Extinct mid-ocean ridges record past plate boundary reorganizations, and identifying their locations is crucial to developing a better understanding of the drivers of plate tectonics and oceanic crustal accretion. Frequently, extinct ridges cannot be easily identified within existing geophysical data sets, and there are many controversial examples that are poorly constrained. We analyze the axial morphology and gravity signal of 29 well-constrained, global, large-scale extinct ridges that are digitized from global data sets, to describe their key characteristics. Additionally, the characteristics of a representative collection of active spreading centers are analyzed to review the present-day variation in the bathymetry and gravity signal of ridges in different tectonic settings such as backarc basin ridges, microplate ridges, and large-scale plate boundaries with varied spreading rates. Uncertain extinct ridge-like structures are evaluated in comparison with the signals of well-defined extinct ridges, and we assess whether their morphology and gravity signals are within the range seen at extinct (or active) ridges. There is significant variability in extinct ridge morphology; yet we find that the majority of well-defined extinct ridges have a trough form and a negative free-air gravity anomaly. We compile available data on the spreading characteristics of extinct ridges prior to cessation, such as their spreading rates and duration of spreading, and find significant differences between ridge subtypes and between oceans. Large-scale extinct mid-ocean ridges persist much longer than extinct microplate spreading ridges and extinct backarc basin spreading ridges before cessation. Extinct fragmented plate and microplate spreading centers have the highest pre-extinction spreading rates, and they have greater median relief at their axial segments, suggesting that different crustal accretion styles could lead to different morphology after spreading cessation. Backarc basin ridges have more pronounced relief when they have been active for longer before cessation, which supports theories of reduced magmatic supply as the basin width increases. Extinct ridges in the Atlantic Ocean have the lowest spreading rates prior to cessation and tend to persist for twice as long as those in the Pacific before extinction. There are a larger number of extinct ridges preserved within marginal basins than expected for their combined area; these ridges may relate to the complexity of the plate boundaries in these regions. Our review of a large number of controversial extinct ridge locations offers some insight into which proposed locations are more likely to have been former spreading centers, and our analysis further leads to the discovery of several previously unidentified structures in the south of the West Philippine Basin that likely represent extinct ridges and a possible extinct ridge in the western South Atlantic. We make available our global compilation of data and analyses of individual ridges in a global extinct ridge data set at the GPlates Portal webpage1.

Extinct spreading centers are tectonic structures in the seafloor that preserve the terminal location of a past divergent plate boundary, after it has ceased both magmatic and tectonic accretion, leading the crust at the boundary to become joined as a single plate. They are valuable sources of information regarding accretionary mechanisms and present a natural laboratory for investigation of spreading ridge characteristics and behavior, without the additional complexity of a thermal anomaly (Jonas et al., 1991; Livermore et al., 2000). Although some extinct spreading centers have a clear signature in bathymetry or gravity maps, there are many proposed extinct ridges that are uncertain in location and/or origin. In some areas, asymmetric ocean floor (Müller et al., 2008), magnetic anomalies, or tectonic reconstructions are used to infer the location of an extinct ridge; yet there may be no obvious structure that can be interpreted as an extinct ridge. This difficulty is compounded in remote regions where marine ship-track data and sampling are incomplete or of poor quality and when ocean crust was formed during magnetic “quiet zones.” Therefore, we seek to better describe the general characteristics of extinct ridges, in order to test whether these can be used to provide an alternative means of assessing ridge-like tectonic structures within ocean basins that have ambiguous or uncertain modes of formation. Locating extinct ridges is essential to the development of accurate regional and global reconstructions that can assist in understanding the evolution of plate boundaries and the geodynamic processes that control these changes.

To improve identification of these structures we catalogue large-scale tectonic structures that have been proposed as extinct ridges previously (Fig. 1; Tables 1A1C) and rank these with respect to how well studied and constrained these features are. To better understand the variability of extinct ridges, we review the characteristics of a collection of active spreading centers that represent different tectonic settings (Table 1D), to permit appropriate comparison of extinct examples of various subtypes. We analyze the characteristics of the “well-defined” extinct ridges, particularly their axial signature, using two-dimensional profiles from global satellite-derived bathymetry data (Weatherall et al., 2015; GEBCO_2014_1D, version 20141103) and gravity data (Sandwell et al., 2014; V23.1 gridded global gravity data). We combine these data with a compilation of information on potential controlling factors for the evolution of each ridge, such as the spreading rate of the system prior to cessation, the type of spreading center, the duration of activity at the ridge system, and the time since spreading cessation (i.e., age of crust formed at the extinct axis). From the characteristics of well-resolved extinct ridges, we revisit ridge-like features with more uncertain origin and further evaluate several oceanic features that may represent previously unreported extinct ridges.

Physical Characteristics of Extinct Ridges

Extinct ridges have experienced deceleration of their spreading rate to zero (Mammerickx and Sandwell, 1986); so it is expected that their morphology will be typical of that seen at slow- to ultraslow-spreading centers, and these characteristics have indeed been reported at a number of extinct spreading centers (Mammerickx and Sandwell, 1986; Jung and Vogt, 1997; Livermore et al., 2000). Yet some variability is also noted, with Okino and Fujioka (2003) reporting that several segments of the Central Basin extinct spreading center in the West Philippine Basin were unusual in that they have fast spreading characteristics such as smooth topography, compared with other segments of the ridge. They attribute this variation to the influence of a proximal hotspot in the west of the basin that was limited in its regional influence (Okino and Fujioka, 2003).

Previous investigations have reported a characteristic negative free-air gravity anomaly at extinct ridges (Jonas et al., 1991; Louden et al., 1996; Livermore et al., 2000; Greenhalgh and Kusznir, 2007). Observation of the anomalous gravity signal permitted Matthews et al. (2011) to refine the locations of 26 extinct ridges using the minima of their gravity signal using global vertical gravity gradient (VGG) data. Jonas et al. (1991) compared the gravity signature of the Labrador Sea, Coral Sea, Magellan Trough, and West Philippine Basin extinct spreading centers and tested various potential explanations for the gravity anomalies observed in these locations. They found that the gravity anomaly across the extinct ridge axis varied according to the paleospreading rate. Specifically, they reported free air gravity lows of 25–45 mGal, with wavelengths 20–45 km, at extinct spreading centers with low paleospreading rates and smaller magnitude gravity lows of 5 mGal, with wavelengths of 10–15 km, for centers with faster spreading before cessation. We seek to increase our understanding of extinct ridges by investigating a global, more extensive data set using high-resolution satellite gravity (Sandwell et al., 2014) and global bathymetry (Weatherall et al., 2015) data. We consider whether there is a characteristic morphology and gravity signal at extinct ridge axes by quantifying the across-axis relief and gravity anomaly at individual segments of extinct ridges.

Classification of Spreading Ridge Subtypes

Global studies of active spreading centers have investigated specific features of the global mid-ocean ridge (MOR) system, such as the variability in morphology related to spreading rates (Small, 1998), abyssal hill form (Goff, 1991) and segmentation (Schouten et al., 1985; Batiza, 1996; Carbotte et al., 2015). As yet, however, there has been no synthesis of the variability of spreading ridges relative to different ridge subtypes. Therefore, we first review the different tectonic settings in which spreading centers occur. We consider microplate spreading ridges, backarc basin spreading ridges, and large-scale mid-ocean ridges as distinct subtypes of spreading centers and a small class of transient spreading ridges that evolve in the context of plate fragmentation. This evaluation provides useful information regarding active spreading ridge variability that permits a more meaningful comparison for extinct ridges of different subtypes.

