Allis shad and European flounder were selected as representative of two distinct functional uses of the marine environment. Allis shad belongs to the Clupeidae family and is distributed along the northeastern Atlantic coast (Baglinière et al., 2003). It is an anadromous species with an average life expectancy of 4-6 years. Eggs are spawned in rivers and larvae/young juveniles migrate downstream to the sea where they grow for 3 to 6 years in the pelagic environment . When sexual maturity occurs (between 4 and 6 years), adults return to spawn and most of them die in their birth river (Thorrold et al., 2001).

European flounder belongs to the Pleuronectidae family and is a demersal species particularly common along the southern and eastern coasts of the North Sea (Cabral et al., 2007). European flounder typically lives for around ten years and makes regular migrations between estuarine and marine environments as part of its life cycle, except during its first 2-3 years when it remains in the estuarine environment. Eggs are spawned at sea in winter and early spring, and larvae enter estuaries after a few days of passive migration (Campos et al., 1994). Then, juveniles settle on the seafloor to feed and grow until their sexual maturity (i.e., about 3 years for females and 2 for males - Ramos et al., 2010). European flounder is often described as having an estuarine or semi-catadromous resident life stage (Elliott et al., 2007). As it may not fully align with the conventional catadromous definition according to Potter et al. (2015), it’s important to note that its life cycle also involves the use of the non-estuarine part of the river.

The study area extends from the Portuguese coasts to southern Scandinavia (from 36 to 62°N and from -12 to 10.2°E) to cover a large part of the species’ distribution ranges (Figure 1). The Mediterranean and Baltic seas were excluded from the analysis because of their narrow mouths to connect with the Atlantic Ocean that likely prevent large-scale dispersal and favor local adaptations (Hemmer-Hansen et al., 2007). In order to enhance the accuracy of our models, we established a maximum depth of 150 m and 100 m for Allis shad and European flounder, respectively, in accordance with their known preferences for shallower habitats (Taverny and Elie, 2001b).

The marine habitats of Allis shad identified in this study are consistent with previous observations of the species being present in the Bay of Biscay, but sparsely abundant in the catchments of the North Sea, and possibly absent from northern Europe (De Groot, 2002; Baglinière et al., 2003). Our findings are also in agreement with previous species distribution models for Allis shad which identified the northern Bay of Biscay as highly suitable throughout the seasons (Trancart et al., 2014; Elliott et al., 2023). Nonetheless, our findings for the Iberian coast sub-region are in contrast with the common perception that Allis shad is abundant there (Nachón et al., 2016).

The results obtained for the European flounder are in line with previous studies that indicated the species as abundant in the southern and eastern coastal North Sea (Campos et al., 1994; Skerritt, 2010). Marine habitats can also be found, in a smaller proportion, along the coast of the Bay of Biscay but are absent along the Iberian coast although several studies mentioned the species’ presence in adjacent Iberian rivers (e.g., Minho, Lima, Vouga and Mondego - Cabral et al., 2007; Mota et al., 2016; Nachón et al., 2016). Portugal is currently reported as the southern limit of the species’ distribution (Pédron et al., 2017).

The present study acknowledges that the model’s predictive ability for European flounder and Allis shad along the Iberian coast may be inaccurate due to the scarcity of occurrence data in that area among the input dataset (0.1% and 0.4% of presence data, respectively, Table S1). Although the model’s performance was not evaluated based on the proportion of absence, we recognize the importance of achieving a balanced representation of absence data (0.9% for European flounder and 0.8% for Allis shad, Table S1). It is well known that cold and saline bottom waters rise to the surface (i.e., upwelling) in summer along the Iberian coast (Nykjær and Van Camp, 1994).These specific thermal and salinity conditions may not have be “included” in the species response curve, preventing the model from predicting marine habitats in this sub-region of the study area. Given the scarcity of occurrence data for the Portuguese and Cantabrian coasts, it would be valuable to employ a temporally more precise modeling approach, such as a monthly time-scale, to capture the environmental conditions associated with upwelling and more accurately predict species’ distributions along the Iberian coast sub-region. The environmental variables in this region differ considerably from those in other parts of our study area, and thus, a higher number of occurrence records is necessary for their full integration into the model and more reliable prediction of the species distribution in the area. Furthermore, the classification of highly suitable habitat might be stringent, as it relies on the predicted probability across the entire geographical range. Such an approach could be deemed justifiable under the assumption that the sampling strategy and techniques are entirely comparable throughout the geographical range.

