https://orcid.org/0000-0002-8815-1959
Figures
Abstract
Spatial heterogeneity linked to fish assemblages has been key to advancing conservation plans and assessing marine resources in remote ecosystems. The Juan Fernández Archipelago is a remote and highly heterogeneous geologic hotspot located on the Nazca plate (SE Pacific), comprising three islands, numerous seamounts and guyots, and a highly rugged coastal marine shelf. Hundreds of rocky reefs of various sizes characterize this insular coastal shelf, supporting diverse finfish populations of interest for the Robinson Crusoe Island artisanal fishing fleet. Two hydroacoustic surveys were conducted on the insular shelf of Robinson Crusoe and Santa Clara Islands. Reef distribution and morphometric attributes (length, height, depth strata, and geographic zones) were characterized across the shelf seascape. To identify traits that correlated with the presence and acoustic backscatter of demersal fish aggregations, a set of statistical models was implemented. Results indicate that reefs were present up to ~200 meters in depth around the islands, with the northern region hosting small and tall reefs, while the southern region had the fewest reefs but the largest. The findings indicated that 75% of the fish aggregations observed outside of reefs were within 280 m of them. Also, there was a higher probability of fish aggregations occurring in depths up to 150 m (~90%) given the presence of reefs, compared to the relatively flat seafloor (~20%). Reef–scale analysis showed that the longer and taller the reef, the higher the acoustic backscatter from fish. These findings highlight the ecological value of reefs for fish populations throughout the insular shelf seascape. In addition, this study highlights the usefulness of small–scale hydroacoustic surveys as a monitoring tool to assess the association between ocean–floor spatial heterogeneity and demersal fish assemblages, and the implementation of effective periodic surveys within the Juan Fernández Marine Protected Area.
Citation: Tapia B, Ernst B, Gauthier S, Monnahan CC, Parada C, Rivara P, et al. (2026) Understanding spatial heterogeneity of rocky reef demersal fish assemblages: Links to conservation plans in the Juan Fernández Archipelago Marine Protected Area in the Southeastern Pacific. PLoS One 21(10): e0358459. https://doi.org/10.1371/journal.pone.0358459
Editor: Judi Hewitt Hewitt, The University of Auckland - City Campus: University of Auckland, NEW ZEALAND
Received: October 30, 2023; Accepted: September 1, 2026; Published: October 9, 2026
This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Data Availability: The complete, cleaned, and post-processed dataset underlying all findings, statistical analyses, and figures in this manuscript is publicly available in the Figshare repository via the following doi: https://doi.org/10.6084/m9.figshare.33868393.
Funding: This research was supported by UNOPS program (Grant GEF–UNDP–PIMS 4147 to BE); Monitoring program for crustacean fisheries of the Juan Fernández Archipelago (IFOP project 2016 to BE); FONDECYT project (Grant 1191606 to CP); Climate Action Planning (Grant R20F0008 to CP); VRID Project (Grant 2024001184INV to CP); and Millennium Nucleus Ecology and Sustainable Management of Oceanic Islands (ESMOI) (to BE). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interest exist.
Introduction
The study of landscapes has made significant progress in terrestrial ecology, identifying critical aspects of habitat definition at different scales, an interplay among geological, environmental, and biological features [1–3]. Similarly, this interplay can be observed at seascapes capable of structuring populations and communities, such as intertidal habitats, seagrasses, seamounts, and reef seascapes [4–6]. In particular, insular shelves around oceanic islands often harbor numerous geographic features and oceanographic characteristics that create high spatial habitat heterogeneity, shaping the local distribution of populations and communities [7–9].
The rugged coastal shelf of the Juan Fernández Archipelago (JFA) off the Chilean coast (Fig 1) constitutes an ideal seascape for studying spatial ecological patterns underlying the marine environment. From volcanic matter deposits associated with geologic hotspots, and the subsequent cooling of the tectonic plate [8,11], a conglomerate of oceanic islands and guyots emerged, constituting the Juan Fernández Ridge. This seamount string is oriented perpendicular to the southwestern edge of the South American continent (SE Pacific, Fig 1B; [12]). JFA presents numerous reefs on its shelves, modulated by geomorphic processes, basaltic composition [13], and sedimentary deposits [11]. This region shelters an extensive reservoir of endemic fish that extends its distribution to the island shelf break (~200 m), forming dense assemblages in both the pelagic and benthic environments [14,15]. Thus, the JFA constitutes a vulnerable marine ecosystem and reservoir for numerous endemic species [16,17], which is embedded in a Multiple Use Coastal Marine Protected Area (AMCP–MU) entitled “Juan Fernández Sea” (Supreme Decree N°10 of 2018, Chile [18]).
(A) corresponds to the study site around Robinson Crusoe–Santa Clara Islands. Numbers (1–5) indicate the zones defined for hydroacoustic sampling. Thin blue lines within the polygons correspond to 2015 (dashed) and 2016 (red continuous) hydroacoustic transects. (B) depicts the Juan Fernandez Ridge, from which emerge the Robinson Crusoe–Santa Clara and Alejandro Selkirk Islands. (C) depicts the southeastern Pacific. The map was created by the authors using GMT v6 [10] and bathymetric data from NOAA NCEI (Public Domain: https://data.ngdc.noaa.gov/ngdcinfo/privacy.html).
Geomorphological features (e.g., reefs, bathymetry), remote/local oceanographic processes, and biophysical interactions on the shelf have been shown to determine species distribution patchiness [19–23]. While rockfish communities are often used as an example, showing strong correlations among abundance, aggregation levels, and geomorphological reef features [7,24,25], other studies have also demonstrated fish–reef associations across diverse ecosystems, linking reef structures with fish density, distribution, and biomass using hydroacoustic approaches [25–29]. In the JFA, aggregated multispecies demersal fish assemblages have been described in association with conspicuous reef systems of various sizes (Table 1; [14,15]). These reefs, in turn, are strongly correlated with traditional fishing grounds used by artisanal fishers around the islands [16,23].
For over a century, the JFA ecosystem has sustained a valuable small–scale artisanal fishery operating under a traditional territorial tenure system [30–32]. This traditional system has brought economic stability and social well–being to the Robinson Crusoe Island local community [33,34]. Despite the importance of fish communities to the JFA ecosystem and local economy, little is known about their conservation status, especially for morwong (locally known as breca, Nemadactylus gayi), an endemic species that supports artisanal fisheries and consumption [16]. Understanding the scales of natural variability in their distribution, behavior, and particularly their association with the seascape requires scientific surveys for evidence–based decision–making regarding the implementation of conservation (e.g., Marine Protected Areas [35]) and management plans for species of high commercial interest [36]. An essential step is thus to develop conservation plans and fishery management actions based on knowledge of the state of fish communities and their relationship with the seascape spatial heterogeneity [35], such as specific zones and depth with a significant presence of reefs and fish patches.
The study of fish communities with non–invasive and non–destructive sampling methods, such as hydroacoustic and optical technologies, can provide essential scientific information, even in hard–to–reach areas [25,37], while conserving the local ecosystem’s fragility. Previous studies in JFA were methodologically restricted to scuba diving surveys and, therefore, limited to shallower depths (<30 m). The focus of these studies was the taxonomic characterization of fish species, their association with the subtidal environment [38,39], and trophic ecology [40]. While these inventories provided valuable baseline information, advances in hydroacoustic methods have proven effective for characterizing seafloor habitats and linking geomorphological features with fish distribution and biomass in other ecosystems [28,41]. Multibeam and split–beam systems have been successfully applied to describe habitat complexity and its relationship with demersal fish communities, providing both abundance estimates and ecological information on the habitats that sustain fish populations [29]. This, in turn, improves the interpretation of backscatter patterns in relation to benthic structures [27, 28].
In this work, we use hydroacoustic technology to monitor deeper zones (up to 200 m) and to understand which features might influence the presence and acoustic backscatter of fish aggregations surrounding JFA [14,15]. Specifically, this study aims to characterize the morphology of the reef system on the Robinson Crusoe–Santa Clara Islands shelf and examine how reef height, length, location, and depth influence fish presence and acoustic backscatter, providing insights relevant to conservation and ecosystem–based management strategies.
Materials and methods
Study area and hydroacoustic surveys
The JFA comprises a group of volcanic islands around 6 million years old [42]. It is located perpendicularly to the southwestern edge of the South American continent, 360 nautical miles (nmi) to the west of the province of Valparaíso, Chile (Fig 1C; [12]). Three oceanic islands grouped into two subsystems form the JFA: Robinson Crusoe and Santa Clara Islands (RC–SC) and Alejandro Selkirk Island, ~ 90 nmi farther west (Fig 1B; [43]).
In this study, data were obtained from two daylight hydroacoustic surveys conducted around RC–SC subsystem in 2015 and 2016 (Fig 1A). Sailing permits for these operations were granted by the Chilean Navy within the JFA. Each survey had its own sampling designs [14,15]. The first survey was conducted between November 6th and 9th, 2015, using a 12–m diesel–powered vessel at a navigation speed of 4–6 knots. SIMRAD EK60 split–beam transducers of 38 and 120 kHz (ES38–B and ES120–7C; Kongsberg Simrad AS, Kongsberg, Norway) were installed on a deployable metal arm two meters below the surface. The downward–looking echosounders were calibrated following standard procedures [44] and configured with a pulse duration of 1024 µs and a ping rate of 1.0 s-1. Sampling followed a zig–zag design with transects of 4.6 km long separated approximately by 1.1 km between apices (Fig 1A). This sampling strategy was developed as part of a 2015 exploratory monitoring project to understand the distribution and composition of the fish community on the RC–SC Islands coastal marine shelf [14] using hydroacoustic and optical sampling techniques.
The data on the distribution and composition of the fish aggregations from the first survey allowed us to refine the sampling design on the second survey, which presents a more precise sampling design, bound to the coastal fish species distribution (Table 1). The second survey was conducted on a 9–m artisanal fishing boat (with the same navigation speed of 4–6 knots as in the previous survey) between October 6th and 12th, 2016. In this case, only the 38 kHz echosounder was used, with an ES38–10 transducer model, maintaining the same configuration as the 2015 study but mounted in a tow–body suspended off the vessel’s side and deployed one meter below the surface. The sampling design implemented quasi–parallel (4.3 km long) transects, configured orthogonally to the coast, and spaced 0.93 km apart.
Both survey designs correspond to sampling strategies traditionally used for insular ecosystems [37]. In both surveys, the sampling area was stratified into five zones around the islands (Table 2 and Fig 1A). The zone selection was based on the captain/fishermen’s knowledge of fishing grounds, which are characterized by particular wind and wave conditions generating less or more sheltered zones for the sampling based on prevailing meteorological conditions.
Acoustic data processing
Acoustic records at 38 and 120 kHz obtained during the 2015 survey were initially used to qualitatively discriminate the acoustic signal of fish assemblages from zooplankton sound scattering based on backscatter intensity, aggregation shape, depth, and location in the water column. However, only the 38 kHz frequency was used quantitatively in this analysis because acoustic records were available for both the 2015 and 2016 surveys. The use of this frequency is further supported by its reliability in deeper waters, because at ~200 m depth–related attenuation and changes in swim bladder target strength are minimal (< 3 dB; [37,44]), ensuring consistent detection of fish schools [37].
Data was processed using Echoview® 8.0.105 acoustic analysis software [45]. Acoustic backscatter processing involved configuring the survey transects’ start and end, removing transect sections located outside the 5 zones established around the islands, as well as intertransect data, and delineating the water column from 7 m below the surface to 1 m above the detected seafloor. The upper limit was set to exclude surface noise, e.g., bubbles from waves and other disturbances, while the lower limit avoided bias from the near–bottom echoes (where discrimination from reflected ocean floor can be problematic) and potential side–lobe interference. Discrete aggregations of fish with discernible edges were identified along transects using a threshold in volume backscattering strength (Sv; [46]) of −70 dB re 1 m-1 to create user–defined regions (polygons) with assigned names (IDs). We separated multispecies demersal fish aggregations from pelagic ones based on morphology, depth, and the aggregations’ positions in the water column. These aggregation types were identified and characterized through optical sampling [14,15]. Underwater camera systems (ROV and stereo cameras) used during 2015 on echo traces indicated that fish aggregations located in midwater and within the upper layer of the water column were constituted mainly by Juan Fernández trevally, while the fish aggregations linked to the seafloor (forming columnar and bottom associated aggregations) were composed of mixed species [14,15], which constitutes the focus of this study (Table 1).
The integration of fish acoustic backscatter values per unit area “NASC” (Nautical Area Scattering Coefficient, units of m2 nmi-2; [46]) was exported in Elementary Distance Sampling Units (EDSUs) covering 50 m along each transect, spanning the entire water column. NASC represents the sum of volume backscatter (Sv) within an area and is proportional to the biomass or abundance (number of individuals) and type of organisms producing this backscatter [37,44]. Although NASC does not directly depict absolute fish density, it is also used as a proxy of acoustic density, i.e., a relative fish density indicator. For each EDSU, acoustic backscatter of fish aggregations and its subsequent transformation into presence–absence data, bathymetry, and geographic location were recorded.
The characterization of a reef’s shelf entailed the identification of features indicated by significant variations in bathymetric relief and acoustic backscatter intensity observed in analyses of bottom echoes (see Fig 2). Reefs were isolated within the echograms by selecting locations (delimited by regions and IDs) where the bottom height was conspicuously above the average seafloor height (>2m), measured from a manually projected baseline from the detected seafloor, as illustrated in Fig 2B. Based on these measurements, a reef’s presence–absence, reef mean height, length, and depth attributes were extracted for each EDSU. Importantly, these attributes were only measured when reefs were present; EDSUs without reefs were not assigned values for these variables. Region–specific IDs were retained when integrating these features across transects to account for fish aggregations and reefs spanning more than one EDSU.
(A) Bottom profile of the southeastern insular shelf of Robinson Crusoe Island, showing temperate reefs and reef aggregations. (B) shows an echogram of fish aggregations and temperate reefs of this region recorded during the 2016 hydroacoustic survey, and (C) a fish school depicted from stereo cameras (see Ernst et al., 2016 for further details).
Characterization of rocky reef structural attributes.
Four attributes were used to characterize rocky reef structure: average reef height, total reef length, geographical locations (zones), and reef depth stratum. The average reef height was calculated by averaging all height observations within each EDSU, with the reef’s baseline defined as a projected straight line based on bathymetric slope, as shown in Fig 2B. The total reef length was calculated from the Euclidean distance between the initial and final geographic coordinates of the identified reef from the echograms (Fig 3). The reefs were also geographically referenced to the five zones around the RC–SC Islands (Fig 1A) defined in section 2.1. Finally, the reefs were classified into four depth strata across the shelf: 0–50 m, 50–100 m, 100–150 m, and 150–200 m.
Conceptual diagrams of reef and fish aggregation metrics used during the post–processing and (A) analysis of acoustic data at the Elementary Distance Sampling Units (EDSU) of 50 m and (B) reef scales.
Spatial distribution of rocky reefs.
Survey and reef coverage areas based on acoustic backscatter data were used to obtain spatial rocky reef statistics computed by zone and depth, such as average reef density (number of reefs/ km2), proportion of reef area coverage, and expected number of reefs.
Given that transects were generally inshore–offshore oriented and beam width varies with depth, a trapezoid was used to estimate the areas. Acoustic beam diameters (trapezoid bases) were calculated using transducer beamwidth and local bathymetric depths, where nominal beamwidths at half–power points (ES38–B and ES38–10) corresponded to 7 and 10 degrees on hydroacoustic surveys in 2015 and 2016, respectively. The Euclidean distance between initial and end coordinates was used to estimate the surveyed distance or reef’s length (trapezoid height). These measurements (diameters for trapezoid bases and length for height) were combined to calculate the survey transect (a) and the reef coverage (r) areas.
The reef density was estimated from the amount of reef detected in each transect segment (si,j,k) within zone ‘i’, depth stratum ‘j’ and transect ‘k’. Transect segments (si,j,k) were defined based on the 50 m EDSUs and could include one or more EDSUs, or a portion thereof, within the same zone and depth stratum. This approach ensures that reef densities account for all observed reefs and the effective survey area, while also accommodating differences in transect length across zones and depth strata. The densities (di,j,k) were then weighted by the proportion of surveyed area , reflecting variation in transect lengths due to distance to depth contours (isobaths), where ‘nsi,j’ is the number of transect segments by zone and depth stratum. The average reef density by zone and stratum, (
) was calculated as:
The proportion of reef area coverage by zone and depth stratum () was obtained through the ratio of total reef area coverage (
) to the total sampled area (
):
Finally, the number of reefs ‘e’ by zone and depth stratum, was estimated based on average density estimates (di,j) and the total area (Ai,j), assessed using the geographic processing software ArcGIS (Environmental Systems Research Institute, ESRI, Table 2). This approach assumes approximate homogeneity of reef density within each depth stratum and zone (di,j).
Analysis of demersal fish assemblage data
The analysis of demersal fish data consisted of modelling both the presence–absence of fish aggregations (a discrete response variable) and their acoustic backscatter (a continuous response variable).
Discrete response variable: presence–absence of fish aggregations.
