Jökull - 01.12.1984, Qupperneq 16

coastal plain below. The lake level may drop 100
m or more during these bursts, or jökulhlaups,
which last from several days to a few weeks. This
part of the Grímsvötn phenomena has been
accounted for in a model by H. Björnsson (1974,
1975). The principal parameters affecting the
behaviour of the jökulhlaups are the subglacial
topography, thickness of ice and the geothermal
heat flux. Less well understood, however, is the
association of volcanic eruptions in Grímsvötn
with the jökulhlaups. According to the compila-
tion of Sigurdur Thórarinsson (1974) evidence
can be found for 21 eruptions in Grímsvötn since
1332, and all of them have been accompanied by
jökulhlaups. Conversely, many but not all the
jökulhlaups have had eruptions associated with
them. This coincidence of eruptions and jökul-
hlaups has led to two hypotheses regarding their
causal relationship:
1. An eruption begins in Grímsvötn triggering
a jökulhlaup by disturbing the release
mechanism.
2. A jökulhlaup begins when the critical lake
level is reached, triggering an eruption by the
sudden release of pressure.
The eruptions have usually been seen several
days after the beginning of the jökulhlaup, even
after the flood has reached its maximum. This
amongst other observations has led several
authors to favor the second hypothesis, i.e. that
the eruptions are triggered by the pressure drop
(see e.g. Thorarinsson 1974). The 1983 eruption
adds a new aspect to this discussion. It is the first
known eruption in Grímsvötn that is not directly
associated with a jökulhlaup, i.e. eruptions can
occur at a high stand of the lake level and without
the triggering effect of a jökulhlaup. The erup-
tion was small, and was never directly observed
from the inhabited lowland. It was discovered
from its seismic signature and observed from
aeroplanes especially diverted to the Grímsvötn
area for that purpose. Before the installation of
seismographs similar eruptions could have occur-
red without being noticed or recorded.
SEISMIC INSTRUMENTS AND DATA
The seismic network in Iceland consists of
about 35 seismograph stations, most of which
have one vertical, short period seismometer with
natural frequency of 3 Hz. Magnification is in the
range 105 — 106 , highest in the frequency band
3—30 Hz. The seismograms are written with pen
and ink on paper, with a time resolution of 90
mm per minute. The direct, visible recording is a
necessary feature of monitoring networks in vol-
canic regions. The data hardly allow frequency
analysis of the seismic signals, but arrival times
can be read with ± 0.1 s accuracy. Many of the
stations in North and South Iceland were instal-
led in 1974 and 1975, lowering the detection
threshold and improving the location accuracy in
Grímsvötn significantly. Another improvement
was achieved in 1976 and 1977 when the stations
AB, MI and KV (Fig.l) were installed in SE-Ice-
land. The detection threshold of located earth-
quakes is now slightly above magnitude 2. A
telemetered station has been in intermittent
operation on the Grímsvötn caldera rim (GF in
Fig.l) since the spring of 1982. It was not in
operation during the period of the present study.
Epicentral locations are done with the location
routine HYPOINVERSE (Klein 1978), using the
velocity structure of the crust and uppermost
mantle under central Iceland derived by Geb-
rande et al. (1980). Standard error of the
epicentral locations is better than 3 km for all
epicenters shown in the maps, most locations are
considerably more accurate. Station corrections
were determined using data from one of the shot
points of Gebrande et al. (1980) in central Ice-
land, 60 km west of Grímsvötn.
Magnitudes of earthquakes reported here are
determined from the duration of the signal, from
the arrival of the P-wave till the amplitude decays
below a certain level. Magnitude scales have
been found by comparison with local magnitude
scales of the station in Reykjavík and the
WWSSN station in Akureyri.
THE LONG TERM SEISMICITY
ANOMALY
Under normal conditions the Grímsvötn vol-
cano does not seem to be particularly active
seismically. During the period 1975—1982, for
which detailed epicentral locations are available,
only 10 earthquakes were located in and near the
Grímsvötn caldera, whereas a few hundred
events occurred in the volcanic areas to the north
and west, such as Bárðarbunga and Hamarinn.
This pattern changed significantly in late 1982
and early 1983. Epicenters in the Vatnajökull
area during the period January 1,1982 to May 27,
1983 are plotted in Fig. 2. In this period, espe-
cially in the last third of it, a large majority of the
14 JÖKULL 34. ÁR