Despite the rich fossil record of Neogene chondrichthyans (chimaeras, sharks, rays and skates) from Europe, little is known about the macroevolutionary processes that generated their current diversity and geographical distribution. We compiled 4,419 Neogene occurrences comprising 102 genera, 41 families and 12 orders from four European regions (Atlantic, Mediterranean, North Sea and Paratethys) and evaluated their diversification trajectories and paleobiogeographic patterns. In all analysed regions, we found that the generic richness increased during the early Miocene, then decreased sharply during the middle Miocene in the Paratethys and moderately during the late Miocene and Pliocene in the Mediterranean and North seas. Origination rates display the most significant pulses in the early Miocene in all regions. Extinction rate pulses varied across regions, with the Paratethys displaying the most significant pulses during the late Miocene and the Mediterranean and North seas during the late Miocene and early Pliocene. Overall, up to 27% and 56% of the European Neogene genera are now globally and regionally extinct, respectively. The observed pulses of origination and extinction in the different regions coincide with warming and cooling events that occurred during the Neogene globally and regionally. Our study reveals complex diversity dynamics of Neogene chondrichthyans from Europe and their distinct biogeographic composition despite the multiple marine passages that connected the different marine regions during this time.

Data.—We gathered chondrichthyan occurrences at the genus level from the Neogene (23 to 2.6 Ma) of Europe based on a comprehensive literature quest that consisted in searching for the terms "chondrichthyans", "fossil", "Neogene" and "Europe" in Google Scholar (https://scholar.google.com). This resulted in 122 journal articles, unpublished theses, conference abstracts and books. This information was complemented with data downloaded from Paleobiology Database (PBDB; https://paleobiodb.org) and from museum online collections databases (Table S1). Additionally, collections housed in the Natural History Museum of Vienna and the State Museum of Natural History of Stuttgart, Germany, were examined. In total, we collected 4,419 occurrences (Fig. 1; Table S1), which we assigned to four regions: Atlantic (n=433), Mediterranean Sea (n=750), North Sea (n=614), and Paratethys (n=2622), based on the paleogeographic reconstructions proposed by Rögl (1999). The regional stratigraphic stages (i.e., Paratethys Sea) were updated based on more recent studies (Heckeberg et al. 2010; Grunert et al. 2010; Hohenegger et al. 2014; Kovac et al. 2018). Ambiguous records with unclear taxonomic names (i.e., non-valid synonyms) or localities (i.e., assigned only to country-level) were excluded from the database. All taxonomic names were updated according to the most recent taxonomic reviews (Cappetta 2012; Pollerspöck and Straube 2021).

Analyses.—To reconstruct diversification trajectories (i.e., genus richness, origination and extinction rates), we used the first and last appearance of each genus based on their occurrences distributed in 1 Ma time bins in each region (i.e., Mediterranean, North Sea and Paratethys). We excluded the Atlantic region from the analysis due to its low number of occurrences (n=440) and the lack of information on the stratigraphic ages of many localities. Genus richness was calculated per time bin using two approaches: 1) “boundary-crossers” (i.e., number of taxa that cross the boundary of the interval; Foote 2000); and 2) shareholder quorum subsampling (i.e., fixed coverage of the frequency curve of genus occurrences, Alroy, 2010). We used 1,000 iterations and quorums of 0.4, 0.6 and 0.8. This last approach was implemented because it accounts for differences in sampling effort, unlike the other approaches, which are prone to sampling biases (Alroy, 2010). However, a strong positive correlation between genus richness (estimated using boundary-crosser method) and the subsampled genus richness (estimated using SQS method) might suggest that sampling bias is relatively systematic in time (Table S7). Origination and extinction rates were estimated as described in the per capita rates of Foote (1999)