the pattern of interactions between plants and their pollinators. This study provides valuable information on pollination communities from the West Indies and on how the pollinator fauna relates to climatic factors in tropical communities.
ACKNOWLEDGEMENTS
We would like to thank Martina Stang and an anonymous referee, for helpful comments which improved the manuscript. Jens-Christian Svenning provided valuable help with the analysis. We are grateful to the Forestry and Wildlife Division, Dominica, Dept. of Forestry and National Parks, Ministry of Agriculture, Grenada, and Dept. de Recursos Naturales y Ambientales (DRNA) of Puerto Rico, for research permission. We thank Elvis Stedman (Dominica Rainforest Aerial Tram), Dean Jules (Natl. Forestry Dept., Grenada), Rubén Padrón (DRNA) and Gary Breckon (University of Puerto Rico, Mayagüez) for help with plant identification. Nico Franz (University of Puerto Rico, Mayagüez) kindly identified the Puerto Rican coleopterans and Robert Powell (Avila University, Kansas) the lizards from Grenada. Also thanks to Nancy G. L. Osler (Archbold Tropical Research and Education Centre at Springfield Plantation, Dominica), David Stemple and the Matthew family for help and advice. The project was financed by the Faculty of Natural Sciences at Aarhus Univ. (AMMG, BD, AT and LHA), Sven Fiedler and Wife Foundation (AMMG), Augustinos Foundation (BD), Knud Højgard Foundation (BD), the Danish National Science Research Council and WWF-Denmark/Novozymes (JMO).
Appendix 1. Plant species present at each site.
Appendix 2. Pollinator species and morphospecies present in our sites grouped by order and family whenever the information was available. Morphospecies names are given within each site, e.g. ‘Apidae sp. 1’ in Dominica lowland is not necessarily the same ‘Apidae sp. 1’ in Grenada lowland. Vouchers are kept at Aarhus University.
