Aim: So far, latitudinal body size-clines have been primarily discussed in the context of thermoregulation, sensu Bergmann. However, body size patterns are ambiguous in ectotherms and this heterogeneity remains poorly understood. We tested whether Bergmann’s rule and the resource availability rule which states that energetic requirements determine species’ body size, apply to damselflies and dragonflies (Odonata). Furthermore, we hypothesised that the contrasting effects of thermoregulation and resource availability (e.g. productivity) can obscure the overall gradient in body size variation.

Location: Global

Time period: Contemporary

Major taxa studied: Odonata

Methods: Using data for 43% of all odonate species described so far, we tested our hypotheses in phylogenetically and spatially comparative analyses at assemblage and species level. For the distribution data, we integrated expert range maps and ecoregional ranges based on all available occurrence records. To distinguish between long-term versus evolutionarily recent responses of environmental drivers in body size, we constructed a phylogenetically informed classification of all odonate species and decomposed the body size into its phylogenetic and specific component for our subset of species.

Results: We documented a weak positive relationship between body length and latitude but found strong and contrasting effects for temperature between dragonflies and damselflies and consistent positive effects for productivity that explained 35%–57% of body size variation. Moreover, we showed a strong phylogenetic signal in sized-based thermoregulation that shaped the distribution of dragonflies, but not of damselflies.

Main conclusion: We concluded that temperature, productivity, and conservatism in size-based thermoregulation synergistically determine the distribution of ectotherms, while the taxon-specific importance of these factors can lead to contrasting results and weak latitude–size relationships. Our results reinforce the importance of body size as a determinant of species distributions and responses to climate change.

For our main analysis, we combined three types of information.

1) Body size

We compiled body size data from measurements of museum specimens and from the literature for 2,802 odonate species worldwide. For the main analyses, we used only data of body length for adult male individuals but mobilized other proxies of body size if available for imputation of the body length and supplementary analysis of body shape differences between Anisoptera and Zygoptera. Specifically, we measured the body and hindwing length (excluding terminal appendages) from images of 724 individuals of African odonates provided by the Naturalis Biodiversity Center (RHNM, Leiden, The Netherlands) and 487 specimens of African species from the Senckenberg Natural History Museum (SNHM, Frankfurt, Germany). For the images of African species from the Naturalis Biodiversity Center, European species from Dijkstra and Lewington (2006), and North American species from Needham et al. (2000), we calculated the body length, hindwing length, and body area as previously described (Pinkert et al. 2017; Zeuss et al. 2017) using the R-package ‘png’ (Urbanek 2013). In short, the number of pixels of the body from the head to the distal end of the abdomen, that of the hindwing from its base to the tip, and the number of all pixels of the body were calculated. The pixel estimates were transformed to metric units through the product of the scale (provided or measured on the images) and image resolution. The body area and body length data from image-based measurements of 1,146 individuals were used to test for the difference in the body shape of both suborders.

3,612 additional length measurements were extracted from species descriptions provided in 19 literature and 2 internet sources (Table S1). Because of sexual dimorphism, we did not use females in our study if measurements from literature differentiated between sexes. If sources reported descriptive body size statistics, we collected the minimum and maximum values to calculate means to aid the integration of data across sources. For 305 individuals, we predicted the body length from the provided hindwing length with a linear mixed effects model that included a random slope for genus nested in family and suborder (n = 810, conditional R² = 0.92). Finally, the 5,128 individual measurement values (1909 singletons including those where only average values were provided) of 2,802 species were aggregated to average values of body length (‘body size’ hereafter) per species.

Because of this very strong phylogenetic signal in the body size of odonates, we partitioned the total variance of average species body size into a phylogenetic and specific component, using Lynch’s comparative method (Lynch 1991) in the R-package ‘ape’ (Paradis et al. 2004). The different aspects of body size variation in species-level, as well as assemblage-level analyses (i.e. averaged across species co-occurring within a 100 km × 100 km grid cell), are named ‘BL_mean’ (unpartitioned), ‘P’, and ‘S’, respectively. The P component represents the variation in body size predicted by the phylogenetic relationships between species. The S component represents residuals from these predictions and hence the species-specific deviation from the phylogenetically predicted part.  

2) Distribution data

We combined two types of distributional information: expert range maps and ranges derived from intersections of occurrence records with the terrestrial ecoregions of the world. We downloaded expert range maps from IUCN.org (IUCN 2021) and digitised range maps that cover the entire ranges of European odonates from Boudot and Kalkman (2015). The data were taxonomically harmonised and intersected with grid cells of approximately 100 km × 100 km (military grid reference system [mgrs]). However, many of the IUCN range maps were incomplete or were delineated by political borders instead of factual species ranges (Hughes et al. 2021). Except for the range maps from Boudot and Kalkman (2015), we used ecoregional ranges to extend and complete the dataset characterizing the distribution (Pinkert et al. 2022a).

3) Environmental data

To investigate the environmental drivers of body size variation in Odonata assemblages, we used two variables associated with geographic patterns of temperature (mean annual temperature [‘Bio_1’] and elevation [‘Elev’]) along with the enhanced vegetation index (‘Annual_EVI’) as a proxy for productivity. The data were downloaded from the CHELSA (Karger et al. 2017, 2018; chelsa.org, current condition records) and EarthEnv (Amatulli et al. 2018) databases. The EVI layer was cropped to the extent of the climate variables (1 km × 1 km).

For our main analyses, these data were aggregated(averaged) to the assemblage level. In addition to the trait and environmental data, the dataset included coordinates of the centroid of each grid cell. For detailed methods and further descriptions see Mähn et al. (Beyond latitude: Temperature, productivity, and thermal niche conservatism drive global body size variation in Odonata).