Jökull - 01.01.2017, Blaðsíða 14
Microearthquakes on the eastern flank of Katla volcano, S-Iceland
Figure 6. Histogram of residuals derived by relative location after the initial inversion with uncleaned data
(family 1). The right frame is an enlargement of dashed area in the left panel. – Súlurit sem sýnir dreifingu
tímafrávika hrárra mæligagna eftir afstæðar staðsetningar (skjálftaflokkur 1). Myndin til hægri er stækkuð
mynd af þeim hluta myndarinnar til vinstri sem liggur neðan við rauðu brotalínuna.
tion corrections starting from no corrections and iter-
ating until converging to stable corrections. We used a
7 by 7 km2 area around the center, extending down to
7 km depth, with a grid resolution of 50 m in all direc-
tions. The location and station-correction estimation
processes converged after five iterations with a final
rms correction of 0.1 sec.
The combined probability density of the hypocen-
ter locations for family 1 is shown in Figure 7. By
combined probability density we mean the sum of the
probability densities for all individual events. There-
fore, the distribution of this density combines the dis-
tribution of the hypocenters and their uncertainties.
The density is distributed over an area of 1000 x 700
m2 in the horizontal dimensions and around 3.5 km in
depth. The average uncertainty of individual events is
around 200 m in the horizontal and 1200 m in depth.
The upper bound one-standard-deviation contour of
combined probability density lies at approximately
1.8 km depth, therefore, the depth of the cluster is
significantly different from zero and we conclude that
the events are located at depth and not at the surface.
Similar analysis of the events in family 2 (not pre-
sented) leads to the same conclusion. The locations
of the template events for the two families lie within
their error ellipsoids and, therefore, cannot be distin-
guished. They are both located at about 3.5 km depth.
Relative Locations
In order to extract more details about the hypocen-
tral distribution and obtain further information on the
size and shape of the cluster, we also applied a rel-
ative location technique using the differential times
obtained with cross correlation. The relative loca-
tion method can be applied to clustered events if the
hypocentral separation between them is small com-
pared to the event-station distance and scale length of
velocity heterogeneities. In such geometry, the ray
paths between the source and a common station are
similar and the time delays between events recorded at
the same station depends primarily on the spatial off-
set between the hypocenters (Fréchet, 1985; Got et al.,
1994; Slunga et al., 1995; Waldhauser and Ellsworth,
2000).
The relative location problem is solved using a
1-D velocity model, by formulating the problem in
a very similar way to the double-difference (DD)
strategy of Slunga et al. (1995) and Waldhauser and
Ellsworth (2000). The DD algorithm uses both ab-
solute and differential arrival times between pairs of
events recorded at a common station to relocate events
in a relative sense, and their center of mass in an ab-
solute sense. Here we used differential arrival times
only and computed the locations of the events relative
to the template event. Similar to the inversion strat-
JÖKULL No. 67, 2017 9