In depths of 155–263 m in the Gippsland basin region of south-east Australia, the studied offshore field possesses ~22 km of flowlines and umbilicals (herein termed ‘flowline’), one manifold and seven wells. Most subsea infrastructure was installed between 2006-2008 with sections of flowline trenched in 2011 ( Table 1 ) to enable trawl fisheries to re-access the area. This survey examined ROV video imagery obtained by Industry during their routine operational surveys which included imagery of nine flowlines, the manifold and three wells ( Figure 1 ). Industry ROV surveys typically operate 24-hours a day and, as such, we considered time of day in our analyses (actual time). For simplicity we refer to surveys as occurring either in the night (20:00 – 6:30) or day (6:30 – 20:00), with raw Australian Eastern Daylight Time (AEDT) retained in statistical analyses.

The purpose-built software program, TransectMeasure (SeaGIS, 2020), was used to record broad classifications of benthic habitat on structures using an adaptation of the CATAMI (Collaborative and Annotation Tools for Analysis of Marine Imagery; Althaus et al., 2015) classification scheme. An additional specialized program, EventMeasure Stereo (SeaGIS, 2020), was used to annotate mobile fauna from ROV video. Both software programs were designed specifically to allow fast, and efficient analysis of biological information from video sequences and have been used previously to analyze ROV videos (McLean et al., 2017; McLean et al., 2018; Bond et al., 2018a; McLean et al., 2019; ).

A total of 10,343 individual animals were observed and comprised 69 taxa including bony and cartilaginous fishes, mammals, and invertebrates ( Supplementary Tables 2, 3 ; Supplementary Figures 1, 2 ). There were 16 species observed of commercial fisheries interest, and three species of conservation value including, handfish (Brachionichthyidae spp.), stingaree (Urolophus spp.), and foxfish (Bodianus frenchii); although these are tentative identifications unable to be verified from the imagery obtained. Australian fur seals (Arctocephalus pusillus doriferus) were observed multiple times along flowlines in 2009, including juvenile males and adult females ( Supplementary Data Table 2 ).

More species of fish were recorded on flowlines (n = 28 species) compared to wells and manifold (n = 19 species). An average of 295 ± 45 individual fish were recorded per km of pipeline. Considering the comprehensive 2020 dataset on its own, 16 species of fish were recorded on the flowlines that were not observed on the wells or manifold, including species closely associated with the benthos e.g. banded cucumberfish (Paraulopus balteatus), ghost flathead (Hoplichthys spp.). Eight species of fish were unique to the wells and manifold which included more mobile, schooling species such as Australian sandpaper fish (Paratrachichthys macleayi) and redfish (Centroberyx affinis). Eight species were observed on both infrastructure types, including commercial fishery species such as ocean perch (Helicolenus spp.), jackass morwong (Nemadactylus macropterus) and pink ling (Genypterus blacodes).

A total of 2,066 individual invertebrates comprised of 32 species were observed on flowline surveys across all survey years ( Supplementary Data Table 2 ). Invertebrate taxa were identified from four phyla, of which Arthropoda and Cnidaria had the highest abundance. A total of 27 individual invertebrates were observed on the wells and manifold across all years and comprised seven taxa ( Supplementary Data Table 3 ). Pancake urchin (Araeosoma thetidis) was the only mobile invertebrate species recorded on both types of infrastructure.

Using the 2020 data only, forty one percent of flowlines surveyed were classed as ‘buried’. This was most prevalent on Umbilical 7 and 8, and Flowline 9 where 97- 100% was buried. Where flowlines were classified as exposed; 32% of observations were less than 50% exposed, 15% were greater than 50% exposed and 11% were completely exposed and spanning. There were no observations of any flowline with a span of greater than 0.5 m above the substrate.

Using the 2020 data only, epibenthic cover of flowlines was predominantly abiotic, i.e. sand (>80% cover) for all flowlines except flowline 3 which was a mix of sand (49%) and pebble/gravel (50%). The most prevalent biotic cover was ‘biofilm’ accounting for 3% of the total cover. Overall, all sponge classes accounted for only 0.3% of cover. Among all flowlines, most epibenthic communities were encrusting (61–91%; negligible in height), with very few of low (9–39% values) and moderate (0–4%) complexity. Infauna burrows were observed beside all flowlines, generally in low densities; between 9% (Umbilical 7) and 39% (Umbilical 4) sparse coverage (<25% cover).

For surveys prior to 2020, poor video quality had the largest impact on benthic assessments because it affected our ability to identify epibenthic organisms. As such, large amounts (49%) of historic imagery was classed as either “unidentifiable” (biota too blurry to identify; 20.3%), “not useable” (biota/flowlines not visible; 7.9%) or “open water” (where the ROV view was at an oblique angle, causing some quadrat points to be placed in the water column; 21.5%) ( Figure 2 ), thus, further analyses on these biota metrics was not undertaken. Sand was the dominant habitat scored on flowlines through the time series, but an overall increase in biofilm was also observed among flowlines over time.

Similar to historic epibenthic habitat classifications of flowlines, a substantial amount (58%) of habitat classifications on wells and the manifold for all years were classed as either “unidentifiable” (37%; video too blurry to identify), “not useable” (0.7%; benthos and flowlines not visible), “fish” (0.9%; where a quadrat point has landed on a fish and the underlying benthos cannot be identified), or “open water” (12%; video where the quadrat point is in the water column). Biofilm was the most common class of biota able to be scored, making up 32% of all classifications ( Figure 2 ). Other types of biotas that were observed in lower densities included sponges (8%; encrusting and massive), black/octocorals (6%) and bryozoans (0.5%). A greater cover of black/octocorals and massive sponge classes were observed in 2020 compared to previous years, although whether this is a true increase rather than a product of improved visibility is not known.

Epibenthic cover was predominantly biotic for all wells and dominated by biofilm for the manifold (61%), Well 2 (31%) and Well 3 (31%). Well 4 had a comparatively high proportion of black/octocorals (22%) and encrusting sponge communities (22%; Figure 2 ). Black/octocorals, bryozoans and ascidians were not observed on flowlines ( Figure 2 ).

A linear regression of total percent cover of colonizing epibenthic communities on wells and the manifold sampled for this study and reported by McLean et al. (2019) showed weak positive correlation with well and manifold age, suggesting an increase in invertebrate epibiota with increasing time in-situ ( Figure 3 ). The addition of samples from this study reduced the R² reported by McLean et al. (2019) from 0.43 to 0.29 ( Figure 3 ), noting that McLean et al. (2019) reported on tropical fish assemblages which may exhibit different patterns to those in temperate ecosystems such as those surveyed here. A stronger negative correlation existed between the depth of the well or manifold and the total cover of epibenthos ( Figure 3 ). Similarly, adding samples from this study to the model, reduced the R² reported by McLean et al. (2019) from 0.74 to 0.57.

Across the 17.8 km of flowlines assessed in 2020, 5,697 individual fish were identified from 24 taxa equating to a relative abundance of 295 ± 45 individual fish per kilometer ( Supplementary Table 4 ). The most abundant species were also the most ubiquitous (>80% of the observations) and included: Parapercis sp2 (species complex, not including P. allporti); ocean perch (Helicolenus spp.); slender sandburrower (Creedia haswelli); barred grubfish (Parapercis allporti); banded cucumberfish (Paraulopus balteatus); and trevally (Carangidae spp.). Six commercial fishery species were observed along flowlines in 2020 including the aforementioned Carangidae spp. and Helicolenus spp., along with blue grenadier (Macruronus novaezelandiae), jackass morwong (Nemadactylus macropterus), pink ling (Genypterus blacodes), tiger flathead (Platycephalus richardsoni). The relative abundance of commercial fishery species across the flowlines varied, ranging from 0.4–79.1 individuals/km, with Flowline 5 having the highest, mainly comprised of Helicolenus spp. Species of ecological interest included the rarely seen broad duckbill (Enigmapercis reducta) and handfish (anglerfish) within the family Brachionichthyidae. The majority of fish species observed on flowlines were <20 cm in size (snout to tail fork) with frequency substantially declining with increasing size classes ( Supplementary Figure 3 ). Commercial fishery species accounted for 83% of all length observations over 20 cm. For commercial fishery species, 98% of length estimates in the <20 cm class were of Helicolenus spp. ( Supplementary Figure 3 ).

Spatial distribution of total abundance and fish species richness along flowlines ( Figure 4 ) in addition to the abundance distribution of ubiquitous species ( Figure 4 ) show some patterns according to positions along flowlines. Individual species show affinity to particular sections of flowline. Parapercis allporti and C. haswelli were present in higher abundance in northern sections of flowlines, compared to flounder (Bothidae spp.) and Parapercis sp2 which had high abundance at the southern, deeper sections of Umbilical 8 and Flowline 9 ( Figure 4 ).

For the 2020 survey, a total of 1,466 mobile invertebrates from 31 taxa were identified, equating to 71 ± 12 individuals/km of flowline ( Supplementary Table 2 ). Mobile invertebrate taxa were identified from four phyla with Arthropoda and Cnidaria dominating the assemblage, in particular, hermit crabs (n = 443) and anemone sp2 (n = 328). Invertebrates of commercial importance included the Tasmanian giant crab (Pseudocarcinus gigas, n = 47), cuttlefish (Sepiidae spp., n = 93), octopus (Octopodidae spp., n = 44), arrow squid (Nototodarus gouldi, n = 1), and Balmain bug (Ibacus peronii, n = 35).

Total abundance and species richness of mobile invertebrates varied spatially, with higher abundance and diversity below 200 m depth ( Figure 8 ). Individual species show affinity to particular sections of flowline. For example, I. peronii, P. gigas, anemone sp2, Phlyctenanthus sp2 and Sepiidae spp. were observed in comparatively high abundance at the southern section of Umbilical 8 and Flowline 9 ( Figure 8 ).

Prior to 2020, the majority of ROV imagery of flowlines (75% of transects) was of poor video quality, while 25% was of moderate quality. In contrast, 100% of 2020 ROV imagery was of good quality, although the downward nature did make identification of some fish species difficult ( Supplementary Table 1 ). Only one commercial fish species was observed in historical video imagery (2014) but not also observed in 2020; the gemfish (Rexea solandri, n = 3). All others that were observed in historical imagery were observed in 2020 ( Supplementary Table 2 ).

For fish and mobile invertebrates, the same sections of five flowlines were assessed over time enabling evaluation of change to these communities ( Table 1 ; Flowline 5, Flowline 6, Umbilical 4, Umbilical 8, Flowline 9). For these repeated measures assessments, the ‘video quality’ covariable had a significant impact on assemblages on all flowlines ( Table 3 ). Even considering large differences in video quality, fish, and mobile invertebrate assemblages along two flowlines (Flowline 5 and Umbilical 8) changed across years ( Table 3 ). Assemblages observed in 2020 had higher abundance and species richness than those in previous years for Umbilical 8, while Flowline 5 only had higher abundance for fish in 2020 ( Figure 9 ), but not for species richness or for mobile invertebrate abundance or richness, with these remaining relatively constant over time ( Figure 9 ).

Separation of assemblages by Year is evident for both Flowline 5 and Umbilical 8 ( Figure 10 ), with a moderate level of distinction between years for Flowline 5 (67% allocation success rate) and a good level for Umbilical 8 (89% allocation success rate). Species correlated with the CAP axes for Flowline 5 were distinct for each year with Australian fur seal (A. pusillus doriferus), pancake urchin (A. thetidis), and P. balteatus correlated with 2009, Decapoda spp. and P. allporti correlated with 2012, Octopodidae spp. with 2017, and Helicolenus spp. with 2020. While for Umbilical 8, Sepiidae spp. were correlated with 2014, and A. thetidis with 2017. Multiple species were correlated with 2020 including Decapoda spp., squat lobsters (Galathea australiensis), Coelorinchus spp., Phlyctenanthus spp., and Parapercis sp2 for the section >1500 m along the flowline that matched with the area surveyed in 2014 and Pseudophycis spp., M. novaezelandiae, Pycnogonidae spp., and C. haswelli, in the section <1500 m along the flowline that matched with 2017.

Flowlines possessed an array of demersal fishes and mobile invertebrates known to exist in the region. However, fish assemblages recorded here represent only a small proportion (~10%) of the diversity of fishes known to occur across a diversity of habitats in the Bass Strait region (200 species; Butler et al., 2002). Species relatively common across the region but absent on ROV imagery of flowlines include dories, dogfish, frostfish, burrfish, warehou, leatherjackets, demersal shark species such as gummy, school and draughtboard sharks, sawsharks and elephantfish (Knuckey, 2006). This is perhaps not surprising as research has shown that ROV lights, sound, and speed can impact fish behavior in different ways, with variability across species (Popper, 2003; Trenkel et al, 2004; Stoner et al., 2008; Ryer et al., 2009; Schramm et al., 2020). Using industry ROV imagery of a pipeline on the north-west shelf, Bond et al. (2018c) theorized that the absence of regionally common Lethrinus punctulatus (blue-lined emperor) and Lethrinus erythropterus (crimson snapper) on surveys was likely a result of these species fleeing with the approach of the ROV. Schramm et al. (2020) found that in soft sediment habitats in north-west Australia, baited video techniques were more effective at sampling a diversity of species than ROV suggesting bait and behavior-related responses. Such a response to the ROV here may account for the absence of other species which are known to occur in similar depths across this region (Williams and Bax, 2001) or lower abundances of some species on flowlines (e.g. jackass morwong N. macropterus). Alternately it may be that the flowlines may not provide a suitable habitat for these species, or there are some other environmental factors specific to this location that do not suit their presence. We note that several of these absent species (elephantfish, draughtboard sharks, leatherjackets) were observed in low numbers around platform infrastructure in the Gippsland region to the east and in shallower waters <100 m depth (Sih et al., 2022).

Fish assemblages were surprisingly distinct on each flowline with diversity varying between 8–18 taxa per flowline (out of 24 total species in 2020), and assemblages varied even for flowlines occupying similar areas. The reasons for these spatially explicit communities could be due to a variety of factors. In this study, depth, the position of the pipeline relative to the seafloor and the percent cover of biofilm, sand and burrows were all associated with variation in the structure of marine communities. Depth is a common environmental driver in structuring marine communities (Baldwin et al., 2018; Stefanoudis et al., 2019) and can be tied to other environmental changes such as differences in temperature, dissolved oxygen content, and nutrient and plankton availability. Here, the Balmain bug (Ibacus peronii) and the whiptail (Coelorinchus spp.) both increased in abundance with increasing depth, concurring with previous research on the ecology and distribution of these species (Haddy et al., 2005; McLean et al., 2015). In general, the number of mobile invertebrate species and total abundance also increased with depth, however this relationship was not strong. I. peronii also tended to be more abundant where burrow numbers were higher although it is unknown whether the burrows observed were created by this species or not. Balmain bugs are known for spending a good proportion of their time in burrows (Faulkes, 2006). The cocky gurnard (L. modesta) increased in abundance as depth became shallower. This species is widely distributed between southern and eastern Australian waters, particularly in depths <50 m (Hyndes et al., 1999; Park et al., 2017). Parapercis allporti was also commonly observed in the space underneath the flowlines (46% of observations) and 86% of observations were recorded when the flowline was not completely buried. Recent research in tropical north-west Australia has documented higher abundances of many species where pipelines span the seafloor (gap between bottom of pipeline and the seabed) linked to species behavior and higher cover of epibenthic communities (McLean et al., 2017; McLean et al., 2020a). Here, however, there were very few spans and biotic cover was minimal.

Low epibenthic cover observed across the flowlines (such as sponges) and high proportion of biofilm and sand meant that these two latter habitats featured as variables of importance for structuring communities observed on flowlines. Slender sandburrower (C. haswelli), red cod (Pseudophycis spp.), and blue grenadier (M. novaezelandiae) were each associated with areas of higher percent cover of sand. Each of these species are sand-affiliated, C. haswelli burrows into sandy and loose gravel bottoms and Pseudophycis spp. (likely Pseudophycis bachus) feeds on fishes, cephalopods, crabs in muddy and sandy areas (Horn et al., 2012). M. novaezelandiae also occur in sandy, muddy regions, but undertake vertical migrations at night, up to within 50 m of the surface, to feed on other fishes, decapods, krill, and squid (Bulman and Blaber, 1986). Interestingly, M. novaezelandiae were observed in greatest abundance at night on flowlines when it might be expected that they would not be present due to feeding activity up in the water column. Absence of fish species at night along pipelines on the north-west shelf has been documented by Bond et al. (2018a) and predicted this to be linked to species departure to feed in surrounding areas. Here, it seemed that nocturnally active species were observed in higher abundance on night surveys, including hermit crabs (Decapoda spp.), sea spiders (Pycnogonidae sp.), several Parapercis (grubfish) species, the slender sand burrower (C. haswelli), banded cucumberfish (P. balteatus) and red cod (Pseudophycis spp.). This has clear implications for the future timing of ROV surveys of marine communities around infrastructure. Based on these results, and those of Bond et al. (2018a), we predict that many fish and mobile invertebrate species will exhibit diurnal patterns of habitat (infrastructure) usage. Knowledge of these patterns will facilitate improved understanding of residency rates around structures (required to inform estimates of production on structures) in addition to nutrient transfer among infrastructure and surrounding ecosystems, i.e., how connected are artificial and natural ecosystems and how is this influenced by diurnal movements? Additional diurnal investigations are needed to better understand the role that infrastructure plays in the day-to-day behaviors of fish and invertebrate species and how this might differ between tropical and temperate ecosystems.

A number of commercial fishery species were observed along flowlines with the most numerically abundant in 2020 including octopus (Octopodidae spp; n = 46), Tasmanian giant crab (P. gigas; 47), cuttlefish (Sepiidae spp.; 93), Balmain bug (I. peronii; 35), ocean perch (Helicolenus spp.; 1207), trevally (Carangidae spp.; 335) and blue grenadier (M. novaezelandiae; 104). The most abundant commercial fish, Helicolenus spp. could be either H. barathi (bigeye ocean perch) or H. percoides (reef ocean perch) however the two could not be reliably distinguished from imagery. Helicolenus spp. comprised the majority of commercial taxa size estimates on flowlines (98% in the <20 cm range, n = 1127) with few individuals (n = 78) 20-30 cm in length. Females of both H. barathri and H. percoides are believed to reach sexual maturity at 15-20 cm (5 years) while males mature at 19-25 cm (5-7 years) (Paul and Francis, 2002 cited from Morrison et al., 2014). It is therefore quite likely that a high proportion of individuals observed here are not sexually mature and therefore not available to the fishery which take individuals predominantly in the 20-40 cm size range in this region (AFMA, 2020). Helicolenus spp. were also among the most common and abundant commercially fished species observed on pipelines to the east of this study in the Gippsland region (McLean et al., 2022; Sih et al., 2022).

The phenomena of pipelines acting as ‘corridors’ has recently been noted for various species around the globe. Previous research has shown that benthic foraging marine mammals and seabirds benefit from seafloor infra-structure as they act as artificial reefs which provide habitat for prey species. Australian fur seals (A. pusillus doriferus) have previously been observed to exhibit foraging behavior along pipelines in the Bass Strait (Arnould et al., 2015), with similar behavior observed in other species in the North Sea (Todd et al., 2016) around windfarm turbine piles and pipelines (Russell et al., 2014) and platforms (Todd et al., 2020). A. pusillus doriferus forage on a large variety of prey observed on the infrastructure (see list of 40-60 species in Kirkwood et al., 2008), but here were observed chasing schools of redbait (E. nitidus). Observations of seals only occurred in 2009 with the time of year of image capture (October) coinciding with heightened foraging activity before breeding season. A. pusillus doriferus are protected under the Environment Protection and Biodiversity Conservation Act 1999 (EPBC Act) and represent a significant predator biomass in south-eastern Australia (Kirkwood et al., 2010). In Arnould et al. (2015) pipelines and cable (electricity and telephone) routes were the most visited and most influential structures associated with foraging locations despite such features having limited vertical scope and habitat complexity (and, thus, diversity in prey habitat) in comparison to wells and shipwrecks. They hypothesized, pipelines/cable routes may represent greater overall area and provide habitat connectivity for prey species potentially making them more profitable sites to exploit.

Two additional species of conservation value were observed on the seafloor above buried flowlines (Brachionichthyidae spp., handfish, and Urolophus spp., stingaree). The poorly studied handfish are considered the most threatened of all marine bony fishes, especially at greater depths (Stuart-Smith et al., 2020). Seven of 14 species were recently listed as Critically Endangered or Endangered by the International Union for Conservation of Nature (IUCN) and this family has the only marine bony fish to be recognized as Extinct – the Smooth handfish (Sympterichthys unipennis) (Stuart-Smith et al., 2020). Imagery obtained of handfish was not sufficient to accurately identify to species level the 10 individuals observed; therefore, their conservation listing is not known. The unidentified stingaree is likely one of three species of Urolophus recorded from the region in a previous study, of which two are listed as Vulnerable (IUCN), wide stingaree (Urolophus expansus) and greenback stingaree Urolophus viridis (Knuckey, 2006) and one listed as least concern (banded stingaree Urolophus cruciatus).

In general, wells had a less diverse community of fishes than flowlines but possessed a high density of individuals per occupancy space. Topographic complexity that the structures provide is known to be a major surrogate for diversity and biomass in previous studies (Wedding et al., 2008; Wines et al., 2020). Areas of high complexity are also commonly related to higher fish abundance and species richness than less complex areas (Galaiduk et al., 2017). The Christmas tree and tree cap assembly were structurally complex and possessed the highest abundances of fish but were also sections where the ROV spent proportionally more time. The most abundant fish observed on wells was the Australian sandpaper fish (P. macleayi), a slimehead endemic to south-eastern Australia. Little is known about this species, with no published scientific works describing its ecology. We noted 144 individuals on a single well (conservative MaxN estimate) which is a higher density than any fish species observed on well infrastructure on the north-west shelf (McLean et al., 2018). While the species does not feature in commercial catches for the region, it is in the same family as highly targeted roughies (Trachichthyidae) (Knuckey, 2006).

Of the 16 fish species observed across the structures in 2020, six are retained by commercial fisheries operating in this region (Boag and Koopman, 2021); redfish (C. affinis, n = 3), redbait (E. nitidus; 164), pink ling (G. blacodes; 1), ocean perch (Helicolenus spp.; 14), striped trumpeter (L. lineata; 1) and jackass morwong (N. macropterus; 140). While E. nitidis was observed in high numbers, this was due to the presence of one school only on Manifold 1. The jackass morwong, N. macropterus, is a commercially important groundfish that is common in coastal and continental shelf waters of southern Australia and New Zealand commonly caught in depths from 80–170 meters with major nursery grounds for this species identified in Bass Strait (Bruce et al., 2001). Structural complexity of habitat has been found to be an important in shaping fish assemblages (Wedding et al., 2008) with the vertical relief and complexity of the wells and manifold likely to provide food and shelter from predators in an area of seafloor where we would expect low relief in the majority of surrounding natural habitat.

Similar epibenthic communities were observed on each well when first surveyed in 2009, but there was subsequently some variation through the years. Time series data showed more complex habitat forms in subsequent years (e.g. black/octocorals on Well 4 in 2014 and 2020). This suggests that this benthic community has recruited to and developed on this structure over time, aligning with previous research by McLean et al. (2018) that documents a strong positive relationship between the cover of epibenthic communities and well age. Here, this relationship was comparatively weak, which could be due to variations in sampling effort, depth of structures and/or reflect a different pattern of establishment for temperate ecosystems and locality such as oceanographic parameters influencing recruitment. Further research is required to assess how epibenthic communities on infrastructure in the Bass Strait establish and change through time.

Industry would benefit from incorporating HD scientific surveys (e.g. stereo-ROV) into offshore ROV campaigns (see McLean et al., 2019) where quantitative data can be obtained in a systematic way to better inform decommissioning.

Advances in our understanding of the impacts of oil and gas infrastructure in marine ecosystems will require dedicated science programs. For ecology, this includes research to examine:

There is also an opportunity to increase the potential benefits of industry ROV data collection for biodiversity assessments. However, it is acknowledged that they would need to be taken into consideration regarding their primary use for inspection, maintenance, and repair campaigns. This would include: