1.

INTRODUCTION

Gyali Island is a small, uninhabited island, located between the islands of Kos and Nisyros in the Southeast Aegean Sea (Figure 1). The geological composition of the island consists of a thick layer of rhyolitic pumice to the south and rhyolitic lava formations to the north^{1, 2}. Those rhyolitic units are separated by an isthmus of loose coastal deposits at the center of the island. Northwest of Nisyros lies a magma pocket which in the past had activated a tectonic discontinuity near the Mandraki fault (in Nisyros) leading to seepage of H2S on Gyali Island³. The Gyali offshore area has not been studied extensively in the past and the data, though limited, concerns mostly offshore bathymetry, morphotectonics, and general stratigraphy and not the shallower marine environment and ecological features/habitats. Only Daskalopoulou et al.⁴ studied the geochemistry of shallow water fluid leakages (<10 m depth) observed in three sandy areas to the west and north of the island, whereas Nomikou et al.³, while surveying the seafloor morphology of the eastern Aegean volcanic arc, discovered biogenic assemblages locally blanketing the volcanic hard substrate and a loose rippled sedimentary cover at the eastern part of the island, using an ROV, however not reporting positions or depth distribution. Yet, shallow and inshore habitats around Gyali Island had never been the focus of any research, although the wider area belongs to the NATURA 2000 network (Figure 1), and two mining companies actively extract pumice and perlite in the area. Only Topouzelis et al.⁵ present information regarding the distribution of seagrass, as part of an effort to map the seagrass meadows around the entire Greek coasts using object-based image analysis in Landsat-8 images and machine learning. Thus, this contribution aims (i) to investigate the shallow morphology around the island and aid knowledge on the distribution of marine habitats, (ii) to assess potential pressures and human imprints induced due to the onshore extraction activities and (iii) to evaluate and compare data obtained from Topouzelis et al.⁵ with the seagrass distribution recorded by sonar data in the present study.

Figure 1.

Location of the study area (red boxes) also showing Marine and Land Natura 2000 sites around Gyali and Nisyros Islands (top left), and aerial photo of the extraction facilities on Gyali Island (bottom left). Filed observations and data collection locations (right).

2.

METHODOLOGY

Data acquisition took place in September 2023, using (a) a high frequency hydrographic SonarMite v5 Echo-Sounder (Ohmex) (0.025 m depth accuracy - RMS) with GNSS RTK TopCon HiPer support (~0.5 m accuracy) and (b) a Humminbird Helix 10 multi-parametric sonar, being capable to simultaneously capture bathymetry (180-240 kHz) and map seafloor morphology operating in a side scan sonar (SSS) mode in 455 kHz. Both instruments were utilized to obtain bathymetric data. A dense grid of 58 parallel and 120 vertical survey lines was conducted, covering 6 km² (Figure 1). About 249,500 bathymetric points were recorded, filtered, and interpolated using a Multilevel B-Spline method for the creation of the bathymetric map in QGIS (Figure 2). The SSS images were processed and composed into a sonar mosaic (Figure 2) using the SonarWiz 7 software, while the seabed-habitat mapping was performed manually due to the morphological complexity of the area. The anthropogenic imprints were also mapped manually. A drop Go-Pro 4K camera was used for visual inspection of the seafloor and for ground-truthing of the SSS images through 82 videos. Also, 41 sediment samples were collected using a Van Veen grab. 37 out of 41 samples consisted mainly of coarse sediments (>0.0625 mm) and were analyzed by dry sieving according to Folk’s⁶ method. 4 samples consisted entirely of seagrass, and they were not analyzed. The Gradistat software⁷ was used to calculate the samples’ granulometric (statistic) parameters. Once the habitat mapping and ground-truthing were completed, the recorded seagrass distribution from the sonar data was compared with the available satellite data.

Figure 2.

Bathymetry of the study area (left) and the side scan sonar mosaic with hydrothermal seepage areas (right).

3.

RESULTS

The bathymetric survey extended to a maximum depth of ~67 m (Figure 2). The seafloor morphology appeared uneven, with outcrops and one extensive reef to the south. To the east there is an extensive zone of low water depths with a micro-relief compared to the western side of the island where depths increase more abruptly. The slopes ranged from 0 to 80%, being lower (5-10%) in the shallow areas, also, off the two small beaches of the isthmus in the center of the island. Areas of hard substrate attain locally slopes higher than 45%. In general, between 5 and 15 m the slopes are smooth (~5%), from 15 to 50 m they increase abruptly up to 45%, with outcrops being up to 80% steep. Beyond 50 m the slopes increase more smoothly. The reef at the south extended between 5 and 20 m and displayed mean slopes between 5-15%.

