The main policy driver of this study is the Marine Strategy Framework Directive (2008) and its aim of achieving GES especially in the qualitative descriptors Biodiversity, Food-webs, the relevant part of Eutrophication and Seabed Integrity. Earlier Directives of relevance include the Urban Waste Water Treatment Directive (UWWTD: 91/271/EC) and the Water Framework Directive (WFD, 2000), which aims to bring “water bodies” to “Good Ecological Status.”

The WFD requires the identification of a “type-specific reference condition” for each type of water body. These are defined (see Appendix V in WFD) as the values of hydromorphological, physico-chemical and biological quality elements under undisturbed conditions—i.e., where there is minimum anthropogenic alteration. If such conditions could be identified, they would provide an objective state from which change could be measured. However, this has proven problematic for WFD coastal water types (Bald et al., 2005; Borja et al., 2012). Reasons include the difficulty of finding undisturbed coastal waters and the dynamic nature of marine ecosystems. For example, Devlin et al. (2009) only had data from one non-impacted site to provide reference conditions for phytoplankton in six UK coastal water types.

The MSFD takes a broader view, Article 3.5 defines “Good Environmental Status” as (in part) where “the structure, functions, and processes of the constituent marine ecosystems, together with the associated physiographic, geographic, geological and climatic factors, allow those ecosystems to function fully and to maintain their resilience to human-induced environmental change.” It has been argued (Tett et al., 2013) that fully-functioning and resilient ecosystems can be found even where there is human impact.

We equate MSFD pelagic GES with a state of good ecosystem health (Tett et al., 2013) and an absence of “undesirable disturbance” (Tett et al., 2007). Some changes from the zero-pressure state cannot automatically be seen as detracting from GES: for example, a pelagic ecosystem might be fully-functioning at a level of production enhanced by moderate nutrient enrichment.

Finally, the MSFD's definition of GES includes the condition that human use of ecosystems is sustainable. Most ecosystem biological services are provided by high trophic levels and are outside the topic of the present paper. Nevertheless, a part of our assessment of the pelagic habitat lies in showing that it provides an adequate environment for commercial fisheries and protected marine vertebrates.

The MSFD draws on scientific understanding to define GES as the condition of a fully functioning ecosystem, resilient to anthropogenic activity, and in which “diverse biological components function in balance.” Ecologists hold a range of views about ecosystems and their holistic properties (Tett et al., 2013). Here, we interpret the concept ecosystem, using General Systems Theory (Von Bertalanffy, 1972), as referring to a bounded open system. Resilience is the system's ability to resist or recover from disturbance and is a property that emerges from ecosystem organization and vigor (Costanza, 1992; Mageau et al., 1995; Costanza and Mageau, 1999). Production is a component of vigor, and excess corresponds to eutrophication and may lead to undesirable disturbance. Conversely, lack of vigor constrains a system's abilities to respond to pressure. Organization is defined as “the types and arrangements or interconnections of the components of a system” (Tett et al., 2013), with the relevant components of ecosystems being functional groups. The pelagic habitat is a sub-system within a marine ecosystem, its functional groups are the main lifeforms of plankton, and its vigor depends on primary production and its coupling to higher trophic levels. Thus, assessment of plankton and its environment needs to consider both vigor and organization.

Although, the MSFD requires ecosystem assessment at the level of large marine sub-regions, the sea is oceanographically heterogenous at this scale. The Directive refers to “associated physiographic, geographic, geological, and climatic factors,” which are expected to determine marine ecosystem structure and function. In the case of terrestrial environments, biome theory relates characteristic lifeforms of primary producers to climatic and geomorphological conditions (Clements, 1916; Woodward et al., 2004). The equivalent theory for marine systems is that of ecohydrodynamics (Tett et al., 2007), a term introduced by Nihoul (1981) to emphasize the interaction between hydrodynamic and ecological processes. Hydrodynamic processes such as turbulent mixing and seasonal stratification characterize the physical dynamics of water bodies in which phytoplankters live and to which species are adapted (Margalef, 1978; Pingree et al., 1978; Bowman et al., 1981; Jones and Gowen, 1990). Regional differences in zooplankton result in part from the seasonal patterns of phytoplankton production and hydrodynamic conditions (Backhaus et al., 1994; Dickey-Collas et al., 1996a; Fromentin and Planque, 1996; Gowen et al., 1998a). Parsons (1979), Cushing (1989), and Legendre and Rassoulzadegan (1995) consider how such differences in phytoplankton and zooplankton might combine into distinctive food-webs.

Defining an ecohydrodynamic type is, therefore, the first step in establishing environmental status, because it provides a basis for what is to be expected in a fully-functioning ecosystem of a given type. Van Leeuwen et al. (2015) used 50-year hindcasts with GETM-ERSEM to map persistent ecohydrodynamic regimes in the North Sea (Table 1). Their definitions of regimes, based on the reliable duration of stratification, provides a basis for (i) establishing homogenous marine areas for the purposes of our scheme, and (ii) identifying the ecohydrodynamic type of such areas. There are several ways to define stratification (e.g., Simpson et al., 1978): we follow the method described by Van Leeuwen et al. (2015) using surface-bottom density difference.

Finally, plankton communities are highly dynamic (Gowen et al., 2013). We see seasonal variation in plankton lifeforms as a key part of pelagic ecosystem organization and the analog of succession leading to climax communities in terrestrial ecosystems. In temperate latitudes the seasonal pattern of primary producer biomass and succession is ultimately driven by the annual cycle of illumination and, in some cases an annual cycle of stratification and mixing. The seasonal pattern of zooplankton consumers is largely driven by the producer cycle. Both are perturbed by other factors such as weather, local physical processes, and chaotic dynamics which creates “noise” in the “signal” of the seasonal pattern. Any assessment of the state of the plankton needs to distinguish the two.

