T82 Physics of Volcanic Eruptions: Implications for Hazards

Monday, 19 October 2009
8:00 AM - 12:00 PM
B116

MODELING AND FORECASTING TEPHRA ACCUMULATION WITH TEPHRA2

Paper number: T82-16
Program: Technical Sessions
Time: 11:45 AM – 12:00 PM

Authors

Abstract

TEPHRA2 is a parallel code written in C and MPI and is used to forecast tephra accumulation following explosive volcanic eruptions. The code uses a closed-form solution of the advection-diffusion equation, particle fall velocities that depend on local Reynold’s number, and stratified wind field to forecast tephra accumulation, usually expressed as kilogram per cubic meter, particle size distribution at specific locations from the vent, and maximum clast size expected as a function of distance from the vent. In practice, deposits of specific eruptions can be modeled if sufficient field data are available using TEPHRA2 and a nonlinear inversion method, the downhill simplex algorithm, to estimate best fit eruption parameters. We demonstrate wide applicability of TEPHRA2 with examples from Cerro Negro, Nicaragua (small volume basaltic eruption), Colima, Mexio (short-lived basaltic-andesite vulcanian eruption), and Chaiten, Chile (comparatively small volume Plinian eruption of rhyolite composition). In each case models of the eruption based on inversion of the deposit thickness and/or granulometry compares well with independent observations. An advantage of this inversion procedure is that it is possible to assign numerical uncertainty to estimates of deposit volume, eruption column height, eruption duration, and similar eruption input parameters. TEPHRA2 is used as a forecasting tool with Monte Carlo simulation. Input parameters are specified using probability density functions for eruption mass, column height and grainsize distribution. Wind data are specified using historical records or reanalysis data. Output is cast as a probability density function of tephra accumulation (or maximum grainsize) at a specific point, or the probability that tephra accumulation will exceed some threshold value. One challenge in the continued development of codes like TEPHRA2 involves model parsimony. How complex need a code become in order to forecast activity from a hazard perspective?