The subduction zones at which the tectonic plates beneath the sea thrust into the deep Earth act as gigantic conveyer belts, carrying water, fluids and volatile compounds into our planet. Water in Earth’s interior is released back into the oceans and atmosphere by volcanoes. These inputs and outputs constitute a global deep-Earth water cycle, but quantifying the total water input from oceanic plates has proved difficult. Writing in Nature, Cai et al.1 report that the Pacific plate, which subducts in the Mariana Trench, contains much more water than was previously supposed — a finding that has major ramifications for Earth’s water budgets.
Water is as crucial to the workings of Earth’s interior as it is to Earth’s surface processes: among other things, it triggers magma generation beneath volcanoes, lubricates deep fault zones, and fundamentally alters the strength and behaviour of Earth’s mantle. Sea water seeps into the oceanic lithosphere through fractures and pores, and reacts with minerals in the crust and mantle to form hydrous minerals (such as serpentine) that store water in their crystal structures.
Water infiltration occurs at a couple of key stages of an oceanic plate’s life cycle. The first is at mid-ocean ridges, when water circulates through hot, newly formed oceanic plates2. But at fast-spreading ridges (which are the primary ‘diet’ of the subduction zones that ring the Pacific Ocean), circulation and hydration are mainly restricted to the plate’s upper crust. The accumulation of sediments subsequently seals off most of the oceanic plate from the ocean, but seamounts (underwater mountains) and fracture zones provide pathways for further water input and output, so that circulation continues away from the mid-ocean ridge3. The final infiltration occurs at the ‘outer rise’ of a subduction zone, where the oceanic plate bends before entering the trench. Here, extensional faults form in response to bending, and are thought to enable pervasive, deeply penetrating hydration of the crust and upper mantle4–6.
The evidence for water entering subduction zones is clear, but several knowledge gaps have hindered attempts to quantify the total volume of water going down these hatches, even at individual subduction zones. One unknown is the depth to which water penetrates the oceanic plate. Most constraints on estimates come from controlled-source seismic data, which are produced by measuring seismic waves generated by artificial sources using dense arrays of recording instruments. These data provide excellent constraints on hydration of the crust and shallow mantle, but, with one notable exception7, do not constrain the full depth of hydration. It is clearly not possible to tally the total volume of subducted water without knowing the full hydration depth.
Another important challenge is to untangle all the factors that alter the speed at which seismic waves travel through different parts of the plate; measurements of such seismic waves are one of the primary means of estimating the amount of water in the subducting oceanic plate. Most estimates assume that any reduction in wave speed results from the replacement of olivine (the main mineral found in the mantle) by serpentine. However, in the crust and shallow mantle, water-filled cracks can also contribute to velocity reductions8. To complicate things further, seismic-wave speeds in the upper oceanic mantle are anisotropic — they depend on the direction of propagation. This is because olivine crystals align in the direction in which the sea floor spreads when new oceanic plates are created at mid-ocean ridges. Further anisotropy can result from fractures formed at the outer rise.
Cai et al. tackle all of these issues by presenting constraints on the hydration of the approximately 150-million-year-old Pacific plate as it subducts at the Mariana Trench. The authors analysed seismic waves from distant earthquakes, recorded by an array of seismometers on the sea floor. This allowed them to model seismic-wave speeds to much greater depths (albeit at lower resolution) than is possible using controlled-source seismic data.
The researchers find that, impressively, the full hydration depth of the lithosphere extends to approximately 30 kilometres below the sea floor (Fig. 1). They were also able to examine velocity reductions in deep regions at which the pressure would be sufficiently high to close all cracks, thus allowing them to eliminate the possible contribution of such cracks to velocities in these regions. Finally, because the authors recorded waves travelling in all directions across their array, they were able to account for the contribution of anisotropy to wave speed.
Cai et al. report that more than four times as much water is entering the Mariana subduction zone than was previously estimated9. Old, cold subducting plates such as that entering the Mariana Trench are particularly effective conveyors of water into the deep Earth because hydrous minerals in such plates are stable to greater depths than in younger, hotter plates. If extrapolated globally to other places where old, cold plates subduct, the authors’ result implies that the amount of water entering Earth’s interior greatly exceeds current estimates10 of the amount being emitted by volcanoes, and thus requires a rethink of the global water budget.
Several questions still need to be answered to better constrain estimates of the inputs to Earth’s deep-water cycle and evaluate the implications of the new results. First, how variable is the water content in the oceanic plate at a range of depths and scale lengths along subduction zones? Many studies have reported changes in extensional faulting and in crustal and upper-mantle hydration along subduction zones, and several competing factors have been proposed to contribute to these changes7,11,12. Characterizing this variability throughout the hydrated part of the plate and understanding its causes will be essential for us to tally water inputs and compare them with outputs. It would also be useful to determine whether hydration occurs in focused zones near faults, as has been observed in a different tectonic setting13, or is distributed more evenly, because this might control whether mineral-bound water is released by dehydration or is carried to greater depths14.
Finally, because a substantial volume of water is probably stored in the crust and upper mantle (the regions that are most accessible to seawater infiltration), the thorny issue of whether changes to seismic waves reflect the presence of water-filled cracks or of hydrous minerals still needs to be directly addressed in these areas. A comprehensive understanding of inputs to Earth’s interior will require a multi-pronged approach, including multi-scale marine geophysical studies, drilling of the faults that are thought to be conduits for water into the oceanic lithosphere, and numerical and laboratory studies. But for now, Cai and colleagues’ results have taken us a big step closer to understanding the total water input at subduction zones.
Nature 563, 335-336 (2018)