Submersible Surveys Research Articles

Characteristics of Habitats Used by Enteroctopus dofleini in Prince William Sound and Cook Inlet, Alaska

Abstract. Habitat characteristics associated with the giant Pacific octopus, Enteroctopus dofleini (Wülker 1910), were studied in Prince William Sound and Port Graham, Alaska, from beach walk, SCUBA and submersible surveys at depths of 0 to 197 m. Octopus counts on beach walk transects were positively correlated with soft substrates (sand, gravel or broken rubble), the presence of boulders, and dense kelp cover immediately offshore of the transect; and negatively correlated with depth on SCUBA transects. No significant habitat correlations were found with counts on submersible transects. On beach walks, octopus counts were reduced on hard substrates to 38 % of the counts on soft substrates. Counts increased five‐fold in the presence of boulders over counts in their absence, and increased fifteen‐fold adjacent to dense (> 75 %) kelp cover over counts adjacent to sparse (< 25 %) kelp cover. On SCUBA transects, the average density at less than 5 m depth was over five times that below 5 m. No trends in octopus size or sex ratio were detected with depth. Den use was inversely correlated with depth although there was no indication that den availability declined with depth. Octopuses were found at densities from 0 to 2.5 per 1000 m2. These densities were only 1 to 50 % of densities of the same species recorded in British Columbia in the late 1970s and early 1980s. Few data are available to test recruitment, mortality, and habitat selection hypotheses that would account for differences between habitats. However, the presence of the highest octopus densities in intertidal and very shallow subtidal areas indicates the likely importance of near‐shore, shallow‐water habitats, and highlights the vulnerability of octopus populations to changes in these habitats.

A Comprehensive Investigation of an Offshore Active Fault in the Western Sagami Bay, Central Japan

AbstractOffshore active faults, especially those in the deep sea, are very difficult to study because of the water and sedimentary cover. To characterize the nature and geometry of offshore active faults, a combination of methods must be employed. Generally, seismic profiling is used to map these faults, but often only fault‐related folds rather than fracture planes are imaged. Multi‐beam swath bathymetry provides information on the structure and growth history of a fault because movements of an active fault are reflected in the bottom morphology. Submersible and deep‐tow surveys allow direct observations of deformations on the seafloor (including fracture zones and microstructures). In the deep sea, linearly aligned cold seep communities provide indirect evidence for active faults and the spatial migration of their activities.The Western Sagami Bay fault (WSBF) in the western Sagami Bay off central Japan is an active fault that has been studied in detail using the above methods. The bottom morphology, fractured breccias directly observed and photographed, seismic profiles, as well as distribution and migration of cold seep communities provide evidence for the nature and geometry of the fault. Focal mechanism solutions of selected earthquakes in the western Sagami Bay during the period from 1900 to 1995 show that the maximum compression trends NW‐SE and the minimum stress axis strikes NE‐SW, a stress pattern indicating a left‐lateral strike‐slip fault.

The Sula Reef Complex, Norwegian shelf

Cool-water carbonates in the aphotic zone of deep shelf and continental margin settings in the Northeast Atlantic are produced by the deep-water coral reefs withLophelia pertusa as the major framework builder. Through a compilation of side scan sonar, airgun and manned submersible surveys from several cruises to the mid-Norwegian Sula Reef Complex (SRC), the facies pattern and zonation of one of the largest deep-water reefs in the Northeast Atlantic is described in relation to the overall seabed topography. The late glacial to early postglacial iceberg scour on the crest and shoulder of the Sula Ridge provides settling ground for the scleractinian corals already in the early Holocene. Since then coral growth continues until today but was supposed to be disturbed by an environmental hazard, the so-called second Storegga event. The distinct distribution pattern of individualLophelia reefs on the Sula Ridge has stimulated a discussion on intrinsic environmental controls such as the bentho-pelagic coupling and the alternative hydrocarbon-based nutrition hypothesis.