Commission Decision 2010/477/EU (2010) provides criteria and operational indicators for the determination of GES. However, the development of appropriate thresholds for these indicators, and rules for aggregating them, is proving challenging (Ojaveer and Eero, 2011; Borja et al., 2012, 2014). There is something of a logical conundrum: waters of known GES are needed to calibrate the indicators, so the indicators themselves cannot be used to assess GES.

Reviewing the difficulties encountered during WFD implementation might be helpful. The first two relevant steps proved straightforward, but it is the third step that bore problems. The first two steps correspond, in our approach, to the delineation of sea-areas of single ecohydrodynamic type that is to: (1) Identify water-bodies, defined as: “a discrete and significant element of surface water such as a lake…or a stretch of coastal water”; (2) Develop a typology for these water-bodies. We understand this delineation, also, as part of the typing and mapping of habitats required by Commission Decision 2010/477/EU (2010) for MSFD descriptor Biodiversity at the habitat level. The third step is more problematic since it requires the establishment of type specific reference conditions for each type. The WFD (Annex II.1.3) defines these conditions as corresponding to expectation for undisturbed conditions, and, although not stated explicitly in the Directive, it is understood that “undisturbed” means absence of significant endogenous pressures (Elliott, 2011). The Directive states that identification of type-specific reference conditions “may be either spatially based or based on modeling, or may be derived using a combination of these methods. Where it is not possible to use these methods, Member States may use expert judgment to establish such conditions.”

The spatially based procedure requires data for water bodies that are free of significant human pressures. It is possible to find sea-areas where there is little loading with anthropogenic nutrient (Halpern et al., 2008) even in North-West European shelf waters, but there are few, if any, where there is no fishing pressure (Lassen et al., 2013) with consequent disturbance of the sea-bed and the food-web (Kaiser et al., 2006; Thurstan et al., 2010). This does not mean that there are no areas at GES, only that it is not possible to use absence-of-pressure as the sole criterion to identify waters at GES.

Models that couple river catchments to regional seas have been used to identify reference conditions for coastal phytoplankton, by simulating reduced nutrient loading (Cugier et al., 2005; Lacroix et al., 2007; Lancelot et al., 2009), but full marine ecosystem models such as ERSEM (Blackford et al., 2004; Van der Molen et al., 2013) do not yet seem adequate to simulate, reliably, the impact of reduction of a range of pressures on the plankton as a whole.

Experts experience several difficulties in reaching objective conclusions. Baselines shift (Duarte et al., 2009): an expert's experience of the present-day world might only include ecosystems that an observer working in an earlier century might consider degraded. Expert groups may tend to a “lowest common denominator” consensus or be biased by a dominant point of view. Delphi procedures (Linstone and Turoff, 2002; Donohoe, 2011) can ameliorate the group difficulty, but are slow and expensive. An alternative is to challenge preliminary assessments with counter evidence, and to report the search for such evidence including that relating to conditions observed by earlier generations of observers. This is the approach we use here.

Many of the methods reviewed in the literature on MSFD application, consider a piecewise approach to the assessment of GES, but this gives rise to challenges of aggregating indicators as well as the conundrum that calibrating indicators requires independent identification of waters at GES. We think that such identification requires a holistic rather than a piecemeal assessment, and that, at the current state of knowledge, a suitably controlled expert method is most likely to achieve this.

What the experts—in the case study, ourselves—are to be asked is: Does the pelagic habitat in a given water body correspond to expectation for a fully-functioning ecosystem component? We propose a 6-step approach: (1) Define the spatial location, extent, boundaries and context of the system under study; (2) Identify the system's ecohydrodynamic type in order to determine expectations for community vigor and organization; (3–4) Acquire data and analyse them in relation to vigor and organization; (5) Collate evidence relating to (lack of) anthropogenic pressure(s), and the functioning of higher trophic levels making use of the pelagic habitat; (6) Seek and take account of evidence that runs counter to conclusions from steps 3–5 about status.

In order to structure the assessment and ensure that all available relevant information is taken into account, Table 2 provides a framework for steps 3–5. It is not intended as an algorithmic procedure, i.e., as a set of rules for arriving at an assessment, but as an agenda for a process of “communicative action” during which evidence and claims are recursively tested (Habermas, 1984). The table arranges items in a matrix in which the rows are the relevant MSFD qualitative descriptors and the columns relate to four aspects of the pelagic ecosystem. While this scheme is intended specifically for north-west European salt waters, it should be generally applicable and may be modified with locally-determined criteria where these are available.

Columns 3 and 4 of the table deal with ecosystem vigor. The column dealing with water chemistry mainly concerns concentrations of winter nutrients, for which thresholds for “non-problem areas” have been set by OSPAR COMMISSION (2003, 2005). Toxic pollution of water, which can impact on plankton, is not explicitly dealt with here (see MSFD QD8); its effects on community structure and function are picked up by other items in the table. Low sub-pycnocline oxygen concentrations can occur naturally in certain stratification regimes (see Gade and Edwards, 1980). The question is whether depletion has been enhanced by additional organic sedimentation from anthropogenic activity.

The “function” column corresponds more closely to vigor: there should be sufficient phytoplankton biomass and primary production to support a food web that is in “balance,” but not so much that production is out of balance with consumption. In relation to the Food-webs descriptor of the MSFD, the goal for a “fully functioning” plankton community is that nutrient supplies are adequate to support expected plankton biomass and production; this sets a lower threshold, as distinct from the upper thresholds. The identification of appropriate minima has been little discussed. In principle, we want primary production to exceed a certain minimum value. In the absence of measured production, it should be possible to estimate this, using relationships between spot measurements of primary production and chlorophyll (Joint and Pomroy, 1993; Gowen and Bloomfield, 1996; Capuzzo et al., 2013). For Seabed integrity the goal for function is that meroplanktonic abundance is consistent with expectation and sufficient to renew the benthos.

