Effects of bromeliad flowering event on the community structuring of aquatic insect larvae associated with phytotelmata of Aechmea distichantha Lem . ( Bromeliaceae )

Aim: We aimed to understand how aquatic insect larvae communities associated with bromeliad phytotelmata are affected by plant architecture, predators and resources (local factors), and by geographical distance (regional factors) in two different plant phenological phases. Bromeliad flowering results in plant structural changes, which favours insect dispersal. Considering that local and regional factors may affect the community of aquatic insect larvae, we expected that composition, beta diversity and the importance of those factors would differ in the vegetative growth and flowering phases. Methods: We performed six samplings of the bromeliad associated fauna in 2010, three during the first semester vegetative growth phase and three during the second semester flowering phase. In each sampling, we collected 12 plants along the rocky walls with similar location distribution, with a total of 72 bromeliads studied. Results: Although beta diversity (PERMDISP) did not differ between vegetative growth and flowering, NMDS followed by ANOSIM showed that composition was significantly different in the distinct phenological phases. IndVal results showed that three Diptera morphospecies were discriminant of the vegetative growth phase. In addition, pRDA revealed differences in the relative contribution of local and regional factors to explain insect larvae community structure. During the flowering phase, local factors predominated, while during vegetative growth, regional factors were more important. Conclusion: Differences in dispersal rates between the two phenological phases, likely due to adult insect pollination and further oviposition, influenced community structuring. Therefore, flowering events account for differences not only in the composition, but also in community structuring of aquatic insect larvae inhabiting the phytotelmata of Aechmea distichantha Lem. (Bromeliaceae).


Introduction
Natural communities may be regulated by local factors, such as competition, disturbances, biotic and abiotic variables, and regional factors, such as geographical distance among patches, dispersal capacity among habitats and climate conditions (Hillebrand & Blenckner, 2002;Cottenie et al., 2003;Paradise et al., 2008).These processes, acting in broad temporal and spatial scales, are important in determining diversity patterns and define the aspects of the regional species pool from where local communities are assembled (Caley & Schluter, 1997).
The area in which the species pool is inserted has a fundamental role in the analysis of local and regional factors since it affects richness estimates (species-area relationship) and the scales in which species interact (ecological factors ;Hillebrand & Blenckner, 2002).However, area extension is generally difficult to be determined (Srivastava, 1999), thus, most studies use a large randomly chosen area (Caley & Schluter, 1997) or an area that is naturally delimited by landscape geography (Oberdorff et al., 1998).Therefore, the difficulty in manipulating communities reduces the possibility of testing the predictions on the influence of local and regional factors on organismal diversity (Hillebrand & Blenckner, 2002).
Flowering events are accompanied by structural modifications in the three-dimensional architecture of bromeliad leaf rosette, which fold back their leaves allowing pollinators a better access to the inflorescence (Romero & Vasconcellos-Neto, 2005b).Those modifications affect various characteristics of those microhabitats, such as the amount of water and organic matter accumulated and evapotranspiration rates (Zotz & Thomas, 1999).Therefore, bromeliads become less suitable to some of the associated organisms (i.e.predators: spiders -Romero & Vasconcellos-Neto, 2005a), consequently altering community structure (Srivastava, 2006;Gonçalves-Souza, et al., 2010, 2011).
In addition, flowering events enhances insect visitation, which are pollinators or feed on floral nectar (Frank & Lounibos, 2009).The presence of floral stimuli produced by the plants (floral volatiles) reach long distances attracting pollinators (Reisenman et al., 2010) increasing dispersal rates and colonization among plants (Kneitel & Miller, 2003).During the flowering phase, these stimuli (i.e. floral volatiles and nectar) are critical for insects to choose suitable oviposition sites, leading to higher rates of oviposition in the plants (Reisenman et al., 2010) and decreasing the influence of spatial distance on the structure of the aquatic larvae associated to phytotelmata (Paradise et al., 2008).
Beta diversity (Whittaker, 1960;1972) can be defined as the variation in species composition among sampling units within a region (Anderson et al., 2006;Hill et al., 2017).Several factors such as the habitat structure, degree of isolation, and the dispersal capacity of the organisms may affect beta diversity (Gering & Crist, 2002).In bromeliads, habitat structure influences the available niche space and offers a wide range of essential elements for the animals, such as reproduction, shelter, and foraging sites (Romero & Vasconcellos-Neto, 2005a;Romero, 2006).Although the three-dimensional vegetation structure is recognized for altering the patterns of beta diversity, how this influence occurs remains poorly understood (Zellweger et al., 2017).Dispersal among habitats may affect species composition and diversity in different manners, at both local and regional scales (Kneitel & Miller, 2003).High dispersal rates assure a constant input of new colonizers in new areas (Winegardner et al., 2012;Heino et al., 2015), leading to the homogenization of community structure at a local scale, and consequently, to a decrease in beta diversity (Cadotte & Fukami, 2005).On the other hand, low dispersal rates prevent species to reach all habitats, even the ones suitable for their development (Winegardner et al., 2012;Heino et al., 2015), increasing beta diversity.
Considering that plant architecture, the presence of predators and resources (local factors) and geographical distance (regional factors) may affect the community structure of aquatic insect larvae in bromeliads, we expect that composition, β-diversity and the relative importance of local and regional factors would be different between vegetative growth and flowering phases.We tested the hypotheses that I) insect larvae composition is different between vegetative growth and flowering phases; II) β-diversity (dissimilarity) is lower during flowering, since this event increases dispersal of insects among plants and consequently enhances oviposition, leading to more homogeneous larvae communities; III) during flowering, local factors would be more important for the community structuring of aquatic insect larvae, since dispersal rates would be sufficient to allow an environmental control based on niche differentiation, while in the vegetative growth phase regional factors would be more influential, considering that dispersal limitation would result in spatially structured communities within the bromeliad phytotelmata.

