For many geoscientists and lay people the water cycle is considered to be part of the Earth’s surface system. That is, the cycle of evapotranspiration, precipitation and infiltration involving atmosphere, oceans, cryosphere, terrestrial hydrology and groundwater. Yet it links to the mantle through subduction of hydrated oceanic lithosphere and volcanism. The rate at which water vapour re-enters the surface part of the water cycle through volcanoes is reasonably well understood, but the same cannot be said about ‘recharge’ of the mantle through subduction.
Subducted oceanic crust is old, cold and wet: fundamentals of plate theory. The slab-pull that largely drives plate tectonics results from phase transitions in oceanic crust that are part and parcel of low-temperature – high-pressure metamorphism. They involve the growth of the anhydrous minerals garnet and high-pressure pyroxene that constitute eclogite, the dense form taken by basalt that causes the density of subducted lithosphere to exceed that of mantle peridotite and so to sink. This transformation drives water out of subducted lithosphere into the mantle wedge overlying a subduction zone, where it encourages partial melting to produce volatile-rich andesitic basalt magma – the primary magma of island- and continental-arc igneous activity. Thus, most water that does reenter the mantle probably resides in the ultramafic lithospheric mantle in the form of hydrated olivine, i.e. the mineral serpentine, and that is hard to judge.
Water probably gets into the mantle lithosphere when the lithosphere bends to begin its descent. That is believed to involve faults – cold lithosphere is brittle – down which water can diffuse to hydrate ultramafic rocks. So the amount of water probably depends on the number of such bend-related faults. A way of assessing the degree of such faulting and thus the proportion of serpentinite is analysis of seismic records from subduction zones. This has been done from earthquake records from the West Pacific subduction zone descending beneath northern Japan (Garth, T. & Rietbrock, A. 2014. Order of magnitude increase in subducted H2O due to hydrated normal faults within the Wadati-Bennioff zone. Geology, on-line publication doi:10.1130/G34730.1). The results suggest that between 17 to 31% of the subducted mantle there has been serpentinised.
In a million years each kilometre along the length of this subduction zone would therefore transfer between 170 to 318 billion tonnes of water into the mantle; an estimate more than ten times previous estimates. The authors observe that at such a rate a subduction zone equivalent to the existing, 3400 km long Kuril and Izu-Bonin arcs that affect Japan would have transferred sufficient water to fill the present world oceans 3.5 times over the history of the Earth. Had the entire rate of modern subduction along a length of 55 thousand kilometres been maintained over 4.5 billion years, the world’s oceans would have been recycled through the mantle once every 80 million years. To put that in perspective, since the Cretaceous Chalk of southern England began to be deposited, the entire mass of ocean water has been renewed. Moreover, subduction has probably slowed considerably through time, so the transfer of water would have been at a greater pace in the more distant past.