Improving allometric models to estimate the proboscis length of tropical bees

27 The proboscis length of bees is a key morphological trait shaping communities, pollination 28 networks, and likely their responses to habitat loss. Despite its importance, it is rarely 29 considered in ecological studies because of logistic limitations in obtaining accurate 30 measurements across many different species. In two previous studies, the proboscis length of 31 temperate bee species was estimated based on body size and bee family. However, bee taxa 32 partially occurring in the tropics might deviate from this allometric relationship due to 33 different functional constraints. Thus, we tested if equations developed for temperate bees can 34 accurately predict the proboscis length in Meliponini, Euglossini (both Apidae), and 35 Augochlorini (Halictidae), three ubiquitous and highly important tribes of tropical bees. We 36 measured the intertegular distance (as a proxy of body size measurement) and the proboscis 37 length of 892 specimens of 105 tropical species. We used these measurements to evaluate the 38 previous model and found that its estimations lacked accuracy when applied to tropical bees, 39 particularly to Meliponini and Euglossini. We developed new allometric equations estimating 40 the proboscis length based on the intertegular distance, using (sub-) genera as an additional 41 predictive variable to refine the estimations. We tested our approach by creating a test model 42 for Meliponini, trained with only 80 % of the data, and evaluated this model using the 43 remaining 20 %, resulting in a high accuracy of estimates. Our results shed additional light on 44 the nature of the proboscis length-body size allometric relationship in tropical bees and 45 provide a tool for future studies on the functional ecology of bees and their interactions with 46 plants. 47 48

allometry, intertegular distance, functional morphology, Meliponini, Euglossini, 49 Augochlorini 50 51 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 3

52 Functional traits determine interactions between various trophic levels and within 53 ecosystems (Schleuning et al. 2023). Traits can e.g. shape the distribution (Pollock et al. 54 2012), dispersal ability and niche occupation (Jiang et al. 2018), resource use (Gravel et al. 55 2016), or ability of species to adapt to a changing climate (Heilmeier 2019). To fully 56 understand these mechanisms, an understanding of the underlying traits is crucial (McGill et 57 al. 2006). A common example of species interactions, largely shaped by different functional 58 traits, is pollination. Among the many variables that shape the interaction between pollinators 59 and plants, the shape and length of their mouthparts play a particularly important role: 60 proboscis length affects flower selection (Temeles et al. 2009; Basari et al. 2021; Inouye 61 1980; Haverkamp et al. 2016), foraging efficiency, pollination effectiveness (Haverkamp et 62 al. 2016; Borrell 2007; Harder 1983; Peat et al. 2005), and extinction risk of pollinators and 63 their host plant species (Stang et al. 2007). In a community-wide context, it regulates resource 64 partitioning between pollinator species (Inouye 1978; Ranta 1984; Ranta and Lundberg 1980; 65 Brown and Bowers 1985) and can be a driver of plant speciation (Borrell 2005; Rodríguez-66 Gironés and Santamaría 2007; Vajna et al. 2021). Thus, proboscis length is a key interaction 67 trait that influences not only community assembly but also the structure of pollination 68 networks by niche partitioning and determining patterns of specialization (Harmon /i1Threatt 69 and Ackerly 2013; Stang et al. 2006, 2007; Stang et al. 2009). 70 For taxa such as hummingbirds or hawkmoths, whose long and often widely varying 71 bills and proboscises indicate a clear functional specialization towards long-tubed flowers, the 72 mouthpart length is routinely included in ecological studies that investigate the functional 73 composition of these taxa (Torres /i1Vanegas et al. 2021), their interactions with plants 74 (Guevara et al. 2023; Johnson et al. 2017) or the degree of specialization (Rodríguez-Flores et 75 al. 2019; Nilsson and Rabakonandrianina 1988). Bees, however, despite being one of the most 76 ubiquitous and ecologically dominant group of pollinators worldwide (Potts et al. 2010), are 77 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 4 seldom explored in terms of the functional significance of their mouthparts. The few existing 78 studies are mostly focused on large-bodied species like bumblebees (Goulson et al. 2008; 79 Harder 1983; Stout et al. 2000). However, measuring the proboscis length of smaller bees 80 usually requires dissection of the mouthparts, which can lead to destruction of the specimen, 81 particularly in very small species, which then have to be identified beforehand (Cariveau et al. 82 2016). Thus, measuring the proboscis length of bees often is not feasible and multiple 83 approaches have been taken to forego measuring proboscises in ecological studies, including 84 adopting length categories, i.e., “long-tongued”, which include Apidae and Megachilidae, and 85 “short-tongued” bees, including Halictidae, Andrenidae, Colletidae, Melittidae, and 86 Stenotritidae (Michener 2007). This approach, however, lacks accuracy and brushes over 87 taxon-specific differences (Ostwald et al. 2024). Thus, effort has been made to estimate 88 proboscis length by allometric power functions including body size and bee family (Cariveau 89 et al. 2016; Melin et al. 2019). Allometric functions can be used to describe the relationship 90 between body size and metabolic rate, growth, or the size of specific body parts (Pélabon et 91 al. 2014). Cariveau et al. (2016) and Melin et al. (2019) showed that the proboscis length of 92 bee species increases with their body size in all families except the Australian Stenotritidae, 93 which were not included in these studies. The proboscis length also differed between families, 94 indicating that there is a phylogenetic component affecting the proboscis length of bees 95 (Cariveau et al. 2016; Melin et al. 2019). However, these studies were carried out in temperate 96 and subtropical regions, while data for tropical bees is lacking. Tropical bees may deviate 97 from these allometric relationships due to functional constraints as a mechanism to avoid 98 interspecific competition in these highly diverse ecosystems (Borrell 2005; Ostwald et al. 99 2024). Notably, morphological measurements to calculate functional diversity in tropical 100 regions play an important role in the context of ongoing deforestation and its impact on global 101 biodiversity (Wright and Muller-Landau 2006; van der Sluijs 2020; Alroy 2017; Ostwald et 102 al. 2024). This is especially true for monitoring the effects of forest restoration as patterns of 103 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 5 resistance and recovery of insect communities are often tied to dispersal and interaction traits 104 (D’Astous et al. 2013; Montoya /i1Pfeiffer et al. 2018; Audino et al. 2014; Montoya /i1Pfeiffer 105 et al. 2020; Lichtenberg et al. 2017). Functionally diverse pollinator communities ensure 106 pollination services within the process of tropical forest recovery and conservation. Hereby, 107 bees play an essential role as they are responsible for the pollination of the majority of 108 tropical plants (Michener 2007; Ollerton et al. 2011). It is therefore paramount to streamline 109 and standardize the estimation of proboscis length. 110 In the Neotropics, three of the most abundantly encountered bee tribes are Meliponini 111 (Apidae), Euglossini (Apidae), and Augochlorini (Halictidae) (Michener 2007). Systematic 112 studies on the morphological traits of these bee tribes are scarce. Although the proboscis 113 length is used as a trait to identify Euglossini males (Bembé 2007) and is comparatively easy 114 to measure because of its length, it is rarely considered in ecological studies (e.g., Brito et al. 115 2018; Guevara et al. 2024). There are a few studies which have assessed the morphometrics of 116 single Meliponini species (Basari et al. 2021; Kiatoko et al. 2023), and some have further 117 placed the proboscis length into an ecological context to uncover floral preferences (Laha et 118 al. 2020) or to explore the phenology of plant-pollinator interactions (Ribeiro et al. 2024). 119 The importance of the proboscis length as a community-shaping morphological trait 120 (Harmon/i1Threatt and Ackerly 2013) makes it necessary to get the most accurate values, 121 which is achieved by direct specimen measurements. However, in large biodiversity 122 assessments with a multitude of species, direct measurements might be constrained by time 123 and resources and may need to be replaced by estimates. Expanding the existing models to 124 tropical species might therefore not only allow such estimations, but also add to our 125 understanding of general allometry in bees, which contributes to recognizing species that 126 differ from expected patterns and explaining certain life history traits (Pélabon et al. 2014). 127 We thus aimed to assess the allometric relationship between bee size and proboscis 128 length in these key tropical bee tribes (Meliponini, Euglossini, and Augochlorini) by (i) 129 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 6 testing the applicability of pre-existing models on a large database of bees caught in a diverse 130 lowland rainforest ecosystem in Ecuador and (ii) providing updated model versions for 131 tropical bees that also account for bee tribe and genus. We hypothesized that the proboscis 132 length of the two tropical bee tribes which were not included in the dataset of Cariveau et al. 133 (2016), i.e. Meliponini and Euglossini, could not be accurately predicted by the existing 134 model, while it would provide accurate estimates for the proboscis length of Augochlorini. 135 Based on our findings, we furthermore composed an R package expanding the scope of the 136 previous model to easily estimate the proboscis length of tropical bees using measurements of 137 intertegular distance and taxonomic information. 138 139

140 Data collection 141 Specimen collection was carried out between March and December 2023 in the 142 Reserva Río Canandé (0°31'33.4"N, 79°12'46.0"W) and Reserva Tesoro Escondido 143 (0°32'30.9"N, 79°08'41.9"W), northwestern Ecuador, Chocó-Darien ecoregion. Specimens 144 were collected via fragrance traps aimed at Euglossini males (Ferreira et al. 2013), vane traps 145 with blue and yellow vanes (Prendergast et al. 2020; Renteria and Brehm 2025), and active 146 netting. All bees were killed with chloroform fumes. Methods are described in more detail in 147 Diniz et al. (2025) and Escobar et al. (2024). 148 After collection, all specimens were identified to the lowest possible taxonomic level 149 using specialized keys (Engel et al. 2023; Engel 2000; Bonilla-Gómez and Nates-Parra 1992). 150 Intertegular distance (IT), defined as the distance between the tegulae, was used as a body size 151 proxy (Cariveau et al. 2016; Stemet et al. 2024; Kendall et al. 2019). The proboscis length 152 (PB) was defined as the length of the prementum plus the length of the glossa or the distance 153 from the base of the mentum to the distal point of the labellum (Cariveau et al. 2016) 154 (Supplementary Information, Fig. S1). Measurements from the smaller sized Meliponini and 155 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 7 Augochlorini were done in photos taken with a Wild Heerbrug M7S microscope attached to a 156 Leica MC120 HD camera using ImageJ (Schneider et al. 2012). For Meliponini and 157 Augochlorini, the proboscis was dissected and mounted on slides for photographing. 158 Measurements of the comparatively large Euglossini bees were taken with a caliper on fully 159 stretched and straight proboscises. All measurements were taken on specimens which had 160 been stored in 70 %-Ethanol at – 20 °C. All measured Meliponini were female workers, 161 Augochlorini were females, and Euglossini were males. Cariveau et al. (2016) found no 162 differences in PB-allometry between sexes. 163 164 Data analysis 165 All analyses were performed in R 4.4.1 (R Core Team 2024). First, to test the quality of pre-166 existing allometric model, PB was estimated via IT using the BeeIT package (Cariveau et al. 167 2016). The estimated values were compared to the observed values and model fit was 168 assessed via linear regression using R² and root mean squared error (RSME) as comparative 169 metrics, with a high R² and a low RSME indicating a good fit. 170 Then, four new linear models were created separately for each bee group, including IT 171 and (sub-) genus as predictive variables independently and as interacting parameters. For 172 easier application and to include lower taxonomic levels than family, we adapted the 173 allometric equation as follows: 174 175 ln /g1842/g1828 /g3404 /g1859 /g3397 /g1854 /g1499 ln /g1835/g1846 176 where g = a coefficient specific to the genus, subgenus or tribe and b = an allometric scaling 177 coefficient, representing the slope of the model. The models’ relative fit was evaluated using 178 the Akaike information criterion (AIC). Unlike Cariveau et al. (2016) and Melin et al. (2019) 179 who calculated species trait means, we used all individual measurements to account for 180 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 8 interspecific variation, to increase the overall amount of training data, especially as some 181 species were represented by only one specimen and to account for groups with difficult 182 separation of biological species, e.g. Augochlorini. 