Celtic sea where is

The Celtic Sea had an abundance of marine mammals in the past. The fisheries of the sea are considered to be overexploited. Currently, the four cetacean species that are frequently found in the Celtic Sea include the common bottlenose dolphin Tursiops truncatus , the minke whale Balaenoptera acutorostrata , the short-beaked common dolphin Delphinus delphis , and the harbor porpoise Phocoena phocoena.

The sea has been named after the Celtic heritage of the region, and its name was proposed by a renowned English marine biologist, E. Holt in , during a meeting of fisheries experts from the countries of France, Great Britain, and Ireland. The need for a uniform name of the sea was felt among all the countries due to the common geology, hydrology, and marine biology of the Celtic Sea.

Temperature effects on higher trophic levels and overall changes in low trophic levels productivity had similar weight in the fit. Integrated independently from plankton data, temperature forcing especially increased the hindcast quality for species with thermal preferenda at outside the temperature ranged covered in the Celtic Sea warmer or colder, Figure 3B. Conversely, PP and ZHS integration essentially led to a better fit for functional groups belonging to the pelagic pathway.

Figure 3. B Changes in the SS for pelagic species and species with temperature optimum outside the Celtic Sea temperature range according to the different environmental drivers integrated in the fit. C Trends in the forcing variables — F is the mean fishing mortality time-series for stocks under assessment. With regard to the basal productivity in the Celtic Sea ecosystem, no clear trend was apparent in the Ecosim model over the period — Despite this relative stability, some differences can be observed in CPR time-series within the phytoplankton community.

Most significant changes in observed phytoplankton data occurred during the late s when the abundance of large phytoplankton diatoms and dinoflagellates dropped whereas the one of small phytoplankton increased Figure 4. However, the trends for both large phytoplankton and small phytoplankton were not correctly captured by Ecosim and remained overall constant over the entire period — in the model.

Figure 4. Observed points; CPR vs. The points correspond to observed time-series used during Ecosim calibration. Similarly to PP, the ZSH remained stable over the study period, yet, looking at the data, substantial changes were observed within the zooplankton community, mostly during the late s.

When PP and ZSH were integrated, Ecosim reproduced the overall observed trend in small mesozooplankton abundance, which substantially declined before and was still decreasing until the mids.

By contrast, large mesozooplankton abundance tended to increase over the period — Again for this zooplankton group, the trend predicted by Ecosim correctly matched observations both in terms of trend and of amplitude of variations.

Both trends and interannual variability in zooplankton abundance were not captured when PP and ZSH were not included in the fit Figure 4. The stability in the predicted phytoplankton time-series precludes us from linking changes in mesozooplankton abundance with changes in primary productivity in the Celtic Sea between and Although Ecosim was able to correctly predict the overall zooplankton abundance level over the study period, the predictions did not perfectly match the observed patterns, especially at the beginning of the time-series.

The observed time-series of biomass, abundance and catch of most of high trophic level groups were satisfactorily reproduced by Ecosim. Overall, Ecosim predictions were better for those single-species groups for which stock assessments are carried out than for multispecific fish groups, and they were less consistent for exploited invertebrate groups. In , main target species such as large Gadiformes [e. Driven by an increase in fishing mortality, the biomass of most of these groups declined between and the mids while their catches simultaneously stabilized or decreased.

In , their fishing mortality decreased and Ecosim predicted a stabilization of their abundance that preceded a progressive recovery from onward Figure 5. Figure 5. Fitted biomass trends from the Ecosim model for a selection of demersal and pelagic fish species. The color points indicate the different time-series of observed data during Ecosim calibration.

The pressure exerted on species that were initially less exploited such as smaller flatfish i. Concomitantly, the natural mortality exerted on these species decreased as a result of the depletion of highly commercial groups such as large predatory fish e.

The predicted biomass trends of lower TL fish and invertebrates varied according to their exploitation patterns Figure 5. Dramatic changes in the composition of the pelagic community were also predicted by Ecosim. In parallel to horse mackerel depletion, the biomass of pelagic fishes that feed upon similar ressources such as suprabenthivorous fish, sprat Sprattus sprattus or boarfish Capros aper increased, while these pelagic fishes were initially almost absent from the Celtic Sea ecosystem.

