Jökull - 01.06.2000, Blaðsíða 63

Páll Einarsson and Bryndís Brandsdóttir
several times and was found to converge and give
a stable set of corrections. Locations of known ex-
plosions were accurate to within the calculated error.
Horizontal errors are usually of the order of 1-2 km,
vertical errors larger. The data are all consistent with
shallow hypocenters, depth less than 5 km.
Locating earthquake hypocenters is more difficult
in the Mýrdalsjökull area than most other areas in Ice-
land. This is due to several effects:
1. Geometry of the network is unfavorable. Sev-
eral of the nearest stations were arranged in an
almost circular array around the active area with
no stations in the central part (Figure 1). Phases
from the smallest events, the ones that were
only recorded by these stations, arrive nearly
simultaneously at the stations. This geometry
provides very good epicentral location accuracy
that is relatively independent of velocity struc-
ture, but the depth resolution is poor and the
depth error is large. For better depth resolu-
tion it is necessary to include seismic waves
recorded at more distant stations as well as the
near stations. The distant stations record down-
diving rays, i.e. rays that leave the source below
the horizontal, whereas the near stations record
rays from the upper focal hemisphere.
2. The crust in the area is heterogeneous. Travel
time residuals are unusually large resulting in
high RMS values for the final hypocentral loca-
tions. Part of this problem is solved with station
corrections, but large deviations remain.
3. Seismic waves are greatly attenuated beneath
Mýrdalsjökull. Waves from events in the west-
ern part are well recorded at stations west of
the glacier but poorly at stations east of it, and
vice versa. P-waves are usually quite emergent
as well and we have reason to believe that the
first arriving waves are often missed, particu-
larly for small events. S-waves are also poor
or absent for a large proportion of the events.
These characteristics are more pronounced for
events beneath the western part of the glacier
than for the eastern events.
EPICENTRAL DISTRIBUTION
The difficulties described above resulted in a rela-
tively large proportion of events that did not fulfill the
quality criteria for this study and were thus rejected.
The epicentral map shows two clusters of epicenters
separated by an area of considerably lower activity
(Figures 2 and 3). The centers of the two clusters are
close to 11 km apart. A total of 385 events could be
located with accuracy better than 5 km in the hori-
zontal direction during our study period of which 286
events were located within the eastern cluster and 99
events originated within the western cluster. The east-
ern cluster is larger and extends 10-11 km N-S and 7-8
km E-W. It is located within the Katla caldera, approx-
imately in the area of P-wave delay and high S-wave
attenuation found by Gudmundsson et al. (1994). The
western cluster is centered 2-3 km west of the western
caldera rim, at Goðabunga. It extends approximately
8 km NE-SW and 4 km NW-SE. The annual seimic-
ity within the two clusters is highly variable (Figures
2 and 4) and does not seem to be connected in any
way. Some years the activity is similar in both clus-
ters whereas higher activity in one cluster during other
periods may coincide with lower activity in the other
cluster. A higher number of earthquakes were located
within the Katla region in 1978, 1979, 1980, 1982,
and 1983, whereas more earthquakes were located in
the Goðabunga region in 1981, 1984 and 1985.
Low-frequency, emergent P-waves, and unclear
S-waves are more pronounced in the Goðabunga
cluster than the Katla caldera cluster (Figure 5).
These are usually considered as characteristics of vol-
canic earthquakes (e.g. Chouet, 1996), reflecting slow
source processes, high attenuation or inhomogeneous
velocity structure. In this sense the Goðabunga earth-
quakes look more "volcanic" than the earthquakes
within the Katla caldera.
A few earthquakes were located near the center
of the Eyjafjallajökull volcano, west of Mýrdalsjök-
ull, during our study period. Only three of those
events fulfilled our location criteria (Figures 2 and
3). They all occurred in 1979. The seismic activity
of this volcano was very low, but noticeable. Peri-
ods of increased activity have been documented, how-
62 JÖKULL No. 49