Historical changes in marine communities uncovered in diverse data sources highlight impacts over half a century

The bulletin and dive log extracts

We extracted all references to species at Shiprock from Bulletins and dive logs from 1965 to 2013. These were tabulated and included metadata for date, article, diver, species common name, species scientific name, relative abundance category and descriptive information such as habitat. Abundance categories were as follows: presence, absence, one, few (2–10), many (11–100) and abundant (101+).

Scientific names were checked, verified and updated where necessary by using WoRMS (see www.marinespecies.org, accessed 9 June 2022) and Fishes of Australia (see https://fishesofaustralia.net.au/, accessed 9 June 2022). Species names required considerable interpretation and translation. Organisms were historically described by a blend of common and scientific names; for example, ‘Telesto’ referred to the soft coral Telesto smithii, which is abundant at Shiprock, now most likely Carijoa sp. (see https://www.inaturalist.org/observations/8199698, accessed 10 June 2022) and possibly invasive (Concepcion et al. 2010). ‘Sea pens’ most likely referred to Cavernularia sp., an octocoral that is common in the sand around Sydney, but which may still be undescribed (see https://www.inaturalist.org/observations/7015860, accessed 1 June 2022), in the family Veretillidae. Charonia rubicunda was a marine gastropod, now C. lampas, and Ritterella was a compound ascidian, most likely R. tokioka (Fig. 2).

Reef Life Surveys (RLS)

RLS is a global program that gathers underwater visual census data by using highly trained volunteers (Edgar and Stuart-Smith 2014). At the start of our study, an RLS site had already been established at Shiprock, with surveys being conducted in 2010, 2015 and early 2017 at depths between 8 and 10 m. A further three surveys were conducted by RLS volunteers (including author J. Turnbull) during our study, in 2018, 2019 and 2020. RLS transects run along a constant depth contour parallel to shore, not down from the shallows as in the perpendicular surveys, and avoid sand under the RLS method (Edgar and Stuart-Smith 2014).

Perpendicular transect surveys

Clarrie Lawler (1998) reported using surveys along two transects, A and B, running perpendicular to shore and down the Shiprock depth gradient (Fig. 3). The A transect ran from 1- to 16-m depth over a distance of 46 m, and the B transect ran from 1- to 14-m depth over a distance of 23 m. Species identifications for Lawler (1998) were confirmed in collaboration with the Australian Museum, often by physical samples collected from Shiprock (C. Lawler, pers. comm.). Whereas the survey method was structured, the data available for analysis were unstructured, reported in prose in the Wetlands article (Lawler 1998).

We replicated the original perpendicular transect methods by adapting RLS methods to vertical, rather than along-contour transects. These comprised fish blocks 5 m wide on either side of the transect tape, invertebrate and cryptic fish blocks 1 m wide on either side of the tape, and photo quadrats perpendicular to the substrate every 1–2 m, heading down the historical transect lines instead of across the wall face.

In total, 21 perpendicular transect surveys were conducted by scientists and volunteers between 2017 and 2019 (n = 12 by the authors of this paper; and n = 9 by URG + RLS surveyors), spanning all four seasons in most years. We captured a total of 856 photo quadrats of sufficient quality to enable the classification of sessile organisms during these surveys. Photo quadrats were classified as top (above the vertical wall lip), vertical wall or bottom (beyond the base of the wall, on transect A only). These were annotated in CoralNet (see https://coralnet.ucsd.edu/source/; Beijbom et al. 2012) with random placement of 25 annotation points, by using CATAMI classification to the morphotaxon level (Althaus et al. 2013).

iNaturalist

iNaturalist (see www.inaturalist.org) is a global citizen science platform hosted by the California Academy of Sciences and National Geographic Society (Mesaglio and Callaghan 2021). On 1 October 2017, we created the Revisiting Shiprock project on iNaturalist to encourage citizen scientists to upload their sightings at Shiprock. The project was open to members of the public, with a simple description of ‘Marine species found at Shiprock, Port Hacking, Sydney, Australia’ and a single rule that sightings must be from Shiprock, as denoted by a pin placed on the map. Almost 3 years later, on 21 September 2020, we extracted all research-grade records from iNaturalist, downloading 2358 records with metadata, including date of observation, date uploaded, user, place, species and common name.