A binomial generalized additive model for location, scale and shape (GAMLSS; [47,48]) implemented in R [49] was used to explore, by EDSU ‘l’, the effect of several factors on the probability of occurrence (NASCl = 1) of demersal fish aggregations (nEDSU = 7462). The primary explanatory variables included the presence–absence of reefs (prl), a non-linear P–spline smoother function (; [50,51]) on the seafloor depth (zl), and the distance to the closest reef (distl; Fig 3A):
A stepwise model expansion approach was used to test spatial heterogeneity around the RC–SC Islands coastal shelf based on the full dataset. A series of models was evaluated, beginning from a baseline framework (M0) that included all primary predictors, to a fully saturated model (M4) by a stepwise inclusion of zone as a fixed factor and its interactions with the predictors (prl × zone, (zl) × zone, and distl × zone). To mitigate potential biases and overfitting in the full model [52], an iterative approach was utilized to evaluate various combinations of predictors, ensuring an optimal balance between model accuracy and complexity. The most parsimonious model obtained from this procedure was referred to as M5, selected using the stepAIC() function from the MASS package [53]. The predictive performance of models was compared using Akaike’s information criterion (AIC; [54]), with lower values indicating better model fit [55]. Additionally, the relative importance of each predictor within each zone was assessed through simple regression models with specific spatial interaction term, estimating the proportion of explained variance via squared Spearman rank correlation as a proxy for localized variance explained.
During the 2015 and 2016 hydroacoustic surveys, zones 1 and 2 were the most revisited of the zones (Table 1). This allowed isolation of a spatial–temporal subset to explore the consistency of the presence of demersal fish aggregations across these zones over time (season: 2015–2016). Spatial and temporal consistency was tested by contrasting a baseline model with season as an additive predictor (ST0) against a spatial model featuring triple interaction terms among the linear primary predictors, zones, and survey season (ST1; zone × {prl, (zl), distl} × season).
Continuous response variable: Backscatter (NASC) of fish aggregations.
To identify the reef (h) structural attributes that may affect the relative acoustic density of fish, a reef–scale analysis was implemented. Contiguous EDSUs with reefs and fish aggregations were integrated by matching specific IDs assigned to each school and the reef polygons (n = 276) described in section 2.2. The integrated fish acoustic density value per unit area was modeled as a continuous variable (NASC of fish aggregations) through a GAMLSS with Weibull distribution, which was determined to be the optimal error structure via the fitDist() function. The response variable was modeled as a function of three reef attributes, namely mean depth (Zrh), length (Lrh), and average height (Hrh; Fig 3B). All of the explanatory variables were fitted using nonparametric P–spline smoother function:
The NASC response variable was modeled using a stepwise regression framework. Beginning from a baseline model without spatial structures (M6) that integrated reef depth, length, and height, the model was subsequently saturated by incorporating the interaction between reef attributes and zones (M10). An iterative procedure analogous to that used for the presence–absence models was employed to determine the optimal multiple predictor combinations (M11). Additionally, simple regression models with zone–specific interactions were implemented to identify the local relative importance of each reef attribute within individual zones, using the proportion of explained variance as a proxy.
Results
Spatial heterogeneity of the rocky reefs
Characterization of rocky reefs around the islands.
Bathymetric features of different sizes were found on the coastal shelf of RC–SC Islands. The reefs’ length varied between 14 and 985 m, and height ranged from 2 to 28 m (Fig 4A). Seventy–five percent of the reefs were under ~328 m long and ~9 m high (Fig 4B and 4C). Very long reefs (>760 m) were observed in zones 2, 3, and 4 (Fig 5A), mainly in deep strata 100–150 and 150–200 m (Fig 5B), as well as very high ones in zone 1 (>20 m; Fig 5C) associated with intermediate depths (50–100 and 100–150 m; Fig 5D). These features are observed as discrete spikes in the density contour of reef length and height (Fig 4A).
(A) Density distribution of reef length and height. (B) Cumulative proportion of reef length and (C) height. Dashed lines correspond to the 0.5 and 0.75 percentiles.
(A) Boxplots of reef length by zones (n = 5) and (B) depth strata (n = 4). Panels (C) and (D) depict reef height by zone and depth strata, respectively, around the Robinson Crusoe and Santa Clara Islands.
Spatial distribution of structural rocky reef attributes around the islands.
High densities of reefs were observed in zones 1, 4, and 5 (>100 reefs/ km2; Table 3). Zone 4 had the largest proportion of reef area among the five zones (Table 4), mainly due to larger reefs (length, Fig 5A and S1A). Zones 1 and 5 also exhibited a high proportion of reef areas (Table 4), but this was largely due to the density of reefs rather than their size (Table 3 and S1A Fig). Reef length was the lowest in these zones than in any other (Fig 5A and S1A). Reef height was highest in zones 3, 4, and 1 (Fig 5C).
Spatial distribution of structural rocky reef attributes by depth strata.
The average reef density was higher in the shallow (0–50 m) and intermediate strata (50–100 m), with around 300 and 70 reefs/ km2, respectively, decreasing towards deeper zones to ~20 reefs/ km2 (Table 3 and S1B Fig). Reef height exhibited a similar pattern (Fig 5D), while reef length displayed an opposite trend, increasing with depth (Fig 5B and S1B). The highest proportion of reef area was observed in the shallow and intermediate strata, 0–50, 50–100, and 100–150 m (Table 4 and S1B Fig). Spatial differences in the reef area were affected by reef density and length. In fact, 0–50 m has the greatest density of small reefs, 50–100 m contained relatively small reefs, but it had the second largest density, while 100–150 m hosted the third largest density, but the second stratum with the longest reefs (<150–200 m; Table 3; Fig 5B and S1B).
Number of expected reefs by depth strata and around the islands.
The expected number of reefs was obtained by weighting the averaged reef density by the area of each depth stratum, showing a predominance of reefs at shallow and intermediate strata (0–50, 50–100, and 100–150 m, ~ 6587 reefs), with zones 1 and 5 making the highest contribution (>2000 reefs each). The northeastern and northwestern regions (zones 1 and 5) accounted for 61% (~4501 reefs) of the total number of reefs on the insular shelf (~7351; Table 5).
Linking rocky fish assemblage and reefs
Reef habitats comprised approximately 9% of the total area surveyed, however, empirical cumulative distribution analysis showed that half of the fish aggregations were located within 48 m of the nearest rocky reefs, and 75% within 280 m (Fig 6). This spatial clustering highlights the strong association between fish aggregations and rocky reefs on the RC–SC insular shelf.
Empirical cumulative proportion of observing fish aggregations (a discrete variable) as a function of the distance to the nearest reef. Dashed vertical lines correspond to the 0.5 and 0.75 percentiles.
Modelling discrete response variable: presence–absence of fish aggregations.
The multiple regression analysis showed that the relationships between the presence of demersal fish aggregations and habitat predictors were strongly heterogeneous across the RC–SC Islands coastal shelf. Allowing habitat predictors to interact with specific spatial zones significantly improved model parsimony, as the sequential addition of zone interactions with reef presence (prl), seafloor depth (zl) and distance to the closest reef (distl) reduced 527 AIC units compared to the baseline model (M0; see S1 Table). Consequently, the stepwise selection identified M5 as the optimal model (AIC = 6061; S1 Table), preserving the habitat–zone interactions except for the seafloor depth effect in zone 5 (S2 Table).
Across all zones combined, model 5 (M5 full model) indicated that, in the presence of reefs, there is a ~ 0.9 probability of finding demersal fish aggregations at depths up to ~150 m, dropping to its lowest levels at greater depths and approaching ~0.4 at 200 m. When reefs were absent, the probability of detecting demersal fish aggregations was substantially lower (~0.2–0.3) at depths up to ~150 m and declined outside of this range, following a dome–shaped pattern that reflects the smoother function depth effect’s significance (Fig 7A). This pattern highlights a key depth range for aggregations. The sharp decline in the probability of detecting fish aggregations in areas with and without reefs within the 150–200 m depth range indicates that deeper zones are likely outside the preferred demersal habitat.
Model predictions of the probability of observing demersal fish aggregations based on the full integrated spatial model (M5). Panel (A) represents the aggregated spatial prediction framework. The upper curves represent predictions over temperate rocky reefs (reef presence, distance to reef = 0 m), and the lower curves represent predictions outside reefs (reef absence, distance to reef = 579 m, the mean distance to the nearest reef when all zones were integrated, which may vary by specific zone). Shaded areas around the curves indicate the 95% confidence intervals. Panels (B–F) represent zones 1–5, respectively, which illustrate localized spatial responses shaped by site–specific bathymetric smoothing splines and interaction terms.
However, this ecological pattern exhibits notable local variations across specific areas. In zones 1 and 4 (Fig 7B and 7E), the curves depict the general dome–shaped bathymetric distribution, but the probability of detecting demersal fish aggregations is overall significantly lower (zone1: = −0.649, p = 0.036; zone4:
= −2.13, p < 0.001; see S2 Table). Furthermore, these zones show a weaker positive effect of reef presence on fish detection, as evidenced by the negative interaction terms (prl × zone1:
= −1.873, p < 0.001; prl × zone4:
= −0.9, p = 0.022; S2 Table).
Conversely, zones 2 and 3 display significantly distinct non–linear curves. In zone 2 (Fig 7C), reef presence sustains a high, stable detection probability (~0.95) that extends to deep water before decreasing past 150 m ((zl × zone2):
= −0.006, p = 0.002; S2 Table). In habitats without reef across zones 2 and 3 (Fig 7C and 7D), demersal fish detection probabilities drop to ~0.05–0.10 at intermediate depths (~75 m) before showing a significant increase at ~125 m (
(zl)× zone3:
= −0.027, p < 0.001; S2 Table).
