Species morphology better predicts plant-hummingbird interactions across elevations than nectar traits

Species traits greatly influence interactions between plants and pollinators where floral nectar is the primary energy source fostering this mutualism. However, very little is known about how nectar traits mediate interactions in pollination networks compared to morphological traits. Here, we evaluated the role of morphological and nectar traits in shaping plant-hummingbird interaction networks along an elevation gradient. For this, we assessed patterns in floral phenotypic traits and network properties of plant species across elevations in Costa Rica. We also analysed whether plant species with generalised flower traits are ecological generalists and how morphological trait matching vs nectar traits affect interactions. We found marked variation in floral phenotypic traits and flower abundance of hummingbird-visited plant species across ten sites along the elevation gradient. We did not find evidence for a relationship between flower morphology and nectar traits or between morphological and ecological generalisation of plant species. Plant-hummingbird interaction frequency increased when the lengths of hummingbird bill and flower corolla were similar, indicating morphological matching, whereas nectar traits were unrelated to interactions. While nectar may play a difficult-to-detect secondary role within plant-hummingbird networks, our results reinforce the idea that morphological matching is an important factor structuring ecological communities.

Methods
Next, we include a summary of the methods and analyses we used in our study

Species morphological traits

We included in our analysis morphological traits of interacting species that have been reported to affect plant–hummingbird interactions. For plants, we measured total corolla length, corolla curvature and width of the corolla opening for 59 plant species. For hummingbirds, we measured total bill length (distance from tip to nostril) and curvature of 220 individuals belonging to 22 species (n = 10 per species). Measurements were taken from photos of fresh flowers and mist-netted hummingbird individuals in the field, complemented with museum specimens, and then processed with ImageJ software.

Nectar traits and flower abundance

We quantified daily nectar production and concentration per flower for common plant species visited by hummingbirds in the understory. We sampled these nectar traits from flowers that were bagged with nylon netting (~0.5-1 mm mesh size) prior to anthesis to prevent visitation by animal pollinators during 24 h. After this period, nectar was extracted with microcapillaries (10 and 15 µl capacity), obtaining its volume, and sugar concentration was measured by using a calibrated pocket refractometer (REED R9500 Brix Refractometer; range concentration 0-32%, g sugar per 100 g nectar). To assess patterns in hummingbird floral resources across elevations, in addition to quantifying nectar, we counted open flowers in these plants, identified them to species level and calculated the mean number of flowers per individual across species.

Plant-hummingbird interactions

To record the interactions between plant and hummingbird species in the understory, we used time-lapse cameras combined with automated computer vision review. Along the sampling transects, we placed unattended cameras on flowers that fit the traditional ornithophilous syndrome, i.e. those with tubular-red, -yellow, or -purple corollas, scentless and no landing platforms, roughly in proportion to their abundance. Because hummingbird-visited flowers do not always have these characteristics, we also considered plant species that fit other syndromes (e.g. bat- or insect-pollinated flowers) known to be visited by hummingbirds based on field observations and previous research.

Statistical analysis

(i) Flower morphology as predictor of nectar traits across elevations

We performed phylogenetic generalised least squares (PGLS) to analyse (i) the variation in floral phenotypic traits (morphological and nectar traits) and flower abundance of plant species visited by hummingbirds across elevation and (ii) the extent to which flower morphology predicts nectar traits.

(ii) Morphological and nectar traits as predictors of network metrics across elevations

We analysed how flower phenotypic traits (morphology and nectar) influenced two measures of network structure: ecological specialisation and closeness centrality, which reflect the role and position of plant species in the studied communities, respectively.

(iii) Morphological vs nectar traits as predictors of plant-hummingbird interactions

We analysed the effect of species morphology on plant-hummingbird interactions by calculating the degree of mismatch between corresponding pairs of morphological traits (bill-corolla length and bill-corolla curvature) and considering a barrier to nectar access for each pairwise interaction. To simultaneously evaluate the effect of morphological and nectar traits on species interactions, we: (i) fitted a full generalised linear mixed model with negative binomial error distribution  (GLMM) as suggested for overdispersed count data, where we included plant-hummingbird interaction frequency as the response variable with an offset term on the cumulative sampling hours for each plant species due to its influence in the dependent variable, and seven fixed effects and (ii) conducted automated selection to identify minimal adequate models from the full model that best explain the effect of morphological and nectar traits on interaction frequency.