Skip to content

Model-based lava flow hazard assessment: pre- and syn-eruptive forecasts with uncertainty quantification

Dave M. R. Hyman1,2, Hannah R. Dietterich3, Kyle R. Anderson4

  • Affiliations: 1 U.S. Geological Survey, Science Analytics and Synthesis, Advanced Research Computing, Lakewood, CO, USA; 2 U.S. Geological Survey, Volcano Science Center, Cascades Volcano Observatory, Vancouver, WA, USA; 3 U.S. Geological Survey, Volcano Science Center, Alaska Volcano Observatory, Anchorage, AK, USA; 4 U.S. Geological Survey, Volcano Science Center, California Volcano Observatory, Moffett Field, CA, USA 

  • Presentation type: Talk [Invited]

  • Presentation time: Tuesday 08:45 - 09:00, Room R380

  • Programme No: 6.5.2

  • Theme 6 > Session 5

Abstract

During effusive eruptions, lava flows can displace populations and destroy homes, roads, and other infrastructure. Although the impacts of individual flows tend to be highly localized, lava flows can threaten wide areas over multiple eruptions since they cool and create new topography, inhibiting future flows from reoccupying old flow paths. Accordingly, lava flow hazard assessments can span a wide range of temporal and spatial scales from regional assessments of inundation on geologic time scales to forecasts of individual flows relevant over the coming few days. Increasingly, computational models are being used to inform short- and long-term inundation assessments either to augment or replace traditional assessments based on geological mapping or historical precedent. However, there are major challenges, including constraining the range of model validity, parameterizing models, and characterizing and quantifying uncertainty.  Uncertainty quantification is a particularly large concern for short-term modeling since there is not typically time to run many models or construct a precise statistical model during a crisis. By contrast, long-term assessments typically embed some aspects of uncertainty (e.g., vent location), although comprehensive uncertainty quantification is often impractical at this scale. Here, we detail recent and ongoing model-based lava flow hazard assessment efforts targeting multiple timescales and user needs including forecasts during the 2022 Mauna Loa eruption and planning for future eruptions of Kīlauea and Mauna Loa.  In particular, we focus on tracking sources of error in these assessments enabling estimation of model product uncertainty and meeting user needs for "most likely" and "worst case" products.