In a previous review of “failed rifts,” Batiza (1989) identified mid-ocean ridges that had jumped to a new location over 400 km away after cessation as examples of the largest scale of extinct ridge. Batiza (1989) included within this group ridges that did not have a subsequent spreading center within the region. Within the large-scale structures, we define extinct mid-ocean ridges (XMORs) as a subgroup of extinct ridges that at the time of spreading were situated between two major plates (Fig. 2A). Active examples of large-scale mid-ocean ridges include the Mid-Atlantic Ridge, the Southwest and Southeast Indian ridges, the Pacific-Antarctic ridge, and the East Pacific Rise.

The spreading centers within active microplates, such as the Easter and Juan Fernandez microplates (Searle et al., 1993), have significantly different morphology, structural characteristics, and crustal accretion style than large-scale mid-ocean ridges; therefore, we separate extinct microplate spreading centers (Fig. 2B) into a second subgroup (extinct microplate spreading ridges [XMPRs]). Microplates have been defined as oceanic plates that are typically less than 500 km in diameter (Bird, 2003) and that move relative to a local pole of rotation (Naar and Hey, 1991; Schouten et al., 1993), such as the well-studied, presently active Easter microplate (Naar and Hey, 1991). They form by ridge propagation into existing ocean floor, a process that detaches a piece of the larger oceanic plate. Dual spreading at both the preexisting spreading center and the new spreading center results in the growth of the microplate by accretion of new crust to the core of the microplate—the captured crust (Engeln et al., 1988; Naar and Hey, 1991; Hey, 2004; Matthews et al., 2016). Dual spreading also results in independent rotation of the microplate, with a number of studies relating this motion to drag that develops between the major plate and the microplate (Engeln et al., 1988; Naar and Hey, 1991; Schouten et al., 1993). If spreading ceases at the preexisting ridge, the microplate, along with its extinct ridge, will be captured by the neighboring plate, and spreading continues at the alternative bounding ridge, which now forms a continuous, large-scale plate boundary for the major plate (Tebbens and Cande, 1997; Hey, 2004). For this reason, extinct microplates typically preserve only one spreading center.

Investigations of present-day active microplates suggest that they are likely to be short-lived tectonic features that may play a role in accommodating changes in relative plate motions between major plates (Engeln et al., 1988; Schouten et al., 1993; Cuffaro and Jurdy, 2006) or to facilitate major ridge-jumps (Naar and Hey, 1991). Well-developed microplate spreading centers frequently demonstrate fanning, mostly symmetric, spreading about a central spreading system and a clear spreading axis may be evident (e.g., the Magellan Trough, Tamaki and Larson 1988; the Galapagos Rise/Bauer microplate, Eakins and Lonsdale, 2003; and the Mathematician ridge, Batiza and Vanko, 1985). Microplate spreading centers can be extremely complex and evolve rapidly (Naar and Hey, 1991); therefore, identification of their extinct spreading axes can be more complex than for large-scale spreading systems.

We also identify a class of short-lived spreading centers that are associated with smaller plates (less than 1000 km in width) that are proposed to have formed by the mechanism of plate fragmentation as a downgoing oceanic plate enters a subduction zone at an oblique angle (Lonsdale, 1991, 2005). We refer to these ridges as extinct fragmented plate spreading ridges (XFPR). Present-day examples of fragmented plate ridges (FPR) include the Rivera plate spreading ridge and the Juan de Fuca spreading ridge, while extinct examples include the Magdalena (Michaud et al., 2006), Guadalupe (Mammerickx and Klitgord, 1982; Batiza and Vanko, 1985; Michaud et al., 2006), and Monterey (Lonsdale, 1991) ridges, all situated offshore the western margin of the North America plate.

Although several published regional reviews of extinct spreading centers (Mammerickx et al., 1980; Mammerickx, 1981; Batiza, 1989; Lonsdale, 2005) have not included extinct backarc basin spreading ridges, there are several well-defined preserved examples. Stern and Dickinson (2010) summarized active and extinct backarc basins suggesting there are 18 extinct examples. However, a number of these extinct backarc basins have little data available (Stern and Dickinson, 2010), and in some cases, they do not have a preserved spreading center, particularly where much of the crust has been subsequently destroyed at a proximal subduction zone (Schellart et al., 2006). The most closely studied extinct backarc basin spreading center (XBABR) is the Parece Vela–Shikoku Ridge (Fig. 2C-i), which extends the length of the Shikoku and Parece Vela basins for ∼2500 km and is thought to have ceased spreading at ca. 15 Ma, after chron C5En (Mrozowski and Hayes, 1979; Chamot-Rooke et al., 1987; Sdrolias et al., 2004). Other examples are noted close to the Antarctic Peninsula (Jane Basin, Powell, Dove, and Protector Basins), in the South Fiji Basin (Fig. 2C-ii), and in the West Philippine Basin (Fig. 2). We review the characteristics of preserved backarc basin spreading centers and consider how these may differ from other ridge subtypes.

Outstanding Problems and Questions

In addition to considering if a characteristic extinct spreading center morphology and gravity signal can be described, our classification of different spreading center subtypes allows us to evaluate how the tectonic setting of a spreading center influences its physical characteristics, for both active and extinct examples. This permits a comparison between these two groups and can offer an insight as to whether the final stages of spreading are likely to be primarily magmatic or tectonically driven. Further, in cataloguing all known examples of large-scale extinct spreading centers, we evaluate the spatial and temporal distribution of ridge-jumps to consider why some regions have a greater or lesser number of ridge reorganizations. For example, why are there so many extinct ridges in the Pacific Ocean but few reported in the Atlantic? What factors determine the frequency of ridge reorganizations? Is the number of extinct ridges proportional to ocean floor area? By compiling the spreading rate at extinct ridges prior to cessation, there is also potential to investigate the relationship between regional spreading rates and the likelihood of ridge-jumps.

We consider what factors drive the variability in behavior and life span of spreading-ridge subtypes. Stern and Dickinson (2010) observed that generally backarc basin ridges are short-lived relative to large-scale MORs and proposed that as the backarc basin widens, the availability of melt sourced from the hydrated mantle wedge to the spreading axis is reduced and that this leads to eventual death of the ridge. By reviewing all reported examples of both types of extinct spreading centers and active examples, we evaluate if there is any morphological or geological evidence to support this proposal.

Finally, we investigate the occurrence of late-stage and postcessation volcanism that has been observed at a number of extinct ridge segments. The best studied examples of volcanic ridges or seamounts at extinct ridge axes are Guadalupe Island (Batiza, 1977; Batiza and Vanko, 1985), the eastern South China Sea ridge (Pautot et al., 1990), the Styx volcano on the Wharton Ridge (Hébert et al., 1999), the Phoenix Ridge (Haase et al., 2011a), the Socorro Islands on the Mathematician Ridge (Favela and Anderson, 2000), and the Central Ridge of the Galapagos Rise (Haase et al., 2011b). At many of these ridges, the volcanic edifice was formed after cessation of spreading. Where postcessation volcanic rocks from seamounts seated on extinct ridge segments have been analyzed, they have been found to be alkali basalts or alkali-olivine basalts (Batiza, 1977; Batiza and Vanko, 1985; Pautot et al., 1990; Haase et al., 2011a) that are suggestive of fertile mantle parental magmas produced with a low degree of melting. Thinned crust in the region of an extinct ridge has been proposed to present a conduit for volcanism (Favela and Anderson, 2000) due to locally elevated mantle temperatures and potentially hydrated mantle conditions where hydrothermal activity has penetrated the upper mantle. This could effectively drive postcessation volcanism but is expected to be limited to a short time frame after cessation and would likely require a tensional environment (Favela and Anderson, 2000). We evaluate whether there are any factors in common between the examples of late-stage and postcessation volcanism and what insights can be gained by a wider comparison with active spreading ridges.