Association of Syscenus infelix (Crustacea: Isopoda: Aegidae) with benthopelagic rattail fishes, Nezumia spp. (Macrouridae), along the western North Atlantic continental slope

During submersible surveys along the continental slope (summers of 1991 and 1992, 184–847 m) between False Cape, Virginia, and Cape Hatteras, North Carolina, USA, we observed the aegid isopod, Syscenus infelix Harger, attached to the macrourid Nezumia bairdii (Goode and Bean). This is the first report of S. infelix attached to fishes in the western North Atlantic. The association of this blind isopod with its host appears species specific. The large, conspicuous isopod always attached to a fish in the same location, the dorsal midline, immediately behind the first dorsal fin. Attachment appears to be long term, with the isopod forming a characteristic scar consisting of a distinct discolored oval depression with seven small, dark impressions that coalesce as the fish grows. Only one S. infelix was found on each host fish. The isopod occurred on 23.7% of N. bairdii observed from submersible on the middle continental slope off Virginia and North Carolina, compared with 16.6% of 1236 museum specimens of the same species (based on inspection for scars) collected at latitudes 26°–64°N. Prevalence of the fish–isopod association was not correlated with depth or latitude. We also found identical scars on preserved specimens of N. aequalis (2.6% of 660 specimens), N. sclerorhynchus (1.2% of 86 specimens), and N. suilla (14.3% of 7 specimens), mostly from areas outside the range of N. bairdii. No scars were found on museum specimens of N. atlantica (n=27), N. cyrano (n=57), or N. longebarbata (n=7). The low incidence of isopod attachment on these species suggests that N. bairdii is the preferred host. Infestation by the isopod appears to result in erosion of host fish scales and tissue. We propose that S. infelix is an obligate associate of its host fish and should be considered parasitic.

Prolonged Magmatic and Tectonic Development of the Caribbean Igneous Province Revealed by a Diving Submersible Survey

Using the diving submersible survey NAUTICA we investigated the central part of the Caribbean large igneous province (CLIP) to observe and sample internal portions of this proposed oceanic plateau. Most of the samples are gabbroic and doleritic rocks; basalts are scarce. Radiometric dating by 40Ar/39Ar incremental heating experiments indicate that the intrusive rocks are Campanian in age (81–75 Ma). In some places these intrusive rocks underlie older Santonian (85–83 Ma) extrusive basaltic rocks, suggesting that the Campanian rocks represent a sill injection and an underplating episode. Results of the diving program supplemented by information from ODP and DSDP drilling sites document a 20 m.y. period (94–75 Ma) of igneous activity in the submerged portion of the Caribbean large igneous province (CLIP). In the northern part of the Beata Ridge late Campanian and/or post Campanian uplift is documented by prominent Maastrichtian (71–65 Ma) erosion and the establishment of a Paleocene-middle Eocene (65–49 Ma) carbonate platform. During and after the uplift an extensional period is indicated by seismic images and the subsidence (3 km depth) of the carbonate platform. Paleocene ages (55–56 Ma) determined on some volcanic samples are attributed to localised decompression mantle melting that accompanied the extension. We document a prolonged period of magmatic and tectonic events that do not fit with the current models of short-lived plateau formation during mantle plume initiation but shares many similarities with the constructional histories of other oceanic large igneous provinces.

Over-exploitation of a broadcast spawning marine invertebrate: Decline of the white abalone

Marine invertebrates have long been consideredto be resistant to overfishing. However, agrowing number of exploited taxa have declinedsubstantially and even disappeared from partsof their former range. We consider the case ofthe white abalone (Haliotis sorenseni);the first marine invertebrate proposed for theUS endangered species list. This high-valuespecies was one of five abalones targeted inthe California and Mexico fisheries; it is nowrare and protected from fishing. The biologicalcharacteristics of this deep-living abaloneindicate that it was particularly vulnerable toover-exploitation; reduction of density orgroup size is now known to lead to declines infertilization success and recruitment failure.Warning signs of potential problems existedboth pre- and post-exploitation but were notrecognized. In particular, serial depletion wasnot detected because catch was not analyzedspatially, perhaps because total landings werereasonably stable for the short period ofexploitation. Recent submersible surveys led toestimates that white abalone now number lessthan 2,600 animals or 0.1% of the estimatedpre-exploitation population size. Densities andestimated population sizes are less than 100animals, at all but one location. Alternateexplanations for the decline in abundance wereconsidered and only exploitation-linkedfactors, such as sub-legal mortality andillegal fishing, were likely contributors.Episodic recruitment appears to be acharacteristic of broadcast-spawning,long-lived species and may make themparticularly vulnerable to over-exploitation.Management strategies based on size limits thatallow a few years of spawning prior to reachingminimum legal size are insufficient.Sustainable fisheries will require multipleprotected areas to preserve brood stockaggregations necessary for successfulfertilization.

