Monitoring ecological dynamics on complex hydrothermal structures: A novel photogrammetry approach reveals fine-scale variability of vent assemblages
Author Contribution Statement: J.S. and M.M. conceived this study. J.S., M.M., M.C. and L.V.A. participated in the collection of video sequences. A.A. and M.M. developed the photogrammetry and annotation software. A.A., M.M. and L.V.A. performed photogrammetry reconstruction. L.V.A. performed image annotations for models of 2018 and 2020. G.P., P.L., M.C., J.S., M.M. and L.V.A. performed analyses. L.V.A. wrote the manuscript. All authors reviewed the manuscript.
Abstract
We set out to characterize the fine-scale processes acting on interannual dynamics of deep-sea vent fauna by using a novel approach involving a 5-yr time series of 3D photogrammetry models acquired at the Eiffel Tower sulfide edifice (Lucky Strike vent field, Mid-Atlantic Ridge). Consistently, with the overall stability of the vent edifice, total mussel cover did not undergo drastic changes, suggesting that they have been at a climax stage for at least 25 yr based on previous data. Successional patterns showed consistency over time, illustrating the dynamic equilibrium of the ecological system. In contrast, microbial mats significantly declined, possibly due to magmatic events. The remaining environmental variability consisted of decimeter-scale displacement of vent outflows, resulting from their opening or closure or from the progressive accretion of sulfide material. As a result, vent mussels showed submeter variability in the immediate vicinity of vent exits, possibly by repositioning in response to that fine-scale regime of change. As former studies were not able to quantify processes at submeter scales in complex settings, this pioneering work demonstrates the potential of 3D photogrammetry models for conducting long-term monitoring in the deep sea. We observed that the ability of mussels to displace may enable them to cope with changing local conditions in a stable system. However, the long-term stability of mussel assemblages questions their capacity to withstand large-scale disturbances and may imply a low resilience of these “climax” communities. This suggests that they may be particularly vulnerable to the negative effects of mining activities in hydrothermal ecosystems.
Following the growing interest in seafloor massive sulfide deposits found at hydrothermal vents, potential environmental impacts must be evaluated for conservation purposes (Boschen et al. 2013). The acquisition of ecological knowledge to explain and predict the behavior of ecological systems requires the identification of the patterns, scales, and forces driving variability observed in natural communities over space and time (Sousa 1984). Ultimately, the characterization of functional traits allowing species to cope with that variability to maintain their populations can be used as baseline information for conservation purposes (Van Dover 2014).
Hydrothermal vents are characterized by the emission of heated fluids at the seafloor in ridges or basins displaying tectonic and volcanic activity. Venting can occur in the form of focused black or white smokers or diffuse outflows, each exhibiting different temperatures and concentrations of reduced minerals. Fluid composition and strength of venting are strongly influenced by the geological nature of the sub-seafloor, the geodynamical context of the region, and the trajectory of seawater within the oceanic crust, including chemical reactions and phase separation induced by magmatism as well as by the pathway of percolation and advection that influences mixing processes (Barreyre et al. 2014; Chavagnac et al. 2018). Vent ecosystems support lush endemic communities that feature low species diversity and exceptionally high biomass that thrive despite the harshness of environmental conditions (Tunnicliffe 1991). Vent communities rely on chemoautotrophic microorganisms that harness energy from the oxidation of chemical compounds (H2S, CH4, Fe, and H2) contained in the fluids, a process called chemosynthesis (Jannasch 1985). The vent habitat is highly heterogeneous, offering a variety of small-scale niches created by the vent fluid dilution gradient (Podowski et al. 2009; Lee et al. 2015). Species distributions are mainly driven by their physiological tolerance, nutritional needs, and biotic interactions (Mullineaux et al. 2003; Henry et al. 2008; Podowski et al. 2009).
In addition to this spatial heterogeneity, temporal dynamics interplay and contribute to the observed variability in ecosystems in general (Sousa 1984; Levin 1992). At vents, temporal variations and successions are typically monitored following interannual photographic transects to map megafaunal assemblages over time while avoiding the destructive effect of dredge or grab sampling (Sarrazin et al. 1997; Cuvelier et al. 2012). At intermediate to fast-spreading mid-ocean ridges (> 50–140 mm yr−1), monitoring studies have usually started from a reference time after a broad-scale disturbance event (e.g., volcanic eruption) resetting communities to an early stage of colonization (Shank et al. 1998). In the absence of massive disturbance on ultra-slow to slow-spreading ridges (< 20–50 mm yr−1), temporal series of images have underlined the small spatial scales (i.e., meter to submeter) at which interannual variability of faunal assemblages occurs on sulfide edifices (Copley et al. 2007; Sen et al. 2014; Du Preez and Fisher 2018).