Assessment of Axial Physical Characteristics

We compile data from extinct ridges that have been proposed previously, including moderate to large-scale tectonic features but excluding ridge-jumps of less than 150 km in distance. This cut-off distance is employed to focus on the larger-scale migration events that are most important for incorporation in regional and global reconstructions, within the bounds of the resolution of the gridded data sets that we use and as a practical limit for the study, given the large number of proposed extinct ridges (∼100 examples globally). We group the ridges into three tiers based on how well defined they are, as outlined in detail in Table 2A. The well-defined extinct ridges in the primary group (Table 1A) are used to define the characteristic signal of the extinct ridges. Ages are given based on the timescale by Gee and Kent (2007).

Extinct ridges were digitized from published studies, and locations were refined using updated bathymetry data (Weatherall et al., 2015, GEBCO_2014_1D, version 20141103) and global gravity and vertical gravity gradient (VGG) data sets (Sandwell et al., 2014). These data sets have been used in numerous previous investigations of seafloor tectonic structures (e.g., Matthews et al., 2011; Sandwell et al., 2014), and global gravity grids can resolve features as small as 6 km in size (Sandwell et al., 2014). The GEBCO 2014 bathymetry data set is compiled from a combination of available depth soundings, ship-track data, and satellite-altimetry–derived data and is estimated to be able to resolve features of 12 km width at a depth of 4000 m (Weatherall et al., 2015), with resolution decreasing with increased depth. Global data sets were used in preference to ship-track data because it is possible to extract a greater number of profiles for each ridge segment from the gridded data and to have a more consistent approach between areas with highly variable coverage by marine surveys. A compilation of magnetic anomaly identifications (Seton et al., 2014) was used to confirm that symmetric magnetic anomalies have been identified about extinct ridge axes. To better understand the characteristics of the axial structure of spreading centers, spreading ridge axes were digitized in continuous segments, bounded by offsets of more than 10 km length. The morphology (bathymetric relief) and gravity signal of extinct ridges are evaluated using two-dimensional profiles across ridge axes (Fig. 3). We provide an online database of the digitized extinct ridge locations and profiles in the GPlates Portal (see footnote 1), along with summaries of our compiled data and key studies in these areas.

Following the same methodology, the axial characteristics of 14 active spreading ridges are evaluated, to provide a representative sample of ridge subtypes and spreading rates for comparison with extinct spreading ridges (Table 1D). Active spreading-ridge locations are refined from Bird’s (2003) digitized plate boundaries, using the VGG data set to better locate axial segments. The active spreading ridges of the Easter microplate are based on the spreading segments identified by Bird and Naar (1994), and the spreading ridges of the Juan Fernandez microplate follow the segments identified by Searle et al. (1993).

The morphology and gravity signal of extinct and active ridges are evaluated using two-dimensional profiles across their axes (Fig. 3), sampling the GEBCO bathymetry grid with 30 arc-second resolution (Weatherall et al., 2015, GEBCO_2014_1D, version 20141103), and the V23.1 free-air gravity grid with a 1 min resolution (Sandwell et al., 2014). We stack profiles across the axes and plot the median profile and range of values between the upper and lower confidence interval (Fig. 3), with these plots provided in our Supplemental Materials2. However, for statistical analyses, we compared a representative cross profile from each ridge segment, for 30 km each side of the inferred location of the ridge axis (Fig. 3). A representative profile was chosen that was within the envelope of the median absolute deviation for the stacked profiles for that ridge segment, was similar to the median stacked profile, and did not include anomalous features, such as ridge flank seamounts or intersecting fracture zones. The location of the representative profile is shown on the maps provided in the Supplemental Materials (see footnote 2) and online database (footnote 1). Bathymetric relief in the region of the ridge axis was determined from the difference between the minimum and maximum depth within 30 km of the ridge axis (Fig. 3C). The peak-to-trough gravity signal was measured from the difference between minimum and maximum gravity within 30 km of the ridge axis (Fig. 3D). The horizontal distance between the minimum and maximum depth or gravity anomaly values, reflects the half-width of the axial valley (or ridge) and the half-wavelength of the gravity signal, respectively. In the case of ridge morphology at the extinct ridge axis, the bathymetric relief is positive (for example, Fig. S4-1-1 in .ZIP File S4 of Supplemental Materials [footnote 2], South China Sea segment 3), and in the case of trough morphology, the relief is negative (Fig. 3).

Review of Spreading Characteristics

The time of commencement and cessation of extinct ridges and their pre-extinction spreading rates are extracted from published studies (Table 1). The precision with which pre-extinction spreading rates have been reported varies significantly, according to the amount of data available regionally, whether marine data was acquired at an optimal orientation to the spreading direction and the timing of cessation. The timespans of magnetic reversals are highly variable and can differ by two orders of magnitude (Gee and Kent, 2007), meaning that the longer the timespan over which the pre-extinction spreading rate is calculated, the lower the resolution of the estimate. At lower resolution, the calculated spreading rate is less likely to indicate the latest, pre-extinction spreading rate, and the time of cessation may be less well constrained. Ideally, a consistent measure would be chosen to assess the pre-extinction spreading rate for all extinct ridges, such as the spreading rate between the last recorded reversal and the time of cessation, or a specific time range such as the past 2 or 5 m.y. of seafloor spreading. Unfortunately, due to the variation in timespans of isochrons, this is not possible, and we therefore provide details of the magnetic isochrons used to infer the pre-extinction spreading rate and the time span over which the rate was calculated (Table S1A in Supplemental Materials [footnote 2]). For active ridges, most spreading rates are taken from DeMets et al. (2010) and are calculated from anomaly A1n for fast spreading ridges and from A2A for slow spreading ridges, with exceptions noted in Table S1B in Supplemental Materials (footnote 2).

Descriptive and Exploratory Statistics

Primary-tier ridge characteristics are described using binned histograms and box plots. The correlation between ridge physical characteristics includes bathymetric relief, the maximum depth in the axial region, the width of the axial structure, the peak-to-trough gravity signal, the half-wavelength of the gravity signal, and spreading characteristics, including the time of cessation, the spreading rate prior to cessation, and duration of spreading. Correlations are calculated in R using the Spearman’s rank correlation method (Upton and Cook, 2014), unless otherwise stated, and the rank correlation coefficient “rho” is given for these calculations. The coefficient of determination (r2) represents the shared variance of the ranked variables, and along with the p-value, indicates the probability that the correlation is meaningful and is reported when rho > 0.5. The p-value gives the probability of obtaining a result at least as extreme as the one that was actually observed, assuming that the null hypothesis is true. Traditionally, a p-value of 0.05 or lower is required to reject the null hypothesis (Upton and Cook, 2014).