Recent tectonics of the Blanco Ridge, eastern blanco transform fault zone

Bathymetric, hydro-acoustic, seismic, submersible, and gravity data are used to investigate the active tectonics of the eastern Blanco Transform Fault Zone (BTFZ). The eastern BTFZ is dominated by the ∼150 km long transform-parallel Blanco Ridge (BR) which is a right-lateral strike-slip fault bordered to the east and west by the Gorda and Cascadia Depressions. Acoustic locations, fault-parameter information, and slip vector estimates of 43 earthquakes (Mw≥3.8) that occurred along the eastern BTFZ over the last 5 years reveal that the Blanco Ridge is a high-angle right-lateral strike-slip fault, with a small component of dip-slip motion, where the Juan de Fuca plate is the hanging wall relative to the Pacific plate. Furthermore, the Cascadia and Gorda basins are undergoing normal faulting with extension predominantly oblique to the transform trend. Seafloor submersible observations agree with previous hypotheses that the active transform fault trace is the elongate basin that runs the length of the BR summit. Brecciated and undeformed basalt, diabase, and gabbro samples were collected at the four submersible survey sites along the Blanco Ridge. These petrologic samples indicate the Blanco Ridge is composed of an ocean crustal sequence that has been uplifted and highly fractured. The petrologic samples also appear to show an increase in elevation of the crustal section from east to west along the Blanco Ridge, with gabbros exposed at a shallower depth farther west along the southern (Pacific plate side) BR ridge flank. Further supporting evidence for BR uplift exists in the seismic reflection profiles across the BR showing uplift of turbidite sequences along the north and south ridge base, and gravity and magnetics profiles that indicate possible basement uplift and a low-density zone centered on the ridge's Pacific plate side. The BR formation mechanism preferred here is first, uplift achieved partially through strike-slip motion (with a small dip-slip component). Second, seawater penetration along the fault into the lower crust upper mantle, which then enhanced formation and intrusion of a mantle-derived serpentinized-peridotite diapir into the shallow ocean crust, causing further uplift along the fault.

Volcanic and Seismic Swarm Events on the Reykjanes Ridge and Their Similarities to Events on Iceland: Results of a Rapid Response Mission

On 21 May 1989, a major earthquake swarm on the Reykjanes Ridge at59°44′ N, 29°32′ W at a water depth of about 1000 m andabout 500 km southwest of Iceland was detected on both the WorldwideStandard Seismic Network (WWSSN) and Icelandic seismic networks. As part ofa multi-institutional response to this swarm, the Naval ResearchLaboratory arranged for a P3 Orion Aircraft to deploy sonobuoys and AXBTs inthe immediate vicinity of the swarm activity. The detection of the swarmmotivated a survey of the region in 1990, using the towed SeaMARC IIside-looking sonar system. In 1990–1991 the Russian ShirshovInstitute of Oceanology offered the use of its MIR deep-divingsubmersibles to investigate the rise axis for recent volcanism. During 1992,a scientific team comprised of five US and ten Russian scientists mobilizedthe twin, deep diving Russian submersibles to study the spreading axis ofthe Reykjanes Ridge. The resulting data analyses allows us to conclude thatthe 1989 seismic swarm event occurred adjacent to and east of the largeaxial high in the center of our survey area. The length, width and depthrange of the earthquakes were very similar to major seismic swarm eventsconfined to fissure systems in the Krafla region of Iceland. It is likelythat the earthquake swarm was located on a fresh, well-defined systemof fissures and faults extending south of the northernmost axial highstudied. The earthquake swarm was probably associated with an emanation oflava creating a region of high backscatter, located just to the east of thecentral axial high. In addition, the region of high-backscatterremains unsampled because it lay underneath the nadir of the processedSeaMARC tracks used to plan the submersible survey. However many sampleswere taken and structural studies of the evolving Reykjanes Ridge werecarried out.

