Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Method Descriptions

1.1 Remotely Operated Vehicles

Remotely Operated Vehicles (ROVs) are tethered robotic devices used for exploration, inspection and the execution of specific tasks under water. A typical ROV (Fig. 1) consists of a sturdy frame, a floatation unit to provide buoyancy (generally, ROVs are marginally positively buoyant), a number of thrusters to enable manoeuvrability in three dimensions, and a tether which connects the system to the host ship. The tether can also be connected to a Tether Management System (TMS): a separate non-buoyant unit that is lowered with the ROV on an armoured cable from the ship, and dampens the effect of the ship’s motions on the ROV.

Fig. 1
figure 1

Types of ROV used for seafloor studies: the scientific working class ROV Isis from the National Oceanography Centre, UK (a photo by L. Marsh), the Falcon inspection class ROV from Plymouth University (b photo by V. Huvenne) and the Deep Trekker mini-ROV from the NOC (c photo by Tim Le Bas, NOC)

Full size image

Initially, ROVs were developed in the 1960s for military use (recovery operations, mine clearing, etc.), and were named CURV: “Cable-controlled Underwater Recovery Vehicles” (Christ and Wernli 2007; Ridao et al. 2007). The technology was introduced into the industrial domain in the late 1970s and 1980s, to enable operations beyond depths that could be reached by divers. Nowadays they are mainly used by the Oil and Gas sector, but they are also deployed for scientific research and salvage operations.

Based on the size of the frame, the operational requirements and the vehicle’s depth rating, an ROV can be equipped with a range of sensors, instruments and tools. In the most basic configuration, an ROV will carry one or more video cameras, of which the data is transferred to the on-board operator in real time. In addition, there may be one or two manipulator arms with various levels of functionality, and a ‘basket’ or storage device to store samples or tools. In more advanced configurations, other sensors or equipment can be integrated into the system, ranging from small CTDs, optical sensors or chemical sensors to HD cameras, sector scanning sonars, multibeam echosounders and suction samplers.

ROVs come in a wide range of capabilities (Marine Technology Society 2015; Kernow Marine Explorations Global Limited 2016). Apart from their depth-rating, ROVs are typically classified according to their size: work-class ROVs are large and powerful vehicles (up to 2 m high and 4 m long), that can be used for complex operations and can carry several instruments or a large volume of samples. In most cases they are equipped with two 5- or 7-function manipulator arms. Within the scientific domain, some of the best-known examples include the Jason ROV from the Woods Hole Oceanographic Institution (WHOI), the Isis ROV from the National Oceanography Centre (NOC), and the ROV Victor from the Institut Français de Recherche pour l’Exploitation de la Mer (IFREMER), which can all operate down to 6000 m water depth (e.g. Yoerger et al. 2000; Rigaud 2007; Marsh et al. 2013). Inspection or observation class ROVs are smaller (metre-size), often have only one manipulator arm, and are mainly used for video surveys. Although more compact than the work-class vehicles, they still have sufficient power to work in full marine conditions, in some cases down to 3000 or 4000 m water depth. Besides video surveying, they can often carry out a number of other tasks (Pacunski et al. 2008). Eyeball class ROVs (also known as mini- and micro-ROVs) are even smaller. They can often be packed in a single suitcase, and can be deployed by a single person. They are specifically built for inspection work, generally in relatively calm and shallow waters (<200 m). Assembly kits are now available online that allow users to build their own mini-ROV (e.g. OpenROV 2016).

1.2 Autonomous Underwater Vehicles

Autonomous Underwater Vehicles (AUVs) are unmanned, un-tethered robot submarines that operate fully independently to carry out pre-programmed operations and surveys (Griffiths 2003). AUV endurance typically ranges from a few hours to several days, although rapid technological developments are now bringing long-range operations within the possibilities, with endurances stretching to weeks or even months (Hobson et al. 2012; Furlong et al. 2012). Operational depths range from a few hundred metres for the smaller vehicles (Griffiths 2003; Wynn et al. 2014) to full ocean depth for the larger models (>6000 m; Huvenne et al. 2009).

