The study area extends from the northern coast of the UK to the coast of the Republic of the Congo. It covers six marine ecosystems: the North Sea, the Celtic Seas, the South European Atlantic Shelf, the Western Mediterranean Sea, the Canary Current, and the Guinea Current (Figure 1A). Boundaries of the northern ecosystems in European waters were defined by ICES/International Council for the Exploration of the Sea (Nathalie, 2015). The recent modification of the northern ecosystem boundaries by ICES in 2015 was not applied in this study. The boundaries of the southern ecosystems in the African waters were based upon the LME/Large Marine Ecosystems (Sherman and Hempel, 2009; O'Reilly, 2013).

The zoogeographic importance of these latitudinal areas, especially the regions in the middle, has long been recognized, for they represent the transition zone between the warm-temperate north-eastern Atlantic and the cold-temperate regions (Gomes, 2001). Many species have their southern or northern distribution limits in the central region, which makes the zone an area of great sensitivity for the detection of climate change (Teixeira et al., 2014). This is one of the reasons this area was selected. Despite the migration of several marine fish species into these regions, studies of species distribution in these areas are rare compared to adjacent areas, such as the relatively closed water body of the Mediterranean Sea.

A preliminary study on the evolution of SST in the area of interest showed a decadal trend of SST rise. Thus, numbers of species were calculated over three decadal periods from 1982 to 2012, based on the availability of SST data obtained from satellite observation: 1983–1992 (Period I), 1993–2002 (Period II), and 2003–2012 (Period III).

The number of species could be strongly dependent on sampling intensity, both for sample size and individual density. For standardization, a rarefaction procedure was applied to analyze the number of species in each ecosystem and period. This procedure consists in plotting randomized species number against the sampling intensity to estimate the number of species in an ecosystem. Rarefaction curves report the average values of randomized species number derived from re-sampling without replacement (Rossi, 2011). These curves were used to determine possible ranges in species richness. When the rarefaction curve did not stabilize, the sampled occurrences were considered not representative of true species richness and that curve was excluded from the analysis. The Rich package from R was used to perform this analysis. Further details about the rarefaction procedure can be found in Rossi (2011).

Some species were present in several ecosystems studied. Jaccard similarity analysis was performed to obtain initial information of the closeness of the ecosystems. Jaccard similarity summarizes the number of species shared between ecosystems divided by the total number of species found in the ecosystems. The tool used was Vegdist function from the Vegan package in R.

Numerous environmental parameters influence the range of species distribution, such as temperature, bathymetry, salinity, substrate, coastline exposure, ocean circulation, nutrient availability, and disturbance by human activity. The impacts of these parameters differ in degree, from low to critical. Accounting for all of these parameters is beyond the scope of this study, and the data are not available at the scale of this study. Direct observation through fieldwork is costly and limited to relatively shallow waters. Temperature and bathymetry are generally regarded the main drivers of species distribution (Summerson et al., 2007; Zenetos et al., 2010) and the data for these factors are available at the spatial and temporal scale of this study. Thus, in this study, we limited our analyses to these two limiting factors. The effects of the other environmental parameters will be addressed in the discussion section.

Comprehensive and consistent time series data are necessary to elucidate the effects of continuously changing environmental conditions on species distributions. The only sources of data so comprehensive are satellite sensors.

Analyses of number of species over time in the study area are presented in Figure 4 (see Supplementary Table 2 for details).

In the North Sea, the number of species increased slightly from the first period to the second period following an increase of occurrence records (Figure 4A). In the recent period of 2003–2012, number of species in this region decreased slightly, following a slight decrease of occurrences.

In the Celtic Seas, the number of species decreased over time despite an increase in the number of occurrence records (Figure 4B). This implies there were more individuals of less species.

In the South European Atlantic Shelf, the number of species increased considerably over time following an increasing number of occurrence records (Figure 4C). Similarly, in the Western Mediterranean Sea, the number of species increased over time along with the increase of occurrences (Figure 4D). In the first period, the numbers of occurrences for both ecosystems were insufficient for rarefaction analysis (see rarefaction curves in Supplementary Table 3).