Pictorial 5: Results of Submersible Survey at the Kane-Transform Fault of the Mid-Atlantic Ridge

Pictorial 5: Results of Submersible Survey at the Kane-Transform Fault of the Mid-Atlantic Ridge

Conductive heat flow at the TAG Active Hydrothermal Mound: Results from 1993–1995 submersible surveys

We report 70 measurements of conductive heat flow at the 50‐m‐high, 200‐m‐diameter TAG active hydrothermal mound, made during submersible surveys with Alvin in 1993 and 1995 and Shinkai 6500 in 1994. The stations were all measured with 5‐thermistor, 0.6‐ or 1‐m‐long Alvin heat flow probes, which are capable of determining both gradient and thermal conductivity, and were transponder‐navigated to an estimated accuracy of ±5–10 m relative to the 10‐m‐diameter central complex of black smokers. Within 20 m of this complex, conductive heat flow values are extremely variable (0.1‐ > 100 W/m²), which can only be due to local spatial and possible temporal variability in the immediate vicinity of the vigorous discharge sites. A similar local variability is suggested in the “Kremlin” area of white smokers to the southeast of the black smoker complex. On the south and southeast side of the mound, there is very high heat flow (3.7‐ > 25 W/m²) on the sedimented terraces that slope down from the Kremlin area. Heat flow is also high (0.3–3 W/m²) in the pelagic carbonate sediments on the surrounding seafloor within a few tens of meters of the southwest, northwest, and northeast sides of the mound. On the west side of the sulfide rubble plateau that surrounds the central black smoker peak, there is a coherent belt of very low heat flow (<20 mW/m²) 20–50 m west of the smokers, suggestive of local, shallow recharge of bottom water. The three submersible surveys spanned nearly two years, but showed no indication of any temporal variability in conductive heat flow over this time scale, whether natural or induced by ODP drilling in 1994.

Interpretation of temperature measurements from the Kaiko-Nankai cruise: Modeling of fluid flow in clam colonies

During the Kaiko-Nankai detailed submersible survey, numerous measurements of the temperature gradient inside the sediment were performed on the deepest active zone of fluid venting, which is situated on the anticline related to the frontal thrust, using the Ifremer T-Naut temperature probe operated from the submersibleNautile. We thus obtained the temperature structure below different types of clam colonies associated with fluid venting. We used the finite element method to model the thermal structure and fluid flow pattern of these vents and to determine the velocity of upward fluid flow through the colonies. On a biological basis, four types of clam colonies are defined. Each biological type has distinctive thermal characteristics and corresponds to a particular fluid flow pattern. Darcian flow velocity in the most active type of colony (type A) is of the order of 100 m/a. The total amount of fluid flowing through colonies in the studied area is estimated to be 200 m3 a−1 per metre width of subduction zone. Most of the flow is vented through type A colonies. This value is more than one order of magnitude too high to be compatible with the amount of water available from steady-state compaction of sediments in the whole wedge. Thermal arguments suggest that downwelling of seawater occurs around type A colonies and that seawater is then mixed with upcoming fluids at a depth of 1 or 2 metres. Furthermore, finite element modeling shows that a salinity difference of a few parts per mil between the upcoming fluids and seawater is sufficient to drive convection around the colonies. As water samples from a few vents indicate that the fluid source should actually be significantly less saline than seawater, we propose that the very high fluid flows measured are a consequence of the dilution of the fluid of deep origin with seawater by a factor of 5 to 10.

Mud volcano field seaward of the Barbados Accretionary Complex: A submersible survey

We describe a 0.55 km2 semicircular mud pie and three adjacent diapiric mounds explored using the submersible Nautile near 13°50′N, 12 km seaward of the Barbados accretionary complex deformation front. This diapiric field had been previously explored with an acoustically navigated deep‐towed side scan sonar, thus providing an accurate base map. The mud pie is situated at a depth of 4938 m. In addition to visual observations, water, rock, sediment, and biological sampling, we measured temperature gradients and made geochemical analyses of the samples. Fluid venting associated with chemosynthetic animal communities is widespread on all four structures but is maximal in the central third of the mud pie. The chemosynthetic life distribution on the mud pie displays a concentric zonation which we relate to an increase in venting activity toward the center, as evidenced by thermal gradients. Recent mud flows are present in the centralmost part of the mud pie. The simplest interpretation of the structure is that it is a mud lake covered by a mechanically resistant carbonate crust. We determine that approximately 106 m3 of fluid are advected upward through the crust of the mud lake every year. This large amount of fluid implies large‐scale lateral transport.