The recovery of commercial species in the s is noticeable, particularly the large increases in the biomasses of large predators, including hake whose biomass increased 5-fold and a few other species.

According to the model, these species recoveries translated into an increase in predation pressure from large predators and a decline in the abundance of cephalopods, small pelagic and small demersal fish. The rising temperatures resulting from climate forcing had various effects on the different species represented in the model, and they mainly impacted fish productivity during the mid, when a substantial increase in both surface and bottom temperatures was observed.

The foraging ability of Boreal species, including cod, plaice, whiting, herring and endobenthivorous demersal fish, dropped at the same time that fishing mortality increased, and this accentuated the depletion of the above-mentioned species.

Conversely, for Lusitanian species that have warmer thermal preferences, such as sardine, boarfish, hake and cephalopods, increased productivity in the Celtic Sea ecosystem in the s acted antagonistically to the increasing fishing pressure and synergistically with decreasing competition pressure.

The differential trends in the productivity of warm and cold water affinity species between the start and the end of the study period were responsible for differences in the recovery capacities of commercial groups in the s.

For instance, cod biomass remains quite stable when temperature is not driving productivity, and plaice recovers much slower from fishing pressure release. Conversely, sole depletion in the first part of the time-series is stronger as its recovery is strengthened Figure 5. These variations triggered changes in the fish assemblages. Interannual variability in high trophic level abundance was better reproduced when incorporating PP and ZSH data, especially for pelagic species Figure 5.

Evaluation tests suggested that GAMs Supplementary Appendix SE were good enough to produce yearly foraging habitat maps for the 38 higher trophic level functional groups year , Supplementary Appendix SE. The tests also indicated that GAM fits and predictive capacities were better for monospecific groups than for heterogeneous multispecies groups.

The fit was also lower when the sampling gear employed during the surveys was not adapted to a given functional group e. The type of environmental predictors retained in final GAMs varied across functional groups, as their number, ranging from 3 for the non-commercial invertebrate groups to 9 for hake.

The preference functions estimated from fitted GAMs exhibited diverse shapes. The retained predictors often included bathymetry 32 groups and temperature surface or bottom; 35 groups , two variables of importance in shaping the Celtic Sea environment. Depth in the Celtic Sea drove marked northeast-southwest gradients, while surface temperature led to pronounced latitudinal gradients and bottom temperature resulted in two opposed pools of warm and cold waters in the English and Bristol Channel and in southern Ireland.

Salinity functional responses quantify habitat suitability in the riverine-influenced Bristol Channel. When they were retained in final GAMs, seabed substrate variables resulted in patchy predicted habitats for demersal species.

In those instances, it was possible to make a clear distinction between restricted areas of rocky or mixed substrate and large irregular strands of muddy, sandy or coarse sediments. The other environmental variables considered in the present study were less frequently retained in final GAMs. Habitat foraging capacity maps for the period — are provided in Supplementary Appendix SE. The Ecospace model predicted the biomass distribution of all components of the ecosystem, from the phytoplankton to top predators.

The biomass distributions of functional groups reflected the overall habitat foraging capacities of these groups, hence their ecological niche, but were also influenced by their trophic interactions with prey, predators and competitors Figure 6. Figure 6. Predicted biomass distribution in for functional groups for which statistical habitat models generalized additive models were developed.

Biomass is represented in relative units, i. The color scale represents the relative abundance of the considered group, from zero blue to the maximum red. The distributions of those functional groups for which no GAM was developed were estimated by Ecospace only according to their feeding relationships birds, mammals, bacteria, benthic meiofauna, large pelagic fish.

A wide variety of spatial patterns were exhibited by the functional groups represented in Ecospace. Due to their high mobility in the water column and their eurythermic characteristics, horse mackerel and mackerel were predicted to be the most widely distributed monospecific groups in the Celtic Sea ecosystem Figure 6. Other fish groups that were predicted to occupy a large fraction of the Celtic Sea were multispecific groups such as benthivorous elasmobranchs, epibenthivorous demersal fish, pouts Trisopterus minutus and T.