Analysis

We classified all fish and mobile invertebrates at the species level, where possible. We analysed sessile invertebrates and most algae at the morphotaxa level because of difficulties in identification without physical samples, the taking of which is highly restricted in the Shiprock sanctuary zone. Owing to the sparse nature of data in the 1970s, 1990s and 2010s, and the relatively high abundance of data in the 1960s and 2010s, we aggregated our data into decades for analysis purposes. Our primary analysis focused on the changing presence of species, owing to shortage of specific quantitative abundance data and formal sampling structure in the historic records. Availability of categorical abundance data (such as a single sighting, few, many) did allow some comparison of relative abundances over time. Species were excluded from our analysis if they were (i) unable to be identified under a contemporary scientific name, (ii) difficult to distinguish visually from other species, (iii) highly cryptic and so unlikely to be noticed, or (iv) commonplace and therefore unlikely to be noted in Bulletins, dive logs or the 1998 paper.

These exclusions are detailed in Supplementary Table S1.

To identify species that have either increased or decreased over the study period, we used a ‘losers’ v. ‘winners’ categorisation, comparing the relatively data-rich decades of the 1960s and 2010s, that we considered appropriate given the limitations of our data. The winners v. losers approach has been applied in prior studies of pressures on marine systems (for example, Castro et al. 2021; Smith et al. 2021). We categorised losers as species that were recorded prior to the 1980s, i.e. including species found in the 1960s perpendicular surveys or 1960s and 1970s Bulletins and dive logs, but not recorded since. The 1980s were chosen as the delineator in our study because this period corresponds with a gap in the historical record at Shiprock and is approximately halfway in our historical timeline. Conveniently, it is also the period in which the Shiprock Aquatic Reserve was established. We categorised winners as species that were recorded after 1980, i.e. in the 1990s and 2000s dive logs, 2010–20 Reef Life surveys, 2017–19 perpendicular surveys or 2017–20 iNaturalist sightings, but with no corresponding records prior to the 1980s.

We chose selected statistical analyses where these were viable given the historical data limitations. These included Chi-Square tests comparing the 1960s to the 2010s for winners and losers, fished v. not fished species and northern, north + south and southern species, multivariate analysis of the community at Shiprock over the decades, and univariate analysis of the data from perpendicular surveys in the 1960s and 2010s, which we considered methodologically comparable. Where patterns were detected that were not significant, we applied power analysis using the R package pwr (ver. 1.3-0, S. Champely, see https://github.com/heliosdrm/pwr) to assess the detectable effect size. Multivariate analysis was conducted in R with the vegan package (ver. 2.6-10, J. Oksanen et al., see https://CRAN.R-project.org/package=vegan), using non-metric multi-dimensional scaling plots based on Gower similarity matrices. These data were not pre-treated with a transformation because our abundance scale of 0–4 already represented an approximation of a log transformation. Blank data in the periods of the 1970s, 1990s and 2000s were treated as missing, rather than zeros, owing to the lack of absence data in these periods.

Univariate analysis was conducted in R (ver. 4.4.1, R Foundation for Statistical Computing, Vienna, Austria, see https://www.R-project.org/) using the lme4 package (ver. 1.1–33, see https://CRAN.R-project.org/package=lme4; Bates et al. 2015) for mixed-effects modelling and DHARMa (ver. 0.4.6, F. Hartig, see https://cran.r-project.org/package=DHARMa) for verification of model assumptions. We investigated the interaction between period and range, and period and fished status, as predictors for species abundance, with a random effect of species, and the Poisson distribution. Range was determined from species distributions at https://reeflifesurvey.com/species/ (accessed 10 June 2022), categorising species with a distribution ranging from NSW northward as North, species ranging southward as South, and species limited to NSW or with a range both north and south of NSW as North + South. Fished status was set to ‘yes’ for species listed in the NSW Saltwater Fishing Guide (see https://www.dpi.nsw.gov.au/fishing/recreational/fishing-rules-and-regs/saltwater-recreational-fishing-guide, accessed 10 June 2022)