183 To evaluate the accuracy of our approach, we created a second model by training it 184 with only 80 % of the measured Meliponini data to estimate PB. We then used the other 20 % 185 to estimate model fit and calculated the difference between the predicted and the measured 186 values and the 90 %-quantile of this difference as a metric for the accuracy of our estimates. 187 188

189 In total, 892 specimens from 105 species were measured (1 to 36 individuals per 190 species): 504 Euglossini from 58 species, 343 Meliponini from 36 species, and 44 191 Augochlorini from 11 species (Supplementary Information, Table S2). From 16 species only 192 one individual was available for measurement. 193 194 Evaluation of the pre-existing model 195 When fit to our data, the model from Cariveau et al. (2016) did not provide accurate 196 estimates of proboscis length for the three investigated bee tribes. The proboscis length of 197 Meliponini was overestimated by 0.71 mm on average (sd = 0.49 mm) and residuals ranged 198 from -0.71 to 1.96 mm (max. deviation > 123 % of IT), which resulted in a low R² and a high 199 RMSE (Table 1). The proboscis length of Euglossini was greatly underestimated with a mean 200 residual of -7.08 mm (sd = 6.24 mm) and residuals ranging from -25.83 to 3.48 mm (max. 201 deviation > 759 % of IT), which resulted in a negative R² and a high RMSE (Table 1). Of the 202 three tribes, the model most accurately predicted the proboscis length of Augochlorini with a 203 mean residual of -0.07 mm (sd = 0.35 mm) and a range of residuals from -1.04 to 0.58 mm 204 (max. deviation > 85 % of IT), which resulted in a low RMSE (Table 1). 205 206 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 9 Table 1 . Goodness of fit measures for the model from Cariveau et al. (2016) against the measured 207 data. 208 Family Tribe R² RMSE Apidae Meliponini 0.33 0.87 Apidae Euglossini -0.65 9.43 Halictidae Augochlorini 0.34 0.36 209 Improved allometric models 210 All new linear models showed a better fit in comparison to the pre-existing model of 211 Cariveau et al. (2016). The most parsimonious models for Augochlorini, Meliponini and all 212 Euglossini included IT and genus as independent predictors, while an interaction between the 213 two parameters showed a similar or better fit to the data (Table 2). For Euglossini ( Euglossa 214 excluded) and Euglossa, the models with the lowest AIC additionally included the interaction 215 between IT and (sub-) genus. Including (sub-) genus as a predictive variable notably lowered 216 model AIC in all groups, indicating that the mean proboscis length differs between (sub-) 217 genera. For Meliponini and Euglossini, the models with only (sub-) genus had a notably lower 218 AIC than the models including only IT (Table 2). 219 220 Table 2. Goodness of fit measures of the new linear models for every bee tribe and additionally for 221 Euglossini excluding Euglossa and a model for Euglossa including subgenera. The best model with 222 the lowest AIC is marked in bold. 223 Tribe Model R² AIC Meliponini IT 0.71 48.39 genus 0.82 -96.26 IT + genus 0.83 -114.02 IT * genus 0.85 -112.22 Euglossini IT 0.14 620.45 Genus 0.20 589.81 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 10 IT + genus 0.22 577.56 IT * genus 0.23 578.30 Euglossini (without Euglossa) IT 0.10 105.56 Genus 0.65 -15.03 IT + genus 0.65 -12.90 IT * genus 0.68 -19.45 Euglossini (only Euglossa) IT 0.04 504.94 Subgenus 0.72 47.51 IT + subgenus 0.72 49.27 IT * subgenus 0.76 -8.40 Augochlorini IT 0.51 -42.71 Genus 0.63 -52.60 IT + genus 0.71 -60.73 IT * genus 0.74 -60.39 224 Figure 1A shows the tribe-specific relationship between IT and PB without 225 considering genus, while Figure 1B shows the relationship as indicated by the lowest AIC. 226 The model for all Euglossini showed the worst fit, however, separating Euglossa and the other 227 genera and including the subgenus in the Euglossa model increased the model fit greatly. 228 Means for IT and PB for all species are provided in the supplementary material (S 2). The 229 power function was parameterized to predict PB of Meliponini, Euglossini, and Augochlorini 230 using the estimates of the models with the best fit (Table 3). 231 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 11 232 Figure 1. Relationship between intertegular distance and proboscis length for the three tribes (A) and 233 separated between (sub-) genera (B) with R2-values provided for our models. Each point represents 234 one measured specimen. The dashed lines in A represent the estimates by Cariveau et al. (2016). The 235 full lines represent the estimates by the new models. In B the names of the genera are specified. 236 237 Table 3. Coefficients of the allometric equations for the three bee tribes, parameterized by intercept 238 (g) and slope (b) of the best linear model for each tribe and genus. Genus “All” specifies the model 239 coefficients for the tribe-based model (PB ~ IT) for genera not considered in this study. For Eufriesea, 240 Eulaema, and Exarete the model excluding Euglossa was used. By inserting g and b in the formula 241 , the proboscis length can be calculated. 