Jökull - 01.01.2019, Side 77

Pálsson et al.
Figure 4. A simulation of the magni-
tude distribution of two earthquake popula-
tions, one with a conventional Gutenberg-
Richter power law distribution, the other
with a normal distribution. – Hermun á
stærðardreifingu þar sem saman fara tvö
skjálftamengi. Annað fylgir hinu venjulega
lögmáli Gutenbergs og Richters, hitt er
normaldreift.
The time variation of the b-value, as plotted in
Figure 5, shows a systematic behavior. Before the
eruption and collapse of the caldera the seismicity was
moderate and the b-value was close to normal, 0.83
for the period 2010 to 2014. Low values are seen
all through the collapse period, which is arguable be-
cause the distribution is not quite linear, see above.
Immediately following the end of the eruption and
collapse the seismic activity becomes low and the b-
value increases back to normal values. As the seis-
mic activity begins to increase in the fall of 2015 the
b-value begins to decrease, and remains low at the
time of writing (October 2019). The seismic activ-
ity reached a maximum in 2017–2018 and then began
to decline slowly.
DISCUSSION
The b-values determined in this study are mostly
within the range 0.4 to 1.4. Typical values for seismi-
cally active areas are around 1.0, and in volcanic areas
the values are frequently higher (see e.g. Kulhanek,
2005). Such normal and higher values are found in
our study only during the few months of 2015 imme-
diately following the end of the Holuhraun eruption
when the stress in the caldera may be assumed to be
low. At other times the values are between 0.4 and
0.5. Our abnormally low values are hard to explain.
A large majority of the caldera earthquakes occurred
along the northern and southern section of the caldera
fault, as seen in Figure 1. Parts of the caldera ring-
fault appear to have moved aseismically. The very
low b-value may indicate that the stress is concen-
trated and very high on the seismically active section
of the caldera fault.
The steepening of the frequency-magnitude
curves towards higher magnitude, as seen in Figure 2,
is a commonly observed phenomenon and we see this
behavior in our whole data set. In a volcanic area this
may be understood as a result of the finite extent of
the stress field produced by the volcano. The source
of the stress is in processes taking place in the root
of the volcano, and the stress dies out rather quickly
with distance from a point-like source. This limits the
magnitude of earthquakes that can occur and the dis-
tribution curve becomes steep. Our three-segment fit-
ting to the distribution was meant to take this effect
into account.
76 JÖKULL No. 69, 2019