To compare the two modes of citizen science data collection, namely, structured surveys v. unstructured opportunistic sightings, we analysed the rate at which these methods detected species over time, and the types of ecological information that could be derived from them. We compared the cumulative rate of species detection for structured surveys (RLS and recent perpendicular transects) with that of opportunistic searching by untrained divers (iNaturalist) spanning the four contemporary years of 2017–20. We then compared the abundance and frequency of reporting of the most abundant and frequently reported species, so as to explore population and community perspectives. During this analysis, we also noted any patterns observed in the data, for example, differing proportions of particular species.

Mobile animals

In total, 273 mobile fish and invertebrate species (i.e. excluding sessile invertebrates) were identified with sufficient confidence to allow comparison of historical data to contemporary records, with approximately twice as many fish (183 spp.) as invertebrates (90 spp.) (Table 2). A total of 64 species were excluded as overly common, cryptic or difficult to distinguish visually (Table S1). Of the remaining 209 species, 34% were recorded in both historical and contemporary periods, 11% were losers and 55% were winners.

Significantly more loser species were invertebrates (χ² = 13.43, P < 0.001), whereas significantly more winner species were fishes (χ² = 5.84, P = 0.015) (Table S1). Our analysis of perpendicular surveys showed a significant increase in abundances of mobile animals with a northern range from the 1960s to the 2010s (Est. = 1.23, P = 0.001), alongside a significant decrease in abundances of animals with a southern range and both north + south range (Est. = −1.92, P = 0.003 and Est. = −1.48, P < 0.001 respectively) (Fig. 5; Table 3). Although there was an observable increase in fished species’ abundances and decrease in non-fished species’ abundances from the 1960s to the 2010s, this result was not significant (Est. = 0.295, P = 0.236) (Fig. 5). Power analysis of changes in abundance between fished and non-fished species between these periods showed that the observed effect size is low (0.050). Given the sample size and a power target of 0.8, an effect size of at least 0.2 would be required for detection.

Fig. 5.

Abundance scale means for surveys perpendicular to shore at Shiprock, from the 1960s and 2010s, showing interactions between period and range and period and fished status. Error bars indicate standard errors. Range N, north of Sydney; S, south of Sydney; and NS, both. Fished status: N, not fished; Y, fished, on the basis of the NSW Recreational Saltwater Fishing Guide.

Sessile biota

The 1960s account of Shiprock described abundant Sargassum and Ecklonia along the top of the wall, accompanied by a ‘multitude of animals’, including molluscs, urchins, ascidians and bryozoans (Lawler 1998). The wall was inhabited by ‘hosts of encrusting invertebrates’ including ascidians, bryozoans, hydroids, worms, sponges and soft and hard corals. At the base of the wall were layers of bivalves, a wide diversity of gastropods and urchins (Fig. 1). The sandy bottom was inhabited by sea stars, urchins and sea pens.

Today, the section of Shiprock above the wall is primarily sand, shell and rock, with substantial stands of Sargassum sp., some Ecklonia radiata and a small number of invertebrates (Fig. 6). The vertical wall is dominated by turfing and epiphytic algae and mostly sponges, with some corals, bryozoans and ascidians. Along the base of the wall are primarily sand, rubble and turfing algae with some encrusting invertebrates.

Fig. 6.

Proportion of cover recorded above (top), on and below (bottom) the wall at Shiprock on perpendicular surveys for sessile morphotaxa; abundances below 1% are grouped in Other. In total, 856 annotated images, 2017–19, by using CATAMI classification to morphotaxon level.

Algae

We categorised one species of algae as a loser (Ecklonia radiata, kelp) and one as a winner (Caulerpa taxifolia).