Finally, zone 5 represents an ecological pattern that stands out because it shows a depth–independent relationship, where the stepwise selection criteria omitted the depth–smoothing predictor (S1 Table). Although the probability of detecting demersal fish aggregations within this zone is significantly lower overall (zone5: = −2.15, p < 0.001; S2 Table), it exhibits constant probabilities across the entire bathymetric range, remaining fixed at ~0.65 over reefs and ~0.15 on habitats without reefs (Fig 7F).
The simple regression analysis with zone interactions isolated the local relative importance of each habitat predictor within individual zones (Table 6). These models showed that reef presence (prl) was primarily predicted fish presence around the islands, except in zone 1, where the best predictor was distance to the closest reef (distl; Table 6).
The spatial and temporal effect analysis for zones 1 and 2 tested whether the dome–shaped pattern described in Fig 7B and 7C holds across the 2015–16 survey seasons, using the fully interactive model (ST1: AIC = 4019; see S3 Table). Both zones maintained their general dome–shaped patterns in the probability of observing fish aggregations as a function of the bathymetric depth range, but exhibited significant temporal variations, with zone 1 (Fig 8A) showing a larger temporal contrast than zone 2 (Fig 8B). The interannual differences in zone 1 were primarily driven by a significant increase in the positive effect of reef presence during the 2016 season (prl × season2016: = 1.695, p = 0.002; S4 Table). Conversely, zone 2 maintained highly stable, near saturated detection probabilities over reefs across both seasons, alongside a consistent localized peak at ~125 m depth in areas without reefs (
(zl) × zone2:
= −0.005, p = 0.014; Fig 8B and S4 Table).
Model predictions of the probability of observing fish aggregations based on interactive spatial and temporal model for (A) zone 1 (B), and zone 2. Colored lines represent predictions over temperate rocky reefs (distance = 0 m), and blue lines represent predictions outside them. Dashed lines indicate predictions for the 2015 survey season and solid lines represent predictions for the 2016 season.
Modelling continuous response variable: Backscatter (NASC) of fish aggregations.
In the multiple regression procedure, the fully integrated model with zone specific interactions M10 (AIC = 4958; see S5 Table) showed the best fit to fish aggregations backscatter across the coastal marine shelf of the Juan Fernández Sea. This model indicated that shallower, longer, and taller reefs sustained significantly higher fish backscatter (Fig 9), though the magnitude of these effects varied per zone (S6 Table).
Model predictions of the backscatter of fish aggregations (logNASC) as a function of reef depth, reef length, and height, based on the main–effects baseline model M6 (representing the mean population response across all zones; see S5 and S6 Tables for the best fitting interaction M10). In panel (A) fish backscatter was modeled as a function of reef depth (Zrh), using mean values of reef length (Lrh = 170 m) and height (Hrh = 6 m). In panel (B), predictions were made using mean values of reef depth (Zrh = 104 m) and height. Panel (C) shows logNASC predictions as a function of reef height using mean values for depth and length. The orange shaded areas around the curves indicate 95% confidence intervals.
Reef depth exhibited a consistent and negative effect on fish acoustic density across all zones (Fig 9A), showing a general decrease that was particularly sharp in zones 3–5 ( ≤ −0.023, p < 0.001), and more moderate in zones 1–2 (
≥ −0.011, p < 0.001; see S6 Table). Conversely, reef length had a significant positive effect on fish backscatter around the RC–SC Islands (Fig 9B), especially in zones 1–2 (
≥ 0.006, p < 0.001), but not in zone 5 (p = 0.411; S6 Table). Similarly, reef height increased density across most of the islands (
≥ 0.085, p ≤ 0.017; Fig 9C), except in zone 1, where the effect was non–significant (p = 0.529; S6 Table).
The simple regression analysis showed that, at the scale of individual reefs, fish backscatter was best explained by reef length (Lrh) in zones 1–2 (explained 35% and 29% of variance, respectively), average reef height (Hrh) in zones 3–4 (49% and 13%), and reef depth (Zrh) in zone 5 (28%; Table 7).
Discussion
The insular shelf of the Robinson Crusoe and Santa Clara Islands, a highly rugged seascape with numerous temperate reefs at different depths, constitutes a discrete, spatially heterogeneous environment vital for structuring populations and biological communities [7,26]. This research constitutes a first attempt to identify and highlight the ecological relevance of these rocky structures on the spatial distribution of local fish assemblages through the use of hydroacoustic sampling methods (down to a depth ~200 m). This work particularly contributes to identifying predictors of fish presence and relative density, which are of high interest for conservation plans and seascape ecological studies [56,57]. In addition, it addresses the need to develop monitoring and non–invasive, non–destructive survey programs that match this spatial seascape heterogeneity and provide robust information to strengthen future conservation plans [58].
Reef characterization
The study of seafloor spatial heterogeneity has been approached by measuring reef density (number of patches along transects), size (height and length), and their spatial distribution on the insular shelf (zones and depth strata). Observing the same reef on multiple occasions, e.g., across neighboring transects, could lead to reef density overestimation. However, the observed maximum reef length was 985 m oriented offshore, while around 75% of the reefs were under ~328 m long and ~9 m high. The distance between transects was generally 926–1111 m in the 2015 and 2016 survey designs, respectively. Assuming a relatively symmetric distribution of reefs (relative to the shore), this source of error should be small.
Zones 1, 4, and 5 show a more rugged seafloor. In zones 1 and 5, this heterogeneity was due to the high density of narrow and tall reefs in a shallow, restricted shelf area (NE and NW; [11,13]). This region is less exposed (sheltered) to relevant erosion factors, such as the SW–prevalent wind and waves (Fig 10; [38]), which shape the reef’s geomorphology [56,59,60]. Whereas in zone 4, to the southwest of the shelf, the heterogeneity of the seafloor was mainly due to the large size of the reefs distributed over a deeper and broader area on the insular shelf (SW; [13]). Currents, sea level, winds, and waves are relevant agents of erosion that are typically more intense in the shallower areas [8]. This shallow–depth stratum from zone 4 corresponded to the SW shelf (<50 m depth), an area exposed to erosion and characterized by fewer reef patches than the sheltered regions (NE and NW). The insular shelf configuration, along with the reefs’ spatial distribution (highly correlated to fishing grounds [16,30,32]), has provided key information for the implementation of an effective fishery monitoring around islands (Fig 10). In the northern region (zones 1 and 5), our results suggest implementing sampling designs that cover a coastal and shallow area of the shelf (<100 m depth). While in the southern region (zone 4), a sampling design that is less coastal and deeper is necessary outside of the breaker areas (>50 m depth). Coincidentally, the southern shallow area is characterized by high intensity of winds and waves [38], posing operational constraints to the finfish artisanal fleet [16] and for research vessel navigation.
The left panel indicates the distribution of reefs along with oceanographic processes (sub– and mesoscale eddies, meander flow, von Kármán vortices, and local upwelling) identified around the coastal marine shelf of Robinson Crusoe and Santa Clara Islands. The right panel shows the distribution of reefs based on depth, where coastal reefs are taller and numerous, while reefs located at greater depth are longer and less numerous. Potential small–scale physical processes that could occur linked to reefs are shown in the right panels: tidal front (1), turbulence mixing (2), internal wave mixing (3), transport (4), surface winds (5). Credit: diagram by Felipe Gamonal.
Reef–fish relationship
The study of the fish community on RC–SC Islands has mainly focused on taxonomic characterization [17,38,61], trophic ecology [40], and association with the shallow subtidal environment (<30 m depth; [39]). This study demonstrated that across a bathymetric range up to ~150 m on the insular shelf, reefs strongly influenced the presence of demersal fish assemblages (~0.9 probability), unlike reef–less (or flat) seafloor sites (~0.2–0.3 probability). The link between demersal fish assemblages and reefs aligns with fishers’ traditional and historical knowledge, which predicts a strong spatial correlation between higher demersal finfish catch rates and the presence of temperate rocky reefs [30,32]. Therefore, this finding provides evidence to support the fishers’ local knowledge, which is key for evidence–based decision–making and for enhancing governance in eventual conservation actions and fishery management [62,63].
Detecting spatial heterogeneity while linking reefs to fish distribution in turn facilitates the design of future assessment plans. The conservation status of fish stocks would thus be assessed based on periodic (annual/biennial) hydroacoustic surveys and on biological–fisheries monitoring of the finfish fishery, which has been in place since 2012 but with limited spatial coverage [64,65]. Studies of fish population dynamics and fishery management have shown that incorporating spatial structure is important for implementing effective management actions [66–68]. In the case of the JFA system, its heterogeneous seascape suggests spatial management actions for the exploited resources.
