MILESTONES ON THE ROAD TO ERUPTION PREDICTION (Invited Presentation)

Once considered beyond reach, eruption prediction has become a realistic goal of modern volcanological research. Prior to the 1980 eruption of Mount St. Helens, predicting the behavior of complex magmatic systems was thought to be impractical in most cases. That view changed dramatically when Swanson et al. (Science, 1983) reported that 13 eruptions at Mount St. Helens had been accurately predicted as much as 3 weeks in advance with no false alarms. The basis for this unprecedented achievement was neither a breakthrough in technology or theory, but rather a combination of careful field observations and pattern recognition. Not rocket science, but excellent science nonetheless. Less than a decade later, scientists from the Philippine Institute of Volcanology and Seismology and the U.S. Geological Survey successfully forecast the 1991 climactic eruption of Mount Pinatubo, resulting in the saving of at least 5,000 lives and $250 million in property. In February 2000, a warning was broadcast that Hekla volcano would erupt explosively within 15–20 minutes based on a pattern of seismicity and ground deformation very similar to that seen before the previous eruption in 1993 (Agustsson et al., 2000, EOS). The eruption began about 24 minutes later. Probabilistic event trees based on expert elicitation have been used during several recent volcanic crises (Newhall and Hoblitt, 2002, Bull. Volc.; Newhall and Pallister, 2015, Elsevier) to successfully anticipate outcomes. White and McCausland (2016, JVGR) have proposed a framework for estimating the volume of magmatic intrusions and forecasting eruptions based on patterns of volcano-tectonic earthquakes. The next challenge is to develop models that incorporate the physics of magmatic processes. In one recent example, Anderson and Segall (2013, JGR) modeled the magma reservoir at that fed the 2004-2008 eruption at Mount St. Helens based on 2 years of erupted volume and ground deformation data, ran their model forward, and successfully predicted (in hindsight) the final volume erupted after 3.5 years. These recent successes demonstrate that eruption prediction is no longer a pipe dream, thanks in large part to the pioneering work of Swanson et al. that emboldened the volcanological community to tackle what seemed to be an intractable problem just 4 decades ago.