Evaluation of Secondary-Tier and Controversial Extinct Ridges

We review several previously unreported or poorly known extinct ridges that were identified from the global bathymetry and gravity data sets (Fig. 4). We also describe the physical characteristics of all secondary-tier extinct spreading ridges, including various alternative placements in controversial areas (noted in Table 1C) and evaluate how they compare with well-defined extinct ridges. The criteria applied for assessment of secondary-tier ridges is outlined in Table 2B. A maximum score of 6 is given if the proposed extinct ridge meets all of the criteria. Proposed locations with a score >4 are considered likely to have been a former spreading center. Centers with a score between 3 and 4 are considered “possible” extinct ridges, and scores <3 indicate that the feature is unlikely to be an extinct ridge. Supporting evidence from additional literature review is given a score between 0 and 2 (Table 2B).

A total of 102 extinct ridge locations are assessed and digitized in 317 individual segments. The Pacific Ocean has the greatest number of proposed extinct spreading center locations (38 ridges, with two alternative placements considered), followed by the Indian Ocean (16 ridges, with nine alternative placements suggested), and then the Atlantic Ocean (ten ridges, with one alternative placement suggested). Marginal basins also account for a significant number (27 ridges), with many of these situated in Southeast Asia and in the northwest Pacific Ocean.

We rank 29 ridges in the primary tier (Table 1A, sections 4.2 and 4.3) for use in describing the characteristics of well-defined ridges, with a total of 129 primary-tier segments. The most commonly occurring extinct ridge subtype is XMOR (17 ridges; 71 segments). This is followed by XBABR (six ridges; 32 segments), XMPR (five ridges; 24 segments), and finally, the Guadalupe Trough is the only primary-tier example of a XFPR (two segments). Although the northern segment of the Guadalupe extinct ridge features a large seamount, because this is believed to have formed postspreading cessation, a profile was chosen that sampled the axial trough at the south of the segment.

We define 48 ridges in the secondary tier that have some uncertainty as to their location or their mode of formation (Table 1B, section 4.4). Twenty-six suggested ridge locations (Table 1C) had insufficient constraints on their location, a very weak signal, or were unlikely to represent former spreading centers; thus, they were excluded from statistical analyses. However, our Supplemental Materials (see footnote 2) and online database (footnote 1) provide details of key studies in these areas and profile observations. Physical characteristics of extinct and active ridges are displayed in Figures 5A and 5B, respectively; their spreading characteristics are shown in Figures 6A and 6B; and correlations between physical and spreading characteristics are presented in Figures 7A and 7B, respectively.

Morphology of Well-Defined Extinct Ridge and Active Ridge Axial Segments

Large variations in relief (Fig. 5A-I) reflect extremely variable along-axis morphology for many of the well-defined extinct ridges. There are also significant morphological differences between segments within individual ridge systems, such as an elevated ridge where the other segments may be deep troughs, with eight ridge segments in the primary tier. The clearest examples of these axial ridges are those previously discussed, such as the Socorro Island on the Mathematician ridge, the central ridge of the Galapagos Rise, the eastern segment of the South China Sea ridge, and Guadalupe Island. The majority of well-constrained extinct ridges (67 segments) have trough morphology in bathymetric profile, although these are highly variable in width and depth (Table 3A), and several are juxtaposed on larger-scale ridge structures. Of the primary-tier ridges, 26 axial segments have no expression in bathymetric profile, and a further 28 segments have irregular profiles that cannot be categorized as having either a trough or ridge-like morphology.

The mean bathymetric relief at primary-tier extinct ridge axes was found to be –772 m (standard deviation [SD] = 816 m). Extinct FPR and XMPR exhibit greater negative bathymetric relief at their axes than large-scale XMORs (Fig. 5A-I; Table 3A-I), with median relief of –1075 at XMPR and –1473 at XFPR, which is considerably more than at XMOR (median –641) and XBABR (median –774). However, XMPRs have the widest range, with a number of segments that have positive relief, and this leads to the mean relief that is closer to the XMOR mean relief (Fig. 5A-I). The maximum depth in the region of the ridge axis increases with the time since cessation (rho –0.65; r2 = 0.42; p-value < 0.001), as expected due to the subsidence of oceanic crust with increasing age. There is a weak correlation between bathymetric relief and the time of spreading cessation (rho 0.22; p-value < 0.05); however, this reflects a greater diversity in relief for segments that have ceased spreading more recently, and relief is relatively stable for ridges that ceased before 40 Ma (Fig. 7A-V). For example, a small number of ridges that became extinct more than 100 m.y. ago retain over 1000 m relief between valley floor and flanks (e.g., segments of the Wallaby Ridge, Magellan Trough, and West Magellan extinct spreading centers). Although for all primary-tier extinct segments there is a weak correlation between the duration that the ridge was active and relief at the ridge axis (rho 0.19, p < 0.05), for XMORs, these variables are moderately correlated (rho = 0.50, r2 = 0.27, p-value < 0.001, Fig. 7A-VI). No meaningful correlation is found between bathymetric relief and the published pre-extinction spreading rates (rho < 0.01, p-value = 0.99).

Active MOR and BABR have similar mean relief (Fig. 5B-I) to their corresponding extinct subtypes (Fig. 5A-I), although median values are weaker for extinct ridges (Tables 3A-I and 3A-II). In contrast, active MPR and FPR are less likely to have the prominent negative relief that is seen at extinct examples. Amongst active spreading ridges, BABRs have greater negative mean relief (mean –972 m, SD = 878) than other spreading types; although when median values are considered active, MORs (median –1198 m) are slightly greater than BABR (–1101 m) (Table 3A-III). As expected, extinct ridges are situated at greater depth (Figs. 5A-II and 5A-III) than active spreading ridges (Figs. 5B-II and 5B-III). Extinct and active subtypes have similar relative maximum depth (Figs. 5A-II and 5B-II); for example, XBABR and BABR tend to have greater maximum depth; yet the minimum depth of active ridges is very similar between all subtypes (Fig. 5B-III), while extinct subtypes vary by up to 1400 m (Fig. 5A-III). Overall, the median width of the axial structure of extinct ridges is similar (mean 25 km, SD = 14 km) to the width of active examples (mean 22 km, SD = 10 km), although there are some variations between subtypes (Figs. 5A-IV and 5B-IV).

Peak-to-Trough Gravity Signal at Well-Defined Extinct Ridge and Active Ridge Axes

The peak-to-trough gravity signals for primary-tier extinct ridge segments are strongly correlated with segment axial relief morphology (rho 0.64, r2 = 0.40, p-value < 0.001, Fig. 7A-IV). A characteristic negative gravity signal at extinct ridge axes is evident on many profiles, as seen in Figures 3 and 5A-V. The gravity signal ranges from –146 mGal to 76 mGal, yet the mean signal is moderately negative, with a value of –39 mGal (SD = 30 mGal) (Fig. 5A-V). The mean and standard deviation of the gravity signal for each ridge subtype are summarized in Table 3A and Figure 5A-V. The half-wavelength of the anomalies (estimated from the lateral distance between the minimum and maximum gravity value within 30 km of the ridge axis) ranges from 12 km to 52 km, with a mean of 24 km (SD = 7 km) (Fig 5A-VIII), which is the same value as the active spreading ridges (mean 24 km, SD = 8 km; Fig. 5B-VIII; Table 3A-II). The wavelength of the gravity anomaly is moderately negatively correlated with the pre-extinction spreading rate (rho –0.42, r2 = 0.17, p-value < 0.001). For all primary-tier segments, there is a very weak correlation between the peak-to-trough gravity signal and pre-extinction spreading rate (rho = 0.15, p-value < 0.1) (Fig. 7A-I), with a weak to moderate correlation found for XMOR (rho = 0.30, p < 0.05).