By contrast, some species were predicted to occupy only a relatively small fraction of the Celtic Sea ecosystem. This was particularly the case for benthic groups, which are highly dependent on the seabed substrate. For instance, the distribution of Norway lobster Nephrops norvegicus was restricted to muddy patches in the western part of the central Celtic Sea where the species can more easily bury in mud. Ecospace modeling allowed us to distinguish between groups that have similar trophic niches but different environmental niches.

Thus, sole and plaice have similar diets but showed limited spatial overlap because of their, respectively, warm and cold water affinities; the same was observed for sardine Sardina pilchardus and herring. Species that are taxonomically close such as the gadoids cod, haddock and whiting were shown to regularly coexist in the same areas but could also have substantially different distributions: whiting inhabited shallower areas than cod and haddock; both haddock and cod had affinities for muddy areas, but haddock also frequented sandy substrates whereas cod preferred coarse sediments.

The distributions of adults and juveniles for hake, anglerfish and cod were comparable yet exhibited some differences i. The distribution of functional groups provided a view on fish community assemblages in the Celtic Sea ecosystem. Very coastal areas and bays around Ireland, United Kingdom and France and the inner Bristol Channel were found to be inhabited by functional groups tolerant to low salinities, such as large concentrations of benthic invertebrates, small benthivorous organisms and, upper in the surface, suprabenthivorous and small epipelagic fish i.

Higher trophic levels in these areas were mainly represented by whiting and sea bass. Also, intense primary and secondary pelagic production took place in these areas. Areas next to the shelf edge, located from the tip of Brittany to southwestern Ireland, were characterized by deeper waters with a steeper slope. Here, abundant populations of megrim, hake, anglerfish and other piscivorous demersal fish were found together with widely distributed pelagic fish species such as boarfish, blue whiting Micromesistius poutassou , mackerel and horse mackerel.

Benthic invertebrates in these areas were mainly deposit and suspension feeders. Primary production was observed limited in these areas, but the pelagic habitat was favorable to zooplankton development. In the cold central and northwestern Celtic Sea, the mix of sediment types and the moderate depth was favorable to diverse demersal species, including epibenthivorous demersal fish, hake, anglerfish and three gadoid species, cod, haddock and whiting.

Mackerel and horse mackerel were the dominant pelagic species in this area. The density of benthic carnivores and necrophageous invertebrates was especially high in this area. Finally, the Western English Channel and the outer Bristol Channel differed from the other regions of the Celtic Sea by their relatively warm waters and coarse substrate and their higher plankton productivity sustained by mesoscale structures.

The Western English Channel and the outer Bristol Channel were characterized by the presence of small benthivorous fish, epibenthivorous fish, pouts, flatfish plaice and sole , commercial invertebrates crustaceans and bivalves , cephalopods species, some gadoid species, and a mix of pelagic species including mainly horse mackerel and mackerel but also small epipelagic species.

Overall, relative habitat capacity was constant for most functional groups over the study period. Most changes in relative habitat capacity occurred between the late s and the mids Figure 7. The intensity of changes was variable across functional groups and can be related to the water temperature preferences of functional groups.

The fitted GAMs indeed highlighted that temperature, which is the variable most affected by climatic changes in the area, was an ecological niche descriptor common to almost all components of the ecosystem. Thus, changes in the modeled habitats over the studied period were observed for those functional groups with most extreme thermal preferences in the Celtic Sea such as pilchard Sardina pilchardus , sea bass, medium pelagic fish and squids mainly or exclusively composed of Lusitanian species , as well as cod, haddock, sprat, herring, endobenthivorous fish and whiting mainly or exclusively composed of Boreal species.

Figure 7. Functional groups dominated by Boreal species illustrate biomass distribution changes related to environment Figure 8. Most changes in the distribution of species that prefer cold waters occurred early in the study period, with a slight contraction of the area occupied by these species in the eastern and southwestern Celtic Sea in response to the more intense warming of already warm bottom waters in those areas.