AUVs are typically categorised as either “cruising” or “hovering” vehicles (Fig. 2). Cruising, or survey AUVs are generally torpedo-shaped and driven by a single propeller. They move at speeds up to 2 m/s, and are optimised to cover large distances along pre-designed tracks (Wynn et al. 2014). They form the main type of AUVs used in the commercial world, while some of the most prominent scientific examples include the Autosub series from the NOC, the AsterX and IdefX from IFREMER and the Dorado series from MBARI (Rigaud 2007; Caress et al. 2008; McPhail 2009). Hovering AUVs, on the other hand, have several propellers/thrusters, which allow them to move in any direction and provide them with a high manoeuvrability, much like an ROV. They are designed for precision operations, slow motion surveys (e.g. seabed photography) and work in distinctly 3-dimensional terrains, such as around hydrothermal vents or coral reefs. Among the best-known scientific examples of hovering AUVs are ABE and Sentry from WHOI (e.g. Yoerger et al. 1998; Wagner et al. 2013).

Fig. 2
figure 2

Examples of AUV types: the torpedo-shaped survey AUV Autosub6000 from the National Oceanography Centre UK (a photo by V. Huvenne), and the hovering AUV Sentry from WHOI (b photo by Oscar Pizarro, Australian Centre for Field Robotics, University of Sydney)

Full size image

Similar to ROVs, depending on their depth rating and size, AUVs can be equipped with a range of sensors (CTDs, ADCPs, chemical sensors, photo cameras, sonars, magnetometers, gravimeters etc.) (e.g. Caress et al. 2008; Connelly et al. 2012; Sumner et al. 2013; Williams et al. 2010; Yoerger et al. 1998). However, the lack of tether, and hence of direct power input, limits the sensor power consumption and duration of activity. In addition, AUVs are currently not yet equipped for extensive seabed or faunal sampling, although sampling of the water column has been achieved (Pennington et al. 2016). Overall, AUVs are more suited for survey operations, acquiring sensor data along smooth, pre-programmed tracklines, while ROVs are optimal for high-resolution, highly detailed and interactive work, including HD video surveying and sampling. An extensive review of the use and capabilities of AUVs for geological research was recently published by Wynn et al. (2014).

Over the last 5 years, a number of ‘hybrid underwater vehicles’ have been developed, combining advantages of ROVs and AUVs (Bowen et al. 2013). Undoubtedly the most famous example was the Nereus H-ROV, engineered by WHOI, which unfortunately was lost at sea in 2014 in the Kermadec Trench (Cressey 2014). Using a single optical fibre as tether, the dual-purpose vehicle could be used as an interactive ROV, enabling sampling operations, while it could equally operate as a hovering AUV, without tether. However, as the single fibre only provided data transfer capacity, the power limitations for the vehicle and its scientific payload still remained.

1.3 Using Robotic Vehicles to Study Seafloor Geomorphology

Nested surveys: The availability of underwater vehicles such as ROVs and AUVs has revolutionised our understanding of seafloor morphology and processes. Their free-diving capability means that the activity of seafloor mapping is no longer restricted to measurements carried out from the sea surface, or from towed vehicles. Using AUVs or ROVs, high-frequency sonars can be operated with high precision close to the seafloor, offering the opportunity to create maps of any required resolution. In principle, the altitude of the echosounder above the seabed can be freely chosen (within the limitations of the instruments used), and therefore, through simple geometry, also the resolution of the resulting map. The increased detail achieved in this way provides insights that previously were impossible to obtain. The trade-off for this extra resolution is generally a reduction in the area that can be mapped within a certain timeframe, as a reduction in altitude of the sonar or optical sensor above the seabed results in a reduction in coverage of single swaths or images. Furthermore, AUVs, and certainly ROVs, travel at a slower speed than surface vessels, and therefore can cover less ground. Hence, underwater vehicles are generally used to complement ship-board mapping activities, rather than replacing them. Operations are often planned in a nested scheme: target areas identified on conventional ship-borne multibeam bathymetry maps are surveyed by a cruising AUV equipped with a multibeam or sidescan sonar system (see Chapters “Sidescan Sonar” and “Multibeam Echosounders”). To achieve the highest level of detail, ROV-based multibeam or photogrammetry techniques are then applied to specific features of interest (see below).