242 Tribe (Sub-) Genus (Sub-) Genus- specific coefficient (g) Scaling coefficient (b) Meliponini All 0.28 1.39 Dolichotrigona 0.21 0.55 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 12 Melipona 1.01 0.55 Nannotrigona 0.36 0.55 Nogueirapis 0.67 0.55 Oxytrigona 0.39 0.55 Paratrigona 0.36 0.55 Partamona 0.94 0.55 Plebeia 0.22 0.55 Ptilotrigona 0.62 0.55 Scaptotrigona 0.81 0.55 Scaura 0.16 0.55 Tetragona 0.21 0.55 Tetragonisca 0.42 0.55 Trigona 0.71 0.55 Trigonisca 0.13 0.55 Euglossini All 1.61 0.74 Eufriesea 5.32 -1.94 Euglossa 1.46 0.87 Euglossa sensu stricto 3.15 -0.87 Euglossa (Glossura) 1.18 1.53 Euglossa (Glossurella) 1.96 0.64 Eulaema 2.39 0.35 Exaerete 2.36 0.43 Augochlorini All 0.45 0.51 Augochlora 0.62 0.64 Augochloropsis 0.34 0.64 Pereirapis 0.45 0.64 243 Our test model aimed at predicting Meliponini PB was trained with 274 random 244 measurements, while the other 69 data points were used as the evaluation data set. With our 245 new model, absolute differences between the measured and estimated data points were always 246 below 1 mm and errors were equally distributed around 0 mm (mean = 0.03 mm, sd = 0.32 247 mm). In contrast, differences between measured and estimated value using the Cariveau et al. 248 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 13 (2016) model often exceeded 1 mm and tended to overestimate PB of Meliponini (mean = 249 0.75 mm, sd = 0.46 mm). For our new model, 90% of the absolute errors were below 0.57 250 mm, while the 90%-quantile of the absolute errors for the model of Cariveau et al. (2016) was 251 at 1.33 mm. The new model accurately predicted PB particularly of small Meliponini, while 252 errors were larger for medium-sized individuals. 253 254 255 Figure 2. Error distribution of the new model (left) and the model of Cariveau et al. (2016) (right) 256 when applied to the evaluation data from Meliponini. 257 258 For easier use, we implemented the allometric equations in the R package tropTongue, 259 which can be downloaded from https://github.com/kilian-fru/tropTongue. The parameters 260 used are specified in Table 3. For other tribes, the package uses the family-specific function 261 from the package BeeIT (Cariveau et al. 2016). 262 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 14 263

264 In this study, we showed that a pre-existing allometric model created for bees in 265 temperate environments (Cariveau et al. 2016) could not accurately predict the proboscis 266 length of several tropical bee tribes. New allometric models were created including the 267 intertegular distance as body size proxy and the (sub-) genus. This significantly improved the 268 accuracy of the models when compared to the Cariveau et al. (2016) model. Model outcomes 269 confirmed an allometric relationship between body size, phylogeny and mouthpart length for 270 the tropical bee tribes studied, which has also been found in other nectar-feeding animals, e.g. 271 tropical butterflies (Kunte 2007), tropical Sphingidae (Agosta and Janzen 2005), subtropical 272 and temperate wild bees (Menegus 2018; Cariveau et al. 2016; Melin et al. 2019), birds 273 (Rombaut et al. 2022) and non-nectar-feeding animals, e.g. weevils (Fleurot et al. 2022) and 274 certain tropical butterfly clades (Kunte 2007). 275 We showed that the allometric relationship between body size and proboscis length 276 varies between temperate and tropical bees, at least in Apidae (e.g., Meliponini, Euglossini). 277 Having an overproportionately long proboscis, like in the case of many Euglossini compared 278 to temperate Apidae, is usually considered a competitional advantage as plants with short- and 279 long-tubed flowers both can be used for foraging (Borrell 2005). However, long-tongued 280 insects were found to exhibit longer flower-handling times and higher energy use while 281 foraging (Kunte 2007; Harder 1983; Borrell 2007). They thus likely target longer-tubed 282 flowers which offer a higher amount of nectar while excluding insects with shorter 283 proboscises (Dressler 1982; Johnson et al. 2017). On the other hand, very small bees, like 284 Meliponini, might be able to compensate for their shorter-than-expected proboscis by their 285 small body size, making them able to crawl into narrow-tubed flowers to forage (Engel et al. 286 2023; Michener 2007). 287 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 15 Extreme values for proboscis length like those described but also for other functional 288 bee traits are more likely to appear in the tropics: Tropical ecosystems are the most biodiverse 289 on Earth, harboring not only a high plant species diversity (Myers et al. 2000) but also an 290 even higher functional diverse flora than their species pool would suggest (Swenson et al. 291 2012). The functional diversity of plants increases towards the equator (Lamanna et al. 2014), 292 due to for example a high number of epiphytes and lianas (Spicer et al. 2020), which might 293 affect the functional diversity of bees. Proboscis length is an important interaction trait linked 294 to the morphological matching between plants and pollinators (Goulson et al. 2008). Because 295 of the high (functional) plant diversity, the trait space occupied by tropical bee communities is 296 likely also larger than the space occupied by temperate bee communities, resulting in a wider 297 range of proboscis lengths in tropical bees. 298 Studying functional traits in these highly diverse ecosystems is challenging. The 299 approach to estimate difficult-to-measure morphological traits, like the proboscis length, by 300 allometric equations can greatly simplify data collection (Cariveau et al. 2016; Ostwald et al. 301 2024). However, as we showed in our study, estimated values should be handled carefully. 302 Cariveau et al. (2016) pointed out, that adopting length categories (“long-tongued” / ”short-303 tongued”) to overcome measurements of proboscis length does not provide sufficient 304 accuracy and thus introduced an allometric equation to estimate the proboscis length using 305 body size and bee family, parameterized with measurements of temperate bees. Our results 306 suggest that combining body size measurements at lower taxonomic levels can further 307 improve estimates of bee proboscis length, at least in the tropics (Ostwald et al. 2024). 308 Especially in morphologically highly diverse families such as Apidae (Engel et al. 2021), we 309 suggest to include the tribe or the genus to increase accuracy, wherever it is feasible. 310 Including the subgenus in the equation further improved estimates and model fit in Euglossa, 311 whose subgenera are partly separated by proboscis length (Bonilla-Gómez and Nates-Parra 312 1992; Bembé 2007) suggesting a strong phylogenetical component shaping the differentiation 313 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 16 of proboscis length in this genus. Euglossa represents the largest genus of Euglossini with 139 314 often co-existing species (Engel and Rasmussen 2019). Differentiation in proboscis length 315 might thus also be a strategy to overcome competition between species within Euglossa. 