The correlation of bathymetric relief and gravity signal is even stronger from active ridges (rho 0.92, r2 = 0.85, p-value < 0.001) than extinct examples (Figs. 7A-IV and 7B-III). The peak-to-trough gravity signal at active MOR (mean = –45 mGal, SD = 53 mGal) and BABR (mean = –33, SD = 34; Fig. 5B-V) subtypes are quite similar to the corresponding extinct subtypes, XMOR (mean = 42 mGal, SD = 26 mGal) and XBABR (mean = –35 mGal, SD = 28 mGal) (Fig. 5A-V). Although mean values for the minimum (Figs. 5A-VI and 5B-VI) and maximum (Figs. 5A-VII and 5B-VII) gravity anomalies near the ridge axis are very different when active and extinct ridges are compared, the subtypes tend to plot in similar fields relative to each other. For example, FPRs tend to have more negative mean minimum and maximum gravity anomalies in the axial region, and anomalies at BABR tend to be less negative, and, in many cases, they have positive gravity anomalies near the ridge axis. Large-scale MORs have the broadest range of values for active and extinct examples, but mean values for the minimum and maximum gravity anomalies near the ridge axis tend to be higher than microplate and fragmented plate ridges and lower than backarc basin ridges.

Spreading Rate, Time of Cessation, and Duration of Spreading

We summarize the collated spreading characteristics of the primary-tier extinct ridges in Table 3B-I and Figure 6A and active ridges in Table 3B-II and Figure 6B. Pre-extinction half-spreading rates range between 2 and 80 mm yr–1, with a mean of 25 mm yr–1 (SD = 20 mm yr–1) (Table 3B-I). The variation in pre-extinction spreading rates for extinct ridge subtypes (Fig. 6A-I) is very similar to that of active ridge subtypes (Fig. 6B-I), with the exception of the XFPR, but there is not a representative sample of this group with only one ridge in the primary tier. Active MPR and FPR have higher mean spreading rates, although fragmented plates have a larger standard deviation than all other groups.

Extinct ridges have strong variations in pre-extinction spreading rates within different ocean basins (Fig. 6A-II). Pre-extinction half-spreading rates were slowest in the Atlantic Ocean (median 7 mm yr–1, range 2–15 mm yr–1), with marginal basins generally in the slow to intermediate range (median 16 mm yr–1, range 3.5–60 mm yr–1), with faster spreading rates in the Pacific (median 25 mm yr–1, range 6–72 mm yr–1) and Indian Ocean (median 27 mm yr–1, range 20–80 mm yr–1) (Fig. 6A-II). The variability of spreading rates between different oceans is fairly consistent with present-day spreading rates at active ridges (Fig. 6B-II). The median half-spreading rates for active ridges are between ∼8–15 mm yr–1 faster than extinct ridges in the corresponding regions (Figs. 6A-II and 6B-II), with the exception of the Indian Ocean where the median spreading rate is lower than extinct examples (median 21 mm yr–1, range 7–35 mm yr–1). At the active ridges reviewed, the highest mean spreading rates are in the Pacific, the lowest in the Atlantic; and in the Indian Ocean and marginal basins, spreading rates are in the intermediate range.

Extinct ridge duration of spreading ranged from a minimum of 4 m.y. at the Guadalupe microplate to a maximum of 50 m.y. at the Labrador Sea Ridge (Figs. 6A-III and 6A-IV). Ridge subtype strongly influences the duration of activity, with XMPR and XBABR having shorter durations compared with XMOR (Fig. 6A-III). Extinct ridges in the Atlantic and Indian oceans had longer durations before cessation than those in the Pacific Ocean and marginal basins (Fig. 6A-IV). Ridge extinction events in our data set have occurred regularly over the past 150 m.y. with the time of cessation ranging from 3 to 136 Ma (Figs. 6A-V and 6A-VI); although a greater number of recently ceased extinct ridges are preserved, proportionate to the area of younger aged oceanic crust in global oceans. The duration of spreading of active spreading ridges cannot be directly compared with extinct ridges, because the length of time that the active ridge will continue to spread is not constrained. However, the time of commencement gives an approximate measure of the longevity of active spreading ridges of different subtypes (Fig. 6B-III) and for different oceans (6B-IV), and results are broadly similar to those from extinct examples (Figs. 6A-III and 6A-IV). In contrast to the weak positive correlation between duration and axial relief for extinct ridges (rho 0.19, p-value < 0.05; Fig. 7A-VI), there is a weak negative correlation between the time of commencement for active ridges and bathymetric relief at the axis (rho –0.24, p-value < 0.05), which is moderate for MOR (rho –0.44, p-value < 0.01) (Fig. 7B-IV).

Extinct ridge spreading characteristics are compared with the physical characteristics of primary-tier extinct ridges (Fig. 7A). Gravity anomalies are a little more subdued for older oceanic crust (rho 0.24, p-value < 0.01). There is a weak correlation between the pre-extinction spreading rate and the magnitude of the gravity signal at the ridge axis for XMOR (rho 0.30, p-value < 0.05) and no significant correlation found for other subtypes (Fig. 7A-V). There is also no meaningful correlation between bathymetric relief at the axis and the pre-extinction spreading rate for extinct ridges. This contrasts with active spreading centers, at which the spreading rate is strongly correlated with the bathymetric relief (rho = 0.72, r2 = 0.52, p-value < 0.001) and the peak-to-trough gravity signal (rho = 0.76, r2 = 0.58, p-value < 0.001) (Fig. 7B-I), as has been reported elsewhere. At extinct ridges, there is a weak negative correlation between the minimum depth in the axial region and the pre-extinction spreading rate (rho –0.21, p-value < 0.05), although a moderate to strong negative correlation is found for XMOR (rho –0.63, p < 0.001) (Fig. 7A-II). Active ridges have a stronger negative correlation between spreading rate and minimum depth (rho –0.58, p-value < 0.001), and this is more pronounced for MOR and MPR (rho 0.67, both p-values < 0.005) (Fig. 7B-II).

Evaluation of Secondary-Tier Ridges

The locations of secondary-tier ridges are shown in Figure 1, and references for published studies are listed in Table 1B, alongside collected data. The degree to which secondary-tier ridges match the characteristics of well-defined ridges is evaluated and scored in Tables 4A and 4B, following the criteria detailed in Table 2B. Of the uncertain ridges listed in Table 4A, there were 19 of 22 locations that achieved a score greater than 4 for 50% or more of their defined segments and are, therefore, deemed “likely” to represent former spreading centers. Only three ridges scored below 4 for the majority of their segments (Sao Paulo, Conrad Rise South, and the Angolan basin suggested extinct ridges). From the new locations we propose, these ridges score above 3 out of a maximum of 4 points for their physical characteristics, which are within a standard deviation of primary-tier extinct ridges; however, they receive no score for criteria C due to no supporting studies. Therefore, they are categorized as possible extinct ridges. For alternative locations considered in controversial regions (Table 4B), only three of 20 locations achieved a score greater than 4 for 50% or more of their defined segments. In some cases, such as the proposed locations on the Perth Abyssal Plain, two possible placements scored in the range of probable extinct ridge.

