1 Introduction

Mid-ocean ridges occur where Earth’s oceanic tectonic plates are separating and the gap between them is being filled by hotter upwelling mantle rocks (Fig. 1a). Those hotter rocks have lower density than mantle rocks beneath older oceanic crust, hence their relative buoyancy has created ridges on Earth’s solid surface. In practice, these are only visible as ridges in topographic cross-sections of large vertical exaggeration. At the ridge crest itself (the “axis”), the lithosphere is rifting. This process has produced many normal faults, which form linear escarpments on the seabed. Where the plates are separating slowly, the uppermost mantle has become colder and more rigid than where plates are separating rapidly. In slow-spreading ridges, where the lithosphere is thick, we commonly find a rift valley, a feature that has been ascribed to a steady-state necking of the lithosphere (Tapponier and Francheteau 1978) or to viscous head loss in the asthenospheric mantle rising between the plates (Ribe 1988). At fast spreading ridges, in contrast, the lithosphere is hotter and thinner, so the axial valley is commonly very narrow and shallow or non-existent, commonly replaced with a narrow ridge. Also a result of the plate separation, pressure in the upwelling mantle progressively decreases, causing partial melting. Those ‘melts’ percolate up towards the surface, where they intrude the crust (which hence is largely formed of gabbroic intrusions). Injections of melt into the brittle crust create the dyke complex and occasional eruptions at the solid surface produce a variety of volcanic landforms (lava flows, ridges, cones and seamounts).

Fig. 1
figure 1

a Regional profile of a mid-ocean ridge illustrating seafloor spreading. b Plan-view pattern of orthogonal spreading segments and transform faults

Full size image

The spreading described above is organized into segments of the ridge, which can be viewed roughly as composite rifting volcanoes. Between those “spreading segments”, the movement can be accommodated by strike-slip faults, which commonly lie within deep “transform valleys” oriented perpendicular to the spreading segments (Figs. 1b and 2a). In contrast, in places where the distances between segments are small, the strike-slip movement instead can be accommodated without a through-going fault, in a more complicated way involving both extension and strike-slip movement (by “transtensional” fault movements and/or by oblique volcanic intrusions). These features are called “non-transform discontinuities”.

Fig. 2
figure 2

Bathymetry maps of the central-north Atlantic illustrating our improved view of regional geomorphology that has accompanied technological improvements. a Physiography based on sparse echo-soundings of research vessels (Heezen and Tharp 1961). b Bathymetry obtained by interpolating between echo-soundings with the gravity field derived from sea-surface shape from satellite altimeters (Becker et al. 2009; Smith and Sandwell 1997). (Reprinted from Mitchell 2015, with permission from Elsevier.)

Full size image

Mid-ocean ridges lie above sea level in only a few isolated places (such as on Iceland and in the Afar region of Africa). Elsewhere, their crests lie at an average water depth of 2700 m. Because of their poor accessibility, the development and deployment of new technology has been important in the history of how the mid-ocean ridges have been explored, starting with the early lead-line soundings for telephone cable routes in the 19th Century, which first outlined the regional shapes of the ridges, to modern research using high-resolution multibeam sonars deployed near the bed. Whereas in subaerial geomorphology, hypothesis and enquiry have been prompted by structures that have always been visible, in submarine geomorphology, advances of understanding have typically occurred when new features have been discovered or imaged with instruments at higher resolution. Also because of the limited accessibility, our knowledge of the marine geological processes creating these morphological features has relied mostly on a forensic type of analysis (reconstructing events based on analyses of samples or other geophysical data) rather than by monitoring of processes more directly, although monitoring has been possible in some instances, some of which are outlined below.

Although not untouched by erosion, mid-ocean ridges have been modified much less than continental landscapes sculpted by rivers, glaciers and coastal erosion. They therefore provide excellent examples of how the shape of the seabed provides clues to the processes that formed it. Many of the original maps of ocean bathymetry drafted by Bruce Heezen and Marie Tharp (e.g., Fig. 2a) contributed information on mid-ocean ridge features that needed to be explained ultimately by the seafloor spreading theory. Those maps were created by interpolating between widely spaced lines of echo-soundings from research ship traverses. If a more continental style of erosion of such landscapes had occurred, it is doubtful that the orthogonal pattern of straight axial valleys and transform valleys would have been revealed so easily.

The following Sects. 2–7 review the different components of ridge geomorphology, while providing some history of that development. A personal view of likely developments follows in Sect. 8. For more detailed reading on mid-ocean ridge geomorphology, Searle (2013) is recommended.

2 Regional Geomorphology

Features broader than 10 km are now well revealed by the free air gravity field derived from satellite altimetry measurements (Sandwell et al. 2014). Figure 2b shows the regional bathymetry derived partly using the altimetry measurements where echo-soundings are sparse [exploiting the effect of the seabed density contrast on free air gravity anomalies to derive bathymetry variation (Smith and Sandwell 1997)]. As shown in that figure, the pattern of fracture zones (valleys) crossing the Atlantic is revealed more clearly than in the earlier map (Fig. 2a), as well as features with trends intermediate between the fracture zones and spreading segments, which are produced by migrating spreading segments (Briais and Rabinowicz 2002).

Early in the development of the seafloor spreading hypothesis and as different parts of the oceanic basement were dated using seafloor spreading magnetic anomalies and Deep Sea Drilling Project samples, it was found that the seafloor subsides in proportion to the square root of crustal age (Parsons and Sclater 1977). Where a ridge overlies hotter mantle (near hotspots or mantle plumes), temperature gradients through the lithosphere might be expected to be steeper, so the lithosphere should cool more rapidly. The oceanic crust has indeed been found to subside more rapidly near hotspots than elsewhere, with subsidence rates correlating strongly with the ridge axis depth—Crosby and McKenzie (2009) found >10% greater initial subsidence rate for ridges 2400 m deep compared with those 2800 m deep.

“Morphometric” studies involve characterizing the shapes of structures and determining how the derived shape characteristics vary with parameters representing potential influences on the morphology. The method has been used to investigate the transition from an axial rift valley at slow-spreading ridges to a ridge at fast-spreading ridges, which occurs over intermediate spreading rates of 40–60 mm year−1, though also varies with absolute depth of the ridge axis, as ridges near mantle hotspots (where ridges are elevated) can have axial highs like fast-spreading ridges (Malinverno 1993; Small 1994).

Regional bathymetry maps such as Fig. 2b are now capable of resolving some of the topographic fabric of abyssal hills (small ridges oriented parallel to the main spreading ridge, north-south in Fig. 2b). As described in Sect. 3, these hills were created mainly by the faults at the spreading ridges. Malinverno (1991) showed that the transition from a rift valley to an axial high is accompanied by a change in the relief of the abyssal hills, such that their root-mean-square relief varies inversely with the square root of spreading rate.

The lithosphere at ridges is thinner and therefore weaker than that beneath the older parts of the oceanic plates. Where weights act on the young lithosphere, it is deflected over shorter distances. The weight of Ascension Island, for example, has depressed the young lithosphere around it to ~40 km from the island (Minshull and Brozena 1997) compared with >200 km deflection by the Hawaii islands on Cretaceous oceanic lithosphere. Other deflections can occur because of rifting, where fault movements remove weight of hanging wall rocks from lithospheric footwalls.

3 Faults

First motions of teleseismic recordings of earthquakes generated at mid-ocean ridges have suggested that normal fault movements occur with moderately dipping fault planes, of ~45° dip (Huang and Solomon 1988). This observation has supported a view up until the late 1990s that such moderately dipping normal faults have generated much of the seabed relief observed at ridges, such as illustrated from the interpretation of Deep-Tow bathymetry profiles in Fig. 3a. With continued seafloor spreading, such relief is inherited by the abyssal hills that form the rugosity of basement underlying the majority of the ocean basins.

Fig. 3
figure 3

Figures illustrating evolving ideas on how tectonic extension occurs at mid-ocean ridges. a Profiles of the Mid-Atlantic Ridge interpreted in terms of steeply dipping normal faults creating most seabed relief. (Reprinted from Macdonald and Luyendyk 1977, with permission from Geological Society of America.) b Perspective view of bathymetry from a shipboard multibeam sonar showing a corrugated surface interpreted as the exposed fault plane of a normal fault (Reprinted from Morris et al. 2009, with permission from Elsevier.)

