Cetaceans are good examples of this kind of species. Despite their broad range and their high dispersal capabilities, many cetaceans often show substantial genetic structure at regional or even fine scale, although the extent varies among species [8]. It is generally argued that these patterns, not always correlated with geographic features, are related to a combination of complex behaviours, such as philopatry, specialisations for local resources, or social organisation into kinship groups [8,9]. On the other hand, while the dispersal and segregation of populations of terrestrial mammals are frequently influenced by geographic features or climatic characteristics, few such obvious barriers are expected to restrict cetacean dispersal and gene flow in the world's oceans [10,11]. Variation in oceanographic properties of the water column, such as depth, temperature, currents and winds, are known as important factors in the life of these animals, most obviously in conditioning the availability of their food (for example, see [12]), but their effect on cetacean dispersal and on population structure remains enigmatic.

Small coastal cetaceans such as those of the porpoise family are a model of choice to investigate this issue because they have to face a suite of intrinsic problems not encountered by larger dolphins and whales. Their small size, their demanding reproductive schedule, and their limited ability to store energy force a strong dependency on their food [13,14]. Therefore, we expect that variation in oceanographic features that determine food availability and abundance (i.e., bathymetry, temperature and primary productivity) should markedly affect local density and dispersal of porpoises. If true, their population genetic structure should correlate, at least partly, with oceanographic characteristics. To test this hypothesis, we examined the genetic structure of one the most widely distributed porpoises, the harbour porpoise Phocoena phocoena (L. 1758), with regards to seascape characteristics. Harbour porpoises occur fairly continuously throughout cold coastal waters of the North Pacific and the North Atlantic, with a relict population in the Black Sea separated from the Atlantic range by the Mediterranean Sea where porpoises are nowadays absent [15-17]. We analysed genetic polymorphism at 10 microsatellite loci for an extensive sampling (n = 752) covering the main distribution of harbour porpoises in the central and eastern North Atlantic (Figure ​(Figure1)1) using a combination of recent individual-based landscape genetic approaches [18-21].

The total isolation of harbour porpoises from the Black Sea has long been suggested on the basis of the lack of field observation of porpoises in the Mediterranean Sea [17], of private mtDNA alleles reported in that population [33], and of morphological differences [34]. Our results lend further support to this hypothesis. The pronounced genetic footprint of this isolation left at nuclear and mtDNA loci suggest this is an ancient isolation that might date back to the last Ice Age ([35] and Fontaine, unpublished results). The genetic differentiation detected at microsatellite loci between the Iberian porpoises and those further north was not apparent at the mtDNA control region previously analysed [35]. The lack of mitochondrial lineage sorting and of private microsatellite alleles suggests that the differentiation we observed with microsatellite analyses is caused by a more recent isolation process than that of the Black Sea.

The corollary of these results is the inference of strong barriers to gene flow in the southern Bay of Biscay and in the Mediterranean Sea that isolate almost completely the Iberian and Black Sea populations. These barriers coincide with strong oceanographic changes of similar nature (compare Figure ​Figure55 with Figures ​Figures11 and ​and7).7). To take them in turn, the conditions in the southern Bay of Biscay differ sharply from those at its margins [36,37]. The continental shelf, widely extended in the northern part, narrows considerably to the south and is cleaved asunder by the Cap Breton canyon, which drops to the abyssal plain in the south-east, only 10 km from the shore. Warm and oligotrophic surface water spreads from the Cap Breton canyon to cover half of the southern Bay in summer [36,37]. In contrast, off the Iberian Atlantic coast upwelling becomes evident from late spring to early autumn [38], bringing to the surface cold nutrient-enriched waters that support a rich food-web [39]. On the north side of the barrier, shallow, cold, and nutrient rich waters prevail most of the year from the French waters of the Bay of Biscay northward to the northern North Sea. From a biogeographical point of view, the southern Bay of Biscay is not only a barrier for porpoises but it is also a transition zone between the boreal and subtropical provinces, with many species reaching their southern or northern limit of distribution in that area [40].

