Technical Comments

Comment on “Glacial cycles drive variations in the production of oceanic crust”

See allHide authors and affiliations

Science  04 Sep 2015:
Vol. 349, Issue 6252, pp. 1065
DOI: 10.1126/science.aab2350


Crowley et al. (Reports, 13 March 2015, p. 1237) propose that abyssal hill topography can be generated by variations in volcanism at mid-ocean ridges modulated by Milankovitch cycle–driven changes in sea level. Published values for abyssal hill characteristic widths versus spreading rate do not generally support this hypothesis. I argue that abyssal hills are primarily fault-generated rather than volcanically generated features.

Crowley et al. (1) purport to show evidence that topographic profiles through abyssal hill morphology perpendicular to the spreading ridge axis exhibit periodicities that correspond well to Milankovitch cycles. The argument is that this correspondence could be explained by volcanic output at the mid-ocean ridge (MOR) modulated by the lithostatic pressure variations associated with rising and falling sea level. A contemporaneous paper by Tolstoy (2) reaches the same conclusion. I will refer to this argument as the “CT hypothesis.” Crowley et al. also formulate a sophisticated numerical model to predict seafloor variations that could result from sea-level variability over the past 1.2 million years.

The CT hypothesis makes specific predictions about the variation of abyssal hill horizontal scales in the direction perpendicular to the spreading axis (the “width” of the abyssal hill) as a function of spreading rate; these predictions can be tested with existing published results. In the simplest possible case, the CT hypothesis predicts that abyssal hill width will scale linearly with spreading rate. For example, if we assume that the 100,000-year Milankovitch cycle is the dominant driver for abyssal hill formation, as suggested by Tolstoy (2), then abyssal hill width will be 1 km for every 1 cm/year of half-spreading rate. Thus, widths would be ~1 km at ultraslow half-spreading rates of 1 cm/year and ~7 km at ultrafast rates of 7 cm/year. The numerical modeling by Crowley et al., however, suggests that shorter-period Milankovitch cycles may be more dominant at faster rates. Depending on certain modeling parameters, the 41,000-year cycle might dominate at 4cm/year half-spreading rate, and the 23,000-year cycle might dominate at 7 cm/year half-spreading rate [see figure 1 in (1)]. If so, this would suggest a much lower sensitivity of abyssal hill width on spreading rate, varying from ~1 km at 1 cm/year half-rate to just 1.6 km at 7 cm/year half-rate.

Global, averaged estimates of the characteristic abyssal hill width (3, 4) are presented as a function of half-spreading rate (Fig. 1). The characteristic width is formally estimated by the width of the covariance function but can also be derived with the von Kármán statistical model (5) from the inverse of the corner frequency of a power spectrum modeled as a band-limited fractal (6). The characteristic width describes the dominant visual scale (6). The von Kármán spectral model, being fractal in nature, does not contain periodicities. However, if this model were fit to the spectral functions modeled by Crowley et al. in their figure 1, the corner frequency would correspond to the periodicity with the highest amplitude, because this is where the spectrum transitions from being approximately white (flat) at low frequencies to red (fractal) at high frequencies. Therefore, the characteristic width should represent a reasonable estimate of the dominant abyssal hill width if the abyssal hills were, in fact, periodic. Under this assumption, the characteristic width values shown in Fig. 1 are incompatible with the predictions of the CT hypothesis in two important ways: (i) average widths measured at the slowest spreading rates, ~8 km, are far too large to be correlated with Milankovitch cycles, and (ii) the trend of widths with spreading rate is negative, rather than the predicted positive or flat trend, at least to half-rates of ~3.7 cm/year (or the transition from axial valley to axial high MOR morphology). The characteristic widths also exhibit a large (>2 km) downward step, independent of spreading rate, where the MOR morphology transitions from axial valley to axial high; these values were estimated on the flanks of the Southeast Indian Ridge (3).

Fig. 1 Abyssal hill characteristic widths, averaged by region, plotted as a function of half-spreading rate.

Error bars, mean ± 1 SD. Figure adapted from Goff et al. (3), with additional data from Sloan et al. (4). Black symbols indicate regions formed at axial valley MORs; open symbols at axial high MORs; and gray symbol at transitional MOR.

The negative trend in characteristic width with spreading rate can be seen as consistent with a faulted origin. Fault offset and spacing scale with elastic plate thickness (7, 8), which is dependent on thermal structure (9). Thermal structure is, in turn, dependent on spreading rate (10). As a result of these relationships, colder lithosphere at slower spreading rates should result in larger offsets on more widely spaced faults. Faulting may also be controlled by variations in magma supply (11), which could explain variations in abyssal hill morphology that are independent of spreading rate, such as along the Southeast Indian Ridge (3). Observational studies have shown that abyssal hills are primarily fault-controlled horst-and-graben features, rather than volcanic-controlled constructs, at both fast (12) and slow (13) spreading rates. Indeed, the abyssal hills displayed by Crowley et al. exhibit a strikingly linear, axis-parallel morphology, which is likely most consistent with a faulted origin. Although the correlation between axial volcanism and sea-level variations is certainly plausible, it appears more likely to translate only as a secondary superposition on the primarily tectonic abyssal hill morphology (12).

At half-spreading rates greater than ~3.7 cm/year, where the axial high ridge morphology dominates, the trend of characteristic width with spreading rate is no longer negative and, in fact, may be slightly positive in going from ~2 km at 3.7 cm/year to ~3 km at 8 cm/year (Fig. 1). Goff et al. (3) argued that lack of sensitivity of abyssal hills to spreading rate at these higher rates could be associated with a nearly uniform presence of a weak zone in the lower crust at axial high ridges (14), which would decouple surface faulting from deeper strain. We cannot, however, discount the possibility that Milankovich cycle–driven volcanism is contributing, at least in part, to the increase in measured characteristic with spreading rate at axial high ridges; no other plausible explanation for this observation has been offered, to my knowledge.

References and Notes

  1. Acknowledgments: The author thanks G. Christeson for helpful comments on an earlier draft.
View Abstract

Stay Connected to Science


Navigate This Article