A significant characteristic, which was detected in the bathymetric records, was the irregularly shaped and extensive strong reflections that are distributed from the bottom to the surface of the water column and were attributed to hydrothermal fluid venting. Several isolated and three extensive areas of seepage were recognized, between 5-55 m (Figure 2).

Analysis and interpretation of the individual SSS images and the sonar mosaic revealed six main reflectivity types (RT), each corresponding to a different type of substrate or habitat (Table 1), as also validated by the camera observations and the sediment sampling:

Table 1.

Side scan sonar reflectivity types and camera snapshots of the seabed.

The wider mean size range of RT 1a is due to sand size variations but it may be also attributed to the lower specific gravity of pumice, which affects the way the acoustic energy is reflected and recorded. The placement of the sonar on the side of the boat, as well as the depth variations, may also affect the recorded reflectivity of similarly textured substrate, contributing to the wide mean size range obtained from sediment samples.

The observed sand ripples develop either almost parallel to the coast, indicating a wave origin, or almost vertical, probably indicating combining generation processes of waves and currents. In a particularly morphologically complex area to the south-southwest, significantly larger ripples were observed, indicating more complex dynamic processes.

The majority of the sediment samples that were analyzed contained biogenic fragments, whilst a few of them also seagrasses. The samples consisted exclusively of coarse sediments, gravel, and sand, and can be classified according to Folk⁶ as medium gravel to fine grained sand (mean size: 0.16 to 13.17 mm). The gravel and sand content of the samples ranged from 0-100% and their spatial distribution was nearly complementary. According to the sediments’ granulometric (statistic) parameters they ranged from well to poorly sorted (standard deviation: 1.22 to 3.64 μm) with the worst sorting occurring to the north. The sediments were strongly fine skewed to strongly coarse skewed (skewness: 0.54 to -1.25) and very platykurtic to very leptokurtic (kurtosis: 0.32 to 2.34), indicating the mixing of various sediment sub-classes and of different origins (terrigenous and biogenic).

The observed RTs correspond to certain habitats (Figure 3) which can be categorized based on Directive 92/43/EEC (Network of Protected Areas NATURA 2000) (Table 2). Only RT 6 could not be categorized as it is a transitional type. Also, types 1110 and 1110-A extend from shallow to deep waters, instead of the shallow waters indicated by the Directive, thus they were also grouped under the code 119 which is devoted to soft sandy substrate.

Figure 3.

Habitat map over the studied area (left) and SSS mosaic showing location of anthropogenic targets (right).

Table 2.

Reflectivity types and corresponding habitats of Directive 92/43/EEC (Network of Protected Areas NATURA 2000) found within the studied area.

*

Cymodocea nodosa was also present locally.

Numerous scattered anthropogenic targets (mostly anchor scars, moorings, debris, construction bases, and a linear target) (Figure 4) were detected through the sonar images (mainly within RT 1 (Figure 3)) and were ground-truthed with camera videos, mainly to the east and south, some of them near the extraction companies port facilities, in depths ranging from 5 to 35 m. The majority of the anchor scars were observed in the sandy substrate and/or between the rocky relief (Figure 4b). The construction bases of the port and the loading platform facilities were distinguished and inspected (Figure 4c). Three large containers connected by a chain, most likely acting as mooring, were also found between 35-40 m (Figure 4d). A linear trace with a length of 230 m was detected to the north, extending between 20 and 45 m water depth, corresponding to a submarine power or telecom cable connecting Gyali with Kos Island (Figure 4e). The overall appearance of the habitat distribution and the visual inspection did not suggest considerable pressures from the onshore mining activities; however, a small number of anchor scars ran through the seagrass meadows (Figure 4f), which may be attributed to fishing boats and yachts or bigger commercial ships.

Figure 4.

Examples of anthropogenic imprints (SSS images and camera photos) around Gyali Island. (a) A metallic plate and a tire littering the seafloor (red circles), (b) an anchor scar (black arrows) in a sandy substrate, (c) a construction foundation (blue outline) and various anthropogenic targets (black arrows), (d) moorings (black arrows), (e) the submarine cable trace (black arrows), and (f) an anchor scar (black arrows) and an abandoned anchor through P. oceanica. (scale: 25 m between blue dotted lines in the SSS images).

