Predicting volcanic eruptions

Scarpa, Roberto

'In the past decade, some major explosive volcanic eruptions have been forecast successfully, for example, at Pinatubo, Philippines, in 1991, Rabaul, New Guinea, in 1994, and Soufriere Hills, Montserrat, in 1995. Anomalous geophysical signals were correctly interpreted as early warnings of an impending eruption, and cities at the foot of these volcanoes were evacuated several hours to weeks before the renewal of explosive activity. Nevertheless, volcanic eruptions and their secondary effects, such as lahars or mudflows, still claim many lives, for example, in the eruptions of Nevado del Ruiz, Colombia, in 1985 and Lake Nyos, Cameroon, in 1986, both of which caused thousands of deaths. The reason for this discrepancy is that no rigorous methodology exists for predicting an impending eruption and the evolution of eruptive activity. Early detection of volcanic activity is mainly based on the observation of an anomalous increase in seismic activity and ground deformations. Substantial variations in these factors, which are a result of increased pressure of the magmatic system and magma motion, certainly constitute a basic tool for issuing a warning (1), but not all volcanoes behave as expected, and existing data are too limited for reliable prediction on a global scale. That may soon change. In the past two decades, the temporal and spatial resolution of ground deformation data has increased rapidly through use of the Global Positioning System (GPS) and satellite radar interferometry. In addition, seismological observations now provide information about the underground structure and the magmatic chambers at a scale of a few hundred meters. Waveform inversion from broadband seismic instruments is emerging as a powerful tool for modeling the source geometry of the magma feeding system and its dynamics. An excellent example is the report by Kumagai et al on page 687 of this issue (2). The wealth of data provided by these techniques and by other broadband instruments such as strainmeters (see the figure) is beginning to put eruption prediction on a quantitative footing.'

Predicting volcanic eruptions

Series