Still further north, depth increases towards Nordic Seas (Figure ​(Figure1),1), but waters remain cold and highly productive [41]. However, the bathymetric change does not seem to restrict gene flow in Nordic Seas, consistent with sightings of some porpoises reported far offshore in deep water [42]. While this suggest there are few, if any, potential barriers to dispersal of porpoises from the northern Bay of Biscay up to Arctic waters, the heterozygosity deficit related to the detected IBD shows nevertheless that porpoises do not mate randomly over that extended area and that gene flow is spatially restricted. We observed a north-south variation in the IBD pattern with higher IBD slope at the southern end of this range compared to northern parts (Table ​(Table2).2). One could argue that this north-south variation in IBD pattern might reflect drift disequilibrium [43] in northern areas associated with the postglacial porpoise recolonisation of Nordic waters in contrast to the southern habitats, which likely remained more stable in time. However, simulation-based sensitivity analysis of current Dσ ² estimation to demographic instability in time and space conducted by Leblois et al [44] showed that spatial expansion with constant density does not significantly affect present-time Dσ ² estimation, especially when the spatial expansion occurred 20 or more generations ago, as it is the case for postglacial recolonisation. Consequently, the higher IBD slope detected in the southern area (NAt-3A) compared to that in waters further north (NAt-3B and NAt-3C) most likely represents a lower current-time Dσ ². Although we cannot exclude variation in σ², a lower porpoise density in southern waters is supported by field estimates based on aerial and ship surveys conducted in the North Sea and adjacent waters [45,46]. These variations in density (and maybe in dispersal patterns) likely reflect variation in habitat. The southern part of the 'continuous' population (i.e., the northern part of the Bay of Biscay, the English Channel and the southernmost part of the North Sea) borders the barrier to gene flow detected in the southern Bay of Biscay and should thus display sub-optimal conditions for porpoises while the middle (i.e., the central and northern North Sea) and northern areas (i.e., the Nordic Seas) would be more optimal for a cold water species such as the harbour porpoise.

The Mediterranean Sea displays similar characteristics to those encountered in the southern Bay of Biscay but at a larger scale. The Mediterranean is composed mostly of deep basins and narrow continental shelves with warm oligotrophic surface waters prevailing most of the year [47]. These characteristics are likely quite unfavourable for cold water species and might explain why the harbour porpoise is absent from this area. The oceanographic conditions in the Black Sea are, by contrast, more suitable for harbour porpoises with low salinity, colder and more nutrient rich surface waters than in the Mediterranean Sea [48]. There are however reports of porpoise strandings in the northern Aegean Sea [17]. This can be understood with regard to oceanographic features in that area. The subdivision of the Aegean into two basins has long been recognised. The northern basin is under the influence of cold, low salinity waters that pour out of the Black Sea. This water is entrained into a cyclonic circulation affecting the northern and western parts of the Aegean, causing an ecological isolation of the northern basin from the southern basin [49]. In the southern basin the continental shelf is very limited and the waters become quickly characteristic of Mediterranean waters [50], unfavourable for harbour porpoises.

To summarise, surface water temperature and primary production seem to be the factors that best characterise the nature of barriers to gene flow encountered across the harbour porpoise range, their population structure, and their geographic distribution. It is worth noting however that in oceanography, these two parameters are often linked [51]. Indeed, the sea surface temperature acts as a useful proxy for other physical processes, such as vertical stratification and nutrient contents, regulating the size structure, taxonomic composition, and abundance of the phytoplankton community, and thus the food availability for top predators [52,53]. These results reinforce previous ecological studies on harbour porpoises that reported significant relationships between abundances and movements with sea surface temperature and food availability [54,55]. Although bathymetry can be important in harbour porpoise ecology [56,57], we showed that this factor alone seems not to restrict gene flow in northern waters of the sampling range.