Additional discussion of the evidence reviewed in our assessment of controversial and uncertain ridge locations is contained in the summary pages linked to our online database of extinct ridges (Fig. 8; footnote 1). Overall, the secondary tier extinct ridge axes have weaker peak-to-trough gravity signals (mean –22 mGal, SD = 37 mGal) than that of the primary-tier ridges (mean –39 mGal, SD = 30 mGal), which may contribute to the difficulty in identifying these features in global data sets.

Comparison with Previous Studies of Extinct and Active Ridges

The magnitude of gravity anomalies at extinct spreading ridge axes within our global analysis are consistent with trends observed at four extinct spreading centers investigated by Jonas et al. (1991). We see a weak spreading rate dependence of the axial gravity anomaly for all ridge subtypes, with higher spreading rates prior to cessation correlated with a lower amplitude gravity anomaly (Fig. 7F). Yet there are many outliers, and this relationship is considerably weaker than the correlation between the gravity anomaly at active spreading segments and spreading rate, which has coefficient of determination r2 = 0.58. There are several possible explanations for this finding. The first may be that due to the variable precision of calculated spreading rates for the final stages of spreading, spreading rates for extinct examples are not accurate enough to obtain a reliable result. Alternatively, if the final stages of spreading at the dying ridge are primarily magmatic rather than tectonic, the morphology and gravity signal of the ridge axis is less likely to be influenced by the final spreading rate.

Extinct microplate spreading ridges and XFPR segments have higher-amplitude negative anomalies at their axes than their active counterparts, suggesting that an additional component contributes to the gravity signal for these extinct ridge subtypes, such as a low-density body situated within the crust, as has been suggested by Jonas et al. (1991). Significant hydrothermal alteration to depth, caused by deep fracturing, or alternatively late-stage magmatic emplacement, could both generate significant density contrasts that are likely to be maintained with minimal attenuation over time.

Influence of Ridge Subtype on Ridge Characteristics

Our results indicate that spreading centers in different tectonic settings may vary considerably in their morphology and the magnitude of their gravity signal, but that this variation is not consistent between active and extinct spreading ridges. For example, although XMPR and XFPR have more pronounced negative relief than other extinct spreading ridges (their median relief is 300–400 m and 700–800 m greater than XBABR and XMOR, respectively), the median relief at active examples of these ridge types is about 1000 m less than that seen at active BABR and MOR segments (Fig. 5B-I; Table 3A-II). This could suggest that the processes related to spreading cessation at microplate and fragmented plate spreading centers are fundamentally different than in other settings, leading to a significant morphological change in the final stages of spreading.

Active microplate ridges have unusual depth profiles, as observed by Naar and Hey (1991), who found that the deepest point on the East Pacific Rise was situated at the northernmost point of the active spreading eastern rift of the Easter microplate (depth 5890 m), in close proximity to the shallowest point on the East Pacific Rise (depth 2050 m), located in the south of the eastern rift of the Easter microplate (Naar and Hey, 1991). This results in almost 4 km of depth change over a distance of less than 500 km. We propose that the combination of two processes that are intrinsic to microplate spreading systems may be responsible for the formation of greater relief at their axial segments. Firstly, microplate crustal accretion is dominated by the process of ridge propagation into existing ocean floor, resulting in capture of the major plate crust, rather than by magmatic accretion (Naar and Hey, 1991; Schouten et al., 1993; Tebbens and Cande, 1997; Hey, 2004). Propagating ridge segments are observed to develop away from shallow regions (Hey, 2004), forming “flow-induced depressions” at the segment tip (Morgan and Parmentier, 1985, p. 8603). The tips of propagating segments have been reported in some cases to be deep, narrow troughs (Kleinrock and Hey, 1989; Hey, 2004), and lower-crustal and even upper-mantle rocks can be exposed within these structures (such as at the Hess Deep, on the Cocos-Nazca spreading center; Gillis et al., 2014; Hekinian, 2014). Therefore, the crustal accretion style is likely to be a key influence in the development of relief at MPRs. Additionally, the rotation of microplate can generate high-relief transpressional (Rusby and Searle, 1993) and compressional ridges (Searle et al., 1993), where convergence occurs between the microplate crust and the bounding major plate (Rusby and Searle, 1993). We suggest that the combination of these two processes explains the observed high relief on XMPR. It is possible that active microplate ridges will develop greater relief with continued rotation and become more similar to their extinct counterparts.

In other respects, the spreading characteristics of extinct ridge subtypes vary in a similar way to those at active spreading centers. For example, we found that XMPRs have higher spreading rates prior to cessation and predictably shorter durations of spreading relative to large-scale XMORs (Fig. 6A-I). Lifespans of XBABR and XMPR are less than 15 m.y. in general; whereas large XMORs are more likely to persist for ∼25 m.y. before failure (Fig. 6B) but are significantly more variable. The life span of spreading centers created by plate fragmentation appears to be usually limited to less than 10 m.y., although in the case of the Cocos–Nazca Ridge, after several ridge-jumps, the present-day ridge appears to have established a stable configuration. This observation may suggest that crustal production initiated by plate fracturing cannot be maintained unless an additional force (such as slab-pull or anomalous upwelling) is available to the spreading system.

Extinct backarc basin ridges show a markedly different trend to the other ridge subtypes when the length of activity is compared with relief at the ridge axis (Fig. 7-VI). While XMORs have lower relief after longer durations of spreading, the opposite trend is seen for extinct backarc spreading centers (Fig. 8C). Increasing relief in the backarc basin spreading ridges after a longer duration of spreading provides some physical evidence to support the cessation mechanism proposed by Stern and Dickinson (2010). They proposed that melts derived from the hydrated mantle wedge may become inaccessible as the backarc basin width increases over time (Stern and Dickinson, 2010). Reduced magmatic supply is likely to result in a similar axial morphology to a slow-spreading axis.

The lack of any examples of backarc basin ridges greater than Cenozoic age, despite much older preserved Cretaceous microplates and large-scale XMORs (Fig. 6C), illustrates the lower preservation potential for backarc basin oceanic crust. Several extinct backarc basins are reported to have been partially or fully subducted a relatively short time after their formation (for example, the South Loyalty, Solomon, and Santa Cruz Basins; Schellart et al., 2006; Stern and Dickinson, 2010). Previous studies (Karig, 1982; Faccenna et al., 1999) have suggested that subduction may initiate at the interface of oceanic and continental crust, such as on passive margins or at extinct island arcs. Thinner crust and a higher temperature regime within the backarc environment have also been argued to promote subduction polarity reversals, given that the strength of the lithosphere is reduced and is more prone to “rupture” (Stern, 2004, p. 280). There is therefore a geodynamic explanation for the lower rates of preservation of backarc-basin–formed crust and their extinct ridges over geological periods.

Evaluation of Controversial and Previously Unreported Possible Extinct Ridges

By determining the character of well-defined extinct ridges, it is possible to estimate the probability of some enigmatic, poorly mapped seafloor features being former spreading centers. For this purpose, the magnitude of the gravity anomaly at the ridge axis appears to be the most reliable indicator for evaluating suspect features (Tables 4A and 4B). The bathymetric relief is more likely to be modified by later sedimentation; yet the gravity anomalies of some extinct ridge subtypes are in a similar range to that of active subtype examples, particularly XMOR and XBABR.