For the remainder of the study period, the relative distribution of biomass of Boreal species remained relatively stable. Figure 8. Predicted changes in the total biomass of Boreal demersal species. A Mean biomass distribution t. Relative changes of Boreal demersal species biomass are displayed in B —; C —; D : — and calculated as the ratio between these selected periods and the initial biomass.

In B—D the colors correspond to the ratio of absolute biomass between the predicted biomass and biomass for — Primary production in the Celtic Sea Figure 9A was essentially located in coastal areas, from the southern coast of Ireland to the Western English Channel and western Brittany, and along the southern part of the Celtic Sea shelf break.

Differences in primary production between the lowest and the highest productive areas never exceed a factor of four. The biomass of secondary producers i. The total biomass of high trophic level groups i. The biomass of primary and secondary producers in the Celtic Sea was concentrated around particular areas, mainly around the coasts and in the English Channel, and this available production is efficiently transferred to higher trophic levels and then likely spread horizontally.

Figure 9. Primary production in the Celtic Sea and biomass for trophic levels 2—3. A Primary production; B total biomass for trophic levels 2—3.

The spatialized mean trophic level MTL and high trophic index HTI highlighted that higher trophic levels were mainly located in the western Celtic Sea. In particular, high values of the MTL were concentrated in the northwestern Celtic Sea, whereas the HTI is higher in deeper areas along the shelf-break Figure 10 , left panel. Areas associated with a high MTL were those areas that were occupied by many medium to large demersal functional group including epibenthivorous and piscivorous fish, elasmobranchs, gadoids, anglerfish and hake.

Areas characterized by a high HTI were those where the biomasses of hake, megrim and anglerfish were high. In the eastern part of the Celtic Sea, the MTL was lower, due the large abundance of small demersal fish species feeding on benthic invertebrates e. In the eastern part of the Celtic Sea, top-predators were mainly represented by large pelagic sharks and sea bass.

Figure Mean values, change and variability in four ecosystem indicators in the Celtic Sea over the period — Spatial patterns in functional group diversity were independent from that of biomass levels. Functional group diversity was found to be particularly high at the border between the main Celtic Sea regions identified previously.

It was highest in the northwestern part of the central Celtic Sea, which was inhabited by generalist species and species that can feed on both benthic invertebrates and fish, including gadoids and anglerfish, but not hake which primarily feeds on medium-sized pelagic fish. All the indicators displayed measurable variations over time and space Figure 10 , middle panels.

The trends in the HTI and MTL were in opposed directions, which means that the ratio of low trophic level biomass to total biomass increased overall but that, in parallel, the proportion of species with a TL greater than 4 increased.

These results were mainly driven by the dramatic changes in horse mackerel and hake abundances that occurred over the study period, while, after a deep depletion, the abundance of most mesopredator in the s returned to levels similar to the s.

For the HTI and MTL, the direction of the change was the same among Ecospace cells, partly because the species driving observed changes were widely distributed in the Celtic Sea region. Trophic diversity declined overall over the whole Celtic Sea, especially in coastal areas. While functional groups diversity also declined in the southern Celtic Sea and the English Channel, it increased in the northwestern part of the central Celtic Sea.

The standard deviations of the four indices Figure 10 , right panels highlighted a measurable modification of the whole structure of the Celtic Sea ecosystem over the period —, yet some specific areas were more affected than others. The standard deviation of spatial indices relative to their mean value also suggests a greater amplitude of changes occurred in the Kempton index. The data types integrated in the Celtic Sea EwE model significantly improved its ability to represent the functioning of the Celtic Sea ecosystem by providing a more relevant picture of the food-web structure and by depicting more realistically the drivers that rule its dynamics, and notably the environmental drivers.

To ensure that a food-web model correctly represents the temporal and spatial dynamics of an ecosystem, trophic relationships have to be defined as accurately as possible, through the combined use of classical and novel analyses Pethybridge et al. In this study, the use of outputs of the EcoDiet model Hernvann et al.

Additionally, the substantial collection of data compiled for the present study [ time-series versus 62 in Moullec et al. Though the integration of abundance time-series for multispecific groups in EwE was informative, the dynamic of these groups remained poorly understood compared to the dynamic groups under stock assessment.