Ecklonia radiata was recorded in the 1960s surveys as ‘forming an almost unbroken frieze along the top edge of the submarine cliff’ and as ‘dense growths’ in the shallows. Many kelp were recorded in dive logs in the 1960s and 1970s, and noted as being eaten by sea urchins in the January 1969 Bulletin. These historical observations contrast with our contemporary ones, where we observed no unbroken frieze or dense growth, and just 14.8% coverage of E. radiata at the top of the wall (Fig. 6).

Sargassum now occupies three times the area (44.8%) and algal turf more than double the area (33.5%) of kelp. Sargassum was noted in the 1960s as abundant, and algal turf is unlikely to have been noted by divers, so we did not categorise either of these as winners. However, Caulerpa taxifolia is a notable invasive species that was first recorded in a dive log in January 2009.

Scientific names

Our analysis of scientific names showed substantial changes between the historical and contemporary periods. Of the 114 species identified in the 1998 Wetlands article (17 of the total of 131 taxa were genera or multiple species), one-quarter (29 species) had different or unknown scientific names today (Table S2).

Species detection effectiveness

Overall, we found a similar rate of species detection between structured and opportunistic sampling modes over 4 years. Structured sampling surveys found more species in the first 2 years; however, on completion of the structured surveys, new species continued to be recorded in iNaturalist, exceeding the structured survey tally (Fig. 7).

Fig. 7.

Cumulative rate of species detection over time during two parallel projects; structured surveys (perpendicular transects and RLS) compared with unstructured, opportunistic sampling (iNaturalist).

Structured surveys reported abundances and, because of the standardised survey area, densities, whereas such information could not be derived from the opportunistic sampling data. Fish biomass data could also be calculated from the structured data by using fish size classes and the allometric growth equation (R. Froese and D. Pauly, FishBase, ver. 06/2017, see www.fishbase.org). Opportunistic data included all the top 20 species in the Shiprock marine community, but in very different proportions to the structured surveys (Fig. 8). We observed that the more frequently reported species were generally large, colourful and easy to photograph (Table S3).

Fig. 8.

Data collection comparisons of most frequently reported species at Shiprock (Table S5). Data points with regression lines (yellow) and standard errors (grey bands). Frequency of recording of species and abundances for perpendicular and RLS structured surveys, and iNaturalist opportunistic sampling frequencies, in six combinations.

Climate change

The clearest indication of the impacts of climate change at Shiprock is in the disproportionate arrival of tropical species from the north. Tropicalisation of marine communities has been documented in numerous studies along the eastern coast of Australia and worldwide (Vergés et al. 2016; Vergés et al. 2019). Although large-scale change cannot be generalised from a single site, it is instructive that Shiprock provides evidence of this global change process in the opportunistic data record. Closely observed sites such as Shiprock can function as long-term sentinels of change (Micheli et al. 2020), informing the design of broader studies that can provide more generalisable conclusions.

Several northern arrivals have the potential for substantial ecological impacts. The black rabbitfish (Siganus fuscescens) is a schooling herbivore that can affect the health of kelp forests (Gajdzik et al. 2021) and may become invasive in the future. The congeneric S. rivulatus, for example, is considered invasive in other jurisdictions (Pickholtz et al. 2018) and, together with other rabbitfish species, can severely deplete macroalgae biomass over large areas (Vergés et al. 2014). Surgeonfishes (Acanthurus spp.) can increase herbivory pressure through schooling (Basford et al. 2015), and large-bodied predators from the Epinephelus genus (gropers) can affect communities both through predation and habitat engineering (Stallings 2008; Ellis 2019).

Declines in kelp at Shiprock may also be related to direct impacts of climate change. Kelp declines related to warmer, nutrient-poor tropical waters have been documented in multiple jurisdictions (Smale 2020). Such losses can have flow-on effects through the loss of the ecosystem services that kelp provides, such as shelter, habitat, nutrient cycling and productivity (Steneck and Johnson 2013). Loss of kelp may then provide space for fast-growing, opportunistic species such as turfing algae and invasive species (Filbee-Dexter et al. 2016).

Invasive species

Surprisingly few invasive species were detected at Shiprock, possibly due to the restriction of our study to CATAMI categories for sessile species and lack of awareness of invasive species by citizen scientists. Many invasive species are cryptic, unremarkable and uncharismatic (see https://www.dpi.nsw.gov.au/fishing/aquatic-biosecurity/pests-diseases/marine-pests, accessed 8 June 2022).

The invasive colonial ascidian (Didemnum vexillum) and fanworm (Sabella spallanzanii) have both been recorded in iNaturalist in the estuary but not at Shiprock (www.inaturalist.org, accessed 8 June 2022). An encrusting colonial ascidian most likely to be D. vexillum was found to be abundant at Shiprock by the authors over the course of this study (e.g. https://flic.kr/p/2acViVS, accessed 8 June 2022).

The Pacific oyster (Magallana gigas) had a single sighting at Shiprock in 2020 (Table S1). No other declared marine invasives were recorded in our study except the alga Caulerpa taxifolia, which was promoted to the diving community as invasive in 2002 (see https://www.urgdiveclub.org.au/post/north-harbour-aquatic-reserve-project-summary, accessed 6 June 2022). There are multiple records of C. taxifolia in Port Hacking, including two at Shiprock (see www.inaturalist.org, accessed 9 June 2022).

Although opportunistic data collection has been found to be useful for monitoring invasive species (Crall et al. 2010), our study has highlighted that care must be taken to manage biases arising from (lack of) awareness and detectability of species.

Over-exploitation and marine protected areas

Although Shiprock is a very small Aquatic Reserve, small MPAs can be effective for some species if they are no take (sanctuary zone), well located and supported by the local community (Turnbull et al. 2018). We found an observable but non-significant increase in fished species’ abundance between the historical and contemporary periods, and there was a wide range of fished species that were not recorded at Shiprock before MPA gazettal in the 1980s, but which were now regularly reported, sometimes in substantial numbers. Many winners such as yellow-fin bream (Acanthopagrus australis) and snapper (Pagrus auratus) were recorded on our transects, together with morwong, leatherjackets, drummer, trevally and tarwhine. Fished invertebrates were also recorded for the first time since the 1980s, including octopus, blue swimmer crabs and cuttlefish.

Mulloway (A. japonicus) has not been recorded in recent times, despite having been recorded at Shiprock in the 1960s. Once widely distributed in subtropical and temperate Australian shallow waters (see https://www.dpi.nsw.gov.au/__data/assets/pdf_file/0005/1329611/stock-status-summary-2021-mulloway.pdf, accessed 7 June 2022), mulloway has been the subject of widescale commercial and recreational fishing, with commercial landings declining over the past 50 years to arrive at today’s depleted stock status. The Shiprock Aquatic Reserve does not appear to have been sufficient to restore local mulloway populations, which is most likely a result of the small size of the reserve being inadequate to offer effective protection for this wide-ranging species (Moffitt et al. 2009).

The Shiprock MPA may also have protected aquaria-targeted species (Madrigal-Mora et al. 2022), including colourful fishes such as Canthigaster callisterna, Chromis hypsilepis, Dendrochirus brachypterus, Mecaenichthys immaculatus, Thalassoma lunare, Abudefduf spp., Acanthurus spp. and Chaetodon spp. all listed as winners in our study.

Multiple stressors

Most of our loser species were invertebrates, particularly molluscs. Mobile invertebrates in Australia’s cool latitudes are particularly vulnerable because warming waters from the north squeeze populations against deep ocean barriers in the south, putting over 30% of species at high risk of extinction (Edgar et al. 2023). Nine of our 12 mollusc losers rely on calcium carbonate shells, the cowries, whelks and bivalves, resulting in a possible additional threat from climate change through ocean acidification (Parker et al. 2013).