A proposed ridge between the Conrad Rise and Del Cano Rise (Fig. 4iii; Table 4A; combined materials on GPlates Portal [footnote 1; http://portal.gplates.org/static/html/ExRidges/ExRidges_HTML_pages/2-04_Conrad-DelCano_Ridge.html] and associated map and profile files available in .ZIP Files S3 [file name S2-4a_Maps_Conrad.jpg] and S4 [file names S2-4_Plots_Bathymetry_Conrad.png and S2-4_Plots_Grav_Conrad-DelCano.png] of Supplemental Materials [footnote 2]) in the southern Indian Ocean (Cross et al., 2011; Nogi et al., 2011) has a morphology and gravity anomaly that strongly resembles other extinct spreading centers such as the West Philippine Basin ridge. Recovery of “granitic” rocks (Kobayashi et al., 2013, abstract) and metamorphic rocks from dredges on the Ob Seamount (Nogi et al., 2011) strongly suggests the presence of a continental fragment within the Conrad Rise; this fragment requires a ridge-jump and extinct ridge in this location. The extinct ridge is also supported by the marked asymmetry to the north of the Enderby Basin between Madagascar and Antarctica (Müller et al., 2008). Therefore, this feature is considered a probable extinct spreading center.

We evaluate the origin of several structures that have not been previously classified. Of these, the “Mati” ridge (Fig. 4i), in the western Palau Basin of the southern West Philippine Sea, bears the closest resemblance to an extinct ridge in map view (Fig. 4i), profile (combined materials on GPlates Portal [footnote 1; http://portal.gplates.org/static/html/ExRidges/ExRidges_HTML_pages/4-04_Mati_Palau.html] and associated profile files available in .ZIP File S4 [footnote 2; file names S4-4_Plots_Bathymetry_Mati.png and S4-4_Plots_Grav_Mati.png]), and by the magnitude of the gravity anomalies at axial segments, which are consistent with those of well-defined ridges. However, with limited magnetic anomaly identifications in the western Palau Basin, or other age constraints on the crust in the region of the proposed extinct ridge, additional data are needed to evaluate this structure. The “Palau” feature, to the east of the Palau Basin, is less characteristic of an extinct ridge in map view (Fig. 4-iv; combined materials on GPlates Portal [footnote 1; http://portal.gplates.org/static/html/ExRidges/ExRidges_HTML_pages/4-04_Mati_Palau.html] and associated map file in .ZIP File S3 [footnote 2; file name S4-4-2a_Maps_Palau.jpg]), although it bears some resemblance to the most easterly segment of the West Somali Basin extinct ridge. The inferred axial location is a bathymetric trough; however, the morphology is complex and highly variable, making it difficult to conclude that the segment is an extinct ridge segment. The “Palau” ridge is an intriguing structure for its similarity to the West Somali Basin segment and therefore would be an interesting target for further investigation.

A possible extinct ridge is observed in the eastern Argentine Basin, southwest Atlantic Ocean. It is located north of South Georgia and the Falkland-Agulhas fracture zone and is a subdued feature (Fig. 4-ii) that has minimal expression in bathymetry and gravity cross profiles (Table 4A; combined materials on GPlates Portal [footnote 1; http://portal.gplates.org/static/html/ExRidges/ExRidges_HTML_pages/1-09_North_Falkand_possible_ex_ridge.html] and profile files in .ZIP File S4 [footnote 2; file names S1-9_Plots_Bathymetry_NorthFalkland.png and S1-9_Plots_Grav_NorthFalkland.png). Discontinuous fracture zones to the east of the suspected extinct ridge suggest a plate reorganization (Fig. 4-ii), and a ridge-jump would address asymmetry of ∼250 km on the South American plate. The age of the extinct ridge in this location is estimated to be older than chron C33o (79.9 Ma), which has been identified to the east of the possible axis (Granot et al., 2012; Granot and Dyment, 2015). Three other extinct ridges are proposed in the South Atlantic to the north including the Vema (Pérez-Díaz and Eagles, 2014), Abimael (Scotchman et al., 2010), and Angolan Basin (Sandwell et al., 2014) extinct ridges. Two of these ridges involved similar eastward migration of the spreading center, and therefore there is significant evidence for a complex opening of the South Atlantic (Heine et al., 2013; Pérez-Díaz and Eagles, 2014; Granot and Dyment, 2015). This structure is scored as a possible extinct ridge in the absence of any additional data but presents a possible target for future surveys.

Of the ambiguous oceanic structures formed during the Cretaceous Normal Superchron in the northwest Pacific Ocean that we reviewed, the Hokkaido Trough (Mammerickx and Sharman, 1988) most closely resembles the well-defined extinct ridges in profile (Table 4A; combined materials on GPlates Portal [footnote 1; http://portal.gplates.org/static/html/ExRidges/ExRidges_HTML_pages/3-10_Hokkaido_trough.html] and associated map and profile files are in .ZIP File S3 [footnote 2; file names S3-10_Plots_Bathymetry_HokkaidoTrough.png and S3-10_Plots_Grav_HokkaidoTrough.png]) and is thought to preserve a piece of the mostly subducted Izanagi plate. In contrast, the Emperor Trough has an asymmetric morphology with the eastern flank consistently more elevated by ∼1000 m relative to the western flank (combined materials on GPlates Portal [footnote 1; http://portal.gplates.org/static/html/ExRidges/ExRidges_HTML_pages/3-14_Emperor.html] and profile files in .ZIP File S4 [footnote 2; file names S3-14_Plots_Bathymetry_Emperor.png and S3-14_Plots_Grav_Emperor.png]). The observed asymmetry is characteristic of an oceanic transform fault and is usually generated by differential oceanic crustal age either side of the structure, and therefore we conclude that the trough is unlikely to be an extinct ridge as was previously suggested (Mammerickx and Sharman, 1988).

Evaluation of alternative ridge locations in controversial regions (Tables 4A and 4B) offers some insight as to the probability of uncertain structures being extinct ridges. In several locations, a proposed axial location that is closer to the mean gravity signal, width, and bathymetric relief of the primary-tier extinct ridges can be argued to be more likely to represent a former spreading center. However, the comparison is less helpful where large volcanic ridges and edifices are present at proposed extinct ridge locations that are likely to be the result of postextinction overprinting. It is also possible that a proportion of the secondary and tertiary ridges investigated here, but not included in statistical analyses for reasons discussed above, experienced a different mechanism of cessation. These may not have developed obvious “extinct ridge-like” trough structures, potentially due to a higher spreading rate and rapid transfer to the new location. Candidates in this subset include the Gallego and Roggeveen ridge-jumps that are proposed in the eastern Pacific (Mammerickx et al., 1980; Okal and Bergeal, 1983), and that are required by significant asymmetry of oceanic crust (Müller et al., 2008), yet have left no significant axial signature. The Roggeveen Rise, in particular, is located in a region of chaotic seafloor fabric that is proposed to have resulted from successive formation of overlapping spreading centers and ridge-crest microplates due to high spreading rates (Searle et al., 1995; Matthews et al., 2011).

