Appendix: Supporting Information Drake et al SI Text 1) Trans-Saharan Species Distributions Maps of the fauna of North Africa have been compiled from Van Damme (1) and Le Berre (2) with additional distributional and species information from more up to date sources (3 to 9). This information was evaluated to identify species found north and south of the Sahara, along the Nile, in isolated oases within the desert, and combinations of the above spatial distributions. A comprehensive literature review of these species was then conducted to assess the numerous changes in their status since publication of their ranges. Some species had been subdivided, thus their names and ranges needed to be updated. Sometimes they no longer had a spatial distribution indicative of trans-Saharan migration and these species were discarded. For example Lemniscomys barbarus was originally thought to be found in the Maghreb and the Sahel but was subsequently found to consist of two species, Lemniscomys zebra south of the Sahara and Lemniscomys barbarus to the north (10). Other phylogenetic studies revealed contrasting results indicating that animal populations to the north and south of the Sahara were closely related; thus suggesting their recent dispersal across the desert. For example, the morphological variation in the leaf-nosed bat, Hipposideros caffer and H. rubber, has led to suggestions that this species complex may contain more than two species. Phylogenetic analysis suggests four distinct lineages (11). Two distinct sister clades are found within H. caffer with Hipposideros caffer caffer restricted to southern Africa and Hipposideros caffer tephrus inhabiting the Maghreb, the Sahel and Arabia and thus retaining a trans-Saharan spatial distribution for one leaf-nosed bat species. In some cases comprehensive genetic studies have been conducted on trans-Saharan animal species that requires their wholesale reassessment. It was initially thought Chalcides ocellatus was found both north and south of the Sahara and in refuges within the desert (2). However, phylogenetic analyses found it to be a clade, exhibiting as much depth as the three other main clades of Chalcides, and having at least six main lineages (6). The species is thought to have originated in Morocco where it diverged into a xeric (C. o. ocellatus) and mesic (C. o. tiligugu) unit at about 2.3 1 Ma. An independent lineage spread eastwards into the Tunisian region at about 4.6 Ma and then along the Mediterranean coastal region to Egypt by around 3Ma, where it divided into two subclades at around 1.4 Ma. One of these migrated to Israel, and the other to Turkey and Cyprus. There is little genetic information on the Sub-Saharan species of C. ocellatus, and it is thus not possible to say when they migrated across the Sahara. However, one Saharan clade does suggest recent trans-Saharan dispersal. It consists of southern Egyptian and southern Mauritanian samples of. C. ocellatus humilis, which separated from other species in the C. ocellatus group at about 4.6 Ma. Even though the two samples are 4200 km apart on opposite sides of the Sahara, they are genetically similar, suggesting either a relatively recent spread, or that they were in contact until recently. C. o. humilis now has a fragmented range which may be attributable to its spread across the Sahara in a recent humid period and subsequent isolation by aridification (6). The results of this literature review are presented in Table S1 that lists fauna showing a trans- Saharan species distribution and infers the most likely dispersal route from this distribution. 2) Aquatic Animal Dispersal Mechanisms Although exchange of fish between the Nile, Chad, Niger, Volta and Senegal basins and the waterholes of the Sahara during the Quaternary has long been accepted in order to explain the similarity in their ichthyofauna (12), the method by which this was achieved has not been satisfactorily explained until now. Theoretically it can be explained by fish dispersal overland, through tornadoes transporting water and associated wildlife (13) or by specialist adaptations (14). Clarid fish can employ terrestrial locomotion, while cyprinodonts have drought-resistant eggs that can be transported in mud attached to birds’ feet and fish can be dropped alive by birds, a particularly effective method of dispersal for mouth brooding cichlids whereby a large number of individuals can be transported at one time (14). However, these potential modes of transport do not appear to be at all effective globally because the vast majority of neighbouring basins have significantly different ichthyofauna, as is the case in nearby basins in West Africa (15). Because cyprinodonts can be dispersed by birds they should be the least likely fish to exhibit such spatially restricted species distributions, yet most cyprinodonts have limited distributions (14). Thus, although it is feasible for fish to be dispersed by birds, in practice it is rare. The more likely and most practiced means of dispersal is for fish to migrate from one region to another by water