In the Canary Current, although the number of occurrences decreased considerably over time, the number of species decreased only slightly (Figure 4E). A similar pattern was observed in the Guinea Current (Figure 4F). In the third period, the numbers of occurrences for both ecosystems were insufficient for rarefaction analysis (see rarefaction curves in Supplementary Table 3).

In the northernmost ecosystems (North Sea and Celtic Seas) species occurrences seem to increase over the time periods and this is accompanied by stable or slightly decreasing numbers of species. The central-most region (South European Atlantic Shelf and Western Mediterranean), on the other hand, appears to be characterized by marked increases in both the number of occurrences and the number of species over time. Finally, the southernmost region (Guinea current and Canary current) exhibits a decrease in both occurrences and the number of species. These results seem to point to an overall enhancement in the biodiversity and biomass in the central region to the detriment of the southernmost region.

The cases of S. senegalensis, Diplodus vulgaris and A. tobianus are presented for further analysis on observational history of species migration. S. senegalensis and D. vulgaris are presented as examples of species with a northward migration pattern, whilst A. tobianus is presented as an example of species with a distribution pattern likely related to seasonal migrations and possibly extending its depth range. In addition, these species have a high commercial value and are therefore, well studied.

In the first and second period of study, S. senegalensis and D. vulgaris were only observed in the Canary Current and the Western Mediterranean Sea (green and yellow points on maps in Figure 5). These species were recently observed in the third period of study in the South European Atlantic Shelf (red points on maps in Figure 5).

Ammodytes tobianus was mostly found in the North Sea during winter (green points on map in Figure 6). In spring and summer, this species could be found both in the Celtic Seas and the North Sea. In autumn, it could be observed in the lower latitudes of the Bay of Biscay in the South European Atlantic Shelf (purple points on map in Figure 6).

In the first three seasons, A. tobianus was generally found in areas between 2 and 100 m deep (Figure 7). Fifty percent of its occurrences were found in areas between 20 and 52 m deep, indicating its preferred depth range. In autumn, this species can be found in areas up to 150 m deep.

Figure 8 presents the movement of mean seasonal SST in all of the studied ecosystems from 1982 to 2012. The monthly SST values in each ecosystem are available in Supplementary Tables 4–6.

In winter, large differences of mean SST were observed among ecosystems, except between the South European Atlantic Shelf and the Western Mediterranean Sea. These two ecosystems both have mean SSTs that range from 12.9 to 14.01°C (detail in Table 2).

In spring, the temperature of the North Sea was close to that of the Celtic Seas. The temperature in the South European Atlantic Shelf reached the range of the winter temperature of the Western Mediterranean Sea. The mean SST of the Western Mediterranean Sea during this season was close to that of the Canary Current.

In summer, the temperature increased in all ecosystems except for the Guinea Current, which showed an opposite trend. The mean SST in the South European Atlantic Shelf was in the range of spring temperatures of the Canary Current and of the Western Mediterranean Sea.

In autumn, the mean SST decreased slowly in almost all ecosystems. The temperature in the South European Atlantic Shelf during this season was in the range of summer temperatures of the North Sea and of the Celtic Seas.

The analysis of Jaccard similarities between ecosystems distinguished two general groups of ecosystems: one including the South European Atlantic Shelf, the North Sea and the Celtic Sea, the other including the Guinea Current, the Canary Current and the Western Mediterranean Sea (Figure 9).

In the northern part of the study area, the North Sea shares many species with the Celtic Seas (Figure 9). Some of these species reach the South European Atlantic Shelf, such as Salmo salar and A. tobianus. In the southern part of the study area, the Canary Current shares many species with the Western Mediterranean, such as Coris julis and Pomatomus saltatrix. The Canary Current also shared many species with the Guinea Current, such as A. parkii and Caranx crysos.

Out of the 89 selected species, we found 13 species that extended their area of distribution northward. Among these species, approximately seven species that were initially found in the Canary Current were recently found with marked records in the Western Mediterranean Sea (ID: CW in Figure 10A, Table 3). Six species that were initially in both the Canary Current and the Western Mediterranean Sea were recently found in the South European Atlantic Shelf (ID: CWS).