Cenozoic deformation and Mesozoic abandoned spreading centre in the Tagus Abyssal Plain (west of Portugal): results of a multichannel seismic survey

The seismic stratigraphy of the sedimentary section that fills in the Tagus Abyssal Plain is specified by means of new multichannel seismic data. Six sedimentary sequence units are identified. These, ranging from Late Jurassic to Recent, have been dated by extrapolation of Deep Sea Drilling Project (DSDP) and Ocean Drilling Program (ODP) drilling results and from sampling by dredging and submersible surveys. Based on this stratigraphic calibration and other geophysical data, a geodynamic evolution can be established for the Tagus Abyssal Plain. Faulting and folding in these sedimentary units clearly indicate that both the Portuguese Margin and the adjacent oceanic crust were deformed by a Miocene compressive event. An Oxfordian–Kimmeridgian rifting in the Lusitanian Basin and on the Portuguese Margin may herald a sea-floor spreading in the Tagus Abyssal Plain. The continent–ocean boundary is marked by landward-dipping reflectors similar to those observed along the Canadian margin. The oceanic structure of the Tagus Abyssal Plain is compared with those of the small oceans and (or) slow-spreading-rate centres. The eastern Tagus Abyssal Plain could be underlain by a Latest Jurassic (M-21)–Earliest Cretaceous (M16) spreading centre that was abandoned in Earliest Cretaceous (142 Ma), whereas the western Tagus Abyssal Plain could be underlain by a younger crust (from M10 to J) formed after a ridge jump. This younger oceanic domain may have a western complement in the Newfoundland Basin.

Zenisu Ridge: mechanical model of formation

Seabeam, seismic and submersible surveys took place during the Kaiko Project and revealed significant compressive deformation at the northeastern end of the Philippine Sea plate, related to the recent collision of the Izu-Ogasawara Arc against Central Japan. Intraoceanic thrusting at the base of the Zenisu Ridge, a linear topographic high running a few tens of kilometers south of the Nankai Trough, is supported by tectonic, magnetic and gravimetric data. We investigate the formation of the Zenisu Ridge in terms of compressive mechanical failure of a thin elastic-perfectly plastic plate, subducting at a trench and subject to a regional compressive axial force. The rheological envelope concept is used throughout the numerical calculations. Based on a detailed study of flexure of the present-day bending far from the deformation zone, we evaluate the bending forces involved: the bulge is 120 to 150 m high and the compressive stress all along the Nankai Trough is about −100 MPa. In the Zenisu Ridge area, an additional compressive stress is superimposed due to the nearby collision at Izu-Peninsula. We compute the vertical distribution of the deviatoric stress before failure and find that the deviatoric stress is maximum at a depth of 20–25 km in the trench area, and again at the surface 60 to 100 km seaward, in the vicinity of the bulge. The development of a thrust joining these two maxima through the entire thickness of the lithosphere is discussed. The model predicts that the formation of the Zenisu Ridge did not occur before 4 Ma and is caused by progressive tectonic uplift due to the redistribution of bending stresses as the ridge approaches the Nankai Trough.

Deep-sea submersible survey in the Suruga, Sagami and Japan Trenches: preliminary results of the 1985 Kaiko cruise, Leg 2

Nine submersible dives were made in three trenches off central Japan, between 2990 and 5900 m of water depth. Our observations confirm the interpretation that Daiichi-Kashima Seamount is a Cretaceous guyot formed on the Pacific plate that has traveled into the Japan Trench. We also confirmed the previous interpretation of a large normal fault that splits the seamount in two halves, the lower one being now subducting beneath the Japan margin. Compressional deformation was identified within the lower part of the inner slope in front of the seamount. The pattern of deformation that affects Quaternary sediments is in agreement with the present kinematics of the convergence between the Pacific plate and Japan. Deep-water (5700 m) clam colonies are associated with advection of fluids, driven by the subduction-related overpressures. In the northern slope of the Boso Canyon, along the Sagami Trough system (Philippine Sea plate-Japan boundary), the deformation affecting a thick upper Miocene to lower Pliocene sequence indicates two directions of shortening: a N175°E direction which is consistent with the present relative motion along the Sagami Trough (N285–N300°E) and a N30°E direction which could be related to a more northerly direction of convergence that occured during the early Quaternary and earlier.

