Intermediate Pleistocene configurations and island fragmentation
The long-lasting MSL configuration does not seem to have shaped continental island species richness patterns (H3). LGM configuration on the other hand, emerged as a major predictor for both native and endemic species richness (Figure 4). The maximum, yet short-lasting connectivity achieved during the LGM (Figure 2; Sakellariou and Galanidou 2016) has negatively affected endemic diversity (except SIEs), suggesting that the LGM spatial configuration of the Aegean archipelago has left its imprint on species’ distribution, richness and evolutionary patterns (e.g., Poulakakis et al. 2015, Kougioumoutzis et al. 2017). Quaternary fragmentation, as well as the repeated and rapid paleo-islands fragmentation due to sea-level oscillations during 20-11 Ka may be responsible for the observed continental MIE patterns, apart from the dramatic island area-loss since the LGM (> 70%). This process affects island species richness irrespective of area, via passive dispersal. Island-fragmentation seems more prominent on continental rather than on oceanic MIEs. The contrast in high numbers of islands shared by continental MIEs and the low number of islands shared by oceanic islands MIEs can be simply explained by much lower numbers of paleo-islands in oceanic island settings that became fragmented by Pleistocene sea-level oscillations (Norder et al. 2018) compared to what we see in continental settings (Simaiakis et al. 2017). As a result, Aegean MIEs had typically a much larger distribution on larger pre-existing paleo-islands, sometimes even reaching beyond the Quaternary (Rechinger 1943, Rechinger and Rechinger-Moser 1951, Runemark 1969, 1971, Greuter 1971, Snogerup 1971, Strid 1972). This may indicate that Aegean MIEs may have a pre-Tertiary imprint of the paleogeography, when the sea level was as high as during the Quaternary interglacials and these land-bridge islands were true islands. Those paleo-islands fragmented earlier than the Quaternary by geological processes over millions of years (Creutzburg 1966, Dermitzakis 1990). For instance, pre-Pleistocene paleo-connections existed between the (Pleistocene) true islands Kythira, Crete, Karpathos and Rodos (Dermitzakis 1990, Schüle 1993) or between the south-central Cyclades and Kriti (Kapsimalis et al. 2009, Sakellariou and Galanidou 2016). Pleistocene sea-level oscillations further fragmented the geologically earlier fragmented populations even more, as for instance on the Cyclades (Simaiakis et al. 2017). Both geological pre-Pleistocene breaking up of islands and Pleistocene sea-level rise induced island fragmentation led to the fragmentation of paleo-endemics that perhaps once were single island endemics on larger pre-existing islands. In topographically complex islands, some of the old MIEs probably gave rise to mostly neo-endemic SIEs through allopatric speciation (e.g. Runemark 1969, 1971, Bittkau and Comes 2009, Jaros et al. 2018). For these reasons, the usual classification of endemics into SIEs or MIEs would make little sense in a continental-shelf system that is both fragmented by regionally different geological events and Pleistocene sea-level oscillations and where dispersal distances between islands are short and past land-connections were frequent. This highlights the importance of working with subsets of species data that reflect the common evolutionary history of the species included.
Continental SIEs however, are unaffected by past configurations (Table 2; Figure 4), contrary to oceanic SIEs which are driven by MSL configurations (Norder et al. 2019). Oceanic island speciation and endemism are elevated due to the MSL long-lasting, large island size (Norder et al. 2019). Meanwhile, on continental islands large island size was achieved during the LGM as a result of maximum connectivity. This enabled increased gene-flow, thus leading to lower speciation and endemism levels (e.g. Panitsa et al. 2018 and references therein). Most Aegean SIEs occur on Crete and then on 25 other islands, most of them (18 islands) containing only one SIE. Plant diversification and speciation in the Aegean is mainly driven by random genetic drift (Georghiou and Delipetrou 2010 and references therein) and active species differentiation took place inside numerous islands (Cardona and Cotandriopoulos 1978). Several Aegean endemics occupy precipitous limestone cliffs, which are remnants of mountaintops that were most pronounced during the LGM, when sea levels were up to 140 m lower. These cliffs were virtually inaccessible even to the dominant components of cliff communities in the adjacent mainland regions (e.g., Inula verbascifolia group and Campanula subsect.Quinqueloculares ) and served as refugia during the Quaternary glaciations, leading to high genetic differentiation and systematic isolation (e.g., Linum arboreum – Thomson 2005). Only three islands host more than 10 SIEs: Evvia, Samos and Samothraki, which all were land-bridge islands since LGM; Samos was a land-bridge island also during median sea levels. These three islands are mountainous and topographically most pronounced, suggesting that topography plays an additional important role in SIE formation and ecological isolation might have driven speciation (e.g. Kallimanis et al. 2010, Trigas et al. 2013, Steinbauer et al. 2013, 2016). The topographic complexity and geodiversity of these islands may also explain the ISAR SIE pattern. The large number of SIEs occurring in Crete is due to its large size, long isolation, high topographical and environmental heterogeneity, as well as its paleogeographical history (Greuter 1972, 1975, Kagiampaki et al. 2011, Médail 2017).