Gathering relevant time-series of fishing effort for the period — for multispecific groups would greatly improve the hindcast. The integration of primary and secondary production related data in EwE only slightly improved the fit of the Ecosim model.

The reason for such a low improvement was the absence of clear variation of primary and secondary production in the time-series. This suggested a relative stability of the hydro-climatic conditions in the Celtic Sea between and as the main driver of plankton development. PP and ZHS integration allowed a better hindcast for pelagic functional groups, especially for zooplankton and some planktivorous fish, both for trends and interannual fluctuations, while trends in phytoplankton abundance were poorly reproduced.

This poor fit for phytoplankton groups illustrates the difficulty to represent plankton prey-predator interactions and zooplankton grazing pressure in EwE models and the consequence this may have on the prediction of higher trophic levels.

Here, high productivity and biomass levels of phytoplankton prevent their control by top-down process and limit their variation in abundance to a very low amplitude. Despite this lack of realism on the dynamic of phytoplankton, the information brought on the bottom-up impact of PP and the development conditions for zooplankton well complemented the insight on top-down impact exerced by higher trophic levels on these secondary producers.

Remote sensing data have been increasingly utilized over the last two decades in marine ecology studies Chassot et al. The use of satellite-derived PP estimates is relatively common in marine ecosystem modeling Christensen et al.

The main advantages of satellite-derived PP estimates are that they derived from a direct observation and they are accessible at high spatial resolution, given that regional biogeochemical models are not always locally available. However, satellite-derived PP estimates are only available for recent past the past two decades and, so far, they have not been calibrated to provide values for different size classes of phytoplankton Kramer et al.

In particular, CPR data suggested that the phytoplankton community went through measureable changes in the Celtic Sea over the period — The PCI index, proxy of the total standing biomass, increased, while no significant trend was detected in the diatom and dinoflagellate both included in large phytoplankton group counts, which suggested an increase in the biomass of nano-phytoplankton between and Despite the relatively short temporal availability of the mesozooplankton niche model of Druon et al.

In particular, estimates from the mesozooplankton niche model allowed us to highlight areas of the Celtic Sea that are favorable to primary production but not to secondary production, and to identify regions with lower food availability for high trophic levels or where the grazing pressure exerted on zooplankton is important. Since the early s, several end-to-end ecosystem models have coupled low and high trophic level models i. However, only a few ecosystem modelers have managed to implement an on-line two-way coupling, and it has been more common to develop alternative ways to link low trophic level and high trophic level dynamics Libralato and Solidoro, ; Piroddi et al.

In this context, forcing future ecosystem models by observation-based primary productivity and zooplankton habitat would represent a valuable and conservative alternative approach. The habitat modeling framework adopted in this study that relied on GAMs was used to assess and predict the response of almost all high trophic level groups to warming and to project the abundance and distribution of each group in the Celtic Sea. The habitat-derived functional response to temperature significantly improved the fit of the Ecosim model.

Here again, the improvement of model fit was relatively low, as surface and bottom temperatures appeared to be stable over a substantial fraction of the study period.

Still, this improvement was substantial for those functional groups that can cope with extreme temperatures, i. In this study, we successfully integrated the outputs of a statistical habitat model into a spatial trophic model.

The limited complexity of the GAM smoothers employed here, together with the choice of modeling probabilities of presence and not abundances with the GAMs, are in line with the ecological niche theory Citores et al. The European IBTS survey data employed in this study encompassed a large range of environmental values, hence providing insights into the potential response of species groups to the warmer conditions that could become the norm in the Celtic Sea by the end of the century.

This also stresses the need for improving large scale sampling for these compartments. While calibrating the Ecosim model, several periods of time with a poorer fit were identified for different functional groups.

Early changes in phytoplankton and zooplankton abundances were hardly reproduced by the Ecosim model. Ecosim failed to catch the great decline in small mesozooplankton abundance at the beginning of the time-series, as well as the decline in diatom and dinoflagellate abundances and the increase in nanophytoplankton abundance.

Although it could be due to wrong trends in PP estimated by the biogeochemical model in the absence of satellite-derived data, this could also be due to uncaptured environmental effects.