Like in the RC–SC Islands, the Hawaiian Islands in the North Pacific Ocean basin also feature seafloor high relief [8], which control the density and richness of herbivorous, benthic fish–carnivores and planktivores [69]. However, this study also reports fish outside the reefs on the RC–SC Island shelf (Figs 6 and 7). Identifying the fish distribution on both reef and outside reef habitats has provided clear evidence of the use of these habitats by fish species within the same survey area, indicating a use of the seascape beyond the reefs [70]. In zone 1 (NE insular shelf), for instance, the distance to the closest reef was the best fish aggregation distribution predictor based on a simple regression model approach (see Table 6). Similar results were found in the Coral Triangle region of Australia, where typical coral fish were reported in other areas without reefs, possibly due to the search for new habitats driven by ontogenetic movements of the species, daily feeding, and fishing activity [70,71].
Spatial patterns
The sheltered region to the NE of the RC–SC Islands (zone 1) did not follow the same pattern as the other zones. Although the presence of reefs was also significant, the presence of fish was better explained by the distance to such bathymetric features. Even though the hydroacoustic survey designs implemented were not initially configured to address spatial and temporal aspects in zones 1 and 2, the depth–related pattern of fish presence/absence was similar across both seasons for the different zones, but zone 1 displayed larger variability in the probability of observing aggregations of fish at the reef sites, driven by a strong positive interaction between reef presence and the 2016 season (S4 Table). High seafloor heterogeneity in zone 1 also predicted high reef density, strengthening the spatial link not only between fish and individual reefs but also suggesting a larger spatial influence around the reefs [60,72].
Similarly, the NW shelf (zone 5) exhibited a particularly distinct spatial behavior. While reefs best predicted fish presence, detection probability was independent of depth, remaining remarkably constant across the bathymetric range over reefs (~0.65 probability) and in reef–less areas (~0.15 probability; Fig 7F). The high density of tall, narrow reefs in this area provides high seafloor complexity, maintaining stable habitat availability that uncouples fish presence from the dome–shaped bathymetric pattern. Although seafloor depth is typically considered a primary structuring factor for demersal communities, localized physical complexity across multiple spatial scales combined with key biophysical and oceanographic drivers can override this gradient, relegating depth to a secondary role in shaping fish spatial distribution [73,74].
The scale of oceanographic features impacting reefs has a broad range. For instance, the spatial influence beyond reefs on fish distribution could be based on meso– and sub meso–scale eddies and their coupled biogeochemical processes, which could modulate specific habitat conditions resulting in an attractor for the fish communities [72–76]. Like in seamount habitats, vorticity and Taylor columns [7,77] may influence circulation patterns that create patchiness in reef ecosystems [78]. On the other hand, tidal currents, turbulence, and internal waves affect the physical dynamics at small scales, promoting mixing around reefs [79,80]. These features contribute to the formation of a highly variable and dynamic zone (Fig 10), which also stands out for having the highest fish density around the RC–SC Islands [15]. While disentangling the association between fish aggregations and reefs requires a deeper understanding of the interplay between environmental and ecological processes, conservation and management plans should start considering the relevance of spatial heterogeneity for finfish communities at a broader spatial scale. Consequently, the northern sector of the RC–SC Islands (zone 1 and 5) could, therefore, require a larger sampling and monitoring effort than other zones, not focused only on reefs but on larger areas around them as well.
Modulating factors
Intrinsic characteristics of the reefs, such as height, length, and bathymetric location, were key drivers modulating the backscatter of demersal fish aggregations, though their influence exhibited spatial heterogeneity around the RC–SC Islands. Research indicates that coastal marine shelves adjacent to taller, longer, and shallower reefs exhibit higher fish acoustic relative density. This is attributed to positive correlations between reef height and length with fish density, while average reef depth demonstrates an inverse relationship (Fig 9). Among the above characteristics, both reef length and height were the most important predictors in explaining changes in fish acoustic relative density across zones, similar to what has been observed in other studies showing that metrics related to seafloor structures are strongly associated with fish assemblage structure, including richness [81,82], diversity [58], and abundance [83–85].
Modelling fish presence and acoustic density with respect to covariates, such as depth, indicated a higher probability of detecting demersal fish aggregations and higher densities at shallower reefs than at deeper reefs (Fig 7 and 9A). However, while depth did not constrain fish presence in zone 5, was as the primary predictor of fish relative density (explaining 28% of variance), decreasing sharply toward deeper reefs. This highlights a clear decoupling between fish presence and density on the NW shelf, while the rugged seascape sustains demersal fish presence across all depths, larger aggregations preferentially concentrate on shallower and taller reef structures. The high heterogeneity that characterizes shallower areas offers more protection, which could reduce predation pressure [86,87], an interaction that contributes largely to the natural mortality of reef fish communities [88]. In coastal and shallow areas, mainly in the northeastern region (<26 m depth in zone 1), important catches of trevally/little–trevally are obtained with handline from dense pelagic aggregations (outside of this research focus) or from the upper layer of multispecies aggregations [15,16]. Whereas JF morwong catch, a central species for human consumption and lobster fishery operation, occurs on fishing grounds that match reef sites located on average at ~100 m depth contours around islands [16], where they are caught with vertical longlines laid out in the lower layer of fish aggregations linked to the seafloor. The JF morwong catch in reef sites also shows an increase in mean size as a function of depth [15]. Hence, it is expected that frequent and dense fish aggregations of JF morwong juveniles/recruits prefer habitats with substantial shelter provided by numerous and rugged reefs located in shallow areas, where the main JF morwong catches would not be developed [30].
Oceanographic processes also play a key role in structuring the environment around the seascape at different scales [89,90]. Processes such as von Kármán vortices, local upwelling, island mass effect [19,21], meanders, surface [91], and subsurface eddies [20,92,93] have impacts on island productivity (Fig 10; [23]). However, the interactions of these physical processes with the different structures on the insular shelf (from discrete reefs to larger groups of them) require additional knowledge to fully assess their roles in the ecological connection between fish and reefs. Thus, while productivity and food availability significantly influence fish distributions, their effects interact with reef structure and oceanographic dynamics, highlighting that multiple interacting factors collectively shape fish presence and aggregations across the insular shelf.
Toward management and conservation
The use of hydroacoustic methods has enabled the acquisition of comprehensive, continuous, high–precision spatial and temporal information on the entire demersal finfish community of the insular shelf without compromising the fragile ecosystem. These methods have been key to understanding the structure and heterogeneity of the insular marine shelf landscape at the spatial scale (Fig 10). It also highlighted the ecological value of different reef systems and established the foundations for subsequent study designs, using a combination of hydroacoustic and optical tools [25,37]. High–resolution mapping of reef characteristics could enhance an area–based design, a key to conservation and fishing management plans. More accurate surveys on rocky reefs, however, require more sophisticated methodologies such as multibeam echosounders to scan the seabed [37]. Using stereo cameras coupled with hydroacoustic equipment may result in quantitative estimation of species proportions and the size structure of demersal fish aggregations [25,94], ideally by depth strata. Although only 38 kHz was used in this survey, future work integrating multi–frequency analyses and target–strength estimations could improve species–level discrimination in reef habitats [37,44,95,96]. These methods would provide more robust indicators of abundance and biomass for key endemic fishes [97–99], such as the JF morwong, thereby supporting both conservation and fishery management. This approach will eventually enable monitoring fish abundance across their entire bathymetric range and optimize sampling effort in areas with a high probability of detecting them (above and/or near the reefs; Fig 10).
A cost–effective surveying approach targeting reefs is of particular interest in the remote JFA, as finfish species are highly vulnerable and of high subsistence and economic value [32,33]. This work linked rocky reefs’ spatial heterogeneity to fish distribution, establishing the seascape fundamentals for subsequent survey designs and conservation plans within JFA Marine Protected Area. To enhance the prospects of implementing an ecosystem–based conservation plan, gaining a more thorough understanding of the sources and dynamics of physical, chemical, and biological processes across various spatio–temporal scales (see Fig 10; [5,23]). Furthermore, evaluating their influence on conservation efforts [6] and considering the socio–economic aspects of the Juan Fernández community [100–102] can contribute to a more comprehensive outlook and better management.
Supporting information
S1 Table. Model selection for the presence–absence of fish aggregation.
Results of regression models (GAMLSS) evaluated via stepwise inclusion of zone as a fixed factor and its interactions with primary predictors. Covariates: prl represents reef presence, zl represents seafloor depth, distl represents distance to the reef, and s(·) indicates a P–spline smooth function. Model descriptors: df corresponds to degrees of freedom, AIC represents Akaike information criterion, and ΔAIC to the difference relative to the best model. Model performance was evaluated using AIC, with lower values indicating a better fit. Bold values denote the best–fitting model.
https://doi.org/10.1371/journal.pone.0358459.s001
(DOCX)
S2 Table. Parameter estimates for the best–fitting model (M5).
Model coefficient estimates (β), standard errors (Std. Error), t–values, and p–values for the presence–absence of demersal fish aggregations. Zone 2 was the baseline reference level for categorical comparisons. Statistical significance was set at p < 0.05.
https://doi.org/10.1371/journal.pone.0358459.s002
(DOCX)
S3 Table. Spatial and temporal model selection for presence–absence of demersal fish aggregations.