However, for structures with complex topography, the detection of small-scale changes based on the analysis of two-dimensional (2D) images could be highly biased from a 3D perspective because of the resulting decrease in the accuracy of cover annotations, and the lack of precision of space–time observation tracking. Our ability to detect and explain fine-scale changes all warrant the use of 3D representations (Sarrazin and Juniper 1998). Structure-from-motion photogrammetry (Moulon et al. 2017) is an imaging technique used to build centimeter-scale resolution 3D mesh and textured models from 2D images acquired with a standard camera mounted on an underwater vehicle (Kwasnitschka et al. 2013). As structure-from-motion has become increasingly used for the past decade for underwater ecology (Pulido Mantas et al. 2023), it is employed for high-resolution mapping of deep-sea communities in complex topographical settings, including vent edifices and vertical walls (Gerdes et al. 2019; Robert et al. 2020). Since photogrammetry surveys have been successful at monitoring centimeter-scale changes in terrestrial and marine environments (D'Urban Jackson et al. 2020), their application to long-term monitoring of deep-sea communities could help unravel fine-scale processes of variability still largely unexplored by 2D imagery and integrate them in a broader knowledge of landscape ecology (Paine and Levin 1981).
This work pioneers the use of time series of high-resolution 3D models to retrieve accurate quantitative measurements and integrate different scales of dynamics of a complex deep-sea environment, from the decimeters to hundreds of meter scales. More specifically, this study aims to assess and quantify the temporal dynamics of the vent fauna and habitat on a deep-sea sulfide edifice and to characterize the processes influencing faunal variability.
Materials and methods
Study site
Lucky Strike is a basalt-hosted hydrothermal vent field located south of the Azores Islands on the Mid-Atlantic Ridge (MAR; Fig. 1A; Humphris et al. 2002), where a deep-sea observatory (EMSO-Azores) has been deployed for over a decade. Since then, yearly maintenance cruises have provided additional time series data (Matabos et al. 2022). Lucky Strike hosts the Eiffel Tower in the South-East, a mature hydrothermal edifice of ~ 450 m2 that comprises a central ~ 11-m-high sulfide tower harboring focused venting activity surrounded by a mix of sulfide deposits and a sulfide-indurated slab displaying numerous cracks and flanges (Fig. 1B; Girard et al. 2020).
The Eiffel Tower topography is complex, with 5-m-high bulbous mounds at the base of the main vertical tower and inactive chimney structures located in the NW periphery (Fig. 1B,C). The fauna is characterized by patchy assemblages of vent-endemic invertebrates occupying distinct habitats along the vent fluid dilution gradient. At the Eiffel Tower, assemblages of Mirocaris fortunata shrimps and Peltospira smaragdina gastropods occupy the warmest habitats while colder areas host the symbiotic mussels Bathymodiolus azoricus dominating 90% of the biomass (Husson et al. 2017; Sarrazin et al. 2022). As exposure to the vent fluid declines with distance, the length of mussel individuals decreases, although their density and the richness of associated species increase (Sarrazin et al. 2020). Assemblages can be covered by white filamentous microbial mats dominated by sulfur-oxidizing Beggiatoa Gammaproteobacteria that benefit from the distal exposure to black smokers mediated by bottom currents and edifice topography (Crépeau et al. 2011; Girard et al. 2020). Dense aggregations of non-vent fauna and empty mussel shells are observed in the coldest habitats (Husson et al. 2017; Girard et al. 2020).
Temporal variability of Eiffel Tower vent communities has been studied over various scales. Previous work captured the impacts of infra-daily tidal modulation and biological interactions at the scale of a m2 mussel assemblage (Matabos et al. 2015; Mat et al. 2020), while remarkable stability was observed from monthly to 7-yr time windows (Cuvelier et al. 2017; Van Audenhaege et al. 2022). Pluriannual transects of 2D images from 1994 to 2008 suggested stability of the mussel assemblages at ~ 100 m2 over the edifice (Cuvelier et al. 2011b). Based on this work, we expect high temporal stability of vent faunal assemblages.
Building the dataset
Video sequences used for 3D reconstruction were acquired during three MoMARSAT cruises (Cannat and Sarradin 2010) by carrying out transects with a forward-looking camera mounted on the ROV Victor6000 in April 2015 (HD videos for 4.33 h; Sarradin and Cannat 2015), August 2018 (HD videos for 5.23 h; Cannat 2018), and September 2020 (4K videos for 3.28 h; Sarradin and Legrand 2020). Lateral repetition of vertical video transects aimed to image all sides of the edifice from the summit to the base, by maintaining the ROV—insofar as possible—at a constant pace and distance from the edifice to produce overlap between vertical transects (i.e., at least 70% of image area). Images were extracted from video sequences every 3 s and color-corrected in the Matisse software (v.1.4.0; Arnaubec et al. 2023). These images were used to reconstruct three different textured 3D meshes of the Eiffel Tower for each survey year, using structure-from-motion techniques embedded in Matisse. The 3D reconstructions were georeferenced based on ROV navigation to provide (1) metric measurements of the surface and (2) spatial alignment (i.e., registration) among the 3D reconstructions. Since offsets sometimes larger than 1 m remained between georeferenced meshes, registration was performed by manual alignment of the models of 2018 and 2020 on the reconstruction of 2015 as reference. At least five corresponding points visibly stable through time were matched across models to compute the geometric transformation (“Align” tool of CloudCompare v.2.11). That task was performed on seven model portions of 2018 and 2020 cropped in “sides”, as defined in the model of 2015 by Girard et al. (2020), originally to allow a smooth computer display. The final registration error among sides was assessed by generating 30 ground points equispaced by 4 m on the 2015 model (“Subsample” in CloudCompare). Close to each of the 30 points, a stable seabed feature was annotated in all models to compute offsets as a proxy for the registration error (Cucchiaro et al. 2018).