Column 5, “structure,” concerns, in principle, the “organization” component of Mageau et al. (1995): the pattern of (mainly trophic) relationships amongst plankters. We see this as highly dynamic at the species level (Gowen et al., 2012), naturally varying both inter-annually and seasonally, but always with a requirement for a dynamic balance of functional groups or “lifeforms.” Assessment can thus be based on the relative seasonal patterns of microplankton lifeforms, such as the succession from diatoms to dinoflagellates expected in seasonally-stratifying temperate seas. The timing (“phenology”) of events in the seasonal cycle is important, especially if it leads to mismatch between phytoplankton and zooplankton cycles, or between these and fish larval food needs (Cushing, 1969; Platt and Sathyendranath, 2008). Where data are available, it is desirable to consider a longer list of lifeforms, (e.g., small flagellates, coccolithophores, ciliates, and picoplankton). Finally, mesozooplankton—both holo- and mero- must be included in relation to Food-webs and Seabed Integrity.

The final column in Table 2 concerns the observed consequences of pressure-induced changes in the plankton for other parts of the marine ecosystem, which includes the benthos and higher trophic levels of the pelagic system such as fish and birds. There are three points to consider here. First, does the plankton provide food for the higher trophic levels that would be expected in a fully functioning system? Are the biota of the higher trophic levels present as expected, or, if less or more abundant than expected, can this be traced to food supplies, as has been claimed for kittiwakes (Frederiksen et al., 2004; Wanless et al., 2007)? Second, is there any evidence of an “undesirable disturbance.” This needs to be considered from both top-down pressure (resulting from fisheries as claimed for the Baltic Sea, Österblom et al., 2007) and bottom-up pressure (nutrient enrichment, Tett et al., 2007). Third, is there any evidence that the pelagic ecosystem is resilient to anthropogenic pressures? The latter requires time series of organization, vigor, and pressures.

Assessment requires a comparison between the observed state of the pelagic ecosystem and its expected state given the physical conditions—i.e., given a particular ecohydrodynamic type. A conceptual model (Figure 1) is useful in considering expectation. According to this model, the main driver of the state of the pelagic ecosystem is the astronomical cycle that gives rise to temperate-latitude seasonal cycle of illumination and heating. The illumination cycle is the direct driver of primary production, and the heating cycle drives not only water temperature and directly affects vigor, but more importantly, changes in thermal stratification. This seasonal cycle is the oceanographic “signal.” It is confused by “noise,” i.e., short-term variation that results from weather, local oceanographic processes, and ecosystem dynamics. The physical signal plus noise then gives rise to seasonality in the chemical and biological dynamics of the pelagic habitat and comparison of observed and expected state needs to consider the generalized seasonal biological signal whilst taking account of the noise.

The main way to distinguish signal from noise is to extract statistics of seasonal variation from several years of data, the argument being that the noise is essentially random and hence that an average pattern will tend toward the true cycle as the random component sums to zero. The signal may evolve because of climate change but can be treated as constant in the medium term.

A picture of the mean seasonal cycle, with information on the variability about that cycle, is widely called “climatology” (after the usage of Levitus, 1982). It is the mean cycle that can be compared with biological expectation, since the expectation in principle derives from the astronomical driver by way of hydrodynamics, while the variability provides a basis for assessing whether there is a significant difference between expected and observed climatologies.

Since the early 1990s, regular surveys of the oceanography, nutrients, and plankton of the WIS have been undertaken by the Department of Agriculture and Rural Development (DARD), subsequently the Agri-Food and Biosciences Institute (AFBI) in Northern Ireland. An instrumented mooring was deployed in 1995 at 53°47.20′N 05°37.75′W (station 38) in a water depth of 105 m. In 1997, the mooring was moved to 53°47′0 N and 05°38′0 W (station 38A) at a depth of 94 m (Figure 2). The mooring was equipped with a CTD instrument, remote access water samplers for nutrients (RAS-100) and phytoplankton (RAS-500) and a surface-bottom string of thermistors. We draw on published and unpublished data from these surveys and the instrumented moorings for our analysis (see Table 3 for details and source references).

CTD profiles obtained from research vessel visits to station 38 and 38A were analyzed according to the criteria in Table 1 with stratification deduced from differences between near-surface and near-bed density. The effect of salinity alone was calculated using temperature averaged over the two depths. The thermally-defined surface mixed layer (th-SML) was defined as extending to the depth h_m at which water temperature was 0.5°C less than that at 2 m (Talling, 1971).

Euphotic zone depth was calculated from k_D in two ways (Table 3). For data retrieved before 2002, k_D was calculated from submarine PAR measurements from depths greater than 1 m optical thickness (Gowen et al., 1995, 2000; Gowen and Bloomfield, 1996). For 2002–2012 values for k_D were taken from remote sensing (Capuzzo et al., 2013). The empirical OC5 algorithm for optically-complex (Case-II) waters was used to retrieve the water column components from ocean color (Gohin et al., 2002). Daily values of k_D (PAR) for this case study were extracted from the daily maps.

Samples for analysis for dissolved inorganic nutrients, oxygen, and chlorophyll, were taken from water-bottles attached to a ship-deployed CTD system; nutrient samples were also taken by the automated sampler attached at a notional depth of 12 m to the instrumented mooring. Dissolved inorganic phosphate (DIP), dissolved inorganic oxidized nitrogen (ToxN, nitrite plus nitrate), and soluble silicate (DSi) were determined colorimetrically (Gowen and Stewart, 2005). Oxygen was determined by the Winkler method (Strickland and Parsons, 1968) and after 2008 an oxygen probe attached to the CTD instrument provided additional data. Chlorophyll was determined following the method by Tett (1987). Samples were filtered and either analyzed immediately or stored frozen (−20°C) for later analysis on board the ship or in the laboratory. Comparisons showed no significant differences between “fresh” and “frozen” samples (B. Stewart, AFBI, pers. com.). Chlorophyll standing stock (mg chl m⁻²) was estimated by interpolating sampled values to 1 m intervals and summing over the euphotic zone (Gowen et al., 1995).

