Epistemic uncertainty in the location and magnitude of earthquakes in Italy from macroseismic data
Language
English
Obiettivo Specifico
5.1. TTC - Banche dati e metodi macrosismici
Status
Published
JCR Journal
JCR Journal
Title of the book
Issue/vol(year)
/101 (2011)
Pages (printed)
2712-2725
Issued date
2011年12月1日
Abstract
Three independent techniques (Bakun and Wentworth, 1997; Boxer
from Gasperini et al., 1999; and Macroseismic Estimation of Earthquake Parameters
[MEEP; see Data and Resources section, deliverable D3] from R.M.W. Musson and
M.J. Jimenez) have been proposed for estimating an earthquake location and magnitude
from intensity data alone. The locations and magnitudes obtained for a given
set of intensity data are almost always different, and no one technique is consistently
best at matching instrumental locations and magnitudes of recent well-recorded earthquakes
in Italy. Rather than attempting to select one of the three solutions as best, we
use all three techniques to estimate the location and the magnitude and the epistemic
uncertainties among them. The estimates are calculated using bootstrap resampled
data sets with Monte Carlo sampling of a decision tree. The decision-tree branch
weights are based on goodness-of-fit measures of location and magnitude for recent
earthquakes. The location estimates are based on the spatial distribution of locations
calculated from the bootstrap resampled data. The preferred source location is the
locus of the maximum bootstrap location spatial density. The location uncertainty is
obtained from contours of the bootstrap spatial density: 68% of the bootstrap locations
are within the 68% confidence region, and so on. For large earthquakes, our preferred
location is not associated with the epicenter but with a location on the extended rupture
surface. For small earthquakes, the epicenters are generally consistent with the
location uncertainties inferred from the intensity data if an epicenter inaccuracy of
2–3 km is allowed. The preferred magnitude is the median of the distribution of bootstrap
magnitudes. As with location uncertainties, the uncertainties in magnitude are
obtained from the distribution of bootstrap magnitudes: the bounds of the 68% uncertainty
range enclose 68% of the bootstrap magnitudes, and so on. The instrumental
magnitudes for large and small earthquakes are generally consistent with the confidence
intervals inferred from the distribution of bootstrap resampled magnitudes.
from Gasperini et al., 1999; and Macroseismic Estimation of Earthquake Parameters
[MEEP; see Data and Resources section, deliverable D3] from R.M.W. Musson and
M.J. Jimenez) have been proposed for estimating an earthquake location and magnitude
from intensity data alone. The locations and magnitudes obtained for a given
set of intensity data are almost always different, and no one technique is consistently
best at matching instrumental locations and magnitudes of recent well-recorded earthquakes
in Italy. Rather than attempting to select one of the three solutions as best, we
use all three techniques to estimate the location and the magnitude and the epistemic
uncertainties among them. The estimates are calculated using bootstrap resampled
data sets with Monte Carlo sampling of a decision tree. The decision-tree branch
weights are based on goodness-of-fit measures of location and magnitude for recent
earthquakes. The location estimates are based on the spatial distribution of locations
calculated from the bootstrap resampled data. The preferred source location is the
locus of the maximum bootstrap location spatial density. The location uncertainty is
obtained from contours of the bootstrap spatial density: 68% of the bootstrap locations
are within the 68% confidence region, and so on. For large earthquakes, our preferred
location is not associated with the epicenter but with a location on the extended rupture
surface. For small earthquakes, the epicenters are generally consistent with the
location uncertainties inferred from the intensity data if an epicenter inaccuracy of
2–3 km is allowed. The preferred magnitude is the median of the distribution of bootstrap
magnitudes. As with location uncertainties, the uncertainties in magnitude are
obtained from the distribution of bootstrap magnitudes: the bounds of the 68% uncertainty
range enclose 68% of the bootstrap magnitudes, and so on. The instrumental
magnitudes for large and small earthquakes are generally consistent with the confidence
intervals inferred from the distribution of bootstrap resampled magnitudes.
Type
article
File(s)![]()
cris-layout.advanced-attachment.dc.title
Bakun_Gomez-Capera_Stucchi_2011.pdf
cris-layout.advanced-attachment.size
3.95 MB
cris-layout.advanced-attachment.format
Adobe PDF
cris-layout.advanced-attachment.checksum (MD5)
checksum-tooltip.info.MD5
d427cb38ed61fc6b085aed2a311f7bb4