Results of regression models (GAMLSS) evaluating spatial and temporal consistency across zones 1–2 and survey season 2015–2016. Models evaluate an additive season effect as baseline model (ST0) against a triple interaction among the linear primary predictors, zones, and season (ST1). Covariates: prl represents reef presence, zl represents seafloor depth, distl represents distance to the reef, season represents survey year, and s(·) indicates a P–spline smooth function. Model descriptors: df corresponds to degrees of freedom, AIC represents Akaike information criterion, and ΔAIC to the difference relative to the best model. Model performance was evaluated using AIC, with lower values indicating a better fit. Bold values denote the best–fitting model.
https://doi.org/10.1371/journal.pone.0358459.s003
(DOCX)
S4 Table. Parameter estimates for the best–fitting spatial and temporal model (ST1).
Model coefficient estimates (β), standard errors (Std. Error), t–values, and p–values for the presence–absence of demersal fish aggregations. Zone 2 and sampling season 2015 were the baseline reference levels for categorical comparisons. Statistical significance was set at p < 0.05.
https://doi.org/10.1371/journal.pone.0358459.s004
(DOCX)
S5 Table. Model selection for backscatter of fish aggregation.
Results of regression models fitted to the continuous response variable (fish aggregation backscatter) via stepwise inclusion of zone as a fixed factor and its interactions with reef attribute predictors. Covariates: Zrh represents mean reef depth, Lrh to reef length, and Hrh to average reef height, and s(·) indicates a P–spline smooth function. Model descriptors: df corresponds to degrees of freedom, AIC represents Akaike information criterion, and ΔAIC to the difference relative to the best model. Model performance was evaluated using AIC, with lower values indicating a better fit. Bold values denote the best–fitting model.
https://doi.org/10.1371/journal.pone.0358459.s005
(DOCX)
S6 Table. Parameter estimates for the best–fitting model (M10).
Model coefficient estimates (β), standard errors (Std. Error), t–values, and p–values for the backscatter of demersal fish aggregations. The intercept represents the expected global baseline backscatter when reef covariates are zero. Statistical significance was set at p < 0.05.
https://doi.org/10.1371/journal.pone.0358459.s006
(DOCX)
S1 Fig. Assessing reef coverage, length, and density across zones and depths.
Proportion of reef covered area (reef area/ sampled area), reef median length (km), and reef density [(number of reefs/ km2) × 1000−1] by (A) zone and (B) depth strata.
https://doi.org/10.1371/journal.pone.0358459.s007
(TIF)
Acknowledgments
We greatly appreciate the permanent support of the Juan Fernández Archipelago artisanal fishing community (STIPA, APDEAJF). We thank S. González, R. Salas, E. Salas, D. Paredes, W. González, A. Vergara, M. Recabarren, W. Chamorro, I. Chamorro, P. Manríquez and J. Chamorro for their commitment during the project. We acknowledge technical and institutional support from A. Karstegl (SUBPESCA), J. Castillo, S. Lillo, E. Molina, F. Leiva (IFOP), R. Cifuentes and C. Espinoza (SERNAPESCA), and F. Neira (EchoView). We are also grateful to D. McKelvey, S. Urmy, J. Helyer, and XX anonymous reviewers who helped improve earlier versions of this document. We thank F. Gamonal (IMO) for creating Fig 10.
References
- 1. Dunning JB, Danielson BJ, Pulliam HR. Ecological processes that affect populations in complex landscapes. Oikos. 1992;65(1):169.
- 2. Forman RT. Some general principles of landscape and regional ecology. Landsc Ecol. 1995;10(3):133–42.
- 3.
Turner MG, Gardner RH. Landscape Ecology in Theory and Practice: Patterns and Process. 2nd ed. New York: Springer-Verlag; 2015.
- 4. Paine RT, Levin SA. Intertidal landscapes: Disturbance and the dynamics of pattern. Ecol Monogr. 1981;51(2):145–78.
- 5. Carr MH, Neigel JE, Estes JA, Andelman S, Warner RR, Largier JL. Comparing marine and terrestrial ecosystems: Implications for the design of coastal marine reserve. Ecol Appl. 2003;13(1):S90–107.
- 6. Wedding LM, Lepczyk CA, Pittman SJ, Friedlander AM, Jorgensen S. Quantifying seascape structure: Extending terrestrial spatial pattern metrics to the marine realm. Mar Ecol Prog Ser. 2011;427:219–32.
- 7.
Pitcher TJ, Morato T, Hart PJ, Clark MR, Haggan N, Santos RS. Seamounts: ecology, fisheries & conservation. 1st ed. Blackwell Publishing; 2007.
- 8. Ramalho RS, Quartau R, Trenhaile AS, Mitchell NC, Woodroffe CD, Ávila SP. Coastal evolution on volcanic oceanic islands: A complex interplay between volcanism, erosion, sedimentation, sea-level change and biogenic production. Earth Sci Rev. 2013;127:140–70.
- 9. Fisher EE, Choat JH, McCormick MI, Cappo M. Relative influence of environmental factors on the timing and occurrence of multi-species coral reef fish aggregations. PLoS One. 2018;13(12):e0209234. pmid:30576357
- 10. Wessel P, Luis JF, Uieda L, Scharroo R, Wobbe F, Smith WHF. The generic mapping tools version 6. Geochem Geophys Geosys. 2019;20:5556–64.
- 11. Astudillo Manosalva VI. Geomorfología y evolución geológica de la Isla Robinson Crusoe, Archipiélago Juan Fernández. H.BSc Thesis, Universidad de Chile. 2014. Available from: https://repositorio.uchile.cl/handle/2250/117071
- 12. Vergara H, Morales E. Morfología submarina del segmento central del cordón asísmico de Juan Fernández, Pacífico suroriental. Investig Mar. 1985:25–34.
- 13. Araya-Vergara JF, Vieira R. Destructive coastal and submarine landsforms in the volcanic systems of San Félix and Robinson Crusoe, Southeastern Pacific. Cienc Tecnol Mar. 2005;28(1):5–23.
- 14.
Ernst B, Oyarzún C, Vilches J, Rivara P, Tapia B, Álvarez C, et al. Estudio ecosistémico de las pesquerías del sitio piloto Archipiélago Juan Fernández. Fase Metodológica. GEF-UNPD Hacia un manejo ecosistémico del Gran Ecosistema Marino de la Corriente de Humboldt (GEMCH)-PIMS 4147). Final report. Concepción, Chile: Universidad de Concepción, Departamento de Oceanografía; 2015.
- 15.
Ernst B, Rivara P, Tapia B, Santa Cruz F, Espinoza L, Manríquez P, et al. Evaluación directa de la breca (Nemadactylus gayi, Kner 1865) en torno a las Islas Robinson Crusoe y Santa Clara. GEF-UNPD Hacia un manejo ecosistémico del Gran Ecosistema Marino de la Corriente de Humboldt (GEMCH)-PIMS 4147. Final report. Concepción, Chile: Universidad de Concepción, Departamento de Oceanografía; 2016.
- 16. Ahumada M, Queirolo D. Explotación de peces asociada a la pesquería artesanal de langosta de Juan Fernández (Jasus frontalis). Lat Am J Aquat Res. 2014;42(1):213–23.
- 17. Dyer BS, Westneat MW. Taxonomy and biogeography of the coastal fishes of Juan Fernández Archipelago and Desventuradas Islands, Chile. Rev Biol Mar Oceanogr. 2010;45(1):589–617.
- 18.
Presidencia de la República de Chile. Ministerio del Medio Ambiente: Amplía el Área Marina Protegida de Múltiples Usos “Mar de Juan Fernández” y Crea el Parque Marino “Mar de Juan Fernández”. Ley N°19300. Decreto Supremo N°10. 2018.
- 19. Andrade I, Sangrà P, Hormazabal S, Correa-Ramirez M. Island mass effect in the Juan Fernández Archipelago (33°S), Southeastern Pacific. Deep Sea Res Part I: Oceanogr Res Pap. 2014;84:86–99.
- 20. Andrade I, Hormazábal S, Combes V. Intrathermocline eddies at the Juan Fernández Archipelago, southeastern Pacific Ocean. Lat Am J Aquat Res. 2014;42(4):888–906.
- 21. Medel C, Parada C, Morales CE, Pizarro O, Ernst B, Conejero C. How biophysical interactions associated with sub- and mesoscale structures and migration behavior affect planktonic larvae of the spiny lobster in the Juan Fernández Ridge: A modeling approach. Prog Oceanogr. 2018;162:98–119.
- 22. Cornejo-Guzmán S. Mecanismos de mesoescala y variables biogeoquímicas que determinan la variabilidad de clorofila-a en el Archipiélago de Juan Fernández. H.BSc Thesis, Universidad de Concepción. 2018. Available from: https://www.dgeo.udec.cl/wp-content/uploads/2020/12/HPSCornejo_2018_300dpi.pdf
- 23. Parada C, Ernst B, Cornejo-Guzmán S, Santa Cruz F, Sandoval Belmar M, Rivara P. Local and remote physical processes driving variability of the planktonic system in the Juan Fernández Archipelago: A multidisciplinary framework providing conservation insights. Aquat Conserv. 2021;31:253–72.
- 24. Holbrook SJ, Kingsford MJ, Schmitt RJ, Stephens JSJ. Spatial and temporal patterns in assemblages of temperate reef fish. Am Zool. 1994;34(3):463–75.
- 25. Jones DT, Wilson CD, De Robertis A, Rooper CN, Weber TC, Butler JL. Evaluation of rockfish abundance in untrawlable habitat: combining acoustic and complementary sampling tools. Fish Bull. 2012;110(3):332–43.
- 26. Wilborn R, Rooper CN, Goddard P, Li L, Williams K, Towler R. The potential effects of substrate type, currents, depth and fishing pressure on distribution, abundance, diversity, and height of cold-water corals and sponges in temperate, marine waters. Hydrobiologia. 2018;811(1):251–68.
- 27. Follana-Berná G, Palmer M, Campos-Candela A, Arechavala-Lopez, Diaz-gil C, Alós J. Estimating the density of resident coastal fish using underwater cameras: accounting for individual detectability. Mar Ecol Prog Ser. 2019;615:177–88.
- 28. Domokos R. On the development of acoustic descriptors for semi-demersal fish identification to support monitoring stocks. ICES J Mar Sci. 2021;78(3):1117–30.
- 29. Scoulding B, Gastauer S, Taylor JC, Boswell KM, Fairclough DV, Jackson G. Estimating abundance of fish associated with structured habitats by combining acoustic and optics. J Appl Ecol. 2023;60(7):1274–85.
- 30. Ernst B, Manríquez P, Orensanz JM (Lobo), Roa R, Chamorro J, Parada C. Strengthening of a traditional territorial tenure system through protagonism in monitoring activities by lobster fishermen from the Juan Fernández Islands, Chile. Bull Mar Sci. 2010;86(2):315–38.
- 31. Arana P, Melo CA. La langosta de Juan Fernández: pesca comercial de Jasus frontalis en las Islas Robinson Crusoe y Santa Clara (1971-1972). Investig Mar. 1973;4(5):135–54.
- 32. Ernst B, Chamorro J, Manríquez P, Orensanz JML, Parma AM, Porobic J, et al. Sustainability of the Juan Fernández lobster fishery (Chile) and the perils of generic science-based prescriptions. Global Environmental Change. 2013;23(6):1381–92.
- 33. Plan de Desarrollo Comunal Archipiélago Juan Fernández. Caracterización y diagnóstico comunal 2015-2020. Final report. Ilustre Municipalidad de Juan Fernández, Valparaíso, Chile, 2015. Available from: http://www.comunajuanfernandez.cl/municipalidad/pladeco/
- 34. Steneck R, Parma AM, Ernst B, Wilson JA. Two lobster tales: lessons from the convergent evolution of TURFs in Maine (USA) and the Juan Fernández Islands (Chile). Bull Mar Sci. 2017;93(1):13–33.
- 35. García Charton JA, Williams ID, Pérez Ruzafa A, Milazzo M, Chemello R, Marcos C. Evaluating the ecological effects of Mediterranean marine protected areas: habitat, scale and the natural variability of ecosystems. Environ Conserv. 2000;27(2):159–78.
- 36. Stamoulis KA, Delevaux JMS, Williams ID, Poti M, Lecky J, Costa B, et al. Seascape models reveal places to focus coastal fisheries management. Ecol Appl. 2018;28(4):910–25. pmid:29421847
- 37.
Simmonds J, MacLennan D. Fisheries acoustics: theory and practice. 2nd ed. Oxford: Wiley-Blackwell; 2005.
- 38. Friedlander AM, Ballesteros E, Caselle JE, Gaymer CF, Palma AT, Petit I, et al. Marine biodiversity in Juan Fernández and Desventuradas Islands, Chile: Global endemism hotspots. PLoS One. 2016;11(1):e0145059. pmid:26734732
- 39. Pérez-Matus A, Ramirez F, Eddy TD, Cole R. Subtidal reef fish and macrobenthic community structure at the temperate Juan Fernández Archipelago, Chile. Lat Am J Aquat Res. 2014;42(4):814–26.
- 40. Ramírez F, Pérez-Matus A, Eddy TD, Landaeta MF. Trophic ecology of abundant reef fish in a remote oceanic island: coupling diet and feeding morphology at the Juan Fernandez Archipelago, Chile. J Mar Biol Ass. 2013;93(6):1457–69.
- 41. Brown CJ, Smith SJ, Lawton P, Anderson JT. Benthic habitat mapping: A review of progress towards improved understanding of the spatial ecology of the seafloor using acoustic techniques. Estuar Coast Shelf Sci. 2011;92(3):502–20.
- 42.
Arana P. Historical perspective and projections of the fisheries activity in the Archipelago of Juan Fernández, Chile. In: Castilla JC, editor. Islas oceánicas chilenas, conocimiento científico y necesidades de investigaciones. Ediciones Universidad Católica de Chile; 1987. pp. 319–53.
- 43.
Castilla JC, Oliva D. Islas oceánicas chilenas: Aspectos descriptivos y potencialidades. In: Castilla JC, editor. Islas oceánicas chilenas, conocimiento científico y necesidades de investigaciones. Ediciones Universidad Católica de Chile; 1987. pp. 17–35.
- 44.
Foote KG, Knudsen HP, Vestnes G, MacLennan DN, Simmonds EJ. Calibration of acoustic instruments for fish density estimation: a practical guide. International Council for the Exploration of the Sea (ICES); 1987.
- 45.
Echoview Software Pty Ltd. Version 8.0.105 [software]. Hobart, Tasmania, Australia: 2018. Available from: https://echoview.com/
- 46. Maclennan D. A consistent approach to definitions and symbols in fisheries acoustics. ICES J Mar Sci. 2002;59(2):365–9.
- 47. Rigby RA, Stasinopoulos DM. Generalized additive models for location, scale and shape. J Appl Stat. 2005;54(3):507–54.
- 48. Stasinopoulos DM, Rigby RA. Generalized Additive Models for Location Scale and Shape (GAMLSS) in R. J Stat Softw. 2007;23(7):1–44.
- 49.
R Core Team. R: A language and environment for statistical computing [software]. Vienna, Austria: 2020. Available from: https://www.r-project.org/
- 50. Eilers PHC, Marx BD, Durbán M. Twenty years of P-splines. Sort. 2015;39(2):149–86.
- 51.
Stasinopoulos MD, Rigby RA, Heller GZ, Voudouris V, De Bastianni F. Flexible Regression and Smoothing: Using GAMLSS in R. 1st ed. Chapman and Hall; 2017.
- 52. Whittingham MJ, Stephens PA, Bradbury RB, Freckleton RP. Why do we still use stepwise modelling in ecology and behaviour? J Anim Ecol. 2006;75(5):1182–9. pmid:16922854
- 53.
Venables WN, Ripley BD. Modern Applied Statistics with S. 4 ed. New York: Springer; 2002.
- 54.
Akaike H. Information theory and an extension of the maximum likelihood principle. In: Petrov BN, Csaki BF, editors. 2nd International Symposium on Information Theory. Budapest: Academiai Kiado; 1973. pp. 267–81.
- 55. Burnham KP, Anderson DR. Multimodel inference. Understanding AIC and BIC in model selection. Sociol Methods Res. 2004;33(2):261–304.
- 56. Walker BK, Jordan LKB, Spieler RE. Relationship of reef fish assemblages and topographic complexity on southeastern Florida coral reef habitats. J Coast Res. 2009;25(53):39–48.
- 57. Thorson JT, Rindorf A, Gao J, Hanselman DH, Winker H. Density-dependent changes in effective area occupied for sea-bottom-associated marine fishes. Proc Biol Sci. 2016;283(1840):20161853. pmid:27708153
- 58. Grober-Dunsmore R, Frazer TK, Beets JP, Lindberg WJ, Zwick P, Funicelli NA. Influence of landscape structure on reef fish assemblages. Landscape Ecology. 2008;23:37–53.
- 59. Lirman D. Reef fish communities associated with Acropora palmata: Relationships to benthic attributes. Bull Mar Sci. 1999;65(1):235–52.
- 60. Chouinard P-M, Dutil J-D. The structure of demersal fish assemblages in a cold, highly stratified environment. ICES J Mar Sci. 2011;68(9):1896–908.
- 61. Pequeño G, Saez S. Los peces litorales del archipiélago de Juan Fernández (Chile): endemismo y relaciones ictogeográficas. Investig Mar. 2000;28:27–37.
- 62. Bundy A, Chuenpagdee R, Boldt JL, de Fatima Borges M, Camara ML, Coll M, et al. Strong fisheries management and governance positively impact ecosystem status. Fish Fish. 2016;18(3):412–39.