Primary production data are taken from Gowen et al. (1995) and Gowen and Bloomfield (1996). These authors used the ¹⁴C tracer method (Strickland and Parsons, 1967) to estimate carbon assimilation, the method of Herman and Platt (1986) to calculate gross daily column production and the method of Joint and Pomroy (1993) to estimate daily column production from chlorophyll standing stock. For the purposes of this paper column daily net primary production was estimated by subtracting column respiration, R° according to Lee et al. (2003). New production during the period of the Spring Bloom was estimated by multiplying the column draw-down of ToxN by the Redfield C:N ratio (Table 5).

The plankton was sampled in two ways during the DARD and AFBI studies. Micro-organisms were sampled by automatic sampler (McKinney et al., 1997) or rosette-water-sampler (Scherer, 2012) preserved in acidic Lugol's iodine, and enumerated using inverted visible-light microscopy after sedimentation (Utermöhl, 1958). Planktonic animals were sampled either by vertical hauls with 300 (Gowen et al., 1997) and 280 μm mesh (Gowen et al., 1999) or by a double-oblique tow of a Gulf III sampler with a 280 μm mesh net (Dickey-Collas et al., 1996a). Catches were preserved with buffered 4% formaldehyde and volumetrically subsampled for analysis. Methods for jellyfish and “semi-nekton” such as fish larvae and euphausiids are given by Dickey-Collas et al. (1996a) and Lynam et al. (2011).

Plankters were identified to species whenever possible, otherwise qualified as high-level taxa with additional criterion such as size. The estimates of numerical abundance or biomass were aggregated into the lifeforms or functional groups listed in Table 4 (Scherer et al., 2014). The set of photo-autotrophic lifeforms (including some myxotrophs) is referred to as phytoplankton; all micro-organism lifeforms (including ciliates, heterotrophic dinoflagellates, and phytoplankters) is referred to as “microplankton” (Scherer, 2012). The latter definition draws on the original sense of the term proposed by Dussart (1965) as distinct from the size range of phytoplankters defined by Sieburth et al. (1978) and Lalli and Parsons (1997).

Two methods were used to extract climatologies from the data. One was to compute means ȳ(m) for each month m in the years chosen for analysis, and then to compute climatological monthly means ȳ̄(m) from these, so weighting each year equally. Variability was estimated from the standard deviation of the ȳ(m). The second method, inspired by cyclical nature of the astronomical driver in the conceptual model (Figure 1), was to fit a truncated Fourier series with a base frequency of 1 yr⁻¹ (Brito et al., 2009).

We used data collected between 1992 and 2014 which means some conclusions rely on comparisons of data taken across time: for example microplankton from 2008 to 2009 with copepods from 1992 to 1996. Changing global and regional climate, and changing anthropogenic pressures, might be expected to lead to decadal-scale changes in marine physical and biotic climates in the WIS. Therefore, only data previously examined for temporal trends (e.g., Allen et al., 1998; Gowen et al., 2002, 2008) were considered here using the Mann–Kendall non-parametric test for monotonic trends (Hirsch and Slack, 1984; Hirsch et al., 1991).

Seasons in temperate seas, although ultimately driven by the same astronomical processes responsible for terrestrial seasons, differ from the conventional seasons. We therefore refer to events or periods using the day of (a generalized) year. Table 8 brings these markers together in relation to named seasons relevant to the WIS ecohydrodynamic and biotic regimes.

The extent of the WIS in Figure 2 was based on numerical simulations (Hill et al., 2008) with ample observational justification (Gowen et al., 1995; Horsburgh et al., 2000) of a seasonally stratified region extending around 60 km North-South and 40 km East-West. The mooring station 38A lies toward the southern end of this region.

Gowen et al. (1995) found that the exact region delineated by the potential energy anomaly (PEA, φ) contour varied between surveys. We assume that the deep muddy trough (Hill et al., 1994) corresponds to the long-term mean extent of the seasonally-stratifying waters. The area of this patch is ~3504 km² and with an average depth of 90 m giving the WIS a volume of ~3.15 × 10¹¹ m³.

Bowden (1950) reported the annual mean surface salinity for offshore waters in the WIS to be between 34.12 and 34.14 (recorded as %0) for the period of 1903–1931. Gowen et al. (1995) gave the 1992 annual median as 34.17%0. For the last 10 years salinity has averaged 34.07 (psu) at station 38A, with a typical range of 0.53 between winter maxima and summer minima. Measurements of sub-surface (5 m) water temperature made between 1996 and 2013 show a pronounced seasonal cycle with winter minima of 7–8°C and a summer maxima ~16° C. These data establish the WIS as a euhaline and temperate water-body.

Thermal stratification typically begins end of April ±3 weeks (Bowden, 1955; Slinn, 1974; Gowen et al., 1995; Horsburgh et al., 2000). Although the deeper water is supposed to be isolated as a “cold pool” within the gyre circulation (Hill et al., 1997) it warms slowly during the summer explained in numerical simulations by advection from the north (Holt and Proctor, 2003). Salinity layering has often been observed from February through May (Figure 3) and Gowen et al. (1995) reported that this contributed to the onset of density layering in the WIS. Models (Holt and Proctor, 2003) suggest that the freshwater results from Irish runoff spreading over the winter's inflow of oceanic water from the Celtic Sea.

The salinity contribution to layering might be expected to allow earlier phytoplankton growth than would otherwise be the case. However, water in the Irish Sea, especially in the winter, contains higher loads of Suspended Particulate Matter (SPM; Coombes et al., 1994), than found in other seasonally stratified regimes such as those in the North Sea (Capuzzo et al., 2013). Smoothed monthly mean sub-surface PAR attenuation in 1992 and 1993 fell from 0.30 m⁻² in March to 0.20 m⁻¹ in June–August but had increased to 0.25 m⁻¹ in October (Gowen and Bloomfield, 1996). This is consistent with reports that SPM concentrations decrease in the Irish Sea during the season of stratification, as the material sinks below the pycnocline (Weeks and Simpson, 1990; Weeks et al., 1993).

