We investigated the amphipod communities of two oceanic basins (Fig. 1): (i) the Clarion–Clipperton fracture zone (CCZ, 6×102 km2, 7000 km wide), an economically important manganese nodule field in the NE Pacific, comprising several different contractor licence areas (who to date, have exploration licences only), and nine designated Areas of Particular Ecological Interest (APEIs) as designated by the International Seabed Authority (ISA) (Lodge et al., 2014), and (ii) the DISturbance and re-COLonisation (DisCOL) experimental area (DEA, 11 km2, 4 km wide), a mining disturbance proxy (also containing manganese nodules) in the Peru Basin in the SE Pacific. In 1989, the DEA sediment bed was artificially disturbed using a plough harrow to create 78 track marks (Appendix A, Sect. A1) (Thiel, 1992). These are supposed to simulate the type of disruption that would be caused by a commercial mining operation. This baseline study was a new approach in deep-sea risk assessment and is still ongoing today, providing us with crucial data from this long-term ecological experiment.

In 2015 (26 years after the first impact in the DEA in 1989), two research expeditions with the RV Sonne visited the CCZ (cruise SO239) and revisited the DEA (cruise SO242-1 and SO242-2) to assess if and how the deep-sea faunal communities had recovered within the DEA and to attempt to quantify their recolonization potential.

Amphipod samples were taken from the CCZ and DEA using a free-fall lander (120 cm × 120 cm × 120 cm), to which four plastic traps were attached (two 20 cm × 25 cm × 40 cm traps with 4 cm openings and two 25 cm × 40 cm × 60 cm traps with 8 cm openings), baited for each station with an 800 g mixture of mackerel, squid and shrimp. Using this specially designed deep-sea sampling equipment, more than 60 000 specimens of scavenging amphipods were collected from the CCZ and the DEA sites.

The baited trap was deployed eight times across the CCZ at a depth range of 4116–4932 m (samples C1–C8) and five times in the DEA at a depth range of 4078–4307 m (samples D1–D5; Fig. 1, Table 1). In the CCZ, we sampled within four different contractor licence areas (Table 1) to obtain a pre-disturbance baseline and to then compare it with one of the nine protected APEIs around the CCZ. In contrast, in the DEA sampling was conducted once within the disturbed area (D1), twice 10 km away (D2, D3) and twice 40 km away (D4, D5) from D1 in four surrounding reference areas (see Fig. 1).

On recovery of the lander, all traps were disconnected and placed in pre-cooled (4 ∘C) buckets of filtered seawater. All specimens were washed on board in a dedicated cool-climate laboratory (4 ∘C), morphologically pre-sorted and fixed in molecular grade (95 %) ethanol, before being stored at -20 ∘C.

Detailed sorting and identifications were performed using the morphological species concept (Futuyma, 1998) and the keys of Schulenberger and Barnard (1976) and Barnard and Karaman (1991) to separate the samples into taxonomic “morphotypes”. The larger fraction (>15 mm length) has been identified to the lowest possible taxonomic resolution. Species not assigned with certainty are denoted here by as affiliated species (e.g. genus aff. species) or conferred species (e.g. genus cf. species).

Of the 60 000 specimens, those with a size of less than 15 mm in length were excluded from the analysis because these were mostly juveniles and their morphological differences were not sufficiently pronounced to allow an accurate identification to the species or even genus level. Some pelagic amphipods were collected accidentally and omitted. Finally, genera containing multiple (and as yet) unidentified species have been summarized as “spp.”.

Our null hypothesis (H0) here is that there are no differences in the amphipod biodiversity of the two basins. To test this hypothesis, we firstly calculated the alpha biodiversity of the two basins using the Simpson index (D) (Simpson, 1949) (Sect. A4). Rarefaction was run on the longer than 15 mm fraction using default parameters. Individual-based curves were generated using the rarefy function of the “vegan” package in R 2.3.0 (Gotelli and Colwell, 2001; R Core Team, 2013) to compare species richness across all 13 sampling stations and to test for the completeness of sampling.

Secondly, to compare the beta biodiversity, we estimated the variability of the community compositions between sites. The Bray–Curtis dissimilarity metric (Bray and Curtis, 1957) was used to calculate differences between community compositions based on species densities, and the results were then visualized in 2-D using a non-metric dimensional scaling (NMDS) plot. The ANOSIM function in the vegan package of R (R Core Team, 2013; Taguchi and Oono, 2005) was used to test the statistical significance of the differences in species compositions between the two study areas.

In total, 6916 scavenging amphipods (>15 mm) were collected from the 13 trap deployments in the two study areas, representing a total of 17 morphotypes (Fig. 2). In the CCZ, we collected 3932 individuals, which represent 10 morphotypes. Five of these have been identified to the species level: Abyssorchomene distinctus (Birstein and Vinogradov, 1960), Abyssorchomene gerulicorbis (Schulenberger and Barnard, 1976), Eurythenes sigmiferus (d'Udekem d'Acoz and Havermans, 2015), Paralicella caperesca (Schulenberger and Barnard, 1976) and Paralicella tenuipes Chevreux, 1908. Two are affiliated to a species (Paracallisoma aff. alberti and Valettietta cf. gracilis) and the remaining three are at least affiliated to a genus (Table 2). The 2984 individuals from the DEA represent 15 morphotypes. Six of these have been identified to the species level: Abyssorchomene distinctus (Birstein and Vinogradov, 1960), Abyssorchomene gerulicorbis (Schulenberger and Barnard, 1976), Eurythenes sigmiferus (d'Udekem d'Acoz and Havermans, 2015), Paralicella caperesca (Schulenberger and Barnard, 1976), Parandaniexis mirabilis Schellenberg, 1929, and Tectovalopsis regelatus Barnard and Ingram, 1990. A further five which have been affiliated with a species include Eurythenes aff. gryllus, Eurythenes aff. magellanicus, Paracallisoma aff. alberti, Stephonyx sp. nov. aff. arabiensis and Valettietta cf. gracilis. The remaining four have been identified to at least an affiliated genus (Table 2).

There are eight morphotypes shared between the basins: Abyssorchomene distinctus (Birstein and Vinogradov, 1960), Abyssorchomene gerulicorbis (Schulenberger and Barnard, 1976), Abyssorchomene spp., Eurythenes sigmiferus (d'Udekem d'Acoz and Havermans, 2015), Eurythenes spp., Paracallisoma aff. alberti, Paralicella caperesca (Schulenberger and Barnard, 1976) and Valettietta cf. gracilis (Fig. 2).

Two morphotypes were found only in the CCZ (Hirondellea sp. and Paralicella tenuipes Chevreux, 1908, and seven morphotypes were found only in the DEA (Eurythenes aff. gryllus, Eurythenes aff. magellanicus, gen. aff. Cleonardo, Parandania sp., Parandaniexis mirabilis Schellenberg, 1929, Stephonyx sp. nov. aff. arabiensis, and Tectovalopsis regelatus Barnard and Ingram, 1990 (Table 2).

Due to differences in allocated ship times (the CCZ cruise being 52 d and the DEA cruise being 29 d), the trap deployments were not identical, making it necessary to check the effect of the different deployment times. The resulting catch-per-unit-effort (CPUE) plot (Fig. 3) shows that there is no statistically significant correlation between the length of time the trap was at the seafloor and total number of amphipods caught (p=0.551).

The rarefaction results (Fig. 4) show that the curves for nine stations reach a plateau, indicating that sampling effort was sufficient to assess diversity levels. These include all CCZ stations except C7. In contrast, four of the five curves for the DEA (stations D1, D2, D4 and D5) are unsaturated. A higher number of different species were collected at D1 and D2; however, many of these were singletons or doubletons, with A. gerulicorbis dominating at both the disturbed station and the station closest to the disturbed area (D2).

Figure 5a and b show that the scavenging community in the CCZ is dominated by three species, A. distinctus (Birstein and Vinogradov, 1960) (36 %), A. gerulicorbis (Schulenberger and Barnard, 1976) (18 %) and Paralicella caperesca (Schulenberger and Barnard, 1976) (31 %), whereas, in contrast, the DEA scavenging community is dominated by a single species, A. gerulicorbis (Schulenberger and Barnard, 1976), accounting for almost 60 % of all specimens. The Simpson index (D) for the entire CCZ area is (with 0.73) higher than the 0.62 that was calculated for the whole of the DEA area (Table 3). The biodiversity of each individual station was further explored (Table 3). In the CCZ, the lowest biodiversity was found at C3 and C6 (D=0.23) and the highest at C2 (D=0.67). In the DEA, the lowest biodiversity of D=0.36 was found at station D1 (the site of the actual disturbance) and just south of the disturbance site at D2 (0.21), while the highest biodiversity was observed at D5 (D=0.61) (Table 3).