Annotation of the 3D models
To assess habitat and community variability within Eiffel Tower, the position of three types of observations was recorded over time: (1) substratum and biological cover, (2) vent outflows, and (3) topographic changes. All annotations were performed on the 3D mesh following the interpretation of textures, color, and spatial arrangement by loading the three models simultaneously as layers in 3DMetrics software (v.0.6.0; Arnaubec et al. 2023). ROV video sequences were used to confirm annotation.
Biological assemblages and substrata were annotated with polygons and defined based on existing catalogs already used for annotation of the 2015 model (catalog in Supporting Information Fig. S1; Girard et al. 2020). Substratum categories were interpreted from their color, ranging from “black material” (recent sulfide-bearing deposits) to “white material” (anhydrite [CaSO4] and barite [BaSO4] deposits) or brownish oxidized sulfide-bearing material (hereafter “bare substratum”). Biological assemblages included the symbiotic mussel B. azoricus classified as small (mean length ± SD: 1.1 ± 0.6 cm), medium (3.3 ± 1.5 cm), or large individuals (6.0 ± 1.4 cm), zoanthids and empty shells (Cuvelier et al. 2009; see Girard et al. 2020 for measurements of mussel size classes). No differentiation was made between microcarnivorous sponges (i.e., possibly Asbestopluma sp., Cladorhizidae) and arboramminid foraminifera (i.e., possibly Luffammina atlantica); they were annotated as erect white morphospecies (Desbruyères et al. 2006). The additional presence of microbial mats on top of the assemblage/substratum was labeled. Following this protocol, each polygon was labeled with the time series of assemblage/substratum that the polygon contained in the three models. Annotation of cover stopped when only stable bare substratum remained unannotated. Independent annotations of M. fortunata and P. smaragdina assemblages, and individuals of the crab Segonzacia mesatlantica were not included in analyses, because of shrimp and crab mobility and high infra-annual dynamics, and P. smaragdina low abundance (only one stable assemblage of 0.04 m2). ROV-induced displacement (e.g., avoidance or thruster blowing) and imagery detection difficulties (i.e., low size, hiding in mussel assemblages) may also occur (Cuvelier et al. 2012).
Vent fluid outflows were classified based on plume geometry, intensity, and color observed in ROV video sequences. They were relocated on the 3D models as points for focused vents, if the fluid was released from a discrete vent orifice, as points for diffusion zones, if only shimmering water was observed, and as lines in the case of flanges. Focused vents were sub-categorized based on the color of the venting fluid as (1) “black smoker” or (2) “white smoker.”
Topographic changes were annotated as polygons after the detection of a modification in the mesh surface between 3D models. The process involved was specified from two main criteria: (1) if some material was lost or built up and (2) if the change involved material either exposed to venting and at least recently exposed (black or white material) or that remained inactive (bare substratum). In the case of flange accretion, the rate was determined from the average distance of flange progression, divided by the time separating both observations.
Polygon/point/line label, XYZ coordinates of the vertices and surface/length measured in the mesh were saved in .json files.
Data analyses
Dynamics of cover and succession among assemblages/substrata were monitored by quantifying the total cover through time as well as the amount of assemblage cover transiting from one to another over the edifice.
Spatial patterns of assemblage change were investigated by pooling polygons in tiles, as polygons are not a standardized surface of sampling. Permanent bare substratum and undetermined cover were excluded from pooling. The total surface of the polygons pooled in a tile could not exceed the largest polygon annotated (i.e., 2.4 m2). The pooling was coded to account for spatial proximity among polygons; tiles, therefore, always formed continuous areas over space. Tiles covered 1.8 ± 0.4 m2 (mean ± SD). Within a tile, a given polygon could only experience two types of dynamics: it remained unchanged (indicating “stability”) or changed once (indicating “variability”). Those dynamics were computed by summing the surface of unchanged/changed polygons within a tile and for each assemblage. Tile variability can be further broken down as a combination of “net change” and “null change”. The net change describes a local “gain” and/or “loss” of cover. The net change is obtained by subtracting the assemblage cover in 2015 from that of 2020 for each tile. The null change measures the surface of cover loss or gain of 2020 that was compensated by a cover gain or loss, respectively. The null change is measured by the difference between the total surface of change and the absolute value of the net change within a tile from 2015 to 2020. Those variables can be nested as circles of radius and width proportional to the area represented: the inner circle representing the net change, the inner torus representing the null change, and the outer one displaying unchanged cover. That symbology was plotted over space to assess patterns of stability and variability by locating tiles on the Eiffel Tower with their XYZ centroids.
