HR: 09:20h
AN: V21C-06 [Abstracts]
TI: A Tephra Database With an Intelligent Correlation System, Mono-Inyo Volcanic Chain, CA
AU: * Bursik, M
EM: mib@geology.buffalo.edu
AF: University at Buffalo, SUNY, Department of Geology
University at Buffalo, Buffalo, NY 14260
United States
AU: Rogova, G
EM: rogova@rochester.rr.com
AF: Encompass Consulting, Encompas Consulting, Honeoye Falls, NY 14260
United States
AB:
We are assembling a web-accessible, relational database of information
on past eruptions of the Mono-Inyo volcanic chain, eastern California.
The PostgreSQL database structure follows the North American Data
Model and CordLink. The database allows us to extract the features
diagnostic of particular pyroclastic layers, as well as lava domes and
flows. The features include depth in the section, layer thickness and
internal stratigraphy, mineral assemblage, major and trace element
composition, tephra componentry and granulometry, and radiocarbon age.
Our working hypotheses are that 1) the database will prove useful for
unraveling the complex recent volcanic history of the Mono-Inyo chain
2) aided by the use of an intelligent correlation system integrated
into the database system. The Mono-Inyo chain consists of domes,
craters and flows that stretch for 50 km north-south, subparallel to
the Sierran range front fault system. Almost all eruptions within the
chain probably occurred less than 50,000 years ago. Because of the
variety of magma and eruption types, and the migration of source
regions in time and space, it is nontrivial to discern patterns of
behaviour.
We have explored the use of multiple artificial neural networks
combined within the framework of the Dempster-Shafer theory of
evidence to construct a hybrid information processing system as an aid
in the correlation of Mono-Inyo pyroclastic layers. It is hoped that
such a system could provide information useful to discerning eruptive
patterns that would otherwise be difficult to sort and categorize. In
a test case on tephra layers at known sites, the intelligent
correlation system was able to categorize observations correctly 96%
of the time. In a test case with layers at one unknown site, and
using a pairwise comparison of the unknown site with the known sites,
a one-to-one correlation between the unknown site and the known sites
was found to sometimes be poor. Such a result could be used to aid a
stratigrapher in rethinking or questioning a proposed correlation.
This rethinking might not happen without the input from the
intelligent system.
UR: http://www.volcano.buffalo.edu/mmvz
DE: 8404 Ash deposits
DE: 8409 Atmospheric effects (0370)
DE: 8494 Instruments and techniques
SC: Volcanology, Geochemistry, Petrology [V]
MN: 2004 AGU Fall Meeting