- 63. Cooke SJ, Wesch S, Donaldson LA, Wilson ADM, Haddaway NR. A call for evidence-based conservation and management of fisheries and aquatic resources. Fisheries. 2017;42(3):143–9.
- 64.
Ernst B, Salas Aguayo E, Porobic J, Román C, Chamorro J, Manríquez P. Monitoreo de las principales pesquerías del Archipiélago Juan Fernández, año 2012. 4728-36-LE12. Chile: Subsecretaría de Pesca-Universidad de Concepción; 2013.
- 65.
Ernst B, Parada C, Chamorro J, Pino Aguilera J, Tapia B, Cornejo S. Monitoreo biológico pesquero de las pesquerías y del ecosistema asociado al Archipiélago Juan Fernández, 2021. Chile: Subsecretaría de Pesca y Acuicultura-Universidad de Concepción; 2022.
- 66. Ying Y, Chen Y, Lin L, Gao T. Risks of ignoring fish population spatial structure in fisheries management. Can J Fish Aquat Sci. 2011;69:2101–20.
- 67. Cadrin SX. Defining spatial structure for fishery stock assessment. Fish Res. 2020;221(105397):1–12.
- 68. Han Q, Grüss A, Shan X, Jin X, Thorson JT. Understanding patterns of distribution shift and range expansion/contraction for small yellow croaker (Larimichthys polyactis) in the Yellow Sea. Fish Oceanogr. 2020;30(1):69–84.
- 69. Parrish FA, Boland RC. Habitat and reef-fish assemblages of banks in the Northwestern Hawaiian Islands. Mar Biol. 2004;144:1065–73.
- 70. Sambrook K, Hoey AS, Andréfouët S, Cumming GS, Duce S, Bonin MC. Beyond the reef: The widespread use of non‐reef habitats by coral reef fishes. Fish Fish. 2019;20(5):903–20.
- 71. Pittman SJ, Caldow C, Hile SD, Monaco ME. Using seascape types to explain the spatial patterns of fish in the mangroves of SW Puerto Rico. Mar Ecol Prog Ser. 2007;348:273–84.
- 72.
Hearn CJ. Hydrodynamics of coral reef systems. In: Hopley D, editor. Encyclopedia of Earth Sciences Series. Dordrecht: Springer Netherlands; 2011. pp. 563–73. https://doi.org/10.1007/978-90-481-2639-2_277
- 73. Harborne AR, Mumby PJ, Kennedy EV, Ferrari R. Biotic and multi-scale abiotic controls of habitat quality: Their effect on coral-reef fishes. Mar Ecol Prog Ser. 2001;437:201–14.
- 74.
Pyle RL, Copus JM. Mesophotic Coral Ecosystems: Introduction and Overview. In: Loya Y, Puglise KA, Bridge TC, editors. Mesophotic Coral Ecosystems. Coral Reefs of the World. 2019. pp. 3–27.
- 75. Moum J, Nash J, Klymak J. Small-scale processes in the Coastal Ocean. Oceanog. 2008;21(4):22–33.
- 76. Candy AS, Shannara K, Parkins T, Van Duyl FC, Mueller B, Arts MGI. Small-scale oxygen distribution patterns in a coral reef. Front Mar Sci. 2023;10:1135686.
- 77.
Rogers AD. Chapter Four - The Biology of Seamounts: 25 Years on. In: Sheppard C, editor. Advances in Marine Biology. Academic Press; 2018. pp. 137–224.
- 78. Ma J, Song J, Li X, Wang Q, Zhong G. Multidisciplinary indicators for confirming the existence and ecological effects of a Taylor column in the Tropical Western Pacific Oceanic. Ecol Indic. 2021;127:107777.
- 79. Eggertsen L, Hammar L, Gullström M. Effects of tidal current-induced flow on reef fish behaviour and function on a subtropical rocky reef. Mar Ecol Prog Ser. 2016;559:175–92.
- 80. Reid EC, deCarlo TM, Cohen AL, Wong GTF, Lentz SJ, Safaie A, et al. Internal waves influence the thermal and nutrient environment on a shallow coral reef. Limnol Oceanogr. 2019;64(5):1949–65.
- 81. Hattori A, Shibuno T. The effect of patch reef size on fish species richness in a shallow coral reef shore zone where territorial herbivores are abundant. Ecol Res. 2009;25(2):457–68.
- 82. Wagner ELES, Roche DG, Binning SA, Wismer S, Bshary R. Temporal comparison and predictors of fish species abundance and richness on undisturbed coral reef patches. PeerJ. 2015;3:e1459. pmid:26644988
- 83. Kuffner IB, Grober-Dunsmore R, Brock JC, Hickey TD. Biological community structure on patch reefs in Biscayne National Park, FL, USA. Environ Monit Assess. 2010;164(1–4):513–31. pmid:19399634
- 84. Folpp HR, Schilling HT, Clark GF, Lowry MB, Maslen B, Gregson M. Artificial reefs increase fish abundance in habitat-limited estuaries. Appl Ecol. 2020;57:1752–61.
- 85.
Costa BM, Tormey S, Battista TA. Benthic Habitats of Buck Island Reef National Monument. NOS NCCOS 142. NCCOS Center for Coastal Monitoring and Assessment Biogeography Branch; 2023.
- 86. Carr MH. Habitat selection and recruitment of an assemblage of temperate zone reef fishes. J Exp Mar Biol Ecol. 1991;146(1):113–37.
- 87. Gregor CA, Anderson TW. Relative importance of habitat attributes to predation risk in a temperate reef fish. Environ Biol Fish. 2016;99(6–7):539–56.
- 88. Conell SD. Effects of predators on growth, mortality and abundance of a juvenile reef-fish: evidence from manipulations of predators and prey abundance. Mar Ecol Prog Ser. 1998;169:251–61.
- 89. Kavanaugh MT, Oliver MJ, Chavez FP, Letelier RM, Muller-Karger FE, Doney SC. Seascapes as a new vernacular for pelagic ocean monitoring, management and conservation. ICES J Mar Sci. 2016;73(7):1839–50.
- 90. Quattrini AM, Demopoulos AWJ, Singer R, Roa-Varon A, Chaytor JD. Demersal fish assemblages on seamounts and other rugged features in the northeastern Caribbean. Deep Sea Res Part I: Oceanogr Res Pap. 2017;123:90–104.
- 91. Parada C, Gretchina A, Vásquez S, Belmadani A, Combes V, Ernst B, et al. Expanding the conceptual framework of the spatial population structure and life history of jack mackerel in the eastern South Pacific: an oceanic seamount region as potential spawning/nursery habitat. ICES J Mar Sci. 2017;74(9):2398–414.
- 92. Correa-Ramírez MA, Hormazábal S, Yuras G. Mesoscale eddies and high chlorophyll concentrations off central Chile (29°–39°S). Geophys Res Lett. 2007;34(12):L12604.
- 93. Correa-Ramirez MA, Hormazabal SE, Morales CE. Spatial patterns of annual and interannual surface chlorophyll-a variability in the Peru–Chile Current System. Progr Oceanogr. 2012;92–95:8–17.
- 94. Watson DL, Harvey ES, Anderson MJ, Kendrick GA. A comparison of temperate reef fish assemblages recorded by three underwater stereo-video techniques. Mar Biol. 2005;148:415–25.
- 95. Weber TC, Rooper C, Butler J, Jones D, Wilson C. Seabed classification for trawlability determined with a multibeam echo sounder on Snakehead Bank in the Gulf of Alaska. Fish Bull. 2013;111:68–77.
- 96. Pirtle JL, Weber TC, Wilson CD, Rooper CN. Assessment of trawlable and untrawlable seafloor using multibeam-derived metrics. Methods Oceanogr. 2015;12:18–35.
- 97. Gauthier S. Target Strength of encaged Atlantic redfish (Sebastes spp.). ICES J Mar Sci. 2001;58(3):562–8.
- 98. Rooper CN, Williams K, De Robertis A, Tuttle V. Effect of underwater lighting on observations of density and behavior of rockfish during camera surveys. Fish Res. 2015;172:157–67.
- 99. Shortis MR, Seager JW, Williams A, Barker BA, Sherlock M. Using stereo-video for deep water benthic habitat surveys. Mar Technol Soc J. 2008;42:28–37.
- 100. Melbourne-Thomas J, Constable AJ, Fulton EA, Corney SP, Trebilco R, Hobday AJ, et al. Integrated modelling to support decision-making for marine social–ecological systems in Australia. ICES J Mar Sci. 2017;74(9):2298–308.
- 101. Selig ER, Hole DG, Allison EH, Arkema KK, McKinnon MC, Chu J. Mapping global human dependence on marine ecosystems. Conserv Lett. 2019;:1–10.
- 102. Porobic J, Fulton EA, Parada C, Frusher S, Ernst B, Manríquez P. The impact of fishing on a highly vulnerable ecosystem, the case of Juan Fernández Ridge ecosystem. PLoS One. 2019;14(2):e0212485. pmid:30794609