Results:
Most of the 104 time-series (i.e. 96) were composed of sequences having mainly been sampled in inland locations rather than in coastal environments, suggesting that sequences from dispersal-limited populations (i.e. living in different inland water bodies) may have been sampled at the considered spatial scale. The median number of independent drainage basins sampled across time-series was 3, ranging from 0 (for Sebastes auriculatus andCoryphopterus venezuelae, the only two strictly marine species; Table A1) to 15 (forSqualius cephalus , a common freshwater species in European watersheds; Table A1 and Figure A1-C for an example). Overall, most fish time-series (84/104) grouped sequences sampled at locations situated within at least 2 different drainage basins, hence resulting in time-series representing non-biologically-meaningful groups of sequences originating from potentially non-interbreeding, demographically-independent populations.
The median number of different sampling locations per time-series was high (12), with a quite variable range (between 3-98 different locations). This, combined with the variable number of locations sampledper year, and the variable, but generally low number of sequences sampled per year and per year per location across time-series (see Table A1) illustrates the scattered nature of samples in the time-series included in Millette et al. (2020). Figure 1 in the main text, as well as Figure A1 A-D below, provide visual examples of the spatial distribution of sequences for five time-series. For instance, for Astyanax bimaculatus (see Figure A1 A below), sequences sampled at locations situated within 11 different South-American drainage basins (including the Amazon and the Orinoco river basins) were pooled together. This example, and all other examples in Figure A1 A-D, also show that the clustering algorithm used to build ‘populations’ in Millette et al (2020) may result in ‘daisy chaining’, with sequences from locations far greater than the considered clustering distance (here, 1,000 km) being grouped together.
Finally, the time-span covered by time-series in years was low (7.9 years), and represented an average of just 2.2 generations for the 33 species with generation time estimates available in FishBase, which is an insufficient time-span for most evolutionary changes in IGD.