Introduction

Understanding the factors that mold species distributions and communities has a long tradition in ecology and biogeography (Wallace 1876; Clements 1916; Phillips 1931). Recently, this topic has greatly benefited from technical and statistical developments, notably those that allow the prediction of the nature and distribution of species assemblages under different environmental conditions (Ferrier and Guisan 2006). Given the current perspective of climate change, this matter is critical for yielding realistic forecasts of the responses of species and communities to global change scenarios (Adler and HilleRisLambers 2008). However, whereas a large number of studies have focused on abiotic drivers, such as climatic (Guisan and Zimmermann 2000) or edaphic (Alvarez et al. 2009) factors, it is widely recognized that biotic factors can additionally strongly influence the distribution and assemblage of species (Pulliam 2000; Lortie et al. 2004). For example several studies have emphasized the importance of competitors and facilitators (Leathwick and Austin 2001; Heikkinen et al. 2007, Pellissier et al. 2010a) in delimiting species ranges. Pollination is among the main biotic factors that control the ecology, distribution, and assemblage of vascular plants. Whereas the pollen of gymnosperm species is predominantly dispersed by wind, the majority of angiosperms are dispersed by animal vectors (Barth 1991). Despite the recognition of pollination as a major facet of plant ecology, the importance of pollinators for predicting plant distribution has not been thoroughly investigated in recent decades. There is a strong need to characterize plant–pollinator interactions at large spatial scales and especially with respect to dynamic communities, whose compositions and patterns of relative species abundance vary in time and space.

Recent research on pollinators, while considering spatial variation, has mostly focused on how coevolution with pollinators can generate within-species geographic variation in the morphology of plant species, leading to plant speciation. In contrast, studies on the ecological links between plant and pollinator species have generally focused on a limited number of taxa and sites, with temporal replicates instead of observations across geographic space. Yet, the intensity and nature of interactions between plants and their pollinators often varies spatially, possibly because the ranges of interacting species do not completely overlap (Thompson 1988) and can cause plant fitness variation across its range. For instance, Espíndola et al. (2011) showed, by examining the entire distribution range of the lure-and-trap Arum maculatum, that pollination was not accomplished by a single specialized fly species as previously thought; instead, pollination was achieved by two fly species that showed distinct regional relative densities, despite being sympatric over a large portion of the plant’s distribution. Segregation of sites with either one or two pollinators followed the cline of environmental gradients related to precipitation. But even for less specialized species, changes in pollinators’ density along environmental gradients can, in turn, affect the pollination of the species and cause pollen limitation (Gómez et al. 2010). In our opinion, the lack of spatial replicates in biotic pollination studies can be mostly explained by the large sampling effort required to properly describe the plant–pollinator network throughout the entire distribution range of a plant. Another frequent bias in such studies is the absence of an accurate examination of pollen transfer, which is associated with plant fitness (Alarcón 2010). Observing the biotic vectors that visit flowers does not necessarily provide information on the nature and efficiency of the pollination process (e.g. the visitors could steal nectar without pollinating the plant, or local change in pollinator behavior could modify the intensity of pollination). Although a plant may attract a wide range of flower visitors, only a few groups can act as efficient pollinators (Bawa 1990). Consequently, drawing conclusions about the dependency of plants on particular biotic vectors may be misleading if pollination efficiency is not examined for each floral visitor (Reynolds and Fenster 2008; Kay and Sargent 2009).