Characteristic versus Atypical Ridges

While complex three-dimensional structures may be present, this review demonstrates that the majority of established extinct ridges exhibit the characteristics of slow-spreading centers. They are defined by prominent axial valleys and uplifted flanks, which are accompanied by negative gravity anomalies that have a symmetric signature. The modest bathymetric relief found between extinct rift flanks and axial valleys (Fig. 5A) is similar in morphology to active ridges with full spreading rates of less than 80 mm yr–1 (see for example, Small, 1998; Fig. 5). Yet, a small number of ridges do not have the characteristics of slow-spreading centers and instead are characterized by high-relief volcanic ridges and positive gravity anomalies. The observation of a subgroup of “atypical” extinct ridges suggests that several uncertain volcanic structures, such as the Sonne, Sonja, or Dirk Hartog ridges (Mihut and Müller, 1998; Robb et al., 2005; Watson et al., 2016) could potentially be explained as extinct ridges, although alternative mechanisms for formation are not excluded.

Regional Distribution of Extinct Ridges

The Pacific Ocean contains a greater number of extinct spreading centers (38), than the Indian (16) and Atlantic (10) oceans combined total of 26 ridges, despite the present area of the Pacific being of comparable area to the other two major ocean basins (Fig. 9A). We note that extinct ridges in the Atlantic and Indian oceans were typically active for longer prior to cessation than those in the Pacific Ocean and marginal basins (Fig. 6A-V), which also suggests a regional influence on the longer-term stability of an oceanic ridge. The greater number of ridge reorganizations in the Pacific may be due to the greater influence of time-varying slab pull forces and/or the higher proportion of oceanic crust of younger age in the Pacific (Fig. 9B), since extinct ridges are more likely to be preserved in younger aged crust. However, marginal basins preserve a large number of extinct ridges (27) compared to their present-day area (∼53 million square km) and do not have a higher proportion of younger crust than the other ocean basins (Fig. 9B).

We observe that the number of extinct ridges in each major ocean increases relative to the increasing complexity of the geometry of the plate system and as the oceans have become increasingly dominated by active rather than passive margins. For example, the Atlantic Ocean has the lowest number of proposed extinct ridges and for the majority of its evolution has essentially been controlled by a two-plate system, stable Euler poles, and is surrounded by passive margins. The Indian Ocean is observed to have a greater number of extinct ridges than the Atlantic despite its smaller size, but significantly less than the Pacific Ocean. At the initiation of spreading in the Indian Ocean, a simple two-plate system was present between Africa and Antarctica (Jokat et al., 2003), and the spreading system later increased in complexity. The ocean has generally been dominated by either single or double triple-junction spreading systems, with a lengthy active destructive margin in the north (Gibbons et al., 2013). The Pacific Ocean is composed of a number of oceanic plates, including the Pacific plate, Nazca and Cocos plates, and the remnant Juan de Fuca and Rivera plates (Bird, 2003). During the Mesozoic, two synchronous triple-junction systems were active in the Pacific (Izanagi-Pacific-Farallon and Pacific-Farallon-Phoenix; Winterer, 1991; Nakanishi et al., 1999), and more recently, the fragmentation of the Farallon plate (Lonsdale, 1991, 2005) has contributed to greater complexity in the east of the ocean. It has the greatest relative proportion of subduction zones along its boundaries, and the Pacific plate has experienced frequent changes in the pole of rotation (Wessel and Kroenke, 2008). It may be that a higher proportion of hotspots in the Pacific Ocean also contributes to the propensity for ridge reorganizations because hotspots have been suggested to facilitate ridge-jumps (Nakanishi et al., 1999; Müller et al., 2001).

The increased frequency of large-scale ridge reorganizations in the Pacific is broadly consistent with the observation of higher rates of small-scale spreading system migration in the Pacific, relative to lower levels in the Atlantic Ocean (Whittaker et al., 2015). It is also consistent with increased levels of skew in the Pacific plate spreading directions in the past, represented by an angular mismatch between the paleospreading direction and absolute plate motion, compared with the Indian and Atlantic oceans (Williams et al., 2016). It may be that the combined influence of numerous, active and proximal subduction zones and a greater number of oceanic hotspots contributes to the frequency of ridge reorganizations in the Pacific. We infer that the greater number of extinct ridges in marginal basins relates to their smaller size and that they are shorter-lived (Fig. 6A-III) and thus represent a greater number of spreading systems than the major oceans, despite the fact that many of the marginal basins are bounded by active, destructive plate margins.

Well-defined extinct ridges often have trough morphology, with mean relief –772 m (SD = 816 m) and a negative peak-to-trough free-air gravity signal of ∼39 mGal (SD = 30), with half-wavelength ∼24 km (SD = 7). This is slightly more prominent than the peak-to-trough signal at active spreading ridges (mean = –35 mGal, SD = 46 mGal) and likely reflects a crustal low-density body, generated by extensive alteration of the seafloor or by emplacement of late-stage magmatic bodies. There is considerable variation within individual spreading systems and axial characteristics differ according to ridge subtype, with increased relief observed at axial segments of extinct fragmented plate ridges and microplate ridges (but not their active counterparts) and more pronounced axial relief at extinct backarc basin ridges that were active longer before cessation. Regional differences are evident in the numbers of extinct ridges present, with the Pacific Ocean and marginal basins recording more extinct ridges; these ridges are likely to be a result of a higher proportion of younger aged crust and more complex plate boundaries. Extinct backarc basin ridges and extinct microplate spreading ridges have shorter durations of spreading and faster spreading rates than large-scale mid-ocean ridges. Evaluation of uncertain extinct ridges by comparison with well-defined examples assists in identifying the structures more likely to have been former spreading centers and can improve regional reconstructions. Our catalogue of global large-scale extinct ridge locations provides a resource that improves access to the many individual studies that have been completed at extinct ridges and summarizes quantitative data about the axial characteristics of extinct spreading centers. Our review has identified several new oceanic structures that are possible extinct ridges and represent interesting targets for further study. These include the Mati Ridge in the western Palau Basin, West Philippine Sea, and the North Falkland structure in the eastern Argentine Basin, southwest Atlantic.

S.J.M. was supported by the Australia-India Strategic Research Fund and the University of Sydney Postgraduate Award. K.J.M. and R.D.M. were supported by ARC Discovery Project DP13010946. S.E.W. was supported by the Science Industry Endowment Fund (RP 04-174) Big Data Knowledge Discovery Project. We are grateful for helpful discussions with Dr. Yatheesh Vaddakkeyakath and Dr. Joanne Whittaker in the early stages of this project. We are grateful to William Sager and an anonymous reviewer for their thorough reading of the text and for constructive suggestions that allowed us to improve the manuscript. We thank Laurent Gernigon and an anonymous reviewer for their helpful feedback and comments on an earlier version of this work.

1Online database of the digitized extinct ridge (and analyzed active ridge) locations is in the GPlates Portal: http://portal.gplates.org/cesium/?view=ExRidges. Reference IDs for extinct ridges are listed on the index page, http://portal.gplates.org/portal/ExRidges/, and these link to collated data and summaries of key studies for each ridge that can be cross-referenced with Tables 1A1C.
2Supplemental Materials. Table S1A: Summary of data collected and calculated at extinct ridges and Table S1B the data for active ridges. .ZIP File S2: Digitized locations of extinct ridge axial segments and active ridges analyzed here. ZIP File S3: Bathymetry, gravity, and magnetic anomaly maps displaying inferred topographic locations of extinct ridge axial segments. .ZIP File S4: Plots of profiles generated across extinct ridge axial segments through global bathymetry and gravity grids. .ZIP File S5: Maps and profiles for active spreading ridges analyzed. Please visit http://doi.org/10.1130/GES01379.S1 or the full-text article on www.gsapubs.org to view the Supplemental Materials.

Supplementary data