The seagrass meadows as detected in the sonar records in the present study occupied 1.32 km², while the meadows as observed by satellites⁵ occupied 1.37 km². However, the common area recorded by the two methods (Figure 5) occupied 0.58 km², with only 44% of the total area mapped by satellite data corresponding to seagrass mapped by sonar data. The similarity observed in the size of the recorded areas is misleading, as the majority of the satellite classification was spatially misplaced. The classification conducted using satellite data appeared to encompass three types of errors:

Figure 5.

(left) Seagrass distribution of the present study (top), and according to the satellite data (bottom), and (right) differences in seagrass distribution according to the two classification methods and their overlap.

4.

DISCUSSION AND CONCLUSIONS

Gyali Island is one of the very few open-pit uninhabited islands worldwide, where onshore extraction activities take place^{8, 9, 10}. However, in none of the other open mining places, such as in Alaska⁸ and Indonesia^{9, 10}, ecological or other environmental studies have been reported. Similar studies are also missing for other coastal mines in operation, whilst few investigations report on the marine ecological and habitat status of big ports¹¹, at least in the Mediterranean Sea. Here, the distribution of marine habitats in the bigger part around Gyali, a Natura 2000 site but also an important insular open-pit, was studied in detail using common techniques for habitat mapping in shallow waters¹². The surveyed area presents an uneven morphology with outcrops, reefs, areas with a sharp increase in slope, and shallow smooth areas, all of which are typical characteristics of a volcanic environment. Five main habitats were distinguished, which are sand and rippled sand (1110/119), reefs and rocky outcrops (1170-A), P. oceanica (1120) locally with pockets / small areas of C. nodosa and areas of rhodoliths and coarse-grained sediments (coarse sands to gravels) (1110-B). These are common habitats, which have been observed elsewhere, in shallow coastal areas, but also in the shelf of the Greek seas^{13, 14}. Posidonia oceanica meadows and rhodolith fields are priority habitats protected by the EU Habitat Directive (92/43/EEC).

Submarine fluid seepages, either extended or isolated, most probably of a hydrothermal origin, were observed to occur mainly through rock outcrops, seagrass, and sand. The SSS images revealed the existence of sandy ripples attributed either to wave activity or locally to the intense hydrodynamic conditions also forced by the complex morphology (reefs, rocky outcrops, etc.).

Several anthropogenic targets were identified, some of which were ground-truthed by camera drops. In general, the visual inspection, apart from validating the interpretation of the SSS images, (i) confirmed the areas of fluid seepages, (ii) designated the quality (high clarity/visibility) of the water column, (iii) identified various targets of anthropogenic origin (construction foundations, moorings, anchor scars, litters, etc.). Litters on the seafloor are distributed on sand, in P. oceanica and near rocky outcrops. Although few of them, especially in front of the quarry dock, seem to be associated with the shipping activity there is no evidence regarding pressures induced due to the onshore mining works, even though biodiversity and other environmental studies need also to be implemented to confirm this conclusion.

Satellite imagery has been an important tool for widespread habitat mapping, especially in areas like the long Greek shoreline with numerous islands, where habitat mapping with hydroacoustic means is costly, time consuming, and difficult. However, satellite data presents relatively low resolution, depth limitations, and it requires extensive ground-truthing in order to validate the obtained data¹⁵. Ground-truthing is important in order to obtain reliable spatial data especially about protected and priority habitats like P. oceanica and coralligenous (rhodolithic) formations. In this study, the comparison of the seagrass distribution with that recorded by Topouzelis et al.⁵ revealed significant differences, with only 44% of the total area mapped by satellite data corresponding with seagrass mapped in the current work. The errors observed exceeded 115% and ranged from inability to distinguish smaller targets to mistaking habitats like rocky relief/reefs for seagrass, as well as inability to discern the deeper limit of seagrass meadows. These significant errors and the small percentage of successful discrimination of seagrass show that classification through satellite data still presents many uncertainties and could not produce fully trustworthy results for this but also other coastal areas in the Greek seas with similar morphological and habitat characteristics, arising concerns but also dictating similar methodological evaluations to be adopted in order to achieve reliable outcomes.