316 Further model adaptations might focus on enhancing data availability and thus quality 317 for tropical bees. Additional data would be especially useful in evaluating such allometric 318 models. The evaluation of our approach with the test model showed an increased accuracy of 319 our model for Meliponini, however, using independent data is crucial to confirm our and 320 future findings. Further improving data quality through e.g. incorporating data from other 321 studies will require a standard protocol to measure proboscis length of bees (Keller et al. 322 2023). For example, we would have liked to add data from Ribeiro et al. (2024), but they used 323 another measuring protocol preventing comparison of measurements. To improve data 324 availability we used individual measurements instead of species means, in contrast to previous 325 models (Cariveau et al. 2016). This approach might bias our allometric equations towards 326 more abundant species, which were measured most frequently. It does however increase the 327 overall quality of the model, especially for groups like Augochlorini, where species 328 identification is difficult and not many individuals were available, rendering means even less 329 accurate. Studies on tropical Augochlorini are scarce and taxonomic keys are missing, even 330 though they might be important indicators of forest loss as they appear to depend on 331 unforested habitats for nesting (Brosi et al. 2007). Additionally, individual-based 332 measurements were shown to produce the same results as species means while decreasing 333 measuring effort and enhancing data availability (Beck et al. 2024). 334 The most accurate method to obtain values of proboscis length is to measure them 335 manually. However, in large biodiversity assessments, for large species pools, very small or 336 rare species or species with unclear taxonomic status, like often found in the tropics, this is 337 not feasible. Our model enables researchers to use body size measurements to additionally 338 infer proboscis length, when it could not be manually obtained. Body size is measured 339 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 17 comparatively often in ecological studies on bees (Osorio-Canadas et al. 2022; Lichtenberg et 340 al. 2017; Montoya/i1Pfeiffer et al. 2020) but the proboscis length is rarely considered. 341 Our results shed additional light on the nature of the proboscis length-body size 342 allometric relationship in tropical bees and may serve as an additional tool for future 343 ecological studies that want to include proboscis length to assess bee (functional) diversity, 344 morphology and allometry in the tropics. The proboscis length of bees is related to various 345 aspects of bee ecology such as flower selection (Basari et al. 2021) and competition (Ranta 346 and Lundberg 1980), determines patterns of distribution and habitat preferences 347 (Harmon/i1Threatt and Ackerly 2013) and might be linked to pesticide uptake by bees (Kopit 348 and Pitts-Singer 2018; Borrell 2007). Therefore, it directly affects the conservation of tropical 349 bee communities which are mostly endangered by habitat loss in particular deforestation and 350 pesticides (Toledo-Hernández et al. 2022). Furthermore, proboscis length has a strong effect 351 on the pollination services of bees (Chase et al. 2023), making it an important trait to consider 352 when assessing tropical forest restoration and conservation, which is largely influenced by bee 353 pollination (Ollerton et al. 2011). We, hereby, emphasize the importance of including the 354 proboscis length in much needed research on tropical bee functional ecology, which is crucial 355 in understanding the effects and underlying patterns of tropical deforestation, forest 356 restoration, biodiversity loss and climate change. 357 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 18

358 This work was funded by the Deutsche Forschungsgemeinschaft (DFG) funded Research Unit 359 REASSEMBLY (FOR 5207; sub-projects LE2750/12-1 and KE1742/13-1). We thank the 360 Ministry of Environment of Ecuador for granting research and collection permits through 361 Contrato Marco MAE-DNB-CM-2021-0187, Sebastián Escobar for handling exportation 362 permits, Martin Schaefer (Fundación Jocotoco) and Citlalli Morelos-Juarez (Fundación 363 Tesoro Escondido) for allowing us to work in their reserves, and the staff of both reserves: 364 Katrin Krauth, Julio Carbajal, Jender Vélez, Bryan Tamayo, Lady Condoy, Leonardo de la 365 Cruz, Jefferson Tacuri, Yadira Giler and Adriana Argoti. 366 367 AUTHOR CONTRIBUTIONS 368 KF, KK, SDL and UMD conceptualized the research. SDL, AK and GB acquired and 369 managed the funding. UMD, SDL and JW performed fieldwork and data collection. KF, KK, 370 MP, JW and UMD did sample processing and data curation. CR, KF, MP, JW and UMD 371 identified insects. KF performed data analysis and wrote the manuscript draft. All authors 372 contributed critically to the last manuscript draft. 373 374 CONFLICT OF INTERST 375 The authors declare no conflicts of interest. 376 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 19 Publication bibliography 377 Agosta, Salvatore J.; Janzen, Daniel H. (2005): Body size distributions of large Costa Rican 378 dry forest moths and the underlying relationship between plant and pollinator morphology. In 379 Oikos 108 (1), pp. 183–193. DOI: 10.1111/j.0030-1299.2005.13504.x. 380 Alroy, John (2017): Effects of habitat disturbance on tropical forest biodiversity. In 381 Proceedings of the National Academy of Sciences of the United States of America 114 (23), 382 pp. 6056–6061. DOI: 10.1073/pnas.1611855114. 383 Audino, Lívia Dorneles; Louzada, Julio; Comita, Liza (2014): Dung beetles as indicators of 384 tropical forest restoration success: Is it possible to recover species and functional diversity? In 385 Biological Conservation 169, pp. 248–257. DOI: 10.1016/j.biocon.2013.11.023. 386 Basari, Norasmah; Ramli, Sarah Najiah; Abdul-Mutalid, Nur Adawiyah; Shaipulah, Nur 387 Fariza M.; Hashim, Nur Aida (2021): Flowers morphology and nectar concentration 388 determine the preferred food source of stingless bee, Heterotrigona itama. In Journal of Asia-389 Pacific Entomology 24 (2), pp. 232–236. DOI: 10.1016/j.aspen.2021.02.005. 390 Beck, Jan; McCain, Christy M.; Brehm, Gunnar (2024): Can you trust comparative trait data 391 based on singleton species? In Frontiers of Biogeography 16 (1). DOI: 392 10.21425/F5FBG62354. 393 Bembé, Benjamin (2007): Revision der Euglossa cordata-Gruppe und Untersuchungen zur 394 Funktionsmorphologie und Faunistik der Euglossini (Hymenoptera, Apidae). In Entomofauna 395 (Supplement 14), pp. 1–146. 