Full size image

Faults on land have been studied with traditional surveying and space-based techniques (InSAR and GPS) to record how the deformation that occurs during individual earthquakes contributes to the development of rift topography. Unfortunately, for most mid-ocean ridges, inaccessibility and attenuation of radar waves in seawater has prevented a similar approach, but there have been successful efforts to record deformation in the Afar and Iceland. For example, Stein et al. (1991) recorded the elevation changes associated with a volcanic rifting event, in which movements of normal faults accompanied intrusion of a dyke. Those displacements partly mimicked the shape of the inner rift, suggesting how individual events contribute to the geomorphology. Mechanical modelling suggested that activated faults had steep dips of 60°–80°, supporting the view of steep faulting at ridges mentioned earlier. Significant work has been carried out in the Afar and on Iceland to characterize such events. However, as Afar and Iceland both overlie mantle hotspots where the crust is unusually thick, the rifting process may not be exactly analogous to the deeper mid-ocean ridges.

Contradicting the above view of steep faults, a closer examination of data from hull-mounted multibeam sonars by Cann et al. (1997) uncovered shallow-dipping corrugated surfaces at the seabed. The example in Fig. 3b was identified by them and shows a rounded surface with corrugations aligned with the seafloor spreading direction, as though a fault footwall has slipped at low angle or moderately steeply in the crust and rotated towards a horizontal inclination at the seabed. Paleomagnetic measurements on samples recovered by scientific drilling (at the open star symbol in Fig. 3b) indeed showed evidence for ~46° rotation (Morris et al. 2009). Although such faults were previously suspected to be associated with weak serpentinite rocks, the drilling recovered mainly gabbros in the footwall. Escartín et al. (2008) estimated that around half of the northern Mid-Atlantic Ridge is spreading by such ‘detachment’ faults, the remainder spreading by intrusion of dykes and high angle fault movements.

The evidence of strike-slip earthquakes in transform valleys, inferred from fault plane solutions (Sykes 1967), was important support for the seafloor-spreading hypothesis (Fig. 1b). Although seismic moment release rates of transform valleys are 1–2 orders of magnitude greater than those of spreading segments, seismic coupling is nevertheless ~<10% (Rundquist and Sobolev 2002), so most fault movements do not generate detectable earthquakes. Furthermore, the large and varied vertical relief of these valleys implies that there must be some dip-slip fault movements in transform valleys. Indeed, escarpments produced by such movements have been sites of submersible traverses investigating the structure of the crust (Karson 1998; Mitchell et al. 2000).

4 Landslides

Bathymetry data collected with multibeam sonars installed in the hulls of research vessels have poor resolution and can reveal only the largest landslides at ridges (Tucholke 1992). However, higher resolution seabed images from sidescan sonars towed closer to the ocean floor have revealed talus ramps and cones formed from the degradation of fault scarps (Kong et al. 1988; Searle et al. 1998). Such features develop within only a few millions of years of the crust forming, possibly from rock slopes destabilized by frequent earthquakes. Following this initial stage, scarps apparently continue degrading over 10s of millions of years, forming trellis channel networks (Tucholke et al. 1997).

Scientific drilling on the flanks of the Mid-Atlantic Ridge has recovered inter-layered serpentinite rock breccias and lavas. Lithological sequences of rocks observed in submersible dives and in rock dredges also do not obviously follow the simple “classical” oceanic crustal structure of lavas successively overlying diabase (dykes), gabbro and mantle rocks (Mitchell 2001). Landslides may explain some heterogeneity, by mixing up the lithologies locally; high-resolution maps of serpentinite-rich basement (Cannat et al. 2013) have revealed landslides in serpentinite, such as those shown in Fig. 4. The breccias created by landsliding can then be over-erupted by lavas, leading to the lithological sequences found in the drill cores.

Fig. 4
figure 4

Perspective views of bathymetry collected with a high-resolution multibeam sonar installed on an ROV revealing slumps and other landslide features (Cannat et al. 2013) (Reprinted from Cannat et al. 2013, with permission from Wiley.)

Full size image

5 Volcanic Geomorphologic Features

Early detailed investigation of volcanic landforms was carried out with shipboard multibeam sonars and manned submersibles during a joint French-American study on the slow-spreading Mid-Atlantic Ridge (Ballard and Van Andel 1977), and during a similar study on the medium-spreading Galapagos Spreading Centre (Ballard et al. 1979). Since then, there have many volcanic investigations involving submersibles, ROVs, AUVs and deeply towed devices, centred mainly on the Mid-Atlantic Ridge and fast-spreading East Pacific Rise.