The pelagic or benthic behavior of the two species makes them vulnerable as by-catch to fisheries of the main commercial species such as mackerel, herring or sole operating in the area (Trancart et al., 2014). Thus, occurrences are mainly from by-catches among commercial fishing data. Fishery data are potentially biased because fishermen select areas where the targeted species is likely to be present with the aim of maximizing yields and minimizing costs by finding hotspots (Dell et al., 2011). Despite this limitation, we used these data to develop our models, as they provided the richest source of information on the distribution and abundance of bycatch species. Additionally, we compared our results to those of previous studies that incorporated observation modeling (i.e., taking into account for the imperfect detection and its variability among gears or other factors) into their models (Elliott et al., 2023) and found them to be largely consistent, which indicates the robustness of our modeling approach.

Professional fishing in the Northern Atlantic is typically focused on coastal areas, which could result in an underestimation of the suitability of offshore environments from modeling approaches based on fisheries-related data (Moustakas et al., 2006). However, this bias may not significantly affect our results given that the two species studied are known to primarily inhabit depths shallower than 200 meters (Summers, 1979; Taverny and Elie, 2001b). Misidentification of species is a common source of errors in studies using fishery data. For instance, Allis shad and Twaite shad (Alosa fallax), as well as European flounder and European plaice (Pleuronectes platessa), share similarities in morphology, and may be misidentified. Nevertheless, the MigrenMer database was carefully curated to reduce the risk of confusion, such as by renaming Twaite shads larger than 50 cm as Allis shad (Elliott et al., 2023).

As our approach adheres to conventional frequentist statistical analysis, we were compelled to omit certain environmental variables due to their relatively weak associations. Nevertheless, it is imperative to recognize that high correlation does not necessarily imply strong causation. There are potential remedies to take into account the dynamic nature of variable correlations, with one such option being the adoption of Empirical Dynamic Modeling (EDM), accessible through the rEDM package in R (Sugihara et al., 2012; Sugihara et al., 2020). EDM has demonstrated its efficacy in numerous fish stock assessments, presenting an innovative standpoint worthy of contemplation (Sguotti et al., 2020). While it may not serve as a universally applicable solution, its suitability hinges on considerations such as species life history and data availability. In this study, we have made the most of the limited occurrence data at our disposal. Additionally, the algorithms we employed for ensemble modeling necessitate non-correlated variables to converge, further compounded by our access to environmental variables over a relatively abbreviated time frame.

Characterizing diadromous species’ habitats functionality is essential to improve their management and maintain population sustainability. Habitats comprise waters and substrates essential for fish to spawn, feed and/or grow to sexual maturity (Dambrine et al., 2021). The ability to identify diadromous species’ habitat functionality depends upon knowledge of the habitats used by fish. European flounder and Allis shad have different Essential Fish Habitats (EFHs) due to their distinct life-history traits. Whereas Allis shad spends the majority of its life at sea for growth and maturation, European flounder migrates to sea once a year to reproduce (Baglinière et al., 2003; Skerritt, 2010). Hence, the marine habitats identified for flounder and shad could be assumed to be potential nurseries and spawning grounds, respectively. However, it would be simplistic to assume that our approach only highlights these specific potential habitats. For instance, it is still unclear whether Allis shad remains in the same marine habitats for successive years or moves between different growth habitats. Here, we assumed the simplest postulate that shad remains in the same marine habitats for five years. Valdimarsson et al. (2012) have identified the presence of Allis shad in Icelandic marine waters, located several hundred kilometers away from the closest known spawning grounds in the British Isles and the European mainland. These catches far from catchments with functional populations suggest a larger dispersion than expected. In addition, the findings of Nachón et al. (2020) suggest that European eel from the Garonne population exhibited a lack of homing behavior during their oceanic phase, as indicated by telemetry data and subsequent capture at sea. The individuals appeared to exhibit a more random wandering rather than a targeted migration pattern to specific areas. These results contrast with previous studies suggesting homing behavior in diadromous fish populations, highlighting the importance of studying population-specific behaviors to understand the ecology of these species.

Furthermore, the European flounder data used for calibration were probably not solely collected during the species’ spawning migration period. Due to limited winter data it was not possible to select winter occurrence only. Additional information on the behavior of both species at sea would help refining the methodology to more accurately identify and highlight marine habitats. It would be possible to construct a model to identify spawning areas by considering only the winter and spring months of a year during the marine phase.

This definition of marine habitats can be adapted to the ecology of other diadromous species. For example, the Atlantic salmon is an anadromous species that lives from 9 to 11 years with several reproduction events during its life cycle, except during their first 2-5 years when they remain in the rivers (McCormick et al., 1998). We could extend our study by investigating recurrence of the highly suitable environments over a decade for this species. In addition, growing areas of this species are known (i.e., Greenland and Norwegian Sea) so it would be possible to refine the projection area for marine habitats. However, salmon migration is of long duration, so characterization of corridors will remain important.