396 Bonilla-Gómez, M. A.; Nates-Parra, G. (1992): Abejas Euglosinas De Colombia 397 (Hymenoptera: Apidae) I. Claves Ilustradas. In Caldasia 17 (1), pp. 149–172. 398 Borrell, Brendan J. (2005): Long Tongues and Loose Niches: Evolution of Euglossine Bees 399 and Their Nectar Flowers. In Biotropica 37 (4), pp. 664–669. DOI: 10.1111/j.1744-400 7429.2005.00084.x. 401 Borrell, Brendan J. (2007): Scaling of nectar foraging in orchid bees. In The American 402 Naturalist 169 (5), pp. 569–580. DOI: 10.1086/512689. 403 Brito, Thaline F.; Contrera, Felipe A. L.; Phifer, Colin C.; Knowlton, Jessie L.; Brasil, 404 Leandro S.; Maués, Márcia M.; Silva, Daniel P. (2018): Effects of habitat type change on 405 taxonomic and functional composition of orchid bees (Apidae: Euglossini) in the Brazilian 406 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 20 Amazon. In Journal of Insect Conservation 22 (3-4), pp. 451–463. DOI: 10.1007/s10841-018-407 0073-9. 408 Brosi, Berry J.; Daily, Gretchen C.; Ehrlich, Paul R. (2007): Bee community shifts with 409 landscape context in a tropical countryside. In Ecological Applications 17 (2), pp. 418–430. 410 DOI: 10.1890/06-0029. 411 Brown, James H.; Bowers, Michael A. (1985): Community Organization in Hummingbirds: 412 Relationships between Morphology and Ecology. In The Auk 102 (2), pp. 251–269. DOI: 413 10.2307/4086767. 414 Cariveau, Daniel P.; Nayak, Geetha K.; Bartomeus, Ignasi; Zientek, Joseph; Ascher, John S.; 415 Gibbs, Jason; Winfree, Rachael (2016): The Allometry of Bee Proboscis Length and Its Uses 416 in Ecology. In PloS one 11 (3), e0151482. DOI: 10.1371/journal.pone.0151482. 417 Chase, Marissa H.; Fraterrigo, Jennifer M.; Harmon/i1Threatt, Alexandra (2023): Bee 418 functional traits and their relationship to pollination services depend on many factors: A 419 meta/i1regression analysis. In Insect Conservation and Diversity 16 (3), pp. 313–323. DOI: 420 10.1111/icad.12635. 421 D’Astous, A.; Poulin, M.; Aubin, I.; Rochefort, L. (2013): Using functional diversity as an 422 indicator of restoration success of a cut-over bog. In Ecological Engineering 61, pp. 519–526. 423 DOI: 10.1016/j.ecoleng.2012.09.002. 424 Diniz, Ugo Mendes; Böttger, Dennis; Brehm, Gunnar; Lalama, Sabine Fernanda Viteri; 425 Frühholz, Kilian; Pitz, Maximilian et al. (2025): Stratification along tropical forest succession 426 enhances pollinator diversity via functionally unique canopies. In BioRxiv. DOI: 427 10.1101/2025.01.23.634515. 428 Dressler, Robert (1982): Biology of the Orchid Bees (Euglossini). In Annual Review of 429 Ecology and Systematics 13, pp. 373–394. 430 Engel, Michael S. (2000): Classification of the Bee Tribe Augochlorini (Hymenoptera: 431 Halictidae). In Bulletin of the American Museum of Natural History 250 (1), p. 1. DOI: 432 10.1206/0003-0090(2000)2502.0.CO;2. 433 Engel, Michael S.; Rasmussen, Claus (2019): Corbiculate Bees. In Christopher K. Starr (Ed.): 434 Encyclopedia of Social Insects. Cham: Springer International Publishing, pp. 1–9. 435 Engel, Michael S.; Rasmussen, Claus; Ayala, Ricardo; Oliveira, Favízia F. de (2023): 436 Stingless bee classification and biology (Hymenoptera, Apidae): a review, with an updated 437 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 21 key to genera and subgenera. In ZooKeys 1172, pp. 239–312. DOI: 438 10.3897/zookeys.1172.104944. 439 Engel, Michael S.; Rasmussen, Claus; Gonzalez, Victor H. (2021): Bees. In Christopher K. 440 Starr (Ed.): Encyclopedia of Social Insects. Cham: Springer International Publishing, pp. 93–441 109. 442 Escobar, Sebastián; Newell, Felicity L.; Endara, María-José; Guevara-Andino, Juan E.; 443 Landim, Anna R.; Neuschulz, Eike Lena et al. (2024): Reassembly of a tropical rainforest 444 ecosystem: A new chronosequence in the Ecuadorian Chocó tested with the recovery of tree 445 attributes. 446 Ferreira, R. P.; Martins, C.; Dutra, M. C.; Mentone, C. B.; Antonini, Y. (2013): Old 447 Fragments of Forest Inside an Urban Area Are Able to Keep Orchid Bee (Hymenoptera: 448 Apidae: Euglossini) Assemblages? The Case of a Brazilian Historical City. In Neotropical 449 entomology 42 (5), pp. 466–473. DOI: 10.1007/s13744-013-0145-1. 450 Fleurot, E.; Venner, S.; Pélisson, P-F; Débias, F.; Bel-Venner, M-C (2022): The 451 morphological allometry of four closely related and coexisting insect species reveals 452 adaptation to the mean and variability of the resource size. In Oecologia 200 (1-2), pp. 159–453 168. DOI: 10.1007/s00442-022-05249-x. 454 Goulson, Dave; Lye, Gillian C.; Darvill, Ben (2008): Diet breadth, coexistence and rarity in 455 bumblebees. In Biodiversity and Conservation 17 (13), pp. 3269–3288. DOI: 456 10.1007/s10531-008-9428-y. 457 Gravel, Dominique; Albouy, Camille; Thuiller, Wilfried (2016): The meaning of functional 458 trait composition of food webs for ecosystem functioning. In Philosophical transactions of 459 the Royal Society of London. Series B, Biological sciences 371 (1694). DOI: 460 10.1098/rstb.2015.0268. 461 Guevara, Esteban A.; Bello, Carolina; Schleuning, Matthias; Pellissier, Loïc; G, Tatiana 462 Santander; Tobar, Francisco; Graham, Catherine H. (2023): Asymmetric effect of 463 deforestation on the functional roles of interacting plants and hummingbirds. DOI: 464 10.21203/rs.3.rs-2527763/v1. 465 Guevara, Jennifer; Calvache, Esteban; Ortega/i1Andrade, H. Mauricio; Pazmiño/i1Palomino, 466 Alex (2024): Body size and abundance patterns of male orchid bees in a fragmented 467 Neotropical landscape. In Insect Conservation and Diversity, Article icad.12787. DOI: 468 10.1111/icad.12787. 469 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 22 Harder, Lawrence D. (1983): Flower handling efficiency of bumble bees: morphological 470 aspects of probing time. In Oecologia 57, pp. 274–280. 471 Harmon/i1Threatt, Alexandra; Ackerly, David D. (2013): Filtering across spatial scales: 472 phylogeny, biogeography and community structure in bumble bees. In PloS one 8 (3), e60446. 473 DOI: 10.1371/journal.pone.0060446. 474 Haverkamp, Alexander; Bing, Julia; Badeke, Elisa; Hansson, Bill S.; Knaden, Markus (2016): 475 Innate olfactory preferences for flowers matching proboscis length ensure optimal energy gain 476 in a hawkmoth. In Nature Communications 7, p. 11644. DOI: 10.1038/ncomms11644. 477 Heilmeier, Hermann (2019): Functional traits explaining plant responses to past and future 478 climate changes. In Flora 254, pp. 1–11. DOI: 10.1016/j.flora.2019.04.004. 