Monitoring over 25 yr
To extend the monitoring up to 25 yr, we used 2D annotations obtained by Cuvelier et al. (2011b) on pluriannual photomosaics of the seven sides of the Eiffel Tower from 1994 to 2008. The surface of each 2D polygon (measured in pixels) was extracted from .psd annotation files in Python (v.3.7.4) and was summed by assemblage and substratum. Because covers were in pixels (1994–2008) and in square meters (2015–2020), they were standardized by dividing the cover annotated by the total surface annotated for each side of the edifice to obtain proportional and relative covers. A univariate regression tree (URT) was applied to detect breakpoints in the time series of mussel and microbial mat covers, with the lowest possible relative error (RE) and cross-validated relative error (CVRE; Borcard et al. 2018).
Results
Registration accuracy
The distance among pairs of 3D models reached 0.06 ± 0.05 m (mean ± SD) and rarely exceeded 0.15 m (percentile 95% = 0.14 m).
Habitat changes
Changes in topography—rates of topographic change were 10.4 m2 between 2015 and 2018 and 5.0 m2 from 2018 to 2020. In total, this change accounted for only 4.4% of the surface annotated (Table 1). Over the whole time series, 77.9% of topographic change was related to active or recently active substratum. Overall, the majority of topographic change (77.3%) involved the building up of material; predominantly related to 107 hydrothermal features, either the directional growth of active geological features such as outcrops or flanges overhanging or blanketing the seabed (Table 1; Supporting Information Figs. S2–S4). Rate of flange accretion reached 0.05 ± 0.04 m yr−1 (mean ± SD) with a maximum value of 0.18 m yr−1 (Supporting Information Fig. S3). Voluminous changes at the base of large bulbous outcrops were related to the accretion of material in areas from which shimmering water was emitted (e.g., Supporting Information Fig. S5). Only vertical spires underwent repetitive growth and collapse (Table 1). Change in the inactive bare substratum was sporadic (Table 1) and was related to the fall of boulders from vertical faces (up to 0.8 m2 for one observation).
Net change | Substratum characteristics | Surface of change [m2] (n = number of features) | Process inferred from observations | |
---|---|---|---|---|
2015–2018 | 2018–2020 | |||
+ | Active | 0.4 (11) | 0.4 (9) | Vertical accretion of spires |
4.0 (31) | 3.4 (26) | Lateral/downward accretion of 3D outcrops | ||
1.6 (19) | 0.7 (11) | Forward and downward accretion of linear features (flanges) | ||
Inactive | 1.4 (19) | 0.0 (0) | Debris build-up at the base of the edifice as scree or chimney fragments | |
Total | 7.4 (80) | 4.5 (46) | ||
– | Active | 0.9 (21) | 0.4 (10) | Chimney collapse |
0.1 (1) | 0.1 (2) | Outcrop detachment | ||
Inactive | 2.0 (18) | 0.0 (0) | Boulder detachment or scree displacement | |
Total | 3.0 (40) | 0.5 (12) |
Changes in the distribution of venting features—hydrothermal outflows concentrated within clusters of ~ 5 to ~ 30 m2 that remained stable over time. Within those clusters, outflow distribution appeared to reposition over submeter scales (Fig. 2). In the South-East, we observed the activation of vent outflow with white material clearly emerging from a soft-sediment talus (Supporting Information Fig. S6).
Activity clusters consisted of a mix of diffusion zones, with white and black smokers often interspaced with flanges (Fig. 2). Counted black smokers were variable from 2015 (n = 31), 2018 (n = 17) to 2020 (n = 39). Overall, the number of white smokers doubled in 2015 (n = 50), stabilizing in 2018 (n = 112) and 2020 (n = 119). The number of flanges slightly increased from 70, 87, and 89 in 2020. Diffusion zones were stable from 2015 to 2018 (n = 33–37) and dropped dramatically in 2020 (n = 17).
Faunal and substratum dynamics
Total cover and successions among assemblages followed similar dynamics over time, between 2015 and 2018 and between 2018 and 2020, respectively (Supporting Information Fig. S7). As a result, we summarized them by summing flows of 2015–2018 and 2018–2020 together and by presenting the net change of total cover from 2015 to 2020 (Fig. 3).
Recently active substrata were predominantly represented by white material (~ 90%) that underwent a loss of 8.8 m2 from 2015 to 2020. On average, the total area covered by all mussel assemblages increased from 254.7 m2 in 2015 to 269.7 m2 in 2020 to represent 59.7% of the total surface of the edifice (Fig. 3). The most abundant medium-mussel assemblages decreased slightly from 2015 to 2020 (−13.2 m2), while the surface occupied by large- and small-mussel assemblages had gradually increased by 2020 of +9.4 and +18.8 m2, respectively. In the periphery, zoanthids erected sessile species, and empty shells barely underwent any change over 5 yr. Microbial mat cover underwent a considerable cover loss of 52.9% from 2015 to 2020 (−66.4 m2; Fig. 3).