Thirty-five species shared their area of distribution with the adjacent ecosystem or moved among ecosystems according to season. Among these species, 10 species were always found in both the Canary Current and the Western Mediterranean Sea (ID: C = W). Twelve species from the Canary Current had ranges that included the Guinea Current (ID: C = G). Thirteen species were present in both the Celtic Seas and the North Sea (ID: C = N), and among those 13 species, 12 also reached the South European Atlantic Shelf (ID: NCS) (Table 3).

Regarding the last 41 species studied, it was unclear whether they are moving or not. Among these species, 27 were always present in all ecosystems and at all periods of the study (ID: ALL in Table 3), and 14 species were present with very low occurrence records (ID: D in Table 3).

The species of Atlantic-origin that have been observed in Mediterranean waters in the last period are among the species listed as exotic for the Mediterranean Sea by the CIESM (2002) (ID: CW in Table 3). Six species originated from the Canary Current and the Mediterranean Sea have been recently found in the South European waters: Chlorophthalmus agassizi, Citharus linguatula, Dactylopterus volitans, D. vulgaris, S. senegalensis, and Sphoeroides pachygaster. These findings highlighted the importance of the South European Atlantic Shelf as a transition zone and as a target zone for southern species when water temperature rises.

All of the species studied were mostly found in the area less than 200 m deep (Figures 10B, 11). The values of depth presented here are the depths of the area where the species was observed. A total of 1801 occurrences with unique depth values were recorded. Ninety percent of these occurrences were found in area less than 166 m deep.

The number of species varies and decreases with increasing latitude, the tropical area being often several times richer in species than temperate ones (Gaston, 2000), and the causes being still under debate (Clarke and Gaston, 2006).

The increasing number of species in an ecosystem was generally followed by the decreasing number of species in the adjacent region. For example, in the Celtic Seas, despite the increase of occurrence records over time, the number of species showed a decreasing trend (Figure 4B); in contrast, the number of species was increasing in the South European Atlantic Shelf, and stable in the North Sea even though the number of occurrences in the last period was decreasing. The studies conducted by Hiddink and Hofstede (2008) in the latter region found that the species richness in the northern part of the North Sea increased over time. This result was also confirmed by Cheung et al. (2009).

Information is sometimes insufficient for political and/or financial reasons in some areas, as is the case for Portugal and Moroccan waters in our study. In contrast, many observed occurrences are available from well-financed, data-sharing countries, even if the real densities may be higher elsewhere (Beck et al., 2014). Our study covers a wide latitudinal area (tropical, subtropical, and temperate waters) to obtain a complete picture of species distribution. Despite the high number of occurrences (above 100,000 records) coming from numerous scientific expeditions in the temperate regions (the North Sea and the Celtic Seas), species diversity is lower than that observed in the tropical regions, where scientific observations are rare and we consequently have fewer than 7000 recorded occurrences: see again Figure 4.

Bias in species occurrences could lead to a distorted view of the pattern of species distributions (Yang et al., 2013). Several factors can cause such biases, such as spatial scale, survey location, sampling techniques, misidentification, and species mobility. The quality issues of some data sources, such as their lack of transparency, have been noted by many authors and publicly criticized (Soberón et al., 2002; Graham et al., 2007; Yesson et al., 2007), especially for records coming from the natural history museums that are rarely collected optimally. Negotiating access to several biodiversity databases that are not yet publicly available is a near-future task, as stressed by Yesson et al. (2007). Thus, data gaps in certain areas could be filled. Nevertheless, many improvements have been achieved and large amounts of good-quality data are already available, are increasingly used and encourage research in many applications.

In the middle and southern ecosystems of the study area, our data in the last period were insufficient for analysis of species richness (Figures 4E,F). Nevertheless, the middle regions of the study area (the South European Atlantic Shelf and the Mediterranean Sea) received many tropical species from the lower latitudes that furthermore change the pattern of species distribution as well as species richness in this area. Zenetos et al. (2010) stated that the presence of tropical-Atlantic species in the central region confirms the phenomenon of “tropicalisation” of both pelagic and demersal species. Although the increasing numbers of species in the South European Atlantic Shelf and the Western Mediterranean Sea was influenced by the increasing number of occurrence records, many studies confirm the increasing frequency and biomass of tropical species in these areas (i.e., Deidun et al., 2011). Bañón et al. (2010) suggested that the presence of tropical species of African origin in the central region of the study area reflects actual natural changes rather than increased sampling effort. The recent observations of tropical-Atlantic species, such as S. senegalensis and D. vulgaris, in the central study area during autumn suggest that these species' ranges have expanded northwards. Borges et al. (2004) and Bañón et al. (2010) opined that these phenomena were probably linked to global warming.