The NMDS shows that the communities of the two basins are dissimilar (ANOSIM: p=0.002); Fig. 6). The disturbed area in the DEA (D1) shows a clear difference from the four reference areas (D2–5). When the communities between the two basins are compared, D2 appears to be most similar to the CCZ community, and more specifically similar to C6, C7 and C8. The reliability of the data ranking is supported by a low stress value of 0.01.

Although the most recent and comprehensive analysis of the animal diversity of the world's oceans estimates a total of less than a million species over all depths (Appeltans et al., 2012), it is not currently known how many species inhabit the deep sea. Over 7000 marine amphipod species have been found below 2000 m. These numbers are reduced to 173 known species, 87 genera and 37 families at depths below 3000 m, and 100 known species, 66 genera and 31 families are known to occur below 4000 m (Vader, 2005; Brandt et al., 2012).

The superfamily Lysianassoidea constitutes an important part of the abyssal amphipod fauna. Also, in our sampling, lysianassoid amphipods were collected in large numbers (99 % of the samples taken in both basins). As a superfamily, they comprise 23 % of all the species found below 2000 m, 35 % of the species found below 3000 m and 31 % of the species found below 4000 m (Brandt et al., 2012).

Many species in the Lysianassoidea occur in multiple abyssal basins and some even have worldwide distributions (Thurston, 1990). Despite the Ocean Biogeographic Information System (OBIS) database containing 615 650 records of Amphipoda, many of these are shelf or pelagic species, with very few records from the CCZ and DEA (OBIS, 2019). Here, we provide additional data for the known bathymetric range of the seven amphipods, which we have identified to species level: Abyssorchomene distinctus (Birstein and Vinogradov, 1960), Abyssorchomene gerulicorbis (Schulenberger and Barnard, 1976), Eurythenes sigmiferus (d'Udekem d'Acoz and Havermans, 2015), Paralicella caperesca (Schulenberger and Barnard, 1976), Paralicella tenuipes Chevreux, 1908, Parandaniexis mirabilis Schellenberg, 1929, and Tectovalopsis regelatus Barnard and Ingram, 1990 (Table 2). In addition, we have found two possibly new species of Eurythenes, previously not known from these basins, which we plan to analyse further in the future.

While we only sampled Hirondellea sp. and Paralicella tenuipes Chevreux, 1908, in the CCZ, Eurythenes aff. gryllus, Eurythenes aff. magellanicus, gen. aff. Cleonardo, Parandania sp., Parandaniexis mirabilis Schellenberg, 1929, Stephonyx sp. nov. aff. arabiensis, and Tectovalopsis regelatus Barnard and Ingram, 1990, only in the DEA, we cannot conclude based on the current data only if these species are unique to their respective basins without confirming these distribution patterns with additional sampling campaigns.

Despite the sampling campaign in the CCZ being twice as long as the DEA, the number of individuals and species collected does not correlate positively with deployment effort. We posit that this is rather an effect of abiotic and organic factors, such as the productivity-driven gradients in the CCZ, which decrease from east to west and from north to south (Hannides and Smith, 2003), and also the productivity differences between both basins.

Figure 5a and b clearly show that the DEA scavenging community has reduced abundances of all species, including A. distinctus (Birstein and Vinogradov, 1960) (1 %) and P. caperesca (Schulenberger and Barnard, 1976) (7 %), and is now dominated by a single species, A. gerulicorbis (Schulenberger and Barnard, 1976), accounting for 60 % of the DEA community. This indicates an interesting resilience and flexibility in the latter species. Similar patterns have been observed in Potter Cove (Seefeldt et al., 2018), where, following glacial retreat, a change in sedimentation rates led to the dominance of a single scavenging amphipod species, Cheirimedon femoratus (Pfeffer, 1888).

The assemblages of the two basins have some overlap in their biodiversity (as is exemplified by the eight shared morphotypes). However, the sampling stations and the two basins as a whole are heterogeneous in their species compositions.

Thus, we can observe some negative influence (possibly attributed to the disturbance in the DEA) on the scavenging amphipod community. This reduced biodiversity is reflected in the higher Simpson index (D) for the CCZ (0.73) as compared to the DEA (D=0.62; Table 3).

