Many factors, including climate changes, habitat loss, habitat fragmentation, diseases, and pollution, pose threats to amphibian species throughout the neotropics (Loyola et al. 2008). Anuran distributions respond to many biotic and abiotic factors, such as the availability of breeding habitats, litter cover, vegetation structure, and the structural diversity of habitats (Indermaur et al. 2010; Ernst & Rödel 2006). In tropical rainforests, topography, soil texture, leaf litter depth, and vegetation are the major factors affecting anuran species distribution (Giaretta et al. 1999; Vonesh 2001). However, anuran species with different developmental modes might respond to habitat disturbance in different ways (Loyola et al. 2008). In Amazon forests, topography and soil features affect the abundance of terrestrial-breeding species (Menin et al. 2007), whereas the distance from streams is the main factor influencing the abundance and occurrence of aquaticbreeding species (Menin et al. 2011).

In general, as in studies with most other taxa, analyses of anuran species distribution have evaluated only environmental constraints. However, studies on a few species have evaluated spatial factors related to dispersal ability (Jones et al. 2006). The neutral theory of biogeography and biodiversity posits that the patterns of abundance and distribution of species can be understood through models that consider individuals as though they were equivalent in birth, death, and dispersal rates, as well as in their competitive abilities. Therefore, species’ spatial distribution patterns, such as the distance decay of similarity in ecological communities, would be the result of stochasticity in dispersal limitation rather than properties of the species niche (Rosindell et al. 2011). Several recent studies on metacommunity dynamics have investigated the role of spatial processes in light of the predictions of the neutral theory (Siqueira et al. 2012), but few have focused particularly on conservation biology.

In the present study, we evaluated the environmental and spatial factors controlling anuran assemblages in an Amazon forest at a mesoscale. In addition to estimating the relative role of these factors in controlling overall assemblages, we separated the species with aquatic reproduction from those with terrestrial reproduction in order to evaluate whether they would respond to the same factors. Because anurans are very sensitive to environmental changes, we hypothesized that environmental control would be more constraining than spatial control, and that species with different types of reproduction would differ in their relationships to environmental and spatial variables. More specifically, we predicted that species with terrestrial reproduction would respond more strongly to spatial constraints, because their distribution is more restricted by limitations on dispersal than by dependence on the availability of water bodies, and that species with aquatic reproduction would respond more to niche factors associated with water availability.

The anuran dataset contained 29 species, of which 20 have aquatic reproduction and nine have terrestrial reproduction. Seven species with aquatic reproduction that are uncommon in the sampling area occurred only sporadically (i.e. in fewer than four plots), and we omitted them from the analyses. Therefore, only 13 species with aquatic reproduction were analyzed (Table S2, supplementary material online).

The environmental and spatial variables explained around 39% of the variation of the anuran assemblages (Fig. 2). Both the environmental [a] and spatial [c] fractions were significant (p<0.001). The spatial component was larger than the environmental component (23.8% and 10%, respectively), and the shared component was 5.3%. The environmental variables related to anuran assemblage distribution were: distance to the nearest stream, soil pH, number of trees in the plot, and soil clay content. We retained nine spatial variables for the partial RDA model.

Fig. 2.

Variation partitioning results for anuran assemblages considering all species, those with aquatic reproduction and those with terrestrial reproduction. Results based on a partial redundancy analysis. Values shown are adjusted R2.

(0.87MB).

Both sets of predictors explained around 34% of the variation in the assemblage of anurans with aquatic reproduction (Fig. 2). Both the environmental [a] and spatial [c] fractions were significant (p<0.001). Environmental variables (distance to the nearest stream and number of trees) alone explained most of the variation (26.6%) in the assemblages of anurans with aquatic reproduction. The only three spatial variables retained accounted for merely 4.9% of the variation. The shared variation was equal to 2.9%.

For the assemblage of anurans with terrestrial reproduction, the spatial and environmental variables explained around 40% of the variation (Fig. 2). Both the environmental [a] and spatial [c] fractions were significant (p<0.001). Contrarily to the anurans with aquatic reproduction, the spatial variables (eight spatial variables retained) better explained the variation in the assemblages of anurans with terrestrial reproduction, and these variables accounted for 30.3% of the variation. The two environmental variables retained (soil clay content and pH) accounted for only 5.4% of the variation; the shared fraction was 3.6%.

Our models explained at least 35% of the between-plot variation in these anuran assemblages. For all species pooled, the spatial patterns were stronger than the effects of the environmental variables included in the analyses. However, the pattern was highly dependent on the reproductive mode of the species. Species with aquatic reproduction responded more to variation in environmental factors (i.e. probable niche control), whereas species with terrestrial reproduction showed spatial patterns that were not associated with the environmental variables included in the analyses. In addition the environmental variables associated with each group differed; soil pH and clay content were associated with species with terrestrial reproduction, while stream distance and number of trees in the plot were associated with species with aquatic reproduction.

Spatial patterns may arise from two main sources. Some spatially structured environmental variables that are important for anurans may not have been measured, and consequently were not included in the models (Landeiro & Magnusson 2011). This implies that the variation in species distribution due to such environmental variable will be attributed exclusively to spatial variables. Another possibility is that species are really not dependent on environmental constraints, and the observed spatial patterns might be attributed to dispersal limitation or other endogenous processes, such as differential reproduction. We measured environmental variables that are commonly found to affect the species distribution of several Amazonian animal groups (Landeiro et al. 2012). Therefore, we believe that the spatial pattern observed in species with terrestrial reproduction is more closely related to intrinsic dispersal processes than to the lack of important environmental variables. Menin et al. (2007) also suggested that dispersal limitation is a potential explanation for distribution patterns of anurans with terrestrial reproduction; however, they did not evaluate the anuran assemblage in relation to spatial predictors. Further studies on genetic data are crucial to confidently attribute differentiation to dispersal limitation, although metacommunity studies generally attribute spatial patterns to dispersal processes (Siqueira et al. 2012).

Assuming the neutral scenario for terrestrial anurans, in which species are equivalent competitors and the environmental characteristics are unimportant, it would be expected that geographic distance is important in predicting the similarity in species composition among sites (Diniz-Filho et al. 2012). Therefore, a search for environmental or biological surrogates/indicators would be unproductive, and geographic distance among sites would be a paramount predictor for this particular neutral assemblage. When species-sorting mechanisms shape local assemblages (Van der Gucht et al. 2007), environmental variables explain compositional similarity and could be reliable proxies for species’ turnover (Ferrier et al. 2007). Therefore, the conservation of anuran species can depend strongly on life-history traits, such as the type of reproduction (Loyola et al. 2008). The inclusion of anuran developmental modes in analyses performed to select priority areas for the conservation of this order could improve comprehensiveness of the selection process. Compared to the usual assessments that do not consider these life-history traits, analyses that include reproductive requirements would improve the conservation of species that breed in aquatic habitats. Otherwise, the procedures used to select priority areas tend to favor species with terrestrial development (Loyola et al. 2008).

Since the severity and frequency of droughts affecting the Amazon region will likely increase (Lewis et al. 2011), together with other potential environmental changes caused by global warming, the effects on anurans might be severe for species that are dependent on aquatic habitats to reproduce (Di Minin & Griffiths 2011), and for those that rely on moist soils for terrestrial reproduction. Most terrestrial-breeding frogs occur in humid areas (Duellman 1988). Our results demonstrated that species with aquatic reproduction respond strongly to environmental conditions, mainly the distance to the nearest water source, indicating that changes in the availability of such sources will affect their occupation of the landscape. In addition, little of the beta diversity observed in terrestrially breeding frog assemblages is associated with habitat variation (Menin et al. 2007).

Zimmerman & Bierregaard (1986) suggested that frogs are not good indicator species for fragmentation studies, because the presence of suitable water bodies for reproduction generally determines their distributions. However, this generalization does not apply to the terrestrial-breeding species (Menin et al. 2007). The effects of climate change are likely to differ in quatic-breeding and terrestrial-breeding frogs. Whereas the distribution of aquatic-breeding frogs across the landscape is possibly influenced by changes in the distribution of water bodies, aquatic-breeding species are not expected to become locally extinct until almost all water bodies are lost, and they could potentially be maintained by artificial water sources. This may explain why aquatic-breeding species are common in the driest areas of Amazonia (Duellman 1988). In contrast, few or no terrestrial-breeding species occur in these areas, suggesting that their limited dispersal abilities, together with their requirement for humid climates, might result in their complete disappearance from the landscape. Considerations about the vulnerability of different species will have to take into account differences in the requirements of the guilds and the scale of the changes.

These findings are relevant to the development of conservation strategies and biological monitoring programs under a metacommunity framework (Siqueira et al. 2012), because species that respond strongly to environmental controls may be more sensitive to specific threats, such as the conversion of riparian areas. Conversely, species that do not have restrictive needs for reproduction, but that present strong associations with humid forests could be better indicators of general environmental degradation associated with climate change or activities such as selective timber harvesting. Therefore, researchers, biomonitoring agencies, and biodiversity managers using different taxonomic groups with different species traits should account for these differences and remain aware of the implications of their results.