Jökull - 01.06.2000, Page 62

Earthquakes in the Mýrdalsjökull area, 1978–1985
3 km (Gudmundsson et al., 1994). The Eldgjá fis-
sure to the north is structurally connected to the Katla
volcano (Robson, 1956; Þórarinsson, 1975), and the
similarity of the chemistry of Katla and Eldgjá lavas
led Sigurdsson and Sparks (1978) to suggest that Eld-
gjá eruptions were fed by lateral flow from the Katla
magma chamber. Several volcanic ridges radiating W,
NW, N, NE and E from the Katla caldera (Björns-
son et al., 1994; this issue), as well as postglacial
craters and small fissures 15-30 km northwest, north
and northeast of the caldera (Jóhannesson et al., 1982)
were most likely fed by lateral intrusions as well. The
Eyjafjallajökull volcano to the west of Katla has a dis-
tinct, E-W striking fissure swarm that merges with the
radial fissure system of Katla. Hence, the volcanoes
appear to be tectonically connected.
Katla eruptions during the last 1100 years are
rather well known from historical records and tephro-
cronological studies (Þórarinsson, 1975; Larsen, this
issue). Twenty eruptions occurred during this time,
the latest in 1823, 1860 and 1918. These were rel-
atively large eruptions, accompanied by substantial
jökulhlaups from the glacier. The eruptions of the
last three centuries have followed a remarkable pat-
tern, occurring around the 20th and the 60th year of
each century with only small deviations (Thorarins-
son, 1960). A large eruption expected around 1960 on
the basis of this pattern has not occurred yet. Instead,
a small subglacial eruption appears to have occurred
in 1955 as suggested by Tryggvason (1960) based
on earthquakes recorded prior to a small jökulhlaup.
No eruption was visible but two subsidence cauldrons
were formed in the ice near the eastern caldera fault.
Similarly, a small and sudden jökulhlaup from the
Sólheimajökull glacier in July 1999 (Sigurðsson et al.,
this issue) is suspected to have been caused by a very
small, subglacial eruption near the southern caldera
rim of Katla (Einarsson, 2000).
The most dramatic evidence for prehistoric activ-
ity of Katla is a layer from about 12.000 years B.P. in
the GRIP ice core, thought to correspond to Ash Zone
1 in Atlantic sediment cores and a deposit formed by
lahars, ash fall and surges on the south flank of the
volcano (Grönvold et al., 1995, Lacasse et al., 1995).
Ash layers from Katla have also been identified in the
GRIP core at 75.400 and 77.500 ice core years (Grön-
vold et al., 1995). Thus it is evident that the Mýrdals-
jökull volcanoes impose a significant hazard in south-
ern Iceland. A violent Plinian eruption is also a threat
to international air traffic in the N-Atlantic region.
The high seismicity of the Katla region has been
known for some time. Prior to instrumental obser-
vation, earthquakes were known to accompany the
initial phase of Katla eruptions. Shocks have been
felt in the neighboring areas 1-7 hours prior to all
eruptions since 1625 (Thoroddsen, 1899; Þórarinsson,
1975; Björnsson and Einarsson, 1981). Tryggvason
(1973) found indications that the Mýrdalsjökull seis-
micity increased in 1952. He also discovered a pro-
nounced annual cycle in the frequency of earthquakes
during the period 1952-1958. A large majority of all
seismic events occur during the latter half of the year.
Brandsdóttir and Einarsson (1992) confirmed this pe-
riodicity for a later period and showed that it is mainly
caused by activity in a separate epicentral cluster near
the western margin of, but outside, the Katla caldera.
In this paper we discuss the hypocentral distribution
in the area and argue that the annual cyclicity is most
likely caused by pore pressure fluctuations in the brit-
tle crust beneath the glacier.
DATA AND DATA ANALYSIS
Data for this study come from the Icelandic Seismo-
graph Network. All stations in this time interval (Fig-
ure 1) had analog recording and absolute timing with
time resolution better than 0.1 s. The network and sta-
tion locations are described further by Einarsson and
Björnsson (1987).
The location program HYPOINVERSE (Klein,
1978) was used for the locations. The velocity model
consists of horizontal layers of constant velocity gra-
dient. An average crustal velocity structure based on
the RRISP-profile (Gebrande et al., 1980) and other
data was used. Station corrections were determined
by successive approximations. A set of well recorded
events was located and the average residuals for each
station used as a station correction for the second run
of the location program. This process was repeated
JÖKULL No. 49 61