To explore whether this reduced diversity in the DEA was truly an artefact of the simulated disturbance, D was also calculated for each sampling station within each basin (Table 3).

In the CCZ, the APEI (C8) shows a moderate level of biodiversity (D=0.44), indicating that it is not optimally placed for representing the biodiversity of the scavenging amphipod community of the CCZ. Additionally, this pre-existing lower biodiversity (in comparison to the contractor licence areas), indicates that the APEI may not serve well as a refugium for amphipods post-disturbance. However, due to the fact that only one of the nine APEIs has been investigated thus far, this APEI along with the remaining eight would need to be sampled.

Within the DEA, the lowest biodiversities are calculated at the site of the disturbance (D1) and south of it (D2; Table 3), indicating that the reduced biodiversity in the DEA could indeed be caused by the simulated disturbance in 1989 (Thiel, 1992).

The highest abundances in the DEA were collected from station D5 (n=717); this station also has the highest Simpson index within the DEA (D=0.61) (Table 3). Bathymetric imaging shows a seamount range to the northwest (NW) of the disturbed area (D1) (Sect. A3). Although the relief change is only 150 m, the range extends laterally for several kilometres, hampering dispersal across barriers such as sills, canyons and ridges (Smith et al., 2006; Blankenship et al., 2006; Etter et al., 2011). However, recent studies have shown that due to their mobile nature, geographic isolation alone would not pose a true barrier to bentho-pelagic amphipod species (Havermans, 2016; Ritchie et al., 2016) and thus cannot explain why such a high number of large scavenging individuals was collected at station D5.

Scavenging amphipods are resilient and dispersive, but most importantly they are highly mobile (Ingram and Hessler, 1983; Lörz et al., 2018). Often driven by their search for erratically deposited feeding opportunities (Smith et al., 1989), they are probably less constrained by local environmental abiotic conditions and seafloor topography.

Beta diversity can be regarded as the dissimilarities in species composition between spatially different communities. As an indication for beta biodiversity, the NMDS (Fig. 6) shows a significant separation in the similarity index between the two basins (ANOSIM p=0.002). However, despite the dispersive and resilient nature of scavenging amphipods, their biodiversity could have been affected by the disturbance experiment, as evidenced by the NMDS (Fig. 6), where the disturbed area (D1) and the area closest to it (D2) show a different Bray–Curtis index to the remaining three reference sites (D3, D4 and D5).

In the CCZ, stations C1, C2, C3, C4 and C5 show a different Bray–Curtis index in comparison to stations C6, C7 and C8 (Fig. 6). The CCZ is a geomorphologically very heterogeneous region, with seamounts of 200 m altitude running from north to south. A barrier of this height would be expected to affect sedimentation rates, nodule presence and currents. Furthermore, the difference in depth from the eastern edge (3950 m) and the western edge (5150 m) is more than 1200 m. These combined factors very likely give rise to different trends in species composition (Glover et al., 2016). However, since it has been established that bentho-pelagic amphipods are less sensitive to such barriers (Havermans, 2016), at this stage other biotic (e.g. the productivity gradient) and abiotic factors causing this separation cannot be excluded as alternative explanations.

Whilst the NMDS (Fig. 6) illustrates a visual separation of the two basins, there is also some similarity in the amphipod fauna between the two areas (as is obvious by the eight shared species), indicating that the dispersal extent for these eight species might be up to at least 3000 km. However, this hypothesis will need to be confirmed with subsequent molecular analyses.

Abyssal amphipods have been shown to be able to travel actively at speeds of almost 4 cm s-1 (Laver et al., 1985), even at temperatures as low as 3 ∘C (Kankaanpää et al., 1995). It is obvious that they are sufficiently strong to swim up weak currents since they can be found several hundred metres above the seafloor searching pelagically for mates (e.g. Eurythenes gryllus occurring up to 1800 m above the seafloor) (Thurston, 1990; Havermans et al., 2013) or following food falls (Baldwin and Smith Jr., 1987).

However, it is apparent that the dispersal of abyssal amphipods is not always contingent on current direction but on passive dispersal. Amphipods can also be carried passively over long distances by strong currents (e.g. the circumpolar current of the Southern Ocean) (Laver et al., 1985), but even weaker deep-sea currents have been suggested as a mechanism for deep-sea dispersal of amphipods (e.g. Eurythenes gryllus; Schüller and Ebbe, 2007). This, coupled with their ability to follow odour plumes (Ide et al., 2006; Premke et al., 2003), significantly increases the probability and extent of their dispersal (Conlan, 1991; Highsmith, 1985). Specialist feeding adaptations for several species found in our assemblages have been reported in Havermans and Smetacek (2018). For example, the semi-tubular flap-like molars of the genera Hirondellea and Eurythenes and the distendible foregut of Abyssorchomene and midgut of Paralicella. It is not clear from our study in the absence of particulate organic carbon (POC) data for the areas of the trap deployments whether the biogeography of these specialized feeders is linked to the productivity gradients in these two basins. The lack of a clear dispersal pattern is obvious from Fig. 6, where station D2 is the station plotted closest with the CCZ basin in terms of species composition despite station D5 being geographically the shortest distance away from the CCZ.

Recent research on Eurythenes gryllus has demonstrated that it thrives in every ocean with a wide (pelagic–hadal) depth range. However, assumptions that individual morphotypes of this species belong to the same genetic lineage have been challenged (Havermans et al., 2013; Havermans, 2016). Ritchie et al. (2016) demonstrated with microsatellite markers heterozygote deficiency in Paralicella tenuipes Chevreux, 1908, which they attributed to cryptic diversity. It is likely that some of the eight shared species between the two basins are cryptic species, which will be tested by ongoing molecular research.

Unfortunately, in the absence of data on deep-sea currents in the study area, especially between the CCZ and DEA, it is not yet possible to fully explain the drivers and mechanisms of amphipod dispersal between these particular basins.

Higher abundances of scavenging amphipods were collected from the CCZ (3932 individuals) as opposed to the DEA (2984 individuals). Yet, we have identified more morphotypes in the DEA (15) than in the CCZ (10), indicating that the DEA is more speciose and thus more biodiverse.

However, although the DEA is more speciose, many of its morphotypes were collected in low abundances, with several of these being singletons or doubletons (collected from one or two sampling stations only). This is reflected in the rarefaction curves (Fig. 4), which indicate thorough sampling in the CCZ with all but station C7 reaching asymptotes. In contrast, four stations in the DEA (D1, D2, D4 and D5) are unsaturated. This pattern suggests first that the less abundant species that are present at fewer stations may not necessarily be rare species and second that there could be as of yet undetected biodiversity in the DEA.

Our preliminary (basin-scale) comparison of the scavenging communities of the two study areas shows that even if the DEA is a small-scale disturbance experiment, it is a very diverse area. Thus, the DEA is a well-chosen site for monitoring the impacts of disturbance and instrumental in its role as a proxy to assess impending mining activities in the CCZ.

At several stations in both basins, we collected amphipods in very high abundances (C1, C8, D3 and D5) (Table 2). Since biotic production is contingent on the sinking flux of particles from the euphotic zone (Sweetman et al., 2017), the biodiversity differences at each of the 13 stations could be driven by POC or erratic whale falls (Smith et al., 1989). However, not all feeding behaviour of scavenging amphipods is based on opportunistic or erratic availability of nutrients (Havermans and Smetacek, 2018). During future sampling campaigns, the POC of these amphipod sampling areas should be monitored, along with experiments on different types of food fall in addition to obtaining side-scan sonar and abiotic data. This will provide a more comprehensive view of the food types required for these species to thrive in the deep sea.

It is not clear from our results whether substrate type (i.e. nodule or non-nodule) has any effect on the amphipod communities (Smith and Demopoulos, 2003) since these kinds of data are only available for stations D3 and D4. To answer this question, resampling of the study areas in combination with an ocean floor observation system (OFOBS) (video or camera) is required.

Although our study only addresses the scavenging amphipod species longer than 15 mm, we have already found indications of a possible disturbance effect in the DEA. It is obvious that scavenging amphipods are only one of several bentho-pelagic impact-indicator groups. Other truly benthic groups such as sponges or less dispersive amphipods (e.g. collected by the EpiBenthic Sledge, EBS) may demonstrate an even more pronounced impact of mining activities and should be investigated in future studies.

With the application of molecular techniques to identify cryptic species (Delić et al., 2017), more realistic estimates of biodiversity can be obtained (Schön et al., 2012), improving our current knowledge of the biodiversity of this area. If these improved estimates of biodiversity also include cryptic species, it is possible that the biological impact of manganese nodule mining on amphipod and other deep-sea faunal communities may turn out to be even higher.