This would be expected as our study period starts in , only a few years before the occurrence of abrupt changes in the state of the Celtic Sea ecosystem e.

As the stability in the plankton-related time-series did not allow us to explain their dynamics, the oceanographic conditions limiting phytoplankton and zooplankton production may not be the main factors of the regime shift. Several hypotheses have been proposed for identifying the main drivers of the severe drop in mesozooplankton abundance that occurred in the Celtic Sea, including climate drivers or changes in ocean circulation, but the mechanisms responsible for these changes mediating factor of phytoplankton-zooplankton interactions, direct mortality factor of zooplankton, changes in zooplankton species composition, changes in phytoplankton composition, etc.

Nevertheless, these developments were not included in this study as: i they did not allow us to better reproduce patterns in plankton abundance in the early years of the period —; ii they may be confounded with environmental factors that may improve, for instance, fish recruitment; and iii they did not provide any indications on the mechanisms at stake. Note that a novel spatio-temporal framework, that integrates both temporal and spatial dimensions, has been developed by Steenbeek et al.

The present approach relies on the sequential development of temporal rather than spatial food-web model and does not treat both aspects simultaneously in the same framework.

Thus, the changes in functional group abundance predicted temporally do not account for eventual mismatch between prey and predators spatial distribution. The consequences on the overall model predictions may remain limited here due to very limited changes in the distribution of functional groups. However, such novel version of Ecopath spatio-temporal framework should be considered to predict long-term changes in the Celtic Sea in a context of climate change, which would imply substantial shifts in species distributions at various rates according to the functional groups.

The Ecospace model highlighted the heterogeneous spatial patterns of biological production in the Celtic Sea that are mainly driven by contrasting water mixing conditions over the study region. The seasonally mixed waters spreading over the shelf of the Celtic Sea lead to lower primary production, which occurs later there and at a lower level than in coastal areas Pingree, ; Hickman et al. Note however, that PP can be locally enhanced offshore by the combined effects of topography and tidally-induced vertical mixing through internal waves at the shelf break; Sharples et al.

The resulting shelf-break front is well developed in South-West Celtic Sea and finer along the western shelf Holligan and Groom, ; Druon et al. In shallower waters, the Ushant front Pingree et al. According to literature, the most productive areas, and especially productivity fronts, were predicted to concentrate biomass over several trophic levels up to top-predators Pade et al. Specific functional groups inhabiting the water column displayed aggregated biomass at their location such as mesozooplankton, mackerel and pelagic sharks, as observed at the Ushant front.

The low level of information on benthos in the present study however, prevents any conclusions about these aspects. The spatial structuration induced by primary production was shown to propagate through the food web. However, except for the above-mentioned functional groups, the spatial contrast in productivity in the Celtic Sea generally dampens with increasing trophic level.

The mismatch between higher trophic level species niche and highest primary productivity areas is more important for predators than for prey. In other words, trophic flows progressively operate a biomass transfer from restricted productive areas to lower productive areas in the Celtic Sea.

Biomass transfers across the shelf are mainly operated through functional groups characterized by wide ecological niches and large dispersal rates one input parameter of Ecospace; Romagnoni et al. For such functional groups, Ecospace enables the representation of locally enhanced biomass production due to a higher availability of resources when accounting for dispersal toward more suitable environments Christensen et al.

Thus, through this modeling framework, we highlighted that high fish densities could be maintained in low productive areas of the Celtic Sea through subsidization by prey and predators movements Polis et al. The species assemblages of the inner Bristol Channel and coastal areas corresponded to those described by Ellis et al.

The central Celtic Sea assemblages identified in this study match those reported in Ellis et al. As the present study, Ellis et al. The large abundances of pelagic species such as mackerel, horse mackerel, boarfish and blue whiting on the Celtic Sea shelf-edge that were also predicted in this study are consistent with the findings of Reid and Trenkel et al. In agreement with Rees , we found that the Western English Channel displays a particularly diverse benthic assemblage that includes large crustaceans and commercial bivalves [ Kaiser et al.