Submersible surveys of benthos near a turbidity cloud

Seabed impacted by settling solids from a turbidity plume can be surveyed by submersible using a ‘bounce’ technique to determine depth of the turbidity front, and hence depth above which benthos observations are practical. The protocol adopted was that after frontal depth was determined by a preliminary dive, the site for the benthic survey was selected. On arrival on the seabed a reconnaisance was made for the observers to agree on the identification of visible benthos. Then 100 × 1 m 2 transects were surveyed using a 1 m 2 quadrat attached to the submersible in view of the port-side observer. At the site investigated, two shallow water stations (depths 13–15 m, and 16–18 m) showing light deposits of mine tailings were seen to support associations of large epifauna and infauna. The first station had an almost single-species stand of the burrowing anenome Pachycerianthus fimbriatus at approximately 100 · 100 m −2, and the second a diverse species association with several starfish and other species at 1–2 · 100 m −2. Burrow holes (up to 3–5 cm diameter) of various sizes indicated that large infaunal species were present. At a deeper station (36–27 m) with heavy tailings and almost no burrow holes, a single Dungeness crab, Cancer magister, and an unidentified shrimp were seen in the 100 m 2 transect. Each dive site was inshore (shallower) from a routinely monitored benthos station shown repeatedly to support a population of small infauna in the tailings. Burrow holes have the potential for determining the identity and abundance of large infaunal species present if an identification system can be developed. The technique of epifaunal and burrow hole surveys (by submersible and scuba diver), combined with infaunal and sediment core surveys and contemporary theory on infaunal succession, provides the potential for a procedure to map the distribution in shallow water of tailings impact and benthic recovery (succession).

Depth limits of Bermudan scleractinian corals: a submersible survey

Depth distribution, zonation pattern and growth morphology of 17 hermatypic and 4 ahermatypic coral species were investigated at eight different locations along the Bermuda platform with the research submersible GEO and by SCUBA diving in August–September 1983. Hermatypic coral growth occurs to a depth of 50 to 70 m, with a single Montastrea cavernosa growing at 78 m. Dominant forms in shallow-water coral communities are Diploria sp. and Porites astreoides, while M. cavernosa, Agaricia fragilis and Scolymia cubensis occur in deep-water associations below 60 m. Vertical visibilities (up to 178 m) and distribution of the photosynthetically active radiation revealed good light penetration values (1% level at about 100 m depth), which should favour hermatypic coral growth to a much greater depth than it actually occurs. Nor should the prevailing temperatures limit the depth of coral growth. Most deep-water hermatypes observed grow on remnants of Pleistocene reefs down to about 60 m. The vast areas of large massed rhodolith nodules below 50 to 60 m are unsuitable bottom for coral colonisation. Macroalgae growth seems to be the strongest factor controlling coral growth in deep water. Bermuda stony corals have a low growth form diversity. Various intraspecific morphs may occur at the same as well as at different depths, with a general trend towards flatter shapes with depth. Comparison with a similar study on Red Sea corals suggests that annual distribution of radiant energy on the most northern Atlantic reefs of Bermuda may be responsible for the occurrence of flat and cuplike growth forms in relatively shallow water, and for the shallower depth limits of hermatypic growth.

Submersible surveys of benthos near a turbidity cloud : Ellis, D.V. and C. Heim, 1985. Mar. Pollut. Bull., 16(5):197–203

Submersible surveys of benthos near a turbidity cloud : Ellis, D.V. and C. Heim, 1985. Mar. Pollut. Bull., 16(5):197–203

Subduction in the Hellenic Trench: probable role of a thick evaporitic layer based on Seabeam and submersible studies

Summary A Seabeam survey of the Hellenic Trench in 1978 was followed by a submersible survey in 1980. Fifteen dives were completed between 1500 and 3000 m covering 58 km on the bottom and obtaining 48 samples. Thus, the Hellenic Trench is probably now the most intensively studied trench as previous work included seismic reflection, dredging, coring, bottom photography, side-scan sonar and drilling. Because of the high sedimentation rate and the relatively shallow depth, we did not expect outcrops except along faults and in canyons. Active tectonics with numerous faults and folds have been observed. But we have also discovered vertical cliffs with hard rock exposure over a depth range of 1000–1500 m. In this paper, we discuss the nature of the hard rock outcrops observed both on the inner wall and on two hills that we interpret as diapirs. We conclude that at least part of these hard rocks are probably Messinian evaporites and concentrate on the tectonic role of this evaporitic layer in the Hellenic subduction. It is likely that the evaporites do not subduct but tend to pile up in the trench, to form a floating evaporitic basin.