Navigation: One of the biggest challenges when using underwater robotic vehicles for high-resolution seafloor mapping is to obtain precise and accurate positioning information. Recording datasets with a spatial resolution in the order of 10 cm is a lost effort if the relative positional accuracy of the vehicle between consecutive depth measurements is not at least an order of magnitude more precise. Navigation for ROVs and AUVs is generally based on a combination of acoustic techniques. As a result of spherical spreading and attenuation of the signal in the water, those become less accurate in deeper water and over larger distances. When high-accuracy maps are required and sufficient time and funds are available, a long base-line (LBL) transponder network can be installed on the seafloor, to enable high-precision triangulation of the vehicle position (Milne 1983). However, the time-investment required for such a set-up, especially in deep water, can be extensive. In most cases, ROV positioning is based on ultra-short base-line (USBL) navigation from the host ship (Christ and Wernli 2007). Once close to the seafloor, ROV systems can also use dead-reckoning based on an inertial navigation system and a Doppler Velocity Log (DVL, Kinsey and Whitcomb 2004). This provides a better relative positioning, but may be subject to a cumulative error or ‘drift’ that gradually builds up over longer distances. Regularly re-setting the inertial navigation with reference to the USBL positioning is recommended. AUVs use the same dead-reckoning technology, but are generally only positioned with the help of the ship’s USBL at the start (and sometimes end) of their mission if they operate fully independently from the host ship. To increase accuracy of the mission start position, specific techniques can be applied (e.g. “range-only navigation”, see McPhail and Pebody 2009).

Further corrections to the vehicle navigation can be applied through feature matching algorithms on overlapping sections of the seafloor mapping data. This can be carried out manually in post-processing, or increasingly in real-time using SLAM algorithms (Simultaneous Localisation and Mapping, e.g. Barkby et al. 2009). SLAM can be described as the process whereby the system continuously builds up a database of landmarks and features of the environment in which it is moving, and simultaneously correlates this database with its current observations to determine its location (West and Syrmos 2006). Similar approaches can be used to integrate new seafloor surveys with existing seafloor data, either from previous shipboard surveys, or from previous AUV or ROV missions.

2 Different Applications of ROVs and AUVs for Geomorphological Studies

2.1 High-Resolution Multibeam Bathymetry

By far the most common use of marine robotic vehicles for geomorphological studies is for the acquisition of high-resolution multibeam echosounder data. ROV- and AUV-based surveys can create spectacular digital terrain models that illustrate a wide range of smaller-scale processes which are traditionally overlooked in studies based on shipboard bathymetry alone. One of the most pertinent examples is the illustration of large deep-sea scours in the outflow channels of submarine canyons (Fig. 3)—features that are normally not resolved by shipboard multibeam maps, but that provide an important insight in the occurrence of sediment gravity flows and their erosive strength (e.g. Caress et al. 2008; Huvenne et al. 2009; Macdonald et al. 2011). Apart from scours and erosive features, AUV- and ROV-based multibeam maps of submarine canyons also reveal bedform morphologies and their evolution (e.g. Paull et al. 2013; Tubau et al. 2015), the location of local cliffs and outcrops (e.g. Masson et al. 2011) and detailed gully systems (e.g. Rona et al. 2015).

Fig. 3
figure 3

High-resolution multibeam bathymetry acquired by the AUV Autosub6000 at ca. 4600 m water depth on the margin of the Whittard Canyon outflow channel in the Bay of Biscay, illustrating large scours which cannot be mapped with traditional shipboard multibeam systems (lower right inset). Modified after MacDonald et al. (2011)

Full size image

A similar revolution in the interpretation of deep-sea geomorphology has been achieved through the ROV- and AUV-mapping of landslide scars (e.g. Lee and George 2004), hydrothermal vents and mid-ocean ridges (e.g. Yoerger et al. 2000; Rogers et al. 2012), mud volcanoes (e.g. Jerosch et al. 2007; Dupré et al. 2008) and cold-water coral mounds (e.g. Huvenne et al. 2005; Grasmueck et al. 2006), for example.