The preponderant community of pelagic species in the Western English Channel has been carefully screened for decades Southward et al. The spatial patterns of abundance of most demersal species predicted in this study were consistent with those derived from English Celtic Sea groundfish survey data Warnes and Jones, ; Tidd and Warnes, , even though the latter were not employed for statistical habitat modeling in this study.

The different species assemblages of the Celtic Sea identified in this study provided some new insights to the possible spatial structure of the food web in the Celtic Sea region. While some functional groups are widely distributed, some others are only present in some specific areas of the Celtic Sea and their absence or negligible biomass at particular locations may be interpreted as reflecting local reorganizations of the food web.

Thus, the complex Celtic Sea food-web structure could be interpreted as a meta food web made of local, environmentally-driven, food webs connected through space by functional groups with larger dispersion capacities Kortsch et al. Ecospace reveals to be a useful tool to understand the spatial structure of food webs as a complement to more qualitative approaches for characterizing the trophic structure of marine ecosystems Albouy et al.

Spatial indicators can be produced with Ecospace to determine the spatial structure of the food web in relation to the environment. Our results suggest that shallowest and least saline affected by rivers waters of the Celtic Sea, together with the deepest waters of the region, are characterized by a low number of compartments and groups with relatively low variability in terms of TL.

The two above-mentioned areas are, however, radically different in terms of functioning. The latter is dominated by a specific trophic pathway where top-predators i.

The former, in deeper environments, is essentially represented by low trophic level benthic and pelagic functional groups with non-diversified diets.

The distinction made between the two areas does not exclude the existence of important benthic-pelagic coupling in both areas, but this coupling may be due to the reliance on pelagic primary production rather than to exchanges between upper trophic level species. Intermediate-depth areas of the Celtic Sea were found to be dominated by pelagic functional groups, yet they displayed large concentrations of benthopelagic fish feeding on the benthos and pelagos. Finally, the central Celtic Sea was found to be dominated by predatory and generalist functional groups that feed on both pelagic and benthic compartments, which may render this specific area more stable than the rest of the Celtic Sea.

Spatialized indicators highlighted that boundaries between these different areas constituted ecotones, characterized by higher functional group and trophic diversity as they suited to a wider range of species. The spatio-temporal dynamic model of this study represents the trophic functioning of the Celtic Sea ecosystem averaged over a year.

We proceeded this way because of data availability for model calibration i. Averaging trophic interactions can be questionable, as trophic functioning over the year is strongly related to seasonal variations in plankton production. These variations are mainly associated with hydrographic features and water mixing processes. In addition, seasonal movements, and subsequent changes in spatial distributions and local abundances have been highlighted for both pelagic [widely distributed pelagic fish Trenkel et al.

These movements can notably be related to feeding, spawning and overwintering. Seasonal changes in distributions would explain the variability of fish diets in the Celtic Sea reported in Trenkel et al.

For instance, large Celtic Sea predators consume more blue whiting during the summer and more mackerel during the winter, whereas the consumption of horse mackerel is relatively constant throughout the year. However, modeling changes in fish spatial distributions at a finer temporal scale remains a challenging issue given the data that are available to us. The Celtic Sea Ecosim model best explained the food-web dynamics when it integrated trophic interactions, fishing, environmental conditions controlling plankton production, and warming as a driver of the productivity of higher trophic level groups.

The way the Ecosim model fitted to data when different data types were integrated provided a better understanding of the relative impacts of fishing and environmental changes on the Celtic Sea ecosystem during the three last decades.

Previous work based on catch and effort reconstruction highlighted the large impacts that fishing has had on the Celtic Sea ecosystem since Pinnegar et al. In particular, the dramatic increase in fishing pressure that has occurred in the Celtic Sea region over several decades may have drastically reduced the biomass of large predators and may have triggered the deep alteration of the food-web structure through trophic cascades Hernvann and Gascuel, The impacts of fishing in the Celtic Sea ecosystem suggested by the present study concur with the conclusions of previous EwE modeling works in the ecosystem by Bentorcha et al.

At the beginning of our study period, in , the biomass of main large bentho-demersal fish and predator species had already been severely impacted by fishing and pelagic and invertebrate fisheries started developing in the Celtic Sea region. Then, between and , fishing kept on severely impacting the Celtic Sea ecosystem. Despite the integration of environmental variability in the Ecosim model, our findings suggest that fishing had more profound impacts on the Celtic Sea ecosystem than environmental changes.