The white material predominantly converted into bare substratum (8.9 m2) or was colonized by large (5.3 m2) and medium (6.8 m2) mussels (Fig. 3). The white material also replaced the bare substratum (5.9 m2), large (3.0 m2), and medium (3.8 m2) mussels. Colonization of mussels occurred on bare substratum (large: 6.9 m2, medium: 13.2 m2, small: 5.0 m2). Large mussels were replaced by 8.0 m2 of medium mussels, whereas 11.2 m2 of medium individuals transitioned to larger ones (Fig. 3). Small-mussel assemblages transitioned to medium ones by 26.4 m2 while 7.8 m2 transited the opposite way. Large- and small-mussel assemblages rarely transitioned from one to the other directly (< 1.0 m2) during the 5-yr time interval.
A substantial cover of mussel assemblages underwent a change from 2015 to 2020 for large—(54.6%), medium—(34.6%), and small—(59.2%; Fig. 4A–C). Cover of microbial mats displayed higher variability (71.3%; Fig. 4D).
Change in large- and medium-mussel assemblages occurred mostly within venting areas (Fig. 4A,B). Overall, changes consisted of a mix of anecdotic cover losses and gains interspaced with larger null changes within venting areas. Over the extent of the venting area, this pattern suggested the overall stability of the assemblages despite meter-scale changes. Still, a gain of large (+5.6 m2) and medium mussels (+1.8 m2) was detected in the South-East venting area (Fig. 4a; Supporting Information Fig. S8, S9). Outside venting areas, where large mussels rarely occurred, medium-size mussels were stable, except for a large decrease in the center of the edifice (−11.0 m2; Fig. 4B). That loss coincided with a gain of small mussels (+11.4 m2; Fig. 4C). Overall, small mussels inhabiting areas outside direct venting gained cover over the edifice. That pattern coincided with the general loss of microbial mat cover (Fig. 4D). As for large and medium mussels, microbial mats exhibited local patterns of variability, mostly within venting areas.
Over 25 yr, mussel total cover ranged from ~ 45% to 55% despite high variability among edifice sides (Fig. 5A). The URT analysis depicted a breakpoint between 2008 and 2015 (RE = 0.4, CVRE = 0.6). Microbial cover was less variable among edifice sides than for mussel cover, illustrating lower spatial heterogeneity of cover among sides (Fig. 5B). Microbial mat cover did not display a stable trend, ranging from ~ 15% to 45% on average over 25 yr (Fig. 5B). The URT identified breakpoints between 1998 and 2001 and between 2015 and 2018 (RE = 0.3, CVRE = 1.2). After an increase of ~ 30% of cover till 2001, microbial mat cover decreased by ~ 15% till it stabilized from 2008 to 2015. From 2015 to 2020, microbial mat cover declined by ~ 15% (Fig. 5B).
Discussion
Using a time series of 3D models, this study described the variability of hydrothermal vent assemblage cover through space and time and at different scales occurring over a complex deep-sea sulfide edifice. To our knowledge, this study is the first to apply 3D quantitative techniques to resolve the fine-scale dynamics of vent communities over 5 yr. It also extends the monitoring of the Eiffel Tower edifice up to 25 yr. This work establishes a methodological and analytical baseline workflow to disentangle variability from the broader 100-m2 scale to the submeter scale, the latter having been rarely quantitatively investigated in the past (Cuvelier et al. 2011b; Sen et al. 2014).
Limitations
We estimated an average registration error among 3D models smaller than the decimeter, demonstrating the high spatiotemporal continuity of this study. Similar to registration error reported from terrestrial studies that used unmanned aerial vehicles (e.g., registration error ranging from 0.04 to 0.11 m in Cucchiaro et al. 2018), this result is a real achievement in a deep-sea context where the acquisition of imagery is constrained, not only by time but also by a number of technical challenges. For example, meter-scale drifts in ROV navigation among dives can increase registration errors. Visual relocation of the ROV based on the image time series could ultimately help improve the global registration of image, especially when considering the high stability of the Eiffel Tower edifice (Boittiaux et al. 2023).
Variability in image acquisition throughout years of sampling may also alter the accuracy of temporal monitoring, due to (1) variable distances of the ROV from the edifice across years, (2) differences in image sharpness due to changing turbidity or plume dispersion, and (3) change in camera resolution. This can ultimately lead to variability in photogrammetry model resolution and to difficulties in the identification of small organisms and vent outflows. The use of pre-programmed tracks and visual repositioning may be the only way to overcome sampling irregularities as it started to be implemented for the Eiffel Tower edifice (Boittiaux et al. 2023).