479 Inouye, David W. (1978): Resource Partitioning in Bumblebees: Experimental Studies of 480 Foraging Behavior. In Ecology 59 (4), pp. 672–678. 481 Inouye, David W. (1980): The effect of proboscis and corolla tube lengths on patterns and 482 rates of flower visitation by bumblebees. In Oecologia 45, pp. 197–201. 483 Jiang, Feng; Xun, Yanhan; Cai, Huiying; Jin, Guangze (2018): Functional traits can improve 484 our understanding of niche- and dispersal-based processes. In Oecologia 186 (3), pp. 783–485 792. DOI: 10.1007/s00442-018-4060-3. 486 Johnson, Steven D.; Moré, Marcela; Amorim, Felipe W.; Haber, William A.; Frankie, Gordon 487 W.; Stanley, Dara A. et al. (2017): The long and the short of it: a global analysis of hawkmoth 488 pollination niches and interaction networks. In Functional Ecology 31 (1), pp. 101–115. DOI: 489 10.1111/1365-2435.12753. 490 Keller, Alexander; Ankenbrand, Markus J.; Bruelheide, Helge; Dekeyzer, Stefanie; Enquist, 491 Brian J.; Erfanian, Mohammad Bagher et al. (2023): Ten (mostly) simple rules to 492 future/i1proof trait data in ecological and evolutionary sciences. In Methods in Ecology and 493 Evolution 14 (2), pp. 444–458. DOI: 10.1111/2041-210X.14033. 494 Kendall, Liam K.; Rader, Romina; Gagic, Vesna; Cariveau, Daniel P.; Albrecht, Matthias; 495 Baldock, Katherine C. R. et al. (2019): Pollinator size and its consequences: Robust estimates 496 of body size in pollinating insects. In Ecology and Evolution 9 (4), pp. 1702–1714. DOI: 497 10.1002/ece3.4835. 498 Kiatoko, Nkoba; Pozo, Maria I.; Kasiera, Wendie; Kariuki, Sophia K.; van Langevelde, 499 Frank; Hundt, Baerbel; Jaramillo, Juliana (2023): Body size as a proxy of probing time and 500 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 23 visitation rates on cucumber by two African stingless bees increase fruit quality and seed 501 quantity. In Scientia Horticulturae 309, p. 111671. DOI: 10.1016/j.scienta.2022.111671. 502 Kopit, Andi M.; Pitts-Singer, Theresa L. (2018): Routes of Pesticide Exposure in Solitary, 503 Cavity-Nesting Bees. In Environmental Entomology 47 (3), pp. 499–510. DOI: 504 10.1093/ee/nvy034. 505 Kunte, K. (2007): Allometry and functional constraints on proboscis lengths in butterflies. In 506 Functional Ecology 21 (5), pp. 982–987. DOI: 10.1111/j.1365-2435.2007.01299.x. 507 Laha, Supratim; Chatterjee, Soumik; Das, Amlan; Smith, Barbara; Basu, Parthiba (2020): 508 Exploring the importance of floral resources and functional trait compatibility for maintaining 509 bee fauna in tropical agricultural landscapes. In Journal of Insect Conservation 24 (3), 510 pp. 431–443. DOI: 10.1007/s10841-020-00225-3. 511 Lamanna, Christine; Blonder, Benjamin; Violle, Cyrille; Kraft, Nathan J. B.; Sandel, Brody; 512 Šímová, Irena et al. (2014): Functional trait space and the latitudinal diversity gradient. In 513 Proceedings of the National Academy of Sciences of the United States of America 111 (38), 514 pp. 13745–13750. DOI: 10.1073/pnas.1317722111. 515 Lichtenberg, Elinor M.; Mendenhall, Chase D.; Brosi, Berry J. (2017): Foraging traits 516 modulate stingless bee community disassembly under forest loss. In The Journal of Animal 517 Ecology 86 (6), pp. 1404–1416. DOI: 10.1111/1365-2656.12747. 518 McGill, Brian J.; Enquist, Brian J.; Weiher, Evan; Westoby, Mark (2006): Rebuilding 519 community ecology from functional traits. In Trends in ecology & evolution 21 (4), pp. 178–520 185. DOI: 10.1016/j.tree.2006.02.002. 521 Melin, Annalie; Krenn, Harald W.; Bowie, Rauri C. K.; Beale, Colin M.; Manning, John C.; 522 Colville, Jonathan F. (2019): The allometry of proboscis length in Melittidae (Hymenoptera: 523 Apoidae) and an estimate of their foraging distance using museum collections. In PloS one 14 524 (6), e0217839. DOI: 10.1371/journal.pone.0217839. 525 Menegus, Giulio (2018): Seasonal morphological variation in a pollinator community: the 526 wild bees of Southern Corsica. Available online at 527 https://thesis.unipd.it/bitstream/20.500.12608/22266/1/TesiMagistraleMenegus_compressed.p528 df, checked on 12/21/2024. 529 Michener, Charles D. (2007): The bees of the world. 2nd ed. Baltimore: Johns Hopkins 530 University Press. 531 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 24 Montoya/i1Pfeiffer, Paula María; Rodrigues, Ricardo Ribeiro; Alves dos Santos, Isabel 532 (2020): Bee pollinator functional responses and functional effects in restored tropical forests. 533 In Ecological Applications 30 (3), e02054. DOI: 10.1002/eap.2054. 534 Montoya/i1Pfeiffer, Paula María; Rodrigues, Ricardo Ribeiro; Metzger, Jean Paul; Da Silva, 535 Claudia Inês; Santos Baquero, Oswaldo; Alves dos Santos, Isabel (2018): Are the 536 assemblages of tree pollination modes being recovered by tropical forest restoration? In 537 Applied Vegetation Science 21 (1), pp. 156–163. DOI: 10.1111/avsc.12335. 538 Myers, N.; Mittermeier, R. A.; Mittermeier, C. G.; Da Fonseca, G. A.; Kent, J. (2000): 539 Biodiversity hotspots for conservation priorities. In Nature 403 (6772), pp. 853–858. DOI: 540 10.1038/35002501. 541 Nilsson, L. Anders; Rabakonandrianina, Elisabeth (1988): Hawk-moth scale analysis and 542 pollination specialization in the epilithic Malagasy endemic Aerangis ellisii (Reichenb. fil.) 543 Schltr. (Orchidaceae). In Botanical Journal of the Linnean Society 97 (1), pp. 49–61. DOI: 544 10.1111/j.1095-8339.1988.tb01686.x. 545 Ollerton, Jeff; Winfree, Rachael; Tarrant, Sam (2011): How many flowering plants are 546 pollinated by animals? In Oikos 120 (3), pp. 321–326. DOI: 10.1111/j.1600-547 0706.2010.18644.x. 548 Osorio-Canadas, Sergio; Flores-Hernández, Noé; Sánchez-Ortiz, Tania; Valiente-Banuet, 549 Alfonso (2022): Changes in bee functional traits at community and intraspecific levels along 550 an elevational gradient in a Mexical-type scrubland. In Oecologia 200 (1-2), pp. 145–158. 551 DOI: 10.1007/s00442-022-05248-y. 552 Ostwald, Madeleine M.; Gonzalez, Victor H.; Chang, Carrie; Vitale, Nydia; Lucia, Mariano; 553 Seltmann, Katja C. (2024): Toward a Functional Trait Approach to Bee Ecology. In Ecology 554 and Evolution 14 (10), e70465. DOI: 10.1002/ece3.70465. 555 Peat, J.; Tucker, J.; Goulson, Dave (2005): Does intraspecific size variation in bumblebees 556 allow colonies to efficiently exploit different flowers? In Ecological Entomology 30 (2), 557 pp. 176–181. DOI: 10.1111/j.0307-6946.2005.00676.x. 558 Pélabon, Christophe; Firmat, Cyril; Bolstad, Geir H.; Voje, Kjetil L.; Houle, David; Cassara, 559 Jason et al. (2014): Evolution of morphological allometry. In Annals of the New York 560 Academy of Sciences 1320, pp. 58–75. DOI: 10.1111/nyas.12470. 