connections. The biogeography of fish can thus provide valuable information on palaeo-drainage patterns (12). Therefore the presence of Tilapia zillii, Clarias gariepinus, Hemichromis 2 letourneauxi and Raiamas senegalensis throughout much of the Sahara (Table S1)) suggests widespread hydrological connections. Freshwater molluscs are more diverse in their modes of dispersal (1). Like fish, some molluscs require permanent hydrological connections whereas others can use temporary water interconnections. Many freshwater molluscs require no connections at all, being transported by attachment to birds, amphibians, water insects and perhaps large aquatic animals such as hippopotamus and crocodile (1). Some mollusc species have evolved to be transported by animals; freshwater pulmonates lay desiccation-resistant sticky eggs that adhere to birds and are thus carried by them, whilst hard-shelled snail species can pass through fish digestive systems undamaged and Sphaeriids can to attach themselves to the gills and fins of fish (1). Van Damme (1) concludes that these different dispersal mechanisms can in some cases provide indications of different palaeo-hydrological conditions. Most common Saharan extant and fossil molluscs, such as all Pulmomates and Melanoides tuberculata (Fig. S5)), are readily dispersed overland and do not constitute strong evidence for the presence of major permanent river systems. Thus it is no surprise that these species are found throughout the Sahara and much of the more humid regions that surround it. Given the different dispersal mechanism from fish the similarity in Saharan fish and mollusc biogeographical provinces is surprising (main text Fig. 3 and Fig. S5) and could indicate that the animals that dispersed these molluscs followed the interconnected trans- Saharan river system believed to have been used by the fish. The presence of Bellamya unicolor, Cleopatra bulimoides, Pila and Lanistes, however, are indicative of stable, long-term hydrological links. They are restricted to the southern Sahara (Fig. S13) suggesting that it was only in this part of the Sahara were the links were deep and durable. 3) Palaeo-hydrological Mapping The recent availability of free digital topographic data and satellite imagery for the entire Sahara has allows an integrated assessment of its palaeo-hydrology for the first time. To achieve this we have interpreted 3 arc-second Shuttle Radar Topography Mission (SRTM3) digital elevation model (DEM) (16,17) and Landsat Thematic Mapper (TM) satellite imagery. The SRTM3 DEM was used to identify palaeo-river channels, closed basins that would have been likely places for lacustrine sedimentation, palaeo-lake shorelines and spillways. Interpretation of Landsat TM false colour composite (FCC) imagery provided further information on palaeo-river systems. Remote sensing was particularly useful for mapping lake sediment outcrops, as the resulting limestones and 3 gypsum-rich sediments are readily discriminated from other materials in Landsat FCC imagery. Where the palaeo-hydrology has already been studied we have integrated our interpretations with the wider literature. Our interpretation of the palaeo-river and lake systems in Libya, the western desert of Egypt, the River Nile catchment and parts of the Chad Basin were guided by references 18 to 28, in southern Mali and Mauritania by 29 to 33; in northern Mali and southern Algeria by 34 to 35 and in northern Algeria and southern Tunisia by 36. Palaeolake areas were generally estimated from the shorelines evident in the DEM using a thresholding method (37), but when they were not evident other information was employed. For instance the size of the lake in the basin of the Chotts was estimated from the DEM using the reported altitude of the outflow channel (38), while the minimum size of the Ahnet-Mouydir Megalake in Algeria was estimated from the DEM using the altitude of 230m reported for dated lake sediments (39). The latter is a conservative estimate as lake shorelines can be considerably higher in elevation than some lake sediments due to lake bed bathymetry. To construct the palaeo-hydrological map of the ‘green Sahara’ during the last interglacial the dates (18,38,40) were combined with new data reported in this paper (Fig. S6). Information on locations and descriptions of the sample types and sites was then combined with the palaeo-hydrological map to determine the area of influence the sample represented. 4) Mechanisms for Saharan Palaeo-hydrological Links A number of different mechanisms for the transfer of water-dependant biota between basins have been proposed, most commonly lake overspill, river capture and tectonics (3). It has not generally been recognised that large alluvial fans, which are prevalent in the Sahara, also provide a very effective mechanism. The Sahara desert, and the river basins that feed water into it, contain 17 such palaeo-fans (main text Fig. 1), all of which were very large with seven being larger than the present day Okavango Delta, the largest active sub-aerial fan in the world (56).