Figure 4 shows euphotic zone depth estimated from remote sensing in 2002–2012 and th-SML thickness for 1992–1996 from Gowen et al. (1998b). According to this, the season during which stratification provides favorable conditions for photosynthesis is relatively short: 123 days (from day 124, May 3 to day 247, September 3).

Whereas, there is clear evidence (Gowen et al., 2002) of nutrient enrichment in Liverpool Bay, and Irish coastal waters, concentrations and seasonal cycles of ToxN and DSi have not changed in the WIS since first measurements in the mid 1950s to late 1960s by Slinn (1974) (Gowen et al., 2008). The median winter concentrations of DIP have decreased since 1989 (Gowen et al., 2008).

Recent data (Figure 5) for near-surface ToxN, DIP, and DSi show maxima in January/February (7.0, 0.8, and 7.8 μm, respectively), followed by minima in June/July (0.8, 0.5, and 1.6 μm, respectively). Climatological near-surface ToxN fell to half of the winter maximum on day 121 (21 April) with a time range of ±3 weeks (30 March–23 May). Deep water concentrations remain high, close to winter values, with upwards losses probably replaced by deep-water and benthic mineralization of sinking organic matter, less denitrification (Trimmer et al., 1999; Gowen and Stewart, 2005).

WIS nutrient dynamics are still debated. Background diapycnal mixing seems weak, and the summer flux of nutrients that supports a midwater chlorophyll maximum may result mainly from event mixing (Williams et al., 2013). The initial summer increase in SML nutrients (Figure 5) seems correlated with SML deepening (Figure 4) and hence with entrainment of deeper water during August and September. The rest of the increase to the winter maxima has been variously ascribed to local remineralization (Gibson et al., 1997), freshwater nutrients, and those brought into the WIS from the Atlantic Ocean by way of the Celtic Sea or the North Channel (Simpson and Rippeth, 1998; Hydes et al., 2004).

Freshwater discharges into the Irish Sea are rich in nutrients. Nitrogen and phosphorus enrichment is demonstrably anthropogenic (Nedwell et al., 2002) but silica enrichment is considered to be a natural result of the weathering of silica-rich sedimentary rocks. Consequent on this enrichment, mixing calculations based on salinities (Table 5) imply that WIS water should be enriched compared to oceanic waters, but this is only evident for dissolved silica (Gowen and Stewart, 2005). The mixing calculation is not reliable for phosphate, which is influenced by estuarine processes (Zwolsman, 1994). It has been shown that N is lost from the oceanic source water (presumably as a result of denitrification) before it reaches the WIS, a journey that takes at least a year (Hydes et al., 2004; Moschonas et al., 2016).

Consequently, winter ToxN and DIP concentrations in the WIS are, despite some anthropogenic enrichment via river water, (a) similar to oceanic concentration at the shelf break and (b) below the OSPAR thresholds for enrichment. With the exception of some low salinity waters, N limitation is the expected norm in most temperate shelf seas and coastal waters (Ryther and Dunstan, 1971; Beardall et al., 1978; Tett and Droop, 1988; Gibson et al., 1997; Gowen et al., 2012). Winter N:P ratios are somewhat less than Redfield (Table 5) implying a predisposition to nitrogen limitation, and this is confirmed by DIP in summer, which is not fully depleted.

Silica-enriched freshwater augments Si concentrations in the WIS, so that the N:Si atomic ratio approaches 1 in winter (Gowen et al., 2002) and falls to about 0.4 in summer. Using uptake ratios, Gowen et al. (2000) showed that N ran out before Si in Irish coastal waters. In the WIS there appears to be adequate silica for diatoms and silicoflagellates for most of the year but during summer, the concentration in surficial waters is less than the half-saturation concentration of 2 μM (Escaravage and Prins, 2002) for silica uptake.

Finally, although there are no adequate time-series of oxygen concentration in the deep water of the WIS, there is evidence (Slinn, 1974; Gowen et al., 2008) that a small oxygen deficit (maximum 20% below saturation) builds in the deep water during stratification, as it does in the northern North Sea (Lee et al., 2002; Queste et al., 2013). Such deep-water de-oxygenation is part of the natural functioning of pelagic ecosystems in stratified waters. There is no evidence that the benthic community of the WIS is negatively affected by it. The burrowing decapod Nephrops norvegicus, an important member of the macrobenthos, is sensitive to hypoxia (Diaz and Rosenberg, 1995), and DARD photographic surveys continue to show good populations of the species in the seabed of the WIS (Gibson, 2011).

The expectation for vigor emerges from the conceptual model (Figure 1) of the pelagic ecosystem as an open system driven by fluxes of energy and nutrients. The WIS was characterized above, as a seasonally-stratifying ecohydrodynamic regime in which nutrients accumulate during the autumn and winter. In such regimes, the next phase in the seasonal cycle follows vernal increases in illumination and stratification, which allow phytoplanktonic photosynthesis to exceed respiration (Marshall and Orr, 1927; Sverdrup, 1953; Smetacek and Passow, 1990; Tett, 1990). It is useful to view the outcome as the yield (Gowen et al., 1992) of phytoplankton biomass from the accumulated limiting nutrient, nitrogen. Thus, the expected consequence of stratification is a pulse of increased chlorophyll—the Spring Bloom—with associated new production which is transferred to the benthos by sinking or to the heterotrophic plankton. Thereafter, the pelagic system is expected to tend toward local equilibrium: the condition in which available nutrients are mainly those locally regenerated and little production is available for export (Dugdale and Goering, 1967), except perhaps from a sub-surface chlorophyll maximum supported by new nitrogen. The latter is the case in the central North Sea (Fernand et al., 2013). Finally, a deepening SML, entraining deep-water nutrients, might be expected to generate an Autumn Bloom.

Observations correspond well to this expectation. Chlorophyll concentrations and microplankton biomass in the surface mixed layer show a spring peak (in May) and a lesser peak in August–September (Figure 6). There is some inter-annual variability in the timing of these blooms, but in the climatology made by fitting a TFS to column chlorophyll, the spring peak was reached c. day 150 (end of May) and the autumn peak c. day 230 (late August; Figure 8). A deep chlorophyll maximum is observed in late June (Gowen and Stewart, 2005; Rippeth et al., 2014) and mid-July (Williams et al., 2013) as a frequent feature with concentrations of up to 8.1 mg m⁻³ (Gowen and Stewart, 2005).

The productive season is short. If defined as the period when chlorophyll exceeds 1 mg m⁻³ its climatological duration is from day 120 (1 May) until day 256 (13 September). Near-surface chlorophyll sometimes exceeded 10 mg m⁻³ during the Spring Bloom, fell to about 1 mg m⁻³ for a short period around day 180 (end of June) when nutrient concentrations are lowest, and rose to, at most, 4 mg m⁻³, during the peak in late summer/early autumn.

The deviations from the climatological envelopes in 1999 and 2000 do not seem part of any long term pattern: time-series analysis has failed to show any temporal trend in DARD data from 1992 to 2004 (Gowen et al., 2008). Furthermore, concentrations of chlorophyll during the period after the Spring Bloom match those of 1–4 mg m⁻³ reported for 1968 by Slinn (1974), and for 1980–1982 by Richardson et al. (1985). However, Allen et al. (1998) have related increased phytoplankton biomass to an increasing trend of anthropogenic nutrients based on a long-term dataset west off the IoM.

Based on data from 1992 to 1993, Gowen and Bloomfield (1996) estimated gross seasonal production to be 140 g C m⁻². Reworking these data and including observations from 1997 and 2010–11, led Scherer and Gowen (2013) to conclude that: the production season can last up to 5 months; gross seasonal production is in the range 101–140 g C m⁻² and annual gross production is in the range from 157 to 291 g C m⁻², with a central value of 204 g C m⁻². These values derive from the fitted climatological TFS in Figure 7B, and the range estimates include short-term as well as inter-annual variation; they do not imply that annual production varied between the limits given. The summer deep chlorophyll maximum contributes only a few grams of new production to these totals (Rippeth et al., 2009; Williams et al., 2013).

There is a difference between the season of chlorophyll >1 mg m⁻³ and phytoplankton growth. As evidenced by euphotic zone chlorophyll in Figure 7A, the micro-algal population typically begins to increase in early February, soon after day 30, and it begins its autumnal decrease at ~day 220 (July). It may be that salinity layering allowed some algae to remain in the euphotic zone in late winter, thus encouraging growth. Gowen et al. (1995) concluded that in 1992 a mean th-SML irradiance of between 6 and 10% of daily mean surface PAR was required to trigger an obvious increase in chlorophyll in mid April. At this time the (thermal) mixed layer was still more than 80 m deep but shallowed rapidly, to 20 m in early May. We hypothesize that, as this happens and the Spring Bloom proceeds, population dynamics become more complex: cellular photosynthesis not only needs to exceed respiration, it must also exceed losses due to sinking and grazing, if population growth is to continue. So far as system vigor is concerned, it is, of course, the extra losses that are important, because these are the transfers to higher trophic levels. Gowen et al. (1999) showed transfer to copepods increased during the Bloom in 1997.

Observations allowing calculation of all the terms in the spring budget were made using data collected between 1992 and 2001 (Table 6). Because the terms were independently arrived at, the budget is not constrained to balance. This results in discrepancies: for example, calculated new production exceeds calculated net production. So does the total of losses to grazing and sinking, even if that total is less than calculated new production. One reason could be an over-estimation of respiration. But this is not the place to try to resolve such difficulties; we simply conclude that of the year's major pulse of net production, approximately half sinks, and half is consumed mainly by copepods but with some going to protozoans. Figure 7 suggests a second pulse of production in the late summer, driven by entrainment of deeper water. Less is known about the dynamics of this bloom. In 2008–2009 there were more grazing protozoans at this time. Although copepod numbers typically decreased after the Spring Bloom (Figure 6C) grazing pressure may remain high at this time, because of warmer temperatures and larger individuals.

The sedimentation flux in Table 7 was calculated on the assumption that it should balance the increase in sea-bed oxygen demand (SOD) observed during the Spring Bloom period in 1998 (Trimmer et al., 1999). In 2005, Hill (2007) observed a peak in deep-water chlorophyll within a few weeks of the Spring Bloom in the upper waters confirming that at least part of the transfer is the sinking of living phytoplankters. Using a relationship between water depth and organic flux (Suess and Müller, 1980; Trimmer et al., 1999) estimated the annual input of new production to the benthos as 17.2 g C m⁻², less than some of the Spring Bloom estimates. Studies later in the year, to estimate sedimentation from the Autumn Bloom, are lacking. Nevertheless, it is possible to make rough estimates (Table 7) of the annual flux of organic carbon from phytoplankton through to zooplankton as well as benthos, and then on to fish.

Although, high phytoplankton biomass (16–23 mg chl m⁻³, Gowen and Bloomfield, 1996) and high cell counts in total (1.6 × 10⁶ diatom cells L⁻¹, Scherer, 2012) have been observed during the peak of the Spring Bloom, average chlorophyll concentration for the period between March and September never exceeded the 90th percentile of the OSPAR eutrophication threshold of 10 mg m⁻³ for coastal offshore waters (Gowen et al., 2008) and the average cell abundance never exceeded the OSPAR thresholds for any single taxon of 10⁶ or 10⁷ cells L⁻¹ in total (McKinney et al., 1997; Gowen et al., 2008). The Spring Bloom is rapidly followed by an increase in numbers of planktonic copepods (Figure 6C), the dominant zooplankter in the WIS (Scrope-Howe and Jones, 1985), implying an effective transfer of production from phytoplankton to their mesozooplankton consumers. Gowen et al. (1999) estimated that 22% of Spring Bloom production was consumed by copepods in 1997. Additional primary production likely passes to ciliate and dinoflagellates consumers, supporting the increases shown in Figure 8. There is effective transfer of biomass from near surface to deep waters and the benthos as evidenced by increased sediment oxygen demand (Trimmer et al., 1999; Hill, 2007). The Trimmer et al. (1999) study suggested that this benthic demand required 46% of total spring phytoplankton production or 61% of new production. Based on Trimmer's 17.2 g C m⁻² new production, Scherer and Gowen (2013) estimated that, 4.4 kg C ha⁻¹ y⁻¹ was available to animals at trophic level 3 in the benthos. A Nephrops catch of 1.4 kg C ha⁻¹ y⁻¹ (Gibson, 2011) therefore represents 31% of the available phytoplankton carbon.

Thus, the WIS appears to be a well-functioning pelagic ecosystem that well exploits available resources (the nutrients available at the end of winter) during the Spring Bloom and efficiently transfers Bloom primary production to higher trophic levels. The sub-surface chlorophyll maximum may be an additional source of new production in summer, as may be the late Summer Bloom that occurs in some years.

The argument in previous subsections has been straightforward, because expectations for vigor (nutrients and production) arise deterministically from ecohydrodynamics. This is less clearly the case in respect of organization. The pelagic food web is complex and may have several stable conditions that equate with GES. Links between these conditions and ecohydrodynamics are less obvious, and have been more disputed than those for vigor. There may therefore be a range of expectations for the functional biodiversity and the trophic links in a healthy, fully-functioning, pelagic ecosystem of a seasonally-stratified regime. Furthermore, there is no consensus on the identification of functional groups in the plankton; Table 4 lists those that we have considered for the UK monitoring programme. Finally, also to be taken into account in assessing whether organization is adequate for GES is the extent of functional response diversity (redundancy; Folke et al., 2004; Hughes et al., 2005) within functional groups. We view zooplankton feeding traits (Kiørboe, 2011) as part of functional response diversity insofar as they determine how X catches Y, and as relevant for organization only if they determine whether X catches Y₁ instead of Y₂.

At issue is whether the “pelagic habitat” in the WIS provides good conditions for larger animals including fish and sea-feeding higher vertebrates, and for the young of benthos nekton. There is an interpretational difficulty, as much data concerning fisheries and sea-birds relate to larger scales. Fish stocks and fish landings are mainly reported from ICES fisheries area VIIa, which covers the Irish Sea between 52°N and 55°N, a much larger area (48,000 km², Brander and Dickson, 1984) than the ~3000 km² of WIS, and one that is heterogeneous in respect of ecohydrodynamics. However, an important part of the link between plankton and fish is by way of plankton-feeding fish larvae, and a good deal is known about their distribution and movements.

The western part of the Irish Sea is an important spawning and nursery ground for a variety of fish, many commercially exploited. Spawning begins in Irish coastal waters in early Spring, and young fish move (or are entrained) offshore into the stratified waters of the WIS in June, taking advantage of better food and feeding conditions there (Dickey-Collas et al., 1996a). Four decades ago these fish recruited to productive pelagic and demersal fisheries, the major landings (by weight) being of herring (pelagic) and cod and whiting (demersal) (Brander and Dickson, 1984; Lynam et al., 2011). At that time a smaller amount of the burrowing decapod, N. norvegicus was caught. N. norvegica adults inhabit burrows in the mud that covers much of the seabed beneath the WIS, where they are predated by cod (Armstrong, 1982; Brander and Bennett, 1986). It is thus perhaps unsurprising that a decline in the cod population (Kelly et al., 2006), due to overfishing, has resulted in an increase in the amount of N. norvegica landed: by 2002 the weight (and value) of shellfish landings from ICES sea-area VIIa (the Irish Sea as a whole) greatly exceeded the weight and value of finfish landings. Estimated Nephrops landing into Northern Irish ports in 2008 were 7876 t with a market value of £14.2 million (Gibson, 2011).

The life cycle of Nephrops seems to adapt them to the WIS gyre, which retains their pelagic larvae (White et al., 1988). Settling to the deep mud sea-bed within this gyre, the larvae grow into burrowing adults that exploit sedimenting organic matter. So, although this part of the WIS ecosystem is not in a pristine state, it is nevertheless functioning well in supporting higher trophic levels and in delivering ecosystem services (such as the Nephrops fishery) to human society. The shift from cod to Nephrops might be seen as evidence of the system's adaptive capacity, or resilience in the face of fishing pressure. Furthermore, it makes more efficient use of primary production, prawns occupying a trophic level below that of cod.

Most sea-birds feeding in the WIS have flown there from distant nesting sites, but there are ample data from shipboard observers concerning distributions (MacKey and Giménez, 2004). The feeding preferences of birds can be identify as plankton feeders and those that take larger fish or fisheries discards. The seabirds observed in large numbers over the WIS during the “breeding” and “post-breeding” seasons include the Manx shearwater (Puffinus puffinus), which feeds near the sea-surface on small planktonivorous pelagic fish (Carboneras et al., 2016). Their population is monitored only on the island of Skomer in St George's Channel, showing no discernible long-term trend in chick rearing success or adult survival rates since 1986 (JNCC, 2015). Guillemots (Uria aalge), razorbills (Alca torda), and gannets (Morus bassanus), birds that dive for small shoaling planktonivorous fish and larger plankton, are also common in the WIS. Although, the exact diet of such birds has not been studied in the WIS, the fish on which they prey are likely to include sand eels and young, i.e., small, gadoids, and clupeiods. Existing relevant population data (JNCC, 2015) suggest that birds are breeding successfully, implying good feeding. However, according to ICES Ecosystem Overviews (published in March 2016) trends in the abundance of many breading seabirds are on a broad downwards trend in the Celtic Seas region since the early 2000s, while species that breed elsewhere have been reported to feed in the area (ICES Ecosystem overview, 2016).