The divergent abundance time-series for smaller and larger fish groups, which increased and decreased, respectively, over the period —, were better identified in this study than in the studies that employed previous versions of the Celtic Sea EwE model.

The analysis of the mortality estimates suggested that the driving factor of trends in mortality was predation release on smaller species, due to both the continuously increasing pressure exerted on large predatory fish and the reallocation of fishing effort on medium fish in response to the depletion of to large predators.

These results concur with the hypotheses formulated in community size-structure studies Blanchard et al. Though this phenomenon was only described for demersal species, the large decline in the biomass of horse mackerel in the beginning of the s predicted in this study a 6-fold decline within only a year time frame resulting from a large increase in fishing effort on horse mackerel may have also contributed to the increasing biomass of smaller species.

This may probably be the case horse mackerel leaving a vacant niche for other small pelagic species, such as boarfish to boom in the early s, as observed by Blanchard and Vandermeirsch and Coad and also correlated to changes in the temperature. Thus, fishing appears as the main driver of changes in species assemblages in the Celtic Sea, especially leading to a decrease in the diversity of diets and a general decline in the mean trophic level in the area.

Finally, the biomass recoveries initiated in the Celtic Sea over the recent period were particularly noticeable, and these biomass recoveries seemed to allow for a progressively increasing dominance of demersal species in the Celtic Sea ecosystem and a rise of the predation pressure exerted on smaller demersal fish, crustaceans, and cephalopods as highlighted by the biomass decline of the latter and the increase in the HTI index.

The importance of the fishing driver in the Celtic Sea contributed to mask the impacts of environmental changes on the ecosystem. Indeed, many abundance trends of exploited groups reflected stock depletions and recoveries independently of warming effects. The dominance of the fishing driver in observed time series is consistent with the lack of evidence of any warming effect in the Celtic Sea from studies that analyzed species biomass trends or community size-structure Genner et al.

However, through the use of GAM predictions and temperature forcing, we were able to account for the cumulative impacts of fishing and environment changes in EwE and to disentangle the relative contributions of these different drivers of change.

The mids were characterized by a net warming of the Atlantic Ocean that was attributed to a shift of the Atlantic Multidecadal Oscillation AMO from a negative to a positive phase, which was likely amplified by anthropogenic climate change Ting et al. As a response to this large-scale warming events, both sea surface and bottom temperatures in the Celtic Sea increased.

The Celtic Sea EwE with Ecospace model suggests that these temperature changes led to a decrease in the productivity of most of Boreal species, including herring, sprat, cod, whiting and haddock, and to an increase in the productivity of functional groups with warm water affinity, particularly species with extreme thermal preferenda such as sardine, sea bass, mixed medium pelagic fish [mainly anchovy Engraulis encrasicolus ] and cephalopods, and to a lesser extent, widely distributed pelagic species such as mackerel, horse mackerel and boarfish.

In future years, these trends in species productivity may impair the recovery of some current commercially important species i. If the scale of these changes was limited, it confirmed the response of these species to climate change, which was barely detectable in temporal trends. The niche contraction of demersal Boreal species in the south of the Celtic Sea was associated to an increase in functional groups diversity, related to the largest proportion of biomass represented by species with warmer water affinities.

The latter, the trophic diversity declined in response to change in assemblage, being mainly benthivorous and of low trophic level. The large range of habitats in the region support a diverse fish fauna, including many commercially important species. Many of these species have relatively short migration routes between feeding and spawning areas. The region has a large number of areas attractive to seabirds and waterfowl. The common or harbour seal and the grey seal are widely distributed throughout the region.

The waters around Ireland and to the west of Scotland support a variety of cetaceans, but apart from the population of bottle-nose dolphins in Cardigan Bay, they are only occasionally seen in the Irish Sea. Region III: Celtic Seas The Celtic Seas region contains wide variations in coastal topography, from fjordic sea lochs, to sand dunes, bays, estuaries and numerous sandy beaches.