Many studies have identified temperature as the dominant environmental factor limiting species distributions (Santos, 2001; Borges et al., 2003). The vast majority of fish species do not have the capability to regulate body temperature, thus their ability to avoid or select environmental temperature is one of the critical components of their behavioral repertoire. Temperature controls reproduction, growth and the availability of food (Summerson et al., 2007). Although temperature is not the only forcing parameter, Cheung et al. (2009) stated that the response of marine species to environmental change, such as ocean warming, is expressed by a shifting of their latitudinal or depth range and changes in the pattern of species richness.

In the southern part of the study area, newly recorded tropical species in the South European Atlantic Shelf (ID CWS in Figure 10) could confirm the effect of climate change in shifting species distributions. Six species originating from the Canary Current and the Western Mediterranean Sea extend their distributions to this area (Figures 8, 10). Bañón et al. (2010) stated that the arrival of numerous tropical fish in this region seems to have altered the traditional species distribution patterns in the area, and they have argued that this is probably linked to global warming. Although temperature is not the only parameter influencing species distribution, the difference of mean SST between the middle and the southern ecosystems during summer was reduced (3.6°, see Table 2). This condition could allow the southern tropical species to migrate to the central region (other limiting factors not considered). However, the absence of survey data or low records of species occurrences in the earlier period in this region limit our view and prevent us from definitively attributing this phenomenon to ocean warming. Nevertheless, many studies suggested that increasing temperatures were the major cause of the presence of tropical species, as mentioned previously. Moreover, the steady increase of SST observed in all regions (Supplementary Figure 5) may promote species emigration, especially in the intermediate regions. Thus, their distribution range is expectedly expanding, as stated also by Kaschner et al. (2011).

In the northern part of the study area, the movements of species distribution seem to be strongly correlated with seasonal environmental change. Hiddink and Hofstede (2008) stated that biodiversity at high latitudes in the subtropical and temperate waters seem to be highly sensitive to climate change. A. tobianus is mostly found in the South European Atlantic Shelf during autumn, while in spring and summer, it can be found both in the North Sea and the Celtic Seas. The mean autumnal temperature in the South European Atlantic Shelf was close to summer temperatures in the Celtic Seas and the North Sea (Table 2). The mean SSTs during spring and summer of the North Sea and the Celtic Seas are close to each other (see also Figure 8).

Our study suggests a generally northward route of species migration. We are somewhat confident of the records of tropical-Atlantic species' presence in the middle region of the study area. In this study, seven species from the Canary Current were found in the Western Mediterranean Sea. Zenetos et al. (2010) noted the increasing number of alien fish species in this region, with a parallel increase in the number of the Atlantic migrant fish species. Deidun et al. (2011) noted 33 exotic fish of Atlantic origin present in the Mediterranean Sea. Many of these species have already become well established and expanded their distribution into the entire basin, suggesting also an eastward movement (CIESM, 2002).

Our study showed also that species' preferential depth and seasonal variation are important factors on species movement. Some species could be observed more in the North Sea in certain seasons, while in the other seasons; they would be more observed in the Celtic Seas or in the South European Atlantic Shelf (Figure 6). The northernmost area of the North Sea has deeper waters than the southern area. As a consequence, species with a preferential depth of 100–200 m, such as Lepidorhombus whiffiagonis, are more comfortable in the northern part of the region, while species with preferential depth less than 50 m, such as A. tobianus, are much more often present in the southern part of the region. Hence, if the winter or spring temperature increases, L. whiffiagonis would migrate northwards, while A. tobianus can migrate eastwards rather than northwards considering its preferential depth range (Figure 6). Bakun (2006) mentioned that the success of fish recruitment is strongly influenced by three categories (known as the fundamental triad) of oceanographic processes: enrichment (such as upwelling and river run-off), concentration (occurs in areas such as fronts and river plumes) and retention (such as fronts, currents, and circulation patterns).