2.2 True 3-Dimensional Morphology

In addition to bringing instruments as close to the seafloor as necessary, robotic technology has created new opportunities to develop unusual sensor configurations. By mounting sonar systems in a sideway or even upward-directed orientation, the true 3-dimensional morphology of underwater terrains can be mapped: something that can never be achieved with a traditional downward-facing set-up. Yoerger et al. (2000) carried out pioneering work when they deployed a forward-looking pencil beam scanning sonar on the Jason ROV to create 3D digital models of hydrothermal vents on the Juan de Fuca Ridge. Similarly, Gary et al. (2008) used an array of 54 single narrow-beam sonars on the DEPTHX AUV, to obtain the complete morphology of submerged caves. Expanding the approach to multibeam sonars to achieve higher data density, Wadhams et al. (2006) mounted an upward-facing EM2000 multibeam system on the Autosub II AUV to develop detailed maps of the underside of icebergs and sea ice. Huvenne et al. (2011) placed an SM2000 multibeam system in a forward-looking position on the front of the Isis ROV and moved the ROV sideways in order to map the geological fabric and biological habitat of vertical and overhanging cliffs in a submarine canyon. They also applied the same approach to map the intricate morphology of submarine landslide headwall scarps (Fig. 4; Huvenne et al. 2016).

Fig. 4
figure 4

Application of ROV-based forward mapping of vertical substrates: submarine landslide headwall scarp at ca. 1400 m water depth on SW Rockall Bank. ROV bathymetry nested in the shipboard bathymetric grid (a), and details of the headwall scarp represented as point clouds (b, c), illustrating individual blocks protruding from the headwall scarp, and vertical cliffs fringed by talus. Modified after Huvenne et al. (2016)

Full size image

2.3 Sidescan and Synthetic Aperture Sonar

While multibeam bathymetry naturally forms the primary source of information for seafloor geomorphology studies, in certain cases the use of sidescan or synthetic aperture sonars mounted on ROVs or AUVs may be a better choice. The low grazing angle used by the latter acoustic systems allows identification of subtle changes in the terrain, such as those caused by pockmarks (e.g. Wagner et al. 2013) or small-scale bedforms (e.g. Wynn et al. 2014) that are not easily resolved by multibeam systems (see also Chapter “Sidescan Sonar”). In addition, the backscatter intensity of sidescan or synthetic aperture sonars, registered at higher resolution than what can be achieved with multibeam systems providing the same swath width, may reveal patterns that are either exclusively expressed or at least accentuated in the sediment type or seafloor roughness. The example in Fig. 5 shows a field of iceberg ploughmarks on Rockall Bank (NE Atlantic, see also Robert et al. 2014). Although these features do have a bathymetric expression at the seafloor, their pattern is enhanced in the backscatter response registered by the sidescan sonar system. The characteristic gravel ridges at either side of the ploughmarks create a high-intensity backscatter return, while the finer-grained sediments ponded in the troughs have a much lower backscatter. In addition, the high-resolution sidescan data (400 kHz) allow identification of individual cold-water coral colonies, based on their high backscatter signal and their acoustic shadow.

Fig. 5
figure 5

High-resolution 400 kHz sidescan sonar mosaic collected with the Autosub6000 AUV at ca. 200 m water depth on Rockall Bank, NE Atlantic. Iceberg ploughmarks are represented by the high backscatter gravel ridges (light colours) and low backscatter ponded sediments (darker colours). Cold-water coral colonies (as illustrated) can be recognised as small features on the gravel ridges. Modified after Robert et al. (2014)

Full size image

2.4 Photomosaicking and Photogrammetry

To reach the finest scale in the nested scheme of geomorphological observations, photographic and video techniques are generally used. However, as a result of the attenuation and backscatter of light in the marine environment, the extent of underwater scenes that can be imaged from an ROV or AUV at any single point in time is severely limited. To place individual visual observations within their direct spatial context, photo- and video mosaicking techniques have been developed, where composite images are created from a large number of overlapping individual pictures or frames (e.g. Pizarro and Singh 2003; Singh et al. 2007). Extensive image corrections need to be applied to compensate for backscatter, non-uniform illumination, colour balance and geometric distortions (Morris et al. 2014). Using common features in overlapping images, photographs or frames can then be tiled together and georeferenced to create large mosaics of extensive seafloor areas that cannot be imaged at once (e.g. Yoerger et al. 2000; Jerosch et al. 2007; Escartin et al. 2015). The method can equally be applied to image large vertical structures, such as hydrothermal vent chimneys that are several metres high (e.g. Marsh et al. 2013).