Habitat dynamics
Physical habitat—the Eiffel Tower has been a stable structure through time: less than 5% of the edifice surface underwent topographic modifications. Topographic changes occurred predominantly in areas restricted to the close vicinity of hydrothermal outflows due to the precipitation of minerals. The rapid formation of friable chimneys and spires, their collapse and accumulation on the soft-sediment talus, were only observed sporadically, likely because of their sub-monthly dynamics (Tunnicliffe and Juniper 1990; Van Audenhaege et al. 2022), which was not resolved in this case study. Those rapid dynamics at sub-monthly time scales, combined with the friability of spires and soft substratum of the talus, explain the scarcity of vent mussels in those areas that most likely rely on a secured anchorage with their byssus threads to maintain themselves under appropriate vent exposure. Variability in the physical habitat occurred predominantly through outward accretion of hydrothermal features such as flanges and outcrops. As accretion proceeds at a rate of a dm yr−1 (Van Audenhaege et al. 2022; this study), the apparent switch from white material to consolidated and oxidized material suggests the build-up creation of vacant “bare substratum”, as was also observed in Pacific vents (Du Preez and Fisher 2018). Since the observed collapse of oxidized material remained scarce, that bare substratum may persist for years. At larger decadal scales, the induration of newly formed vent material may contribute to the edifice build-up while supplying a solid hard substratum, opening habitats suitable for an effective attachment of vent organisms.
Hydrothermal activity—venting clustered in a few areas of several square meters, containing flanges, diffusion zones and focused emissions, and was mainly located on the sides of the mound. The stability of the location of those active clusters suggests the absence of major redistribution events of venting activity on and at the periphery of the structure (Sarrazin et al. 1997; Cuvelier et al. 2011b). Considering finer spatial scales, the repositioning of vent exits was predominantly restricted to changes within clusters. The mineral accretion and friable feature collapse, and the progressive subsurface clogging by mineral precipitates could explain submeter scale repositioning of vent fluid emission (Hannington et al. 1995; Sarrazin et al. 2002). Here, directional relocation of vent outflow predominated, following accretion and oxidation of hydrothermal material.
Faunal assemblage dynamics
Mussel dynamics—the megafaunal community was remarkably stable over time, consistent with the lack of dramatic changes in the hydrothermal and physical habitat, corroborating the high stability observed from 1994 to 2008 by Cuvelier et al. (2011b). It is generally assumed that randomness of biological change, as observed at vents in the East Pacific (Sarrazin et al. 1997), indicates a regime of chronic and patchy disturbance as proposed by Horn (1976) for plant communities. However, in the case of the Eiffel Tower, the consistency of an ordered succession model underlines the maintenance of spatial assemblage segregation as disturbances only occurred at small scales (Cuvelier et al. 2011b; this study). Despite that variability, the consistency of assemblage flow suggests an apparent “steady-state equilibrium” of the megafaunal community over the 5 yr of monitoring. In the absence of dramatic disturbance events (e.g., magmatic/tectonic disruption), an abundant mussel population (~ 263 m2) has likely thrived there during at least three decades since the first scientific survey in 1992, supporting the hypothesis of a “climax” stage of a succession of Eiffel Tower communities (Cuvelier et al. 2011b).
The variability of large and medium mussels generally coincided with the meter-scale variability of vent fluid repositioning and topographic changes observed in venting areas of the Eiffel Tower. Other studies under continuous venting regimes have observed similar small-scale dynamics, as they remain inevitably the main drivers of natural variability in the absence of significant and large-scale habitat modifications (Copley et al. 2007; Sen et al. 2014; Du Preez and Fisher 2018). This confirms that the closer vent exits are to one another, the more dynamic the habitat will be and, consequently, the more vent megafaunal assemblages will likely vary. Small-scale change in the venting habitat is likely to trigger (1) a direct disturbance from material accretion or vent opening and (2) a change in conditions that will support or threaten organisms. Notwithstanding, mussels still displayed very low net change despite the high variability observed in the vicinity of vent exits, hence suggesting that any impact on mussel cover was compensated within 2–5 yr. This sustenance of populations over the venting habitat demonstrates that B. azoricus mussels have developed abilities to cope with small-scale changes in the vent habitat. The higher stability of medium mussel and zoanthid assemblages located away from the venting activity may be related to a more stable environment influenced by the cold deep-sea waters that dilute vent fluids and, therefore, buffer any distal influence of vent exposure change. In fact, B. azoricus may benefit from the edifice arrangement since the interaction of bottom currents with complex topography could provide a steady trophic supply by redistributing hydrothermal vent plume, preys, and suspended particles over a large area (Girard et al. 2020). However, a decrease in mussel sizes over the edifice summit may still be compatible with an overall slow decline of the Lucky Strike field activity, as proposed by Barreyre et al. (2012) based on a comparison of the 1996 and 2009 seafloor images. However, this pattern remains to be confirmed as the mussel population seemed relatively stable across 25 yr of monitoring despite a significant increase in mussel cover between 2008 and 2015. Although assemblage cover was standardized, that breakpoint remains difficult to confirm because of the methodological switch from opportunistic 2D data to more accurate 3D annotations.
