For probabilistic forecasting in volcanology, two widely used methods are the Cooke method (Aspinall and others) and BET-EF (Marzocchi and others). The paper noted in the title, in Papale P (ed), Volcanic Hazards, Risks and Disasters. Academic Press, describes a third approach that we call the Method of Multiple Data Sets. It can be applied quickly in times of crises, without need for a large group of experts and formal elicitation. We provide Excel versions of figures 2, 3, and 5 (Electronic supplements 1, 2a,b, and 3) from that paper.
Figure 2 (Electronic Supplement 1) is a generic event tree, constructed in Excel, which can be modified to suit any specific volcano or crisis. Numbers in that spreadsheet are dummy numbers, purely for illustration of how the spreadsheet works. PLEASE NOTE: THREE SMALL BUT IMPORTANT FORMULA ERRORS IN THE ORIGINAL VERSION OF THIS GENERIC TREE HAVE BEEN CORRECTED AS OF SEPT 2017, THANKS TO AN ALERT FROM A USER. IF YOU DOWNLOADED A VERSION PRIOR TO THIS DATE, PLEASE REPLACE IT WITH THE CURRENT VERSION.
Figure 3, in two parts, shows the probabilities that we used in 2004 discussions with the US Forest Service at Mount St. Helens.
Figure 5 shows the probabilities for Garbuna Volcano that we used in early 2006 for discussions among scientists of USGS and the Rabaul Volcano Observatory, and for subsequent discussion with civil defense officials near Garbuna. Note that the data and reasoning behind each probability is given under the Tab labelled NOTES.
In all of these cases, we considered multiple data sets and probabilities at each node. For simplicity, we present here only the preferred values that we discussed with public officials, but we recommend that in your own applications, you track and present numbers from at least several data sets at each node, partly to approximate uncertainty and partly in the interest of full disclosure of your considerations.
We invite those who use this generic tool, or modifications thereof, to send us queries or descriptions of your own application.
Cite this work
Researchers should cite this work as follows: