The Milford Haven Waterway, Pembrokeshire, Wales, UK, at 55 km² is the largest ria-estuary complex in the UK and comprises a central waterway (maximum depth 27.5 m) with numerous shallow embayments, tributaries and pills [1]. The maximum tidal range is 7.76 m, and 30% of the area is intertidal. The waterway has been industrialised in some form since at least the medieval period, albeit at a limited scale [2,3]. However, this changed in the 1960s following a government decision to make it the major deep-water oil port in the country [4,5]. Currently, there are two large Liquefied Natural Gas (LNG) plants, jetties and pipelines, as well as one of the UK’s largest storage terminals for bulk petroleum products and the port handling the most shipments of these products of any port in the UK. Together, these facilities are capable of supplying nearly a third of the UK’s gas needs alone [1,6,7]. This infrastructure is economically important, supporting over 3800 full-time equivalent jobs, equating to ~40% of total employment in the local economy and 7% across the wider region [6].

The transport of large quantities of hydrocarbons renders the waterway vulnerable to environmental accidents, and there have been some comparatively minor pipe leakages and refinery explosions over the years. Since the 1960s, the waters of the Haven have received chronic inputs of hydrocarbons from the refineries and the oil-fired power station sited on its banks, as well as from small oil spills and domestic inputs [8,9], including the Chryssi P. Goulandris spill of >250 t in 1967 [3,10]. However, Little et al. [11] estimated that <240 t of oil enters Milford Haven annually, most of it well dispersed in water and already associated with suspended particles. A major environmental catastrophe did occur when the Sea Empress oil tanker became grounded on mid-channel rocks at the entrance to the waterway on 15 February 1996 en route to the Texaco oil refinery. Over a week, ~72,000 t of its cargo of 131,000 t of light crude oil (Forties Blend) and 480 t of heavy fuel oil were released into the surrounding waters, resulting in the contamination of 200 km of the Pembrokeshire coastline [8,12]. It was Britain’s third-largest oil spill and the twelfth-largest in the world at the time [13]. The cost of the clean-up operation was estimated to be £60 m at the time (£115 m in 2020) and involved the 446 t of chemical oil dispersant, which enhanced the rate of natural dispersion of the oil and prevented an additional 57,000 to 110,000 t of emulsified oil from impacting the beaches [12,14].

Despite the industrial infrastructure, Milford Haven is regarded as having a high conservation value, being designated by the European Habitats Directive as a component of the Pembrokeshire Marine Special Area of Conservation, and is also a Site of Special Scientific Interest [1,15]. The area contaminated by the oil spill in 1996 is well known for its natural beauty and is utilised for various purposes such as tourism, fisheries (e.g., herring; Clupea harengus) and aquaculture. The value of the ecosystem and its history of anthropogenic use have led to numerous monitoring projects since the 1960s (see summary in [3]). More contemporary monitoring has revealed that heavy fuel oil was still detectable in the sediments of Milford Haven in 2010 [9] as were metals, organotins, polyaromatic hydrocarbons and polychlorinated biphenyls in a range of algae and benthic invertebrates [16].

Macrobenthic invertebrates, in particular annelids, molluscs and crustaceans, are major components of estuarine fauna in the UK and globally [17,18] and play vital structural and functional roles in the benthic environment. Through their movements and behaviour, e.g., respiration, bioirrigation and bioturbation, they influence carbon, nitrogen and sulphur cycling and help oxygenate sediments [19,20]. Moreover, by ingesting detritus and becoming prey for many higher trophic organisms such as fish and birds, they are crucial links in the food web [21,22]. Changes in the diversity and community composition of the macrobenthos have been detected along natural environmental gradients, e.g., salinity and sediment grain size, but also in response to a range of deleterious anthropogenic influences, e.g., eutrophication and hypoxia [23,24,25,26]. These predictable and reliable changes in the macrobenthos have resulted in these taxa becoming one of the mainstays of monitoring the ecological condition of waterways [17,27,28]. Many studies have demonstrated the deleterious effects of hydrocarbon spills on invertebrate communities [29,30,31].

The main objective of this study was to use data from the surveillance program in Milford Haven to determine the extent to which attributes of the macrobenthic communities in this waterway are indicative of environmental perturbation and, if possible, to correlate these attributes with environmental variables that might imply cause and effect. A problem with such an exercise is establishing reference conditions against which the status of these communities can be evaluated in the future [49]. Usually, the best-scoring samples, indicating the most pristine state, are used to establish local reference conditions that act as a baseline against which spatial differences and/or temporal changes can be assessed. Several authors have argued against the use of a pristine state as a reference point against which potentially impacted sites or systems can be evaluated [50,51], and this is particularly relevant in the case of estuaries, where all sites might be impacted to some degree and no appropriate reference sites may be available. Furthermore, environmental conditions in macrotidal estuaries are highly dynamic and can vary dramatically and unpredictably at different times of the year [18].

Three of the indices used here, i.e., ABC curves, taxonomic distinctness and the phylum-level meta-analysis, adopt different approaches to setting reference conditions that do not require extensive temporal or spatial data to establish a baseline. The ABC method exploits the fact that when an assemblage is perturbed, the conservative species are less favoured in comparison with the opportunists, and the distribution of biomass among species behaves differently from the distribution of numbers of individuals among species. The three conditions (unperturbed, moderately perturbed or grossly perturbed) are recognisable without reference control samples in time or space, the two curves acting as an internal control against each other and providing a snapshot of the condition of the assemblage at any one time or place. With average taxonomic distinctness (Δ⁺) and variation in taxonomic distinctness (Λ⁺), the permutation test determines the significance of departure from expectation under specific null hypothesis conditions, i.e., that the species present are a random selection from the regional species pool, and these conditions act as a reference against which the status of samples can be assessed. In this case, the concept of spatial or temporal reference sites is replaced by the concept of a reference condition. With the phylum-level meta-analysis, the scale of perturbation is determined by comparison with the 50 samples in the training data which represent a range of types and severity of perturbation, none of which is of necessity pristine. Although all three of these methods provide some slight indications of environmental perturbation in certain years, the overriding conclusion is that the communities are in a healthy state, even in 1996, immediately after the Sea Empress oil spill. Taxonomic distinctness indices do indicate significant community stress at stations 7 and 8, both in comparison with other stations in Milford Haven and with the more extensive CSEMP data at other coastal sites in the UK, as well as in the phylum-level meta-analysis. However, reduced salinity is likely to be implicated here. For example, echinoderms, which are very sensitive to reduced salinity [52], are virtually absent from station 8. ABC curves, which are not taxonomically based, have strongly positive W-statistic values at stations 7 and 8 in all years, except for station 8 in 2010, which is still slightly positive.

The dominance or prominence of benthic “indicator species”, i.e., those that have been recorded in areas polluted or enriched by organic material such as the small opportunistic polychaetes Capitella capitata and certain spionids [25], have been widely used in the assessment of ecological condition. This principle was elevated to a much more sophisticated level by Borja et al. [53] in AZTI’s Marine Biotic Index AMBI by classifying the benthic macroinvertebrate species present into five ecological groups based on their implied sensitivity to environmental stress. However, AMBI is essentially an indicator of organic enrichment and associated reduction in the oxygenation of the sediments, properties that vary naturally and potentially confound any biotic responses to anthropogenic contamination or disturbance [54]. This is a particular problem in estuaries, where reduced and fluctuating salinities and tidal water movement scours the sediment also impose natural stresses on the fauna. Thus, Tweedley, Warwick and Potter [17] found that AMBI did not reflect levels of contamination at the 61 estuarine and coastal sites in the CSEMP study, whereas taxonomic distinctness indices were significantly correlated with contaminant loadings. Similarly, Muxika et al. [55] found that AMBI was a poor indicator for detecting the physical impacts of sediment disturbance, which might be associated with the oil and gas industry, there being no increased abundance of opportunistic species as a result, and was similarly not useful in other naturally stressed communities. Therefore, because of the estuarine environment present at the inner stations in Milford Haven, AMBI was not applied in this study, but suffice to say, none of the species listed by Pearson and Rosenberg [25] as indicators of excessive organic pollution are prominent at any station.

The disappearance of taxa considered sensitive to the effects of environmental perturbation has also been regarded as symptomatic of anthropogenic disturbance, especially oil pollution. For example, Hobbs and Smith [33] reported that the immediate effect of the Sea Empress oil spill on the benthic assemblage of the waterway was a decrease in amphipods, a common feature noted in other European coastal waters. For example, after the 1978 Amoco Cadiz oil spill in the Bay of Morlaix in the western English Channel and the 1992 Aegean Sea oil spill in the Ria de Ares and Betanzos in the northwestern Iberian Peninsula, amphipods disappeared [56,57]. In Milford Haven, amphipod populations showed clear signs of recovery within five years of the oil spill [8]. The Milford Haven data are confusing in this respect. In 1996, stations 1–7 were sampled in March, immediately following the Sea Empress oil spill on 15 February, while station 8 was sampled the following October. Amphipods comprised 6.2% of the total abundance of macrobenthos in that year, followed by 3.5% in 2008, 2.1% in 2010 and 19.3% in 2013. This suggests that amphipod abundance may be an unreliable symptom of perturbation in situations where other indicators imply quite minimal effects.

Relating community composition to potentially causal environmental variables poses a greater problem since no contemporaneous measurements of contaminants such as hydrocarbons or heavy metals were made at the faunal sampling stations. However, it is relevant to note that, in the more extensive survey of 36 stations in October 1996 by Levell, Hobbs, Smith and Law [34], salinity is the overriding factor in determining the community patterns and that sediment granulometry, polycyclic aromatic hydrocarbons and total hydrocarbons were of little importance [32]. Salinity and sediment granulometry in the present study were taken to be temporally stable and thus could not be invoked to explain any differences between the years in which the macrobenthic samples were collected. The sequential change in community composition from the outer station (1) to the inner station (8), resulting from the distributions of individual species along this transect, can, however, be accounted for in terms of these two suites of variables. The strong correlations between community composition and the environmental variables in each of the four years support the conclusion of Levell, Hobbs, Smith and Law [34] that salinity is the major driver, with sediment granulometry being of overall secondary importance. Thus, species gradually declining or increasing in abundance can be related to the gradient of reducing salinity, while those peaking at the intermediate stations may be favoured by the higher silt/clay content there. The decreasing number of species recorded in an upstream direction overall in Milford Haven matches that recorded by Ysebaert, Herman, Meire, Craeymeersch, Verbeek and Heip [23] in the macrotidal Schelde Estuary in the Netherlands and Belgium.

Although the number and total volume of oil spilled annually from ships decreased over the past four decades, catastrophic releases continue to occur from the production and transport of petroleum [29,58]. As such, the data in the current study provided a comprehensive spatial and temporal baseline against which the effects of any future environmental accidents or industrial developments can be assessed.