Our characterization of the spatiotemporal scales of variability can help identify candidate traits contributing to the resilience of vent communities. In this study, we witnessed rates of flange accretion reaching < 0.1 m yr−1 (e.g., Supporting Information Fig. S2, S3) and (de)activation usually occurring within areas < 1 m2 (e.g., Supporting Information Fig. S4). At infra-annual time scales, previous observation evidenced that, as accretion of new substratum proceeds, B. azoricus benefits from the indurated substratum to adapt their position to optimal vent fluid exposure through migration of large individuals (~ 0.1–0.3 m d−1; Van Audenhaege et al. 2022) or, more slowly, of byssus-bounded aggregations (~ 0.1–0.2 m yr−1; Van Audenhaege et al. 2022). As the migration capability of B. azoricus is comparable to the scales of change we observed near vent outflows, it seems reasonable to state that mobility is a fundamental trait ensuring vent mussels cope with local habitat changes in active areas of the edifice, as previously hypothesized for vent mussels and gastropods of the Pacific Ocean (Johnson et al. 1994; Sen et al. 2014). On vent edifices, mobility and the ability to orientate migration may be particularly important to sustain a pole position in a habitat where small-scale changes are persisting, as also observed for Pacific vent gastropods (Sen et al. 2014). Taking advantage of these gradual small-scale changes could even help mussels rapidly gain access to vent exposure while avoiding the cost of intraspecific competition of dense aggregations (e.g., see Bertness and Grosholtz 1985). As dense assemblages of large mussels relocate, competition may be relaxed to allow the settlement of smaller mussels, for which recruitment was suggested to be inhibited by larviphagy, therefore resulting in size-based spatial segregation of mussels (Comtet and Desbruyères 1998; Lenihan et al. 2008). In the case of vent exposure increase, smaller mussels could still hold the capability to modify their endosymbiont community through environmental acquisition (Won et al. 2003). As a result, their growth could be fast enough to possibly avoid being outcompeted and smothered by larger migrants (Nedoncelle et al. 2015). Since Nedoncelle et al. (2014) estimated the mean yearly shell growth rate of B. azoricus at ~ 1.68 cm yr−1, it is plausible that medium mussels (~ 3 cm long) could grow to become large mussels (~ 6 cm) within 2–3 yr. Outside venting areas, a net loss of medium mussels favored the colonization of smaller individuals. As mortality and migration create vacant habitats, those can be colonized through the settlement of smaller individuals, emphasizing the importance of recruitment to maintain the mussel population at the carrying capacity of the edifice. Nevertheless, growth and recruitment rates remain to be confirmed experimentally as a recent post-disturbance monitoring at Lucky Strike rather suggested that the growth of B. azoricus individuals was slow, with only a partial recovery of new habitats through recruitment over 2 yr of monitoring (Marticorena et al. 2021). While experimentation is needed, our cartography of changes delineated dynamic areas where the deployment of time-lapse cameras could provide a sub-annual image set to better discern the processes acting under natural conditions (e.g., South-East of Eiffel Tower for large mussel colonization).
The disturbance regime can be defined by its area extent, magnitude, frequency, predictability, and turnover rate (Sousa 1984). These characteristics have great implications for community composition because they contribute to habitat heterogeneity and influence individual survival. Over evolutionary time scales, the disturbance regime is a factor of selection to which species adapt to maintain their populations through time; it has thus a strong influence on community resilience (Levin 1992). Overall, because of the absence of large-scale habitat disturbance or mortality from predation, and since B. azoricus mussels are well adapted to fine-scale habitat changes, it is not surprising that the mussel population has been maintained at the carrying capacity of the edifice for years (Cuvelier et al. 2011b). The Eiffel Tower edifice harbors a larger talus compared to other sites at Lucky Strike, suggesting high maturity (M. Cannat, pers. com.; see Hannington et al. 1995 for the definition of maturity). Maturity provides high stability, extent, and complexity of the habitat, which allows, in turn, the sustainable settlement of an extensive mussel population. Given that the Eiffel Tower hosts large and stable communities of the foundation species B. azoricus, such mature vent edifices may be keystone features for the regional maintenance of vent populations and biodiversity (e.g., Sarrazin et al. 2020).
Despite Lucky Strike being part of the Azorean marine protected area, our insights enhance our global ecological knowledge on vent ecosystem dynamics at slow-spreading ridges, where seven exploration licenses have been granted so far by the International Seabed Authority (ISA) (i.e., Mid-Atlantic and Indian Ocean Ridges). The observed pluri-decadal stability lines up with other edifice monitoring that depicted decadal stability of vent edifice habitats and associated communities at slow-spreading features (MAR: Cuvelier et al. 2011b; Copley et al. 2007; South-West Indian Ridge: Zhou et al. 2018; Pacific back-arc basins: Sen et al. 2014). In addition to the rarity of eruptive events at slow-spreading ridges (i.e., millennia; Rubin et al. 2012), the cumulative evidences indicate the lack of trait selection providing resilience capacity toward disturbance for the vent fauna, being natural or anthropogenic such as deep-sea mining (Du Preez and Fisher 2018). This underlines the need for a precautionary approach to mining activities until more knowledge is gained to define exploitation guidelines (Van Dover et al. 2018).
