Aim: Biogeographic regionalization is scant for most insect groups due to shortfalls in distribution and phylogenetic information, namely the Wallacean and Darwinian shortfalls respectively. Here, we focused on the European ants and compared new techniques to classical analyses based on regional lists and taxonomic methods. We asked the following: 1) Can grid-based regionalizations using novel distribution data improve biogeographic transitions? and 2) Can phylogenetic approaches reveal new insights regarding ant evolutionary history?

Location: Europe and Anatolia.

Taxon: Ants (Formicidae).

Methods: First, we developed a refined database integrating the occurrences of 747 ant species across 207 regions of Europe and Anatolia, based on newly expert-validated records derived from the existing Global Ant Biodiversity Informatics (GABI) database. Using range estimates for these species derived from polygons and species distribution modelling, we produced species assemblages in 50 × 50 km grid cells. We calculated taxonomic and phylogenetic turnover of ant assemblages, then performed a hierarchical clustering procedure to delineate biogeographic structure.

Results: At both the regional list- and grid assemblage-levels, the Mediterranean has higher turnover and more biogeographic regions than northern Europe, both taxonomically and phylogenetically. Delineations based on grid assemblages detected more detailed biogeographic transitions, while those based on regional lists showed stronger insularity in biogeographic structure. The phylogenetic regionalization suggested a very similar spatial structure but varied affinities between assemblages in comparison to the taxonomic approach.

Main conclusions: Here, we integrated expert-validated regional lists, species distribution modelling, and a recent phylogeny to tackle Wallacean and Darwinian shortfalls for an important insect group by developing a next-generation map of biogeographic regionalization for European ants. The results of this study suggest strong constraints from geographic barriers and potential effects of climatic history on ant distributions and evolutionary history and also provide baseline spatial information for future investigations of regional insect distributions.

1. Regional species lists of the western Palearctic ants: This database is derived from the Global Ant Biodiversity Informatics (GABI, Guénard et al., 2017) but features a higher spatial resolution for the region. Ultimately, we compiled native ant taxa occurrence information for each of the 207 geographic divisions (i.e., regional lists) for the western Palearctic realm, which was previously divided to 57 regions in GABI. Our definition of the western Palearctic realm does not include North Africa and the Arabian Peninsula because of historically poor sampling and the lack of recent taxonomic revisions for species in those regions. Geographical divisions used in the database were delimited based on either administrative region (GADM, version 2.8, accessed 1^st Sep. 2020) or modified areas based on the physical geographic area (e.g., islands and mountains), depending on data availability. Preliminary versions of the dataset were validated by ant experts (co-authors of this study) who identified dubious records and provided additional information (e.g., unpublished or missing records) to complete and provide more accurate ant range maps. Occurrence records were deemed dubious for reasons including nomenclatural changes in recent taxonomic revisions, outdated taxonomy, and misidentifications (which can be numerous in older literature or databases). For all ant taxa in our database, we also verified nomenclature based on AntCat, an online, global catalog of ants (Bolton, 2021), with validation and inclusion of taxa up to July 1^st 2021. Here, we treated valid subspecies as species in our analysis, which resulted in a total of 747 valid native species (including 40 subspecies) for regional lists.

2. Binary range maps of the western Palearctic ants: Ranges were estimated for low-data species (<5 occurrence records) with univalue polygons (either buffered [30 km] points or convex/alpha hull, depending on data availability), and for species with sufficient data (≥ 5 occurrence records) using SDMs. We used the presence-background machine-learning algorithm Maxent to train models over a study extent defined by their polygon range estimate (buffered alpha hull) using 19 bioclimatic predictor variables at 10 arcminute resolution (~20 km at the equator) from Worldclim 2.0 (Fick & Hijmans 2017). We tuned models for optimal complexity (i.e., combinations of feature classes and regularization multipliers) using sequential criteria of cross-validation results (based on the 10 percentile omission rate and validation AUC; Radosavljevic & Anderson 2014) with the R package ENMeval 2.0.0 (Kass et al. 2021). We used these tuned models to make predictions of suitability over the species’ study extents, effectively constraining range estimates to the limits of the occurrence data, and made them binary (presence/absence predictions) by thresholding with the 10 percentile omission value. Range estimates represented by polygons for low-data species were converted to 10 arcminute grid cells to align with the modeled range estimates. Detailed metadata of SDMs can be found in the metadata as the ODMAP protocol format (Zurell et al., 2020).

All those data are part of an unpublished database: the European Ant Distribution (EUAD) database, we are still working on it and its open access. Thus we highly recommend you to contact us first if you want to use this dataset.