is showing signs of activity. Much might also be learned from synthesis of the large and growing volume of data that are available on entire volcanic arcs stretching over thousands of kilometers. Patterns in the timing and volume of eruptions at this large scale have received relatively little attention but could prove important for relating volcanic activity to rates of large-scale tectonic processes that can also be monitored with geodetic and seismic techniques.
If human populations are to live close to active volcanoes with a reasonable degree of safety, geoscientists must be able to (1) assess the risk of eruptive activity based on past history and (2) provide reliable predictions of eruptive potential during times of volcanic restlessness based on eruption precursors. The geological record provides information on the recurrence rates and magnitudes of large volcanic eruptions. More detailed eruptive histories of specific volcanoes have enabled some long-term predictions to be made. For example, the analysis of the history of Mount St. Helens (Crandell and Mullineaux, 1978) was notable for its accurate forecast of another eruption before the end of the century. However, it is not clear whether data from infrequent but much larger eruptions in the past can be compared with data from brief, small, recent eruptions. We know that the magnitude and destructiveness of past volcanoes have greatly exceeded anything in human history (Figure 4.9). For example, supereruptions have produced more than 2,000 km3 of pumice and ash as recently as 75,000 years ago in Indonesia (the Toba eruption; Rose and Chesner, 1987) and 2 million years ago at Yellowstone in the United States (Christiansen, 1984). This is 50 times the amount of material erupted by Tambora volcano, in Indonesia, in 1815, an eruption that caused the deaths of more than 90,000 people and disrupted global climate (Stothers, 1984; Sigurdsson and Carey, 1989). Similarly, the enormous outpourings of basaltic lava (more than 2,000 km3 during single events) in Washington and Oregon about 16 million years ago (Hooper, 1997) dwarf that of the Laki eruption of 1783 in Iceland (about 15 km3), an eruption that killed more than 9,000 Icelanders directly and unknown
numbers of Europeans because of crop failures and starvation (Thordarson and Self, 2003).
The largest explosive eruption monitored by modern techniques was that of Mount Pinatubo in 1991, when the release of only about 5 km3 of magma caused the collapse of the volcano’s summit. Emergency evacuation of surrounding cities and towns before the event was a success for volcanic eruption prediction (Newhall and Punongbayan, 1996). Yet much larger volcanic systems such as Yellowstone have shown signs of restlessness that could, at some point, portend an impending eruption. We do not know whether we can accurately scale up modern instrumental data for Pinatubo-sized eruptions to anticipate events that may be 100 or 1,000 times larger.
Another challenge in predicting volcanic eruptions will be to combine diverse observational data sets (e.g., seismic, geodetic, infrasound, thermal, gas measurements, visual observation via webcams) to track, in real time, not only the movement of magma toward the surface but also changes in the material properties of the magma that affect its explosive potential.