On fast-spreading ridges, erupted lavas tend to have low viscosities and therefore spread out easily, leaving flat-lying (“sheet” and “lobate”) flows, as well as steeper slopes (Soule et al. 2007). Lava lakes containing pillars where the lava became chilled by trapped water venting through them are common, with pillar “bathtub rings” caused by subsequent lava drain back. The more recent high-resolution bathymetry from AUVs has revealed channels, pathways travelled by the distributing lava (Soule et al. 2005).

In contrast, eruptions on slow-spreading ridges produce a more rugged geomorphology. The terrain is commonly described as ‘hummocky’ and in photographs typically contains stacks of pillow lavas, although flat-lying lava flows also occur (Perfit and Chadwick 1998). Hummocky flows can be arranged in ridges and tumuli (Bryan et al. 1994), likely originally erupted over dykes or dyke swarms. Some superimposed cones may represent late-stage focused effusion from the dykes, as occurs in subaerial eruptions (Head et al. 1996). Some cones may be rootless (Bryan et al. 1994) and fed by lava tubes (Smith and Cann 1999). Photographs of both fast and slow ridges typically show abundant fissures from the on-going extension.

Illustrating how the new high-resolution topography available from AUVs and ROVs is helping to study volcanic landforms, Fig. 5 shows data collected from the East Pacific Rise (Deschamps et al. 2014). Despite faster cooling and high ambient pressure, many features such as those shown are comparable with those found in lava flow fields on land and can be interpreted in terms of lava flowage, inflation, cooling and drainback.

Fig. 5
figure 5

Detail of volcanic lava flow structures in high-resolution bathymetry from the northern East Pacific Rise (Deschamps et al. 2014). (Reprinted from Deschamps et al. 2014, with permission from Wiley.)

Full size image

6 Hydrothermal Springs

Emissions of hydrothermal fluids can produce hot ‘black smokers’ (jets of fluids containing precipitating sulphide particles) or clearer diffuse venting. The particles in black smokers can settle to form smooth mounds of hydrothermal sediment. Precipitating silicate and sulphide minerals from the fluids can also produce chimneys many metres to >10 m tall, such as illustrated in Fig. 6. Chimneys are unstable and can topple during earthquakes or otherwise fracture to contribute talus to the sedimentary mounds the chimney fields stand on. Petrological and geochemical analyses of samples obtained from these chimneys have revealed that they were formed of layers of different minerals, suggesting cycles in the hydrothermal fluid composition and temperature, as well as mineral reactions during seawater ingress. While many hydrothermal springs are associated with fluid circulation above the hot magma, some are associated with circulation through ultramafic rocks brought towards the seabed by detachment faults.

Fig. 6
figure 6

Pinnacles in a hydrothermal vent field on the Juan de Fuca Ridge (Kelley et al. 2001) (Reprinted from Kelley et al. 2001, with permission from Geological Society of America.). a Perspective view derived from elevation data collected with a 675 kHz scanning sonar mounted on an ROV. b Mosaic of 73 still images

Full size image

7 Sediment Transport and Deposition

Where ridges lie far from the continents, only pelagic particles accumulate on them, generally at slower rates than other forms of sediment near continents, so that hints of the oceanic basement relief commonly remain visible in seabed morphology even on crust of many millions of years in age. Those pelagic sediments can accumulate like snow with little redistribution so that they drape the underlying basement topography. Commonly, though, some lateral (usually downslope) movement occurs to leave concave up-wards deposits (Mitchell 1995) or flat lying turbidite ‘ponds’ (van Andel and Komar 1969). Where ridges lie near to continental sources of sediment, more frequent turbidity currents can flood the axial valley with sediment, such as the northern Juan de Fuca Ridge (Davis and Lister 1977). Even more exotically, the invading sediments can include flows of evaporites, such as in the Red Sea (Augustin et al. 2014).