In addition to global climate-change pressures, the bivalves Ostrea angasi and Pecten fumatus have been affected by local human exploitation (Flood et al. 2012; Cook et al. 2021). Ostrea angasi is endemic to Australia’s southern waters, but has experienced declines in many locations. It is the subject of recent restoration programs (Pereira et al. 2019), but restoration can be challenging because multiple stressors are at play. In addition to climate change and over-exploitation, habitat loss, sedimentation and nutrient inflows have contributed to population reductions in this socially valuable species (Cook et al. 2021).

These additional stressors may also be at play in other results in our study. Although fishing pressure is moderate in Port Hacking (Steffe and Murphy 2011), even small levels of exploitation can affect populations, and the small size of the Shiprock reserve may limit its effectiveness (Edgar et al. 2014; Turnbull et al. 2018). The majority of the shoreline and catchment in Port Hacking is undeveloped and pollution levels are reported to be low, but pollutants are higher in concentration in northern embayments and so may also be affecting on the community at Shiprock (Alyazichi et al. 2021; Birch et al. 2021).

Threatened, vulnerable and protected species

Several threatened, vulnerable and protected species were recorded at Shiprock (see https://www.dpi.nsw.gov.au/fishing/species-protection/what-current, accessed 9 June 2022) (Table S1). The charismatic blue groper (Achoerodus viridis) is protected from spearfishing in NSW (see https://www.marineconservation.org.au/bluegroper/, accessed 8 June 2022) and has increased in numbers over the period of our study. White’s seahorse (Hippocampus whitei) was first recorded in 1965 (Lawler 1998) then not again for over 30 years, until being photographed in 1998, 2004 and 2008 (see www.inaturalist.org). Another seahorse, H. abdominalis, and the ornate ghost pipefish (Solenostomus paradoxus) have not been recorded since the 1960s. Whereas H. whitei is specifically listed as Endangered under Australia’s Environment Protection and Biodiversity Conservation Act 1999 (see http://www.environment.gov.au/cgi-bin/sprat/public/publicspecies.pl?taxon_id=66240, accessed 9 June 2022), these latter two species fall under more general syngnathid protection (see https://www.legislation.gov.au/Details/C2021C00182, accessed 8 June 2022).

Black cod (Epinephelus daemelii) was only recorded at Shiprock after the establishment of the Aquatic Reserve, in 1999 and again in 2003 URG diver logs, and in recent years on RLS and iNaturalist (Table S1). Black cod is listed as Near Threatened on the IUCN Red List and Vulnerable in NSW after declines caused by overfishing dating back over 100 years (Francis et al. 2016). Slow growing, long lived and a target for spear-fishers, this species has been protected in NSW since 1983, but has been slow to recover (Harasti and Malcolm 2013).

Sampling and technology change

Our conclusions must be considered in light of the historical and structural limitations of the data collection methods and technologies employed. Early data collection at Shiprock required hand-written notes, drawings, memory and physical sample collection. Underwater cameras were not widely available nor affordable. Over the 58 years of our study, there was an observable increase in the ease and volume of data collection, ranging from several species able to be recorded by hand or collected on a dive in the 1960s, to 24–36 photographs that could be taken on a single roll of film, to hundreds of photographs taken per dive with a modern digital camera.

The publication process has also accelerated in efficiency over that time, and contemporary online technologies provide a novel layer of personal motivation and reinforcement (Jennett et al. 2016). Historical data were either rarely published, for example, in personal dive logs and collections, or required extensive manual effort in hand drawing, colouring, stencilling and physical printing and distribution. It is probable therefore that many species may have been present at Shiprock without ever appearing in the historical record.

Structured v. unstructured data collection

Although it is difficult to compare structured surveys to unstructured opportunistic sampling due to the wide variation in sampling effort, our study standardises the spatial scale to a single site and provides a basis for temporal comparison by using parallel structured and unstructured projects. The structured project involved trained professional and citizen scientists conducting surveys over standardised areas, and the unstructured project involved initiating an online effort for anyone with an underwater camera to record species in the same time and space. Despite the divergent methods, the rate at which these two projects identified species was surprisingly similar (Fig. 7).

Frequency of sightings (i.e. presence on a survey) was strongly positively correlated between structured (RLS and perpendicular survey) methods, and moderately correlated between these and iNaturalist (Fig. 8). We found a weak negative correlation between the iNaturalist frequency of sighting (as a possible proxy for abundance of a species) and abundance in the structured surveys, even though the data sets are not directly comparable. This appeared to be due to the most abundant, but non-charismatic species such as T. taeniatus, S. lineolata, T. novozelandiae and A. strigatus being infrequently noticed and recorded in iNaturalist (Table S3).

The potential application of structured v. unstructured, opportunistic data sets therefore varied substantially. Whereas structured surveys provided reliable information beyond a list of species, such as densities, species absence, community structure and change over time, unstructured sightings provided primarily species presence. Structured surveys can also be designed to target inconvenient or inaccessible times and places (Callaghan et al. 2020), particularly if they are part of a broader program such as Reef Life Survey. As opportunistic records represented no systematic search in either time or space, there were no reliable absence records, and sighting frequencies were widely divergent from structured survey abundances and densities.

As part of our analysis, we observed that colourful, photogenic and charismatic species were frequently reported in iNaturalist, despite comprising a small proportion of individuals on a standardised transect (e.g. in the case of S. jacksoniensis, less than 1%). Other studies have noted this bias (Roberts et al. 2022). Although modelling may be used to attempt to compensate for the limitations of unstructured data, such as by mimicking randomness in absence and hypothesising factors such as detectability and observer effort (Brown and Williams 2019), such models require their own set of assumptions. Such assumptions do not consider observer-driven variations in sampling effort, for example, a diver focusing on photographing gobies for a period, which then gives a false signal of change in the opportunistic data record.

There were also notable differences between RLS and perpendicular surveys. RLS places transects along a depth contour on hard substrate, avoiding sand, and at Shiprock the chosen RLS depths were between 6 and 10 m. Perpendicular surveys ran down from the water surface to the deepest point on the site, spanning sand and rubble both above and below the wall and incorporating very shallow areas. Sand- and subsurface-dwelling fish were therefore more abundant on perpendicular transects, for example, G. subfasciatus and A. vaigiensis juveniles respectively. Fish that prefer structured habitat were more abundant on RLS transects, for example, T. taeniatus and O. limenus. It is evident that, even with structured survey methods, it is important to understand methodological foci and limitations.

Overall, we found that structured surveys provided broader community, population and temporal-change information, whereas unstructured sampling provided better recording of rare, threatened and invasive species (Roberts et al. 2022), and the potential for retrospectivity (Table 4).

Relevance to management, governance and sustainability

Our study has highlighted the value and potentially irreplaceable nature of historical ecological information at high stewardship sites such as Shiprock. Such sites represent an opportunity for managers to discover indicators of change spanning retrospective timescales, which are impossible in newly designed forward-looking studies.

Both structured and unstructured data have limitations. For example, loser species may be detected in unstructured, opportunistic data, particularly if they are explicitly searched for in a current project, but winners cannot be conclusively determined without historical structured searches that reliably detect absences. Frequency of opportunistic observation is not a suitable proxy for abundance, and biomass, population and community structural information can at best be modelled using assumptions. Structured surveys are superior for broad-scope, reliable community change information; however, such data are less abundant and so are of very high value where they do exist. Merged, diverse data sets incorporating structured and unstructured data therefore provide the most comprehensive insights.

Our study has shown that a single site such as Shiprock can be a sentinel for change, including detecting declines in foundation species, community shifts relating to global factors such as climate change, and local winner and loser species. However, this depends on an active, engaged local community that takes on the challenge of monitoring and conserving the site. Management actions that encourage such local stewardship can therefore have wide-ranging benefits for the long-term sustainability of the social–ecological system.