The interaction between the Linnean and Darwinian shortfalls affects our understanding of the evolutionary dynamics driving diversity patterns of New World coralsnakes

Aim: In this study, we sought to understand how the Linnean shortfall (i.e., the lack of knowledge about species taxonomy) interacts with the Darwinian shortfall (i.e., the lack of knowledge about phylogenetic relationships among species), which potentially jeopardizes geographical patterns in estimates of speciation rates.

Location: New World

Taxon: Coralsnakes (Serpentes: Elapidae)

Methods: We created an index of taxonomic uncertainty (ITU) that measures the likelihood of current species being split after undergoing future taxonomic revisions. The ITU was used in simulations where species with higher taxonomic uncertainty had a higher likelihood of having their phylogenetic branches split, generating new hypothetical species along their geographic ranges. We estimated the speciation rates before and after the split of taxonomically uncertain species.

Results: We found that a high number of coralsnake species display substantial taxonomic uncertainty, positively correlated with the latitude of the species' geographical range centroid. The estimated speciation rates based on currently available data have a weak relationship with latitude. However, after incorporating taxonomic uncertainty into the phylogeny, we detect a higher positive correlation between speciation rate and latitude.

Main conclusions: The observed change in speciation rates following the incorporation of taxonomic uncertainty highlights how such uncertainty can undermine the empirical evaluation of geographical patterns in speciation rates, revealing an interaction between the latitudinal taxonomic gradient and the latitudinal diversity gradient. Given that taxonomic changes can alter the number of species recognized as valid over time, our study highlights the need to incorporate taxonomic uncertainty into macroecological and macroevolutionary studies, enhancing the robustness of patterns inferred from these data. --

Methods
Data

List of accepted species and their description years: New World coralsnake accepted species and description years were obtained from Silva Jr. et al. (2021a).

Maximum body size: The maximum body size was obtained from Feldman et al. (2015), Silva Jr. et al. (2021b), and the specialized literature.

Number of synonyms: The number of synonyms was obtained from the Reptile Database (Uetz et al., 2022) and Silva Jr. et al. (2021b).

Published papers: The number of published papers for each species was obtained from searches in Web of Science and Scopus using the accepted species name and its synonyms.

Taxonomic uncertainty: We distributed a questionnaire to taxonomists specializing in coralsnakes and asked them what the likelihood was of each species undergoing a splitting. Taxonomists could answer with “very unlikely,” “unlikely,” “likely,” or “very likely.” We assigned a value range from 1 (very unlikely) to 4 (very likely) and obtained the average of the taxonomists’ answers for each species.

Macroecological variables: Occurrence records were gathered from the Global Biodiversity Information Facility (GBIF, https://www.gbif.org/), scientific literature, and vouchers from museums and biological collections. Records retrieved from GBIF underwent a cleaning process. We calculated the number of occurrence records for each species. We used the occurrence records to generate the geographic range for each species using minimum convex polygons and calculated the range size in square kilometers. We also obtained the centroid for each species.

Phylogenetic tree: The phylogeny was obtained from Tonini et al. (2016). Species in Tonini’s phylogeny that were not considered valid by our list of accepted species were removed, and accepted species not present in Tonini’s phylogeny were included based on taxonomy. We explored uncertainty in branch lengths by obtaining one thousand phylogenetic trees.

Data analysis

We calculated the speciation rate using the DR statistic proposed by Jetz et al. (2012) for each species.

We performed a Phylogenetic Generalized Least Squares (PGLS) where the eight variables mentioned above (year of description, number of occurrence records, number of papers, DR, number of synonyms, latitude of the centroid, body size, and range size) were used to model the taxonomists’ responses. The estimated values of the model were scaled to vary between 0 and 1 and were used as our Index of Taxonomic Uncertainty (ITU). Values closer to 1 indicate high taxonomic uncertainty (i.e., high likelihood of species splitting), while values closer to 0 indicate low taxonomic uncertainty (i.e., low likelihood of species splitting).

ITU was used in simulations where species with high ITU had a higher likelihood of undergoing splitting, generating new sister species in phylogenetic trees. For each of the 1,000 phylogenetic trees, 100 new phylogenies were generated. To estimate the effect of latitude on speciation patterns, we performed two PGLS models. The first model considered the DR statistic as a function of latitude (formula: DR ~ latitude), while the second model incorporated a quadratic term to account for non-linear trends in speciation patterns (formula: DR ~ latitude + latitude²).