Fronts are narrow boundary zones that separate water masses (Mason et al., 2005). Fronts have stronger gradients of many water properties and are more productive, with increased nutrient and plankton concentrations. These environments may benefit adult fish as well as larvae and eggs. Fronts are ubiquitous features in the South European Atlantic Shelf or in eastern ocean boundaries that act as barriers and reduce the export of particles from coastal waters to the continental shelf (Bakun, 2006). Thus, they act also as potential barriers to planktonic dispersal and adult migration (Figure 12): see Mason et al. (2005) for details.

Nutrient availability, salinity, dispersal rates, the seafloor profile, and oceanographic force can also influence the patterns of species distribution and species composition. All of these combined factors can lead to fish migrations, especially of highly tolerant and naturally adaptable invasive species (Hiddink and Hofstede, 2008).

The upwelling zones (off the Portuguese coast in the South European Shelf and off the Moroccan and Mauritanian coast in the Canary Current) with nutrient-rich water increase primary production and in turn food for higher trophic levels, hence supporting high fish production (Aggrey-Fynn, 2007). During summer, from June to October, Borges et al. (2004) mentioned that the upwelling system of the Canary Current ecosystem reach Portugal waters and brings colder and nutrient-rich waters to the surface. This could be one of the reasons of the presence of tropical Atlantic waters in the area. Furthermore, Borges et al. (2004) also mentioned that the presence of denser, more saline Mediterranean waters in this transitional zone bring a number of Mediterranean species to the region. Bañón et al. (2010) explains that upwelling off Cape Finisterre from March to October (Figure 12C) is probably the reason for the presence of cool-water species in this area, while they are absent or scarce in the Bay of Biscay. The water in Cape Finisterre is noticeably colder than in the northern Galician coast and the Bay of Biscay. In contrast, warm-water species are present in the Bay of Biscay and absent or scarce off Cape Finisterre.

The sharp north-to-south temperature gradient in the South European Atlantic Shelf, the bathymetry with a narrow continental shelf, and the condition oceanographic in the area mentioned above could restrict the distribution of many species (Borges et al., 2004; Bañón et al., 2010; Figure 12). Other limiting factors such as over-exploitation and habitat loss are also important influencing species diversity pattern (Hiddink and Hofstede, 2008). Disturbance from fishing activities may influence species distributions by reorganizing the fish assemblage (Summerson et al., 2007). Garrison and Jason (2000) stated that the spatial differences in exploitation rates can modify the pattern of species distribution. Fishing activities remove and reduce the regional abundance of predator species, which cascades to lower trophic levels (Martins et al., 2012). If the abundance of the less exploited or less predated species increases, expansion of their distribution as an indirect impact is then possible (Martins et al., 2012). An FAO statistics report (FIGIS, 2016) showed a diminishing trend of catches in European Atlantic regions, whereas an increase of captures was observed in the Eastern Central Atlantic (Supplementary Figure 6) which is the potential source of our migrating species. In the central region of the South European Atlantic Shelf, Borges et al. (2004) and Teixeira et al. (2014) mentioned that the catch of colder marine species has decreased in recent decades, while the catch for warm marine species has increased considerably. Although fishing activities could be important in determining species distribution patterns, fishing effects were not considered the major driver in this study.

In this study, we incorporated several data sources including publicly available databases as well as unpublished databases. We took several measures to minimize the risk of bias, including using recent records (after 1982 only), excluding occurrences lacking time of observation, performing a strict selection of species studied by number of occurrence, and removing suspicious occurrences. These databases enabled the investigation of patterns of species distributions at large spatial and temporal scales to better understand how species extend their area of distribution (mostly northward in the Atlantic European waters and also eastward in the Mediterranean basin in response to environmental change. Further study is needed to gain additional insights on species seasonal movements and depth range across this large spatial extent. Species distribution modeling based on species environmental niches is promising, as results could be used to predict species ranges and species richness. The use of such models could be an alternative or a complementary approach to the limitation in availability of data at large scale, and would lead to a better understanding of species distribution and biodiversity changes, and enable the identification of present and future high-biodiversity areas.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.