other aspects of the process must be resolved before we can fully understand how volcanoes work and extend the ocean ridge model to magma generation in other environments (see Question 9). For example, current models do not explain how magma produced in a broad, 150-km-wide zone under midocean ridges is focused to erupt mainly within a narrow 10-km-wide zone at the ridges. The chemistry and Th-isotope ratios of midocean ridge lavas also do not match model predictions of the depth of the melting region or the way magmas with different compositions and viscosities move and mix under the ridges (Sims et al., 2002). More comprehensive numerical models are beginning to incorporate chemical reactions accompanying magma flow but still suffer from our limited knowledge of the mechanical properties of partially molten rock and our inability to represent the chemical reactions accurately.
A less well understood type of magmatism occurs in association with subduction zones. Although modest in number, subduction zone volcanoes represent nearly all of the explosive volcanoes (Question 9) and the mechanism by which much of the continental crust is produced (Question 5). That volcanoes are located above relatively cold parts of the mantle is evidence that a fundamentally different mechanism(s), perhaps unique to Earth, is responsible for producing magma. Although small-scale convection driven by frictional drag on the slab may cause melting above the slab, water is the melting mechanism invoked most often. Water (in the form of OH− groups in minerals like amphibole) is carried into the mantle by subducting slabs and then is lost as the slabs are metamorphosed (Tatsumi and Eggins, 1995). This water lowers the melting temperature of the mantle by 200°C or more. If the released water moves upward from the cool slab into hotter mantle above, it can produce the magma needed to generate volcanoes. The supply of water by subduction to the magma-producing regions located 100 km or more below the volcanoes is confirmed by the presence of the short-lived isotope 10Be, derived from the atmosphere, in some island arc lavas. The mechanisms by which water- and CO2-rich fluids move in the mantle are poorly understood but central to this puzzle; these mechanisms also influence the chemical and isotopic tracers that subducting slabs carry back into the deep mantle. Other processes may also cause melting above the slab, such as small-scale convection driven by frictional drag on the slab.
In general, our knowledge of volcanic processes is much better for near-surface regions than for deeper regions where magma initially forms. A major objective is to understand volcanism from the bottom up—that is, to learn to predict the volume, composition, and eruptive behavior of volcanoes from models of convection and heat transfer processes in the upper mantle and lower lithosphere. The bottom-up approach contrasts with traditional volcanology, which is motivated by hazard assessment to study volcanoes from the top down (Question 9). Bottom-up volcanology may also benefit from studies of other planets, such as Mars and Io, where boundary conditions are different enough from those of Earth to allow models to be tested. Better models for the deep structure of volcanoes and long-term degassing of planetary interiors will require major leaps in our knowledge of partially molten rock and magma, the role of water in melting, the effect of melting on the viscosity of partially molten rock, and the distribution of volatile elements between solids and liquids.
Exchange in the interior: Subduction and mantle plumes. Subduction occurs when old oceanic seafloor moves slowly away from an oceanic ridge and across the ocean bottom, cools, and sinks into the mantle (Question 5). Cold subducted slabs contain rock that has reacted chemically with ocean water (Box 2.2) and sediment derived from continents and shell-forming organisms in the oceans. Although much of the sediment may be scraped off in the shallow part of the subduction zones, the slabs carry some of it, plus chemical and isotopic traces of reaction with the ocean, down into the mantle. In this way subduction changes both mantle geochemistry and the volume and composition of the oceans (Question 7).
The extent to which subducted slabs are assimilated into the mantle is an open question. Some seismological images have high-velocity tabular features in the midmantle and even at the base of the mantle that are suspected to be former oceanic lithosphere. Numerical models indicate that it is plausible that sinking slabs remain cool and coherent all the way to the base of the mantle, where they pile up in a “slab graveyard” (Figure 2.2; Christensen and Hofmann, 1994). If this happens,