In this subsection are considered arguments against the claim that the WIS is in GES. Significant counter-arguments must show that the WIS is less than the fully-functioning pelagic ecosystem expected under the identified ecohydrodynamic conditions. Evidence of change is relevant only if the change is adverse to full-functioning and the provision of ecosystem services.

The definition of GES in MSFD Article 3.5(a) links it to “the structure, functions and processes of the constituent marine ecosystems” in a marine region or sub-region. Lindeman (1942) defines an ecosystem, as “the system composed of the physical-chemical-biological processes active within a space-time unit of any magnitude.” The state of an arbitrarily-bounded system might however be difficult to establish, and our method therefore requires the system under study to be uniform in respect of ecohydrodynamic regime. The evidence shows that the WIS exists as a system with spatial boundaries naturally defined by bottom topography and the seasonal gyre, although open to “reset” in Winter by inflows from the North Atlantic and Irish coastal waters, and with a typical system-scale process closure time of a year. Thus, although it is only a small part of the Celtic Sea region that is the MSFD unit of reporting and management, the WIS pelagic ecosystem exists as a distinct system in the sense required for our assessment. Furthermore, the system has a distinct ecohydrodynamic regime that has been observed in all years studied and which (we've argued) drives the seasonal cycle. The regime falls in the “seasonally stratified” category although with several special features: early haline stratification aids early growth of phytoplankton, offsetting the high turbidity of these waters.

Lindeman's definition of ecosystem is descriptive and open to different interpretations of the meaning of “the physical-chemical-biological processes” taking place therein. The textbook idea of an ecosystem is one of these interpretations. It has been pointed out (O'Neill, 2001) that “an ecosystem,” as it is customarily understood, is not an observable but a construct, a way of organizing a set of observations. The paradigmatic example is the food web of the herring in the North Sea (Hardy, 1924), constructed from many observations of the stomach contents of fish and plankters.

The MSFD is also a conceptual structure, but it is a model of what should be rather than what is. It is intended to shape human behavior to achieve the objectives set out in the Directive's preamble. Within the MSFD, the idea of GES can be understood as both an objective assessment of the state of the relevant marine ecosystem—i.e., something arrived at independently of the Directive by the procedures of science—and a description of the state that European society wishes marine systems to reach, in order to ensure (MSFD Article 3.5) that “the use of the marine environment is at a level that is sustainable, thus safeguarding the potential for uses and activities by current and future generations.”

Bringing together the two parts—the assessment and the goal—is challenging (Mee et al., 2008; Borja et al., 2013, 2014; Tett et al., 2013). Although we have elsewhere proposed the Plankton Index (Tett et al., 2008; Gowen et al., 2015) as a quantitative tool for tracking change in pelagic community organization, it is not self-calibrating. In our view, there are no value-free methods for evaluating environmental status in relation to the MSFD or any other socially-determined framework for marine environmental protection. Furthermore, the arguments advanced above lead to the conclusion that the difference between not-GES and GES should not be seen as a threshold on a continuum of state. Instead, a status assessment is a judgment about certain deductions that can be made from available information; a marine ecosystem in GES is one that, according to present scientific knowledge, is likely to contribute to the socially desired purpose of ensuring good ecosystem health and sustainable ecosystem services; a non-GES system requires “programmes of measures” in order to achieve those purposes. The judgment is one that should be made holistically, on the balance of evidence, rather than by algorithmic aggregation of component indicators.

The evidence concerning the state of the WIS included vertical profiles of temperature, measurements of photosynthesis, counts of organisms contained in a water-sample or net haul, and so on, which we've analyzed according to the conceptual causal model in Figure 1 and the assessment scheme in Table 2. The scheme is explicitly informed by the criteria given in Commission Decision 2010/477/EU (2010), which expand the relevant qualitative descriptors (1, 4, 5, 6) of the MSFD. Implicitly, it assumes the existence of trophic transfers, which are evidenced by the gut contents data referred to in Section Observations of Zooplanktonic Predators concerning zooplankton predators, and by the experiments of Gowen et al. (1999) on copepod grazing.

On the basis of this and other evidence analyzed in Section Results and First-Order Discussion, we conclude that pelagic ecosystem vigor appears to be in balance with organization and with the system's capacity to channel production to higher trophic levels without undesirable disturbance. Although there is evidence of local human impact (in respect of fisheries effects on higher trophic levels, and in nutrient enrichment) and perhaps of the effects of climate change, these do not appear to have interfered with the organization and full-functioning of the pelagic ecosystem. Because the picture we have presented of pelagic ecosystem vigor and organization has been assembled from data taken over more than three decades, our synthesis relies on the absence of significant long-term trends. Gowen et al. (2002, 2008) show the absence of such trends in nutrients and phytoplankton during the 1990s and 2000s, but there is the counter-evidence of an increase in spring chlorophyll (Allen et al., 1998), and in scyphozoan jellyfish (Lynam et al., 2011). Nevertheless, we conclude that, on the balance of the evidence, it is more likely than not that the pelagic habitat in the WIS is in GES.

Since this conclusion was not algorithmically assembled, we cannot assign a quantitative uncertainty to our finding. It may be best understood in terms of its practical implication, which is that the planktonic component of the ecosystem (the pelagic habitat) is in a state to continue supplying services at the present level without additional management measures. Although, the WIS is only a small part of the Celtic Seas MSFD sub-region, it provides an ideal site for the continued monitoring of the pelagic habitat in relation to the supply of final ecosystem services, being already well-studied, and being comparatively sheltered and close to land. Thus, it could continue to provide a test case for the finding of GES.

Finally, as a next step in research related to monitoring of the pelagic habitat for the MSFD or for any other approach to maintaining a healthy state in marine ecosystems, observations from the WIS can be used to calibrate the Plankton Index (PI) tool. The WIS is in GES, therefore state space envelopes derived from recent plankton monitoring can be used as reference conditions in tracking change in the plankton community away from the desirable condition, whether in the WIS itself or in other sea-areas of similar ecohydrodynamic type.

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.