At the end of this century, Allis shad and European flounder are predicted to gain new marine habitats north of their current distribution range. These results are consistent with the literature that has shown that marine or marine-affiliated species may move northward in response to climate change (Lassalle et al., 2008; Hastings et al., 2020). Diadromous fish displacements have already been observed, for example, in the eastern United States, American shad have been observed to move their spawning distribution northwards in response to warming waters and striped bass is being caught further north than previously in Newfoundland and Labrador (Hare and Able, 2007; Andrews et al., 2019). In addition, Lassalle and Rochard (2009) demonstrated an expansion of the distribution range of three diadromous species, which gained suitable basins predominantly in a northerly direction.

The Bay of Biscay coastal band would contain most of the current Allis shad marine habitats that would remain suitable in the future, while habitats further offshore would be reduced. The expected reduction in deeper habitat suitability should not have a substantial impact on the species. Indeed, although the Allis shad is able to explore up to 150 m depth in summer, this species is predominantly concentrated in shallow coastal areas (<70 m depth; Taverny and Elie, 2001b) and is mostly found at depths less than 50 m depth according to a previous distribution modeling study (Trancart et al., 2014).

The potential reduction in offshore habitats for Allis shad in the Bay of Biscay in the future can be attributed to warming trend of waters (Table 1). Projections indicate that, by the end of the century (under RCP 8.5), the means temperature for the lost marine habitats in the Bay of Biscay could reach 20.1°C, which is higher than the current period’s average of 15.6°C. This rise in temperature is a critical driver of change and is expected to have cascading effects on various ecological processes. Although primary production is projected to decrease in future scenarios, it would not seem to have a pronounced impact on habitat suitability for Allis shad (Figures S1-D). However, the decline in primary production in interaction with the rising temperature could potentially offer an alternative explanation for this habitat loss.

Our results indicate that, by the end of the century, the marine habitats of European flounder in the English Channel and North Sea are expected to become less suitable. This projected decrease in suitability of European flounder marine habitats in the North Sea and Channel regions is likely to result from increasing water temperatures, as indicated by a comparison of mean temperature values calculated for lost habitats in future projections versus current conditions (Table 2). These analyses reveal a substantial temperature increase of up to 3-4°C at the end of the century, with a projected mean temperature of 16.8°C in marine habitats that will be lost in the North Sea under RCP 8.5, compared to 12.4°C in the current period.

This loss of marine habitats is expected to be primarily caused by rising temperatures. Nonetheless, climate change can also have indirect effects on diadromous species by influencing catchment flows that can impact larval entrance into estuaries and various demographic parameters, such as reproductive rates, recruitment, survival, and growth (Waldman and Quinn, 2022). Of particular concern is the potential mismatch created by warming seas if “external” conditions induce fish to spawn at inappropriate times or affect their offspring (Boisneau et al., 2008). According to Mendes et al. (2020), eggs and larvae are tolerant of a wide range of temperatures and can hatch regardless of temperature, but changes in currents associated with temperature modifications may potentially impede their arrival at the appropriate location.

The findings related to the impact of climate change on marine habitats for the species are quite similar, with either more highly suitable or stable environments in the northern range. However, a comparison of the potential implications of these findings for anadromous versus catadromous species may differ depending on the level of connectivity to the continental phase of their respective life cycles. For Allis shad and European flounder, the use of migratory corridors is critical for accessing high-quality spawning areas and delivering juveniles to those areas for successful recruitment of adult populations (Drouineau et al., 2018). Impeded connectivity poses a significant threat to Allis shad, as it can restrict access to the species’ spawning grounds, hindering their reproduction. This is particularly concerning as Allis shad have a homing behavior to their natal rivers (Baglinière et al., 2003). In contrast, European flounder migrate to the sea for reproduction but spend the majority of their time in fresh waters. Baden et al. (2022) suggest that European flounder exhibits a diverse range of spawning areas, including both brackish and marine waters. This adaptability implies that the species may possess greater flexibility in response to the anticipated northward shift of marine habitats resulting from climate change. In contrast, Allis shad, which has a stronger association with the continent for reproduction, may face greater challenges in accessing suitable spawning areas in the new habitats, potentially exacerbating the already observed decline in abundance of this species. Furthermore, the observed decline in the abundance of Allis shad compared to European flounder (Rougier et al., 2012) may be attributed to its more maladaptive responses to environmental changes.

Finally, one potential avenue for further analysis is to explore the differential impact of changing marine habitats on anadromous and catadromous species, with a particular focus on the importance of connectivity to the continental phase of their life cycles. By examining how reductions in connectivity might affect spawning areas and the delivery of juveniles to these areas, it may be possible to gain a more nuanced understanding of how the changing marine environment is likely to affect recruitment success in adult populations of these species. We anticipate that our method could be applied to other diadromous species worldwide, thereby highlighting new challenges for the sustainability of diadromous species populations and promoting the linkage of marine and continental habitats throughout their life cycle.