Study area
The study area is located in the Paraná River basin, between the mouths of Paranapanema and Baia rivers.Sampling was performed at the left bank of the Paraná River (Porto Rico, PR, Brazil -22°45'53.5"S53°15'27.2"Wand 22°43'11.5"S53°10'46"W; Figure 1).The region is characterized by the asymmetry of both sides of the valley, with an elevated left bank marked by rocky walls and sparse flooded areas (Souza Filho & Stevaux, 2004).These rocky walls provide a suitable habitat for the development of several epiphytic plants, such as the bromeliads.
Vegetation of the Upper Paraná River basin is inserted in the Atlantic Forest biome in fragments of semideciduous forests forming strings of riparian vegetation near the border of the Paraná River (Campos & Souza, 1997;Souza & Kita, 2002).Precipitation varies from 1400 to 1600 mm/year (maximum rainfall during summer), relative annual Acta Limnologica Brasiliensia, 2017, vol. 29, e111 humidity is approximately 70% and the average temperature is 24 °C (IBGE, 1990; IAPAR, 2000).

Sampling
Aechmea distichantha Lem.(Bromeliaceae) was chosen because of the high density of this plant in the region.This facultative epiphytic bromeliad is widely distributed in South America, occurring from the sea level to 2,400 m altitude (Smith & Downs, 1979).According to Reitz (1983), this Bromeliaceae is 40-100 cm height with few leaves (15 to 25) which grow in a rosette, forming water-collecting cisterns.The foliage has margins covered with thorns, with an acute ending.Water volume contained in the bromeliad tanks are on average 200 cm 3 .Flowering usually occurs between June and September, a period in which the plant exhibits a densely flowered inflorescence (70 to 330 flowers), lasting for 20 to 30 days (Scrok & Varassin, 2011).
We performed six samplings in 2010, three during the first semester -vegetative growth phase -and three during the second semester -flowering phase.In each sampling, we took 12 plants along the rocky walls with similar location distribution, with a total of 72 bromeliads studied.For each plant, geographical coordinates were taken.Using these data, a distance matrix was constructed, considering the distance (meters) from each plant in relation to the others (Spatial component).Measurements were taken using Google Earth.
Plants were manually removed from the rocky walls, wrapped in plastic bags and taken to Nupelia/University of Maringá Field Station, where the following measurements of the morphometric parameters were taken: height and width of the plant, perimeter, and height of the bromeliad tank and leaf number.Assuming that bromeliads approximated a geometric shape of a cone, we calculated the plant total volume (V plant ) and bromeliad tank total volume (V tank ) as cone volume, according to Armbruster et al. (2002) (Equation 1): where to calculate the total volume of the plant (V plant ), the radius of the plant (r plant ) was obtained by dividing plant width by two, and h plant is the height from the base to the top of the central leave.
To calculate bromeliad tank total volume (V tank ), bromeliad tank height (h tank ) was measured from the plant base to the interlocking of the more internal leaves.Bromeliad tank radius (r tank ) was calculated using the perimeter (plant perimeter/2 × π) (Figure 2).
Leaf number in each plant was considered a proxy of its complexity, since each leaf axil constitutes a discrete water body, compartmentalizing the internal space inside the bromeliad (Armbruster et al., 2002).Water contained within each plant was carefully removed and inspected for macroinvertebrates (including aquatic insect larvae, spiders, ants, among others).Water volume from each bromeliad was further measured with graduated cylinder.
Sampling of the associated bromeliad fauna followed the protocol described by Armbruster et al. (2002): leaves were cut near the base of the plant, removed, individually checked for more macroinvertebrates and then washed with distilled water inside a white tray.Associated fauna was fixed with 70% alcohol.Insects were sorted, counted and identified to larval and adult morphospecies in the zoology laboratory at Maringá State University.Diagnostic features used for assigning morphospecies were wing venation, structure of mouth parts, chaetotaxy, antennal segmentation and structure, tarsal structure and overall body form, according to Armbruster et al. (2002).The classification of insect morphospecies into each order was based on Borror et al. (1989).Although the identification to morphospecies is common in studies on macroinvertebrate fauna associated with phytotelmata, there are some problems related to its use.For example, the adult and larvae of the same species are usually classified as different morphospecies.However, these different development stages may have different ecological niche, which justifies the identification of different development stages as different morphospecies to better comprehend the functional role of these organisms within phytotelmata (Armbruster et al., 2002;Araújo et al., 2007;Jabiol et al., 2009;Montero et al., 2010).
Spiders (predators) were counted and identified at the family level in the Arachnology laboratory of Acta Limnologica Brasiliensia, 2017, vol. 29, e111 Museu Paraense Emilio Goeldi (Belém, PA).Spiders were identified to morphospecies and identification was primarily based on the shape of adult female reproductive organs -epigyne.When spider families were composed only by juveniles, we considered those families to be composed of at least one taxon, and this criterion was used in all statistical analyses.Based on ecological characteristics of each family, spiders were grouped in "web weavers" and "hunters".
To estimate ciliate density, water samples from each plant were analysed in vivo using an optic microscope (for more details, see Buosi et al., 2015).

