Synthetic benchmarking of concentrated pyroclastic current models

Abstract

Validation and benchmarking of pyroclastic current (PC) models is required to evaluate their performance and their reliability for hazard assessment. Here, we present results of a benchmarking initiative built to evaluate four models commonly used to assess concentrated PC hazard: SHALTOP, TITAN2D, VolcFlow, and IMEX_SfloW2D. The benchmark focuses on the simulation of channelized flows with similar source conditions over five different synthetic channel geometries: (1) a flat incline plane, (2) a channel with a sharp 45° bend, (3) a straight channel with a break-in-slope, (4) a straight channel with an obstacle, and (5) a straight channel with a constriction. Several outputs from 60 simulations using three different initial volume fluxes were investigated to evaluate the performance of the four models when simulating valley-confined PC kinematics, including overflows induced by topographic changes. Quantification of the differences obtained between model outputs at t = 100 s allowed us to identify (1) issues with the Voellmy-Salm implementation of TITAN2D and (2) small discrepancies between the three other codes that are either due to various curvature and velocity formulations and/or numerical frameworks. Benchmark results were also in agreement with field observations of natural PCs: a sudden change in channel geometries combined with a high-volume flux is key to generate overflows. The synthetic benchmarks proved to be useful for evaluating model performance, needed for PC hazard assessment. The overarching goal is to provide an interpretation framework for volcanic mass flow hazard assessment studies to the geoscience community.