Where pairs of photographs are available that picture the same scene from slightly different points of view, techniques of photogrammetry can be used to interpret the 3D morphology of the scene in great detail (Fig. 6). This can be achieved either through the use of stereophotographic cameras, or by using consecutive photographs from a single moving camera that have sufficient overlap. Using the former approach, Ling et al. (2016) mapped the fine-scale morphology of coral reefs, identifying the key habitat for invasive urchins. It equally allowed for the quantification of reef rugosity, slope and aspect (Friedman et al. 2012). The second technique, also named “structure from motion”, has seen huge development over the last few years under the influence of the mobile phone market. The technique, however, is just as applicable to the marine environment, and can result in very detailed representations of the seafloor, that allow ‘virtual geological fieldwork’ (e.g. Kwasnitschka et al. 2013).

Fig. 6
figure 6

Example of photogrammetry results. a 3D textured mesh reconstructed from HD video frames (position in green) acquired during an Isis ROV dive on a vertical cliff face in Whittard Canyon at ca. 1300 m depth, illustrating step-like features and overhanging coral framework. Steps involved in the reconstruction: b video frames, c generated point cloud and d zoom on a section of the resulting high resolution textured mesh

Full size image

2.5 Laser Line Scan

In addition to the acoustic and optical techniques described above, a further technology has been developed that potentially can fill a resolution gap between multibeam and sidescan sonar mapping on the one hand, and photogrammetry on the other hand. This technique is generally referred to as “Laser Line Scanning”, and can be used to image the seafloor (e.g. Carey et al. 2003) or to obtain 3D morphological reconstructions (e.g. Tetlow and Spours 1999). In both cases, a strip of seabed across the path of the robotic vehicle is illuminated, either by a sheet laser or by a scanning pencil laser. As the vehicle moves forward, the changing reflectivity and shape of the strip is recorded in a series of (stereo-)photographs, from which a seabed image or 3D reconstruction is created line by line. The method was developed to image man-made structures such as pipelines or mines (e.g. Tetlow and Spours 1999), but has also been applied to map archeological remains (e.g. Roman et al. 2010, 2012; Bruno et al. 2011) or the morphology of hydrothermal vents and benthic communities (e.g. Maki et al. 2011). Recent developments have also expanded the technique for the detection of diffuse seafloor venting (which would equally disturb the path of the laser beam; Smart et al. 2013). Because laser mapping works with narrower beam widths than acoustic approaches, it can potentially provide a finer resolution than most multibeam or sidescan systems (Roman et al. 2010, 2012). At the same time, the narrow light source in the laser pencil or sheet produces less backscattering in the water column, which means the approach is also applicable in fairly turbid waters, where photogrammetry methods would fail (Bruno et al. 2011). A number of laser line scanning systems are now available on the market, although their use is still less common than multibeam, sidescan or photography.

3 Future Directions

Given the wide range of advantages offered by autonomous and robotic vehicles (cost savings for mapping and monitoring, improved resolution, ability to reach inaccessible places, …), the rapid technological developments in the field, their increasing availability and the resulting reduction in cost, it is to be expected that the use of unmanned vehicles will only increase in the future. As they become more common, new applications and configurations will equally become mainstream. This may include new sensors, or engineering solutions for automated seafloor monitoring capabilities and repeat surveys, for example. However, by far the most promising development in the field is the increasing deployment of fleets of vehicles in coordinated operations. Combinations of AUVs and USVs (Unmanned Surface Vehicles) are currently being tested, for example the Innovate UK Autonomous Surface/Sub-surface Survey System (ASSS) project (Research Councils UK 2016), with one or more USVs providing continuous USBL navigation to the submarine AUVs, reducing navigation uncertainty or the need to follow the vehicles with a surface vessel. They will also enable the operator to continuously monitor the AUV position, data and performance from shore, via acoustic and satellite links with the USV. Equally, squads of vehicles may be deployed simultaneously to increase the area covered during single surveys, or to create nested observations of seafloor and water column characteristics in one single operation. Small-scale AUVs that can be rapidly deployed from USVs or manned/unmanned aerial vehicles will in the future provide the ability to rapidly respond to e.g. oil spills and/or map extensive areas of seabed at reduced cost. Finally, there are major developments underway in automated adaptive sampling, whereby AUVs can respond to environmental cues and adapt their survey pattern accordingly without human control, e.g. an AUV chemically detecting a signature of active hydrothermal venting could automatically adjust its mission to then map and image the targeted feature, without having to return to the surface. Thanks to these new multi-vehicle, and increasingly intelligent and adaptive operations, it can be expected that the volume of high-resolution seabed morphology data will increase rapidly in the near future, opening exciting opportunities for new insights in seafloor geomorphology.