Study system
Spiders and ants are among the most commonly arthropods found foraging on Bromeliaceae leaves (Montero et al., 2010;Hénaut et al., 2014) and may play an important role in structuring phytotelmata communities (Romero & Srivastava, 2010;Céréghino et al., 2011).Considered as predators in these microhabitats (Romero & Srivastava, 2010;Gonçalves et al., 2017), spiders and ants may affect communities associated with bromeliad leaves and tanks, through the direct reduction in the abundance of prey (consumptive effects), or by altering the behaviour and use of these habitats by these organisms (non-consumptive effects) (Werner & Peacor, 2003;Hill & Weissburg, 2013).
Associations between spiders and bromeliads are common in Neotropical regions (Romero, 2006) and the presence of these predators on the plants may repel pollinators, reducing bromeliad fitness (Gonçalves-Souza et al., 2008).The presence of spiders in bromeliads could lead to a reduction in invertebrate diversity and abundance, changing the composition of the associated fauna, from a community dominated by aquatic insect larvae to a community dominated by other aquatic invertebrates (i.e.ostracods and oligochaete; Romero & Srivastava, 2010).Therefore, the presence of these predators may affect not only aquatic organisms with complex life cycles (i.e. insects), but also those spending their whole life inside phytotelmata (Romero & Srivastava, 2010).
Several ant species use bromeliads as shelter and sites for nest construction, and the interaction between the ants and these plants may be species-specific (Céréghino et al., 2011;Leroy et al., 2012;Talaga et al., 2015;Gonçalves et al., 2017).Although the presence of these organisms in bromeliads may result in some advantage for these plants, such as protection against herbivory or seed dispersal (Leroy et al., 2012), the predatory behaviour that ants exert on potential pollinators and colonizers may alter the structure of the associated aquatic and terrestrial communities (Gonçalves et al., 2017).
Ciliates have a key role in the flow of energy and matter within bromeliad phytotelmata, being more efficient than metazoans in the nutrient remineralization in the water column (Carrias et al., 2001).In this environment, ciliates can be considered as a proxy of organic matter (Carrias et al., 2001;Petermann et al., 2015b), due to their efficiency in decomposing detritus (Kneitel & Miller, 2002, 2003).Moreover, ciliates are in the intermediate trophic levels of phytotelmata food webs, being grazed by aquatic insect larvae (Kneitel & Miller, 2002).
Considering the importance of the several factors above mentioned for the community structuring of aquatic insect larvae in phytotelmata, spider and ant abundances (predators), ciliate density (resources) and plant morphometric parameters (plant height, plant total volume, leaf number, bromeliad tank height, bromeliad tank total volume and water volume) were considered as local factors (Environmental component) in this study.

Data analysis
To observe the effect of flowering events on the composition of insect larvae associated with A. distichantha (hypothesis I), we performed non-metric multi-dimensional scaling (NMDS; Clarke, 1993) using Jaccard distance (presence/absence data).NMDS rearranges objects in a space with a particular number of dimensions, reproducing the observed distances.Distortion of the two-dimension resolution is expressed by S value (stress): the nearer this value is to zero, the better is the adjustment among original distances and the configuration obtained by the analysis (Legendre & Legendre, 1998).Analysis of similarities (ANOSIM; Clarke, 1993) with 9,999 permutations was performed to verify statistical differences in the composition patterns observed in NMDS.ANOSIM is a procedure of non-parametric permutations based on the ranking of the dissimilarity matrix generated by NMDS, comparing the degree of separation within and between sample groups using R statistics.If R=0 there are no differences in community composition between groups, while R=1 indicates complete distinction between communities (Quinn & Keough, 2002).These analyses were performed Effects of bromeliad flowering event… Acta Limnologica Brasiliensia, 2017, vol.29, e111 using PAST 2.17 (Paleontological Statistics Software Package for Education and Data Analysis - Hammer et al., 2001).
Indicator morphospecies of vegetative growth and flowering phases were identified using Indicator Species Analysis (IndVal; Dufrêne & Legendre, 1997).This analysis combines relative abundance and frequency occurrence of each species, generating an indicator value varying between zero and one.The value is near one when all individuals of a species occur in all plants in a certain phenological phase.Significance (P<0.05) was tested by 1,000 random permutations.
Beta diversity of insect larvae associated with phytotelmata was measured through permutation test of multivariate homogeneity of groups dispersions (PERMIDISP, Anderson et al., 2006).The biological matrix with presence/absence data of insect larvae was transformed into a distance matrix based on Jaccard distance.In this analysis, beta diversity is measured as the mean dissimilarity of a sample in relation to the group centroid (Anderson et al., 2006), in our case, flowering and vegetative growth phases of A. distichantha (hypothesis II).
The relative importance of local and regional factors in the diversity of aquatic insect larvae associated with A. distichantha (hypothesis III) was evaluated through Partial Redundancy Analysis (pRDA; Legendre & Legendre, 1998;Legendre et al., 2005).RDA is an extension of multiple regression analysis, with a dependent variable (insect larvae composition) and different explanatory matrices: local factors (environmental component using morphometric parameters, ciliate density and the abundances of ants, hunter spiders and web weaver spiders) and regional factors (spatial component using geographical distances).A Principal Coordinates of Neighbour Matrices (PCNM - Borcard & Legendre, 2002;Borcard et al., 2004;Diniz-Filho & Bini, 2005;Dray et al., 2006;Griffith & Peres-Neto, 2006) was applied to a distance matrix (meters) and scores were used as explanatory variables of the spatial component.Abundance data was Hellinger transformed (Legendre & Gallagher, 2001).The significance of the components was tested through 9,999 Monte Carlo randomizations (Borcard et al., 1992) and R 2 adjusted values were considered.IndVal analysis, PERMDISP, PCNM and pRDA were performed in R software (R Development Core Team, 2013).

Results
Considering both phenological phases, we collected 3,481 insect larvae associated with A. distichantha, distributed in four orders and 16 morphospecies (Table 1).Diptera and Lepidoptera were the most representative orders, with seven morphospecies each.Coleoptera and Neuroptera showed only one morphospecies.Out of all the morphospecies registered, we verified that nine were common to both phenological phases (vegetative growth and flowering).Flowering showed the highest number of morphospecies (n=13), out of which four were exclusive.During vegetative growth, three morphospecies were exclusive, with a total of 12 morphospecies in this phase.
Regarding the contribution of each order to total abundance, Coleoptera was the most representative, with around 96% total abundance registered (3,349 individuals), consisting of the Scirtidae family (vegetative growth: 1,446 ind.; flowering: 1,903 ind.).Highest values of abundance of the aquatic larvae community were registered in the flowering phase, with 1,959 individuals (56% of abundance).In the vegetative growth phase, 1,522 individuals were collected (44% of abundance).NMDS results indicated a separation of morphospecies of aquatic insect larvae between vegetative growth and flowering phases (Figure 3; NMDS stress: 0.25).ANOSIM results confirmed significant differences in community composition between the phenological phases (R=0.12;p<0.001).
PERMDISP results showed no significant differences in beta diversity of insect larvae between the phenological phases (Figure 4), indicating that the variation in taxa composition is similar between flowering and vegetative growth phases.
pRDA revealed that different components affected insect larvae diversity in phytotelmata during vegetative growth and flowering phases (Table 2).As expected, local factors (environmental component) were more important for larvae community during flowering (19%), with the contribution of plant morphometric parameters (plant height and total volume) together with ciliate density as the main responsible for community structuring in this phenological phase (Table 3), whereas regional factors (spatial component) were not significant.
During vegetative growth, abundance of web weaver spiders and bromeliad tank total volume (environmental component), together with PCNMs 6 and 9 (spatial component) were selected to explain community variation (Table 3).However, only the spatial component (regional factors; 6%) was significant and explained community structuring of aquatic insect larvae during this phenological phase (Table 2).

Discussion
Throughout the study, we found 3,481 aquatic insect larvae in the phytotelmata, distributed in four orders and 16 morphospecies.Species composition was significant different between vegetative growth and flowering (hypothesis I), whereas beta diversity was not significant different between the two phenological phases (hypothesis II).The relative contribution of local and regional   Statistically significant results shown in bold font.factors in structuring the insect larvae community (hypothesis III) was different between vegetative growth and flowering phases.During flowering, local factors predominated, whereas during vegetative growth, regional factors were more important.
Significant differences regarding insect larvae composition between vegetative growth and flowering phases may be explained by the differences in plant architecture and insect dispersal among bromeliads in the two phenological phases.During flowering, the structure of leaf rosette in bromeliads is modified to offer better access to pollinators (Romero & Vasconcellos-Neto, 2005a, b).These modifications alter several plant morphological parameters (Zotz & Thomas, 1999), influencing the attributes of the insect larvae community and their predators in phytotelmata (Armbruster et al., 2002).In addition, insect dispersal enhanced by pollination during flowering, which leads to an increase in oviposition rates (Reisenman et al., 2010), counteracts the limited dispersal during vegetative growth, constituting another factor responsible for differences in species composition between phases.However, those factors do not seem to influence beta diversity patterns which, contrary to expected, were not different between the phenological phases.
Although phytotelmata may harbour many of the main aquatic insect orders, Diptera is considered the most common, with over 20 families present (Williams, 2006).Indeed, we found that three Diptera morphospecies were discriminant in the vegetative growth phase.Several morphological and behavioural traits of this insect order suggest long term association with phytotelmata (Williams, 2006).Diptera larvae are very abundant in bromeliads (Araújo et al., 2007), pitcher plants (Baiser et al., 2011) and tree-holes (Blakely et al., 2012), and are considered the main predators in the phytotelmata of several bromeliad species, where they exert a great influence on microorganisms (Walker et al., 2010;Baiser et al., 2011).Although several studies show that zygoptera larvae are the main aquatic predators within phytotelmata (Petermann et al., 2015a;Romero et al., 2016), in the bromeliads used in our study no larval forms of these organisms were observed.Paradise (2000) suggest that alterations in pH due to the decomposition of organic matter by Scirtidae larvae (Coleoptera) within phytotelmata may lead to a reduction in the mortality and an increase in the abundance of Diptera larvae in this habitat.In fact, we found very high abundances of Scirtidae larvae in A. distichantha phytotelmata.
Insect larvae community structuring was influenced by different factors in the two phenological phases.During flowering, ciliate density (proxy of resource availability) and the morphometric parameters plant height and plant total volume were the most important local factors.Plant parameters are directly related to plant architecture and habitat complexity -the main predictor of arthropod distribution in vegetation (Gonçalves-Souza et al., 2011).According to Lawton (1983), seasonal changes may considerably affect plant architecture.In bromeliads, flowering events cause profound alterations in the threedimensional leaf rosette structure, which opens to guarantee better access of pollinators to plant inflorescence (Romero & Vasconcellos-Neto, 2005a, b).These modifications alter habitat conditions, such as water availability, resource capture (Zotz & Thomas, 1999) and the presence of predators (Romero & Vasconcellos-Neto, 2005b).
By actively participating in nutrient cycling and bacterial population control in aquatic environments, ciliates are considered important components within food webs in those ecosystems (Sherr & Sherr, 2002;Durán-Ramírez et al., 2015).Carrias et al. (2001) point out that these organisms are key components to phytotelmata metabolism, since they are more efficient than metazoan in nutrient remineralization, that are released in the water column of this microhabitat and may be absorbed by the plants during development.This efficiency in organic matter degradation leads to an increase in the abundance of those microorganisms with the large amount of detritus input in phytotelmata (Kneitel & Miller, 2002;2003), thus they are considered as a proxy of organic matter in this habitat (Carrias et al., 2001;Petermann et al., 2015b).Furthermore, since they are in the intermediate level of food webs in those ecosystems, ciliates serve as a resource to insect larvae in phytotelmata (Kneitel & Miller, 2002).
Therefore, insect larvae community structuring during flowering may be explained by the alterations in plant structure in response to the flowering event itself, which changes the physical characteristics of the plants, besides enhancing the capture of organic matter by the bromeliad, increasing ciliate abundance.Thus, the combined effects of those factors lead to modifications in the community of insect larvae in this phenological phase.Acta Limnologica Brasiliensia, 2017, vol. 29, e111 In contrast to the flowering phase, aquatic insect larvae during the vegetative growth phase were structured by regional factors (spatial component).Spatial distribution of plants in the environment may affect the community attributes of associated arthropods (i.e. species richness: Gonçalves-Souza et al., 2011;and abundance: Hanski, 1982), since this distribution directly influences plant exposure to organisms, consequently enhancing colonization (Neuvonen & Niemelä, 1983) and dispersal of associated fauna.
According to Kneitel & Miller (2003), dispersal among local communities may have a variety of effects on species diversity and composition.For instance, an increase in dispersal rates -as we expected during flowering -may lead to an increase in the richness and abundance of the while decreasing variation among communities in a certain area (Kneitel & Miller, 2003).On the other hand, during vegetative growth, reduced dispersal may lead to an increase in the importance of the spatial component in community structuring.
Differences in dispersal rates between vegetative growth and flowering phases could be explained by the increase in resource availability for adult insects, due to bromeliad inflorescence.During the reproductive phase, plants produce floral stimuli (floral volatiles - Reisenman et al., 2010) and offer rewards to pollinators (Nicholls & Altieri, 2013), generally nectar or pollen (Labandeira, 2011).Concerning the interactions between bromeliads and insects, other resources may be offered, such as resin and other material for nests, shelter, foraging, mating and oviposition sites (Kevan & Baker, 1983;Labandeira, 2011).Accordingly, this wide range of resources during flowering becomes a decisive factor for insect larvae fitness, in part due to the choice of oviposition sites by adult insects -active dispersers of this community (Basset & Novotny, 1999;Gonçalves-Souza et al., 2014).During vegetative growth, the absence of those rewards results in a decline of insect movement among plants and oviposition.Thus, we suggest that insect behaviour as a function of bromeliad flowering determines aquatic larvae community structuring in both phenological phases.However, since our samplings were performed in distinct seasons, further experimental studies are necessary to disentangle the effects of seasonal fluctuation and the phenological phase of plants on invertebrate fauna.

Conclusion
The hypothesis that community composition of aquatic insect larvae is different between the phenological phases was corroborated.Variation in this attribute could be attributed to differences in bromeliad architecture, insect dispersal and higher rates of oviposition among plants in the phenological phases, which is likely the result of rewards offered by the plants during the flowering phase.However, those factors may not seem to affect beta diversity patterns, which were not different between vegetative growth and flowering phases, contrary to expected in hypothesis II.Finally, our third hypothesis regarding the relative importance of local and regional factors in structuring the community of aquatic insect larvae was corroborated.During flowering, local factors were more important in community structuring, whereas during vegetative growth, regional factors predominated in explaining community variation.

Figure 1 .
Figure 1.Map showing the sampling sites at the left bank of the Paraná River.

Figure 2 .
Figure 2. Morphometric parameters measured in the plants (A) and bromeliad tank (B).Dashed lines represent plant measures used in our study.

Figure 4 .
Figure 4. Distances to centroid in vegetative growth and flowering phases.

Figure 3 .
Figure 3. Non-metric multi-dimensional scaling (NMDS) ordination for phenological phases (vegetative growth and flowering), based on composition of insect larvae associated with A. distichantha in rocky walls.

Table 1 .
Morphospecies occurrence of aquatic insect larvae found in A. distichantha of the rocky walls at the left bank of the Paraná River (PR, Brazil) during both phenological phases.

Table 3 .
Environmental and spatial variables selected in pRDA analysis during each phenological phase (vegetative growth and flowering) of A. distichantha.

Table 2 .
Variance partitioning showing the relative contribution of [a] pure spatial component, [b] shared component, and [c] pure environmental component.Residuals: non explained variation.X1= Spatial model and X2= Environmental model.