Microbial mat dynamics
We observed a spectacular decline in microbial mat cover from 2015 to 2020, suggesting that microbial mat dynamics may vary on more rapid time scales than vent megafaunal assemblages. This is supported by the results of Van Audenhaege et al. (2022) who reported periods of absence/presence over only a few months in an area of diffuse outflow at the base of the Eiffel Tower over 7 yr of monitoring. However, this decline detected between 2015 and 2018 contrasts with the general faunal and habitat stability. While biotic factors like grazing activities can strongly regulate microbial cover (Micheli et al. 2002; Cuvelier et al. 2011a), microbial mat dynamics may also be sensitive to telluric events, and this may explain that rearrangement at depth of the plumbing system detected in September 2015 coincides with the decline of microbial mat cover detected between 2015 and 2018 (Ballu et al. 2019; this study). As suggested by a switch in the microbial composition of iron-rich mats in 2016 (Astorch-Cardona et al. 2023), the 2015 event could have created temporary conditions also favoring a community change in filamentous mats at Eiffel Tower. Similarly, the breakpoint detected between 1998 and 2001 coincided with a magmatic episode detected by earthquake swarms in 2001 at Lucky Strike, which was followed by an increase in filamentous mats abundance and a subsequent decline till 2008 (Dziak et al. 2004; this study). At the Eiffel Tower, microbial mats are sustained by distal exposure to plumes of focused vent outflows (Girard et al. 2020). Therefore, despite the apparent stability in venting activity, the overall decline of microbial mat could be explained by particular changes in the vent fluid composition (e.g., chlorine increase as Ballu et al. 2019 detected at the central “Lava Lake” and Eiffel Tower edifice of Lucky Strike). Over time, the recurrence of those observations enabled by the multidisciplinary EMSO-Azores observatory, will likely shed light on the mechanisms and ecological implications of arrhythmic change in microbial mat cover.
Conclusion
This study showed that photogrammetry is a suitable method for multi-scale data acquisition in remote underwater environments to investigate the large-scale maintenance of mosaic communities, which is also the objective of the “patch-dynamics” approach in landscape ecology (Paine and Levin 1981; Levin 1992). Our results emphasized that the predominant scales of variability in faunal distribution and venting activity at the Eiffel Tower restricted to the submeter scale. At large scales, however, the lack of significant shifts in the habitat over two decades explains the overall stability of faunal assemblages. In this study, we observed strong similarities in scales and patterns of environmental and faunal variability in the vicinity of vent exits with those recorded from sub-monthly long-term monitoring at the assemblage scale (Van Audenhaege et al. 2022). This consistency between the two studies shows that the mechanisms and processes observed at a single vent outflow with sub-monthly images could be extrapolated over the edifice over at least two decades. Multi-scale data acquisition, made possible with pluri-disciplinary deep-sea observatories, may thus be particularly useful for disentangling the relevant processes shaping community dynamics at deep-sea hydrothermal vents. Finally, the methodology proposed in this study could be reused to optimize standardization for the spatiotemporal monitoring of complex habitats at slow-spreading centers, particularly in the context of deep-sea mining.
Open Research
Data availability statement
Data are stored in the SEANOE database: (1) 3D models of 2015 (doi: 10.17882/79218), 2018 (doi: 10.17882/92215) and 2020 (doi: 10.17882/92219) and (2) image annotations for the 2D annotations from 1994 to 2008 (doi: 10.17882/95808) and 3D annotations from 2015 to 2020 and their HTML displays (doi: 10.17882/96475). The R code to compute figures is available in Supporting Information Materials.
References
Acknowledgment
We firstly thank the crew of the R/Vs Pourquoi pas? and L'Atalante as well as the pilots of the ROV Victor6000 during the MoMARSAT cruises of 2015, 2018, and 2020. We are grateful to M. Ferreira, C. Boittiaux, and more generally, the PRAO research labs of Ifremer DFO department, for the development of high-resolution optical sensors of the ROV as well as reconstruction of 3D models. We are grateful to D. Cuvelier, who provided annotations from 1998 to 2008. We would like to thank F. Girard for sharing her expertise on the annotation of the 2015 3D model. We express our gratitude to the editor, associated editor and two anonymous reviewers, who significantly improved the manuscript with their suggestions. This manuscript was professionally edited by C. Engel-Gautier. This work and LVA's PhD thesis were supported by the European Union's Horizon 2020 research and innovation project iAtlantic under Grant Agreement No. 818123. This output reflects only the authors' view and the European Union cannot be held responsible for any use that may be made of the information contained therein. We also acknowledge the financial support from the EU project EMSO (http://www.emso-eu.org/) and the French observatory EMSO-Azores funded by Ifremer and CNRS.
Conflict of Interest
None declared.