561 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 25 Pollock, Laura J.; Morris, William K.; Vesk, Peter A. (2012): The role of functional traits in 562 species distributions revealed through a hierarchical model. In Ecography 35 (8), pp. 716–563 725. DOI: 10.1111/j.1600-0587.2011.07085.x. 564 Potts, Simon G.; Biesmeijer, Jacobus C.; Kremen, Claire; Neumann, Peter; Schweiger, Oliver; 565 Kunin, William E. (2010): Global pollinator declines: trends, impacts and drivers. In Trends 566 in ecology & evolution 25 (6), pp. 345–353. DOI: 10.1016/j.tree.2010.01.007. 567 Prendergast, Kit S.; Menz, Myles H. M.; Dixon, Kingsley W.; Bateman, Philip W. (2020): 568 The relative performance of sampling methods for native bees: an empirical test and review of 569 the literature. In Ecosphere 11 (5), Article e03076. DOI: 10.1002/ecs2.3076. 570 R Core Team (2024): R: A Language and Environment for Statistical Computing. Vienna: R 571 Foundation for Statistical Computing. 572 Ranta, Esa (1984): Proboscis Length and the Coexistence of Bumblebee Species. In Oikos 43 573 (2), pp. 189–196. 574 Ranta, Esa; Lundberg, Hans (1980): Resource Partitioning in Bumblebees: The Significance 575 of Differences in Proboscis Length. In Oikos 35 (3), pp. 298–302. 576 Renteria, Elizabeth; Brehm, Gunnar (2025): Is blue the most attractive color for bees? 577 Exploring the attractiveness of colors in vane traps. 578 Ribeiro, Caroline; Varassin, Isabela Galarda; Pagioro, Thomaz Aurélio; Souza, Jana Magaly 579 Tesserolli de (2024): Bee and plant traits drive temporal similarity of pollination interactions 580 in areas under distinct restoration strategies. In Arthropod-Plant Interactions 18 (5), pp. 979–581 989. DOI: 10.1007/s11829-024-10064-7. 582 Rodríguez-Flores, Claudia I.; Ornelas, Juan Francisco; Wethington, Susan; Del Arizmendi, 583 María Coro (2019): Are hummingbirds generalists or specialists? Using network analysis to 584 explore the mechanisms influencing their interaction with nectar resources. In PloS one 14 585 (2), e0211855. DOI: 10.1371/journal.pone.0211855. 586 Rodríguez-Gironés, Miguel A.; Santamaría, Luis (2007): Resource competition, character 587 displacement, and the evolution of deep corolla tubes. In The American Naturalist 170 (3), 588 pp. 455–464. DOI: 10.1086/520121. 589 Rombaut, Louie M. K.; Capp, Elliot J. R.; Cooney, Christopher R.; Hughes, Emma C.; 590 Varley, Zoë K.; Thomas, Gavin H. (2022): Allometric conservatism in the evolution of bird 591 beaks. In Evolution Letters 6 (1), pp. 83–91. DOI: 10.1002/evl3.267. 592 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 26 Schleuning, Matthias; García, Daniel; Tobias, Joseph A. (2023): Animal functional traits: 593 Towards a trait/i1based ecology for whole ecosystems. In Functional Ecology 37 (1), pp. 4–594 12. DOI: 10.1111/1365-2435.14246. 595 Schneider, Caroline A.; Rasband, Wayne S.; Eliceiri, Kevin W. (2012): NIH Image to ImageJ: 596 25 years of image analysis. In Nature Methods 9 (7), pp. 671–675. DOI: 10.1038/nmeth.2089. 597 Spicer, Michelle Elise; Mellor, Hannah; Carson, Walter P. (2020): Seeing beyond the trees: a 598 comparison of tropical and temperate plant growth forms and their vertical distribution. In 599 Ecology 101 (4), e02974. DOI: 10.1002/ecy.2974. 600 Stang, Martina; Klinkhamer, Peter G. L.; van der Meijden, Eddy (2006): Size constraints and 601 flower abundance determine the number of interactions in a plant–flower visitor web. In 602 Oikos 112 (1), pp. 111–121. DOI: 10.1111/j.0030-1299.2006.14199.x. 603 Stang, Martina; Klinkhamer, Peter G. L.; van der Meijden, Eddy (2007): Asymmetric 604 specialization and extinction risk in plant-flower visitor webs: a matter of morphology or 605 abundance? In Oecologia 151 (3), pp. 442–453. DOI: 10.1007/s00442-006-0585-y. 606 Stang, Martina; Klinkhamer, Peter G. L.; Waser, Nickolas M.; Stang, Ingo; van der Meijden, 607 Eddy (2009): Size-specific interaction patterns and size matching in a plant-pollinator 608 interaction web. In Annals of botany 103 (9), pp. 1459–1469. DOI: 10.1093/aob/mcp027. 609 Stemet, Lilia; Biddinger, David J.; Naithani, Kusum; Szalanski, Allen L.; Joshi, Neelendra 610 (2024): Intertegular Distance of Wild Bees and its Use in Estimating Proboscis Lengths and 611 Foraging Ranges to Better Understand Bee Conservation Ecology. 612 Stout, Jane C.; Allen, John C.; Goulson, Dave (2000): Nectar robbing, forager efficiency and 613 seed set: Bumblebees foraging on the self incompatible plant Linaria vulgaris 614 (Scrophulariaceae). In Acta Oecologica 21 (4), pp. 277–283. 615 Swenson, Nathan G.; Enquist, Brian J.; Pither, Jason; Kerkhoff, Andrew J.; Boyle, Brad; 616 Weiser, Michael D. et al. (2012): The biogeography and filtering of woody plant functional 617 diversity in North and South America. In Global Ecology and Biogeography 21 (8), pp. 798–618 808. DOI: 10.1111/j.1466-8238.2011.00727.x. 619 Temeles, Ethan J.; Koulouris, Carolyn R.; Sander, Sarah E.; Kress, W. John (2009): Effect of 620 flower shape and size on foraging performance and trade-offs in a tropical hummingbird. In 621 Ecology 90 (5), pp. 1147–1161. DOI: 10.1890/08-0695.1. 622 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint 27 Toledo-Hernández, Erubiel; Pena-Chora, Guadalupe; Hernández-Velázquez, Víctor Manuel; 623 Lormendez, Caleb C.; Toribio-Jiménez, Jeiry; Romero-Ramírez, Yanet; León-Rodriguéz, 624 Renato (2022): The stingless bees (Hymenoptera: Apidae: Meliponini): a review of the 625 current threats to their survival. In Apidologie 53 (1). DOI: 10.1007/s13592-022-00913-w. 626 Torres/i1Vanegas, Felipe; Hadley, Adam S.; Kormann, Urs G.; Jones, F. Andrew; Betts, 627 Matthew G.; Wagner, Helene H. (2021): Tropical deforestation reduces plant mating quality 628 by shifting the functional composition of pollinator communities. In Journal of Ecology 109 629 (4), pp. 1730–1746. DOI: 10.1111/1365-2745.13594. 630 Vajna, Flóra; Kis, János; Szigeti, Viktor (2021): Measuring proboscis length in Lepidoptera: a 631 review. In Zoomorphology 140 (1), pp. 1–15. DOI: 10.1007/s00435-020-00507-z. 632 van der Sluijs, Jeroen P. (2020): Insect decline, an emerging global environmental risk. In 633 Current Opinion in Environmental Sustainability 46, pp. 39–42. DOI: 634 10.1016/j.cosust.2020.08.012. 635 Wright, Joseph S.; Muller-Landau, Helene C. (2006): The Uncertain Future of Tropical Forest 636 Species. In Biotropica 38 (4), pp. 443–445. 637 638 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 27, 2025. ; https://doi.org/10.1101/2025.02.21.639535doi: bioRxiv preprint