8 Remaining Issues and Developments

Our understanding of mid-ocean ridge geomorphology should continue to improve as new technology is developed and deployed. Modern multibeam sonars are capable of imaging acoustic backscattering from the water column as well as the seabed and could be used to record emissions and biota over vent sites along with more optimized acoustic imaging of them (Bemis et al. 2015). Synthetic aperture sonar (adapted from synthetic aperture radar) may further improve seabed imaging. Acoustic backscattering strength in sidescan sonar images depends on seabed micro-roughness at the scale of the sonar acoustic frequency, any sediment cover and incidence angle of the signal with the bed (which varies with bed topography). For example, the low backscattering of smooth sheet flows (Soule et al. 2005) illustrates the effect of roughness. As attenuation in any sediment cover also varies with acoustic frequency (Mitchell 1993), we might expect data collected with multi-frequency sonars to reveal useful information. More generally, increasing use of deeply towed systems, ROVs and AUVs with seabed imaging systems (cameras, sonars and LiDAR) should progressively enlarge the database of observations we have of high-resolution geomorphology.

Knowledge of the subsurface lithologic structure would be helpful, as on land, for working out how the geomorphology develops, in particular, for volcanic landforms. However, structure is difficult to image with geophysical methods and sampling by scientific drilling has been limited because of cost and technical issues (holes in friable volcanic rock are unstable). Illustrating the problem with geophysical datasets, seismic refraction methods can reveal boundaries in the upper crust that are not lithological boundaries exactly, because seismic velocities vary with the porosity of the rock, affecting its bulk rigidity (Carlson 2010), and porosity may vary within units. Alteration also varies velocities, illustrated by reflections in seismic data collected over the detachment fault in Fig. 3b (Collins et al. 2009). Similar issues apply to electro-magnetic measurements, where electrical resistivity varies with connected porosity and pore fluid resistivity (sensitive to conducting brines and magmatic fluids). Nevertheless, further progress will be made by carefully designed experiments combining geophysical experiments with information from scientific drilling and observing rock sequences exposed at fault escarpments (Karson 1998).

Much of the effort on ridges has involved ‘forensic science’, i.e., we have imaged seabed features and attempted to reconstruct the processes that formed them, aided by analyses of samples and geophysical data. There has been less progress in observing geological events or processes as they occur. Although seismicity can forewarn a volcanic eruption in some instances, the times of occurrence of volcanic eruptions, fault movements and landslides are largely unpredictable, making planning of any in situ observation problematic.

This is important for physical volcanology. On land, infrared measurements are routinely made on erupting lavas to monitor temperature; lava viscosity is strongly non-linear with temperature, so the final geometries of flows strongly depend on lava temperature and eruption rate. Temperature measurements in submarine mid-ocean ridges have not been possible. Individual lava flows are the basic units of the uppermost crust and thus important for determining how the crust develops, but mapping their geometries at the seabed is difficult because lava types can vary with effusion rate and gradient so their surface morphology is insufficiently characteristic for distinguishing them (on land a single lava flow can change between a’a’ and pahoehoe types during the same eruption). To overcome this, boundaries to some historical lava flows have been mapped out based on the extent of un-sedimented lava (e.g., Soule et al. 2007). Caress et al. (2012) characterized the shapes and thickness distributions of ~10 m thick lava flows erupted in 2011 on the Juan de Fuca Ridge using repeated high-resolution bathymetry from an AUV. As these studies illustrate, information should accumulate over time with further surveying, encompassing lavas newly erupted since previous surveys were carried out.

Further information on eruptions and crustal deformations will arise from bottom observatories (cabled or autonomous packages of instruments). For example, Chadwick et al. (2012) documented inflation-deflation cycles from pressure sensors on the Juan de Fuca Ridge, including a sudden uplift immediately prior to the 2011 eruption associated with intrusion of a dyke. The same event was recorded with ocean bottom seismometers, which revealed changes in seismicity leading up to the eruption and constrained the eruption duration (Dziak et al. 2012). Cabled observatories off western North America will likely yield more information on eruptions over time.

More generally recording how the seabed changes its topography during earthquakes (Stein et al. 1991), will be challenging as associated depth changes of ~1 m are typically smaller than the depth precision of hull-mounted multibeam sonars. Nevertheless, recordings of earthquake T-waves using hydrophones moored in the oceans can be used to detect and locate small magnitude earthquakes, allowing volcano-tectonic earthquakes to be studied (Smith et al. 2003). Such information provided in real-time may allow opportunistic inspection of earthquake ruptures and deformation in the future. Observatory pressure sensors and acoustic ranging to bottom-moored transponders may also ultimately reveal tectonic movements.

As the above notes illustrate, we have made enormous progress in understanding mid-ocean ridge geomorphology, despite their deep and commonly remote locations, and we can expect further exciting developments to emerge over the decades to come.