Jökull - 01.01.2015, Qupperneq 53

Seismicity beneath Þeistareykir, NE-Iceland
Drouin et al., 2013, Metzger et al., 2013), but more
recently explained by a combination of processes in-
cluding plate spreading, viscoelastic relaxation, and
inflation/deflation of shallow magma chambers be-
neath Þeistareykir and Krafla (Ali et al., 2014). The
base of the fault locking zone, sometimes equated
with the seismogenic zone within which interseismic
strain accumulates (Smith-Konter et al., 2011), is es-
timated from geodetic data to be at 5.1–9.5 km depth
(Metzger et al., 2011; Ali et al., 2014).
Using a short-term seismic network of six stations
Vogfjörð (2001) located several microearthquakes
along a N-S line beneath northern Bæjarfjall at less
than 3 km depth. Most of the events originated
at 0.5–2.1 km depth. An increase in seismicity at
Þeistareykir in 2007–2010 (Hjaltadóttir and Vogfjörð,
2011), coincided temporally with the local geodetic
uplift signal. Local magnitudes (ML) ranged from
-0.6 to 3.2. Shallow earthquakes were found to mostly
follow the geothermally altered surface manifesta-
tions suggesting that active underlying faults facilitate
the transfer of geothermal fluids.
Exploration wells have been used to build a hy-
drothermal model of the upper 2.8 km in the past
decade. Geothermal fluids beneath well ÞG-2 (Fig-
ure 1b) originate from a deep (1600–1700 m depth)
steam-dominated aquifer at ≥300◦C, where fluids
propagate upwards along fractures and partly con-
dense (Ármannsson, 2014). Beneath the vertical well
ÞG-1, surface steam comes from a shallow (600–
700 m depth) water-dominated aquifer at 200–280◦C,
which is fed in turn from the deeper steam-dominated
aquifer. The deepest wells exhibit maximum geother-
mal fluid temperatures of 380◦C (Sveinbjörnsdóttir et
al., 2015) pointing towards an existing heat source
within the upper crust.
Transient electromagnetic (TEM) and magneto-
telluric (MT) data inversions (Yu et al., 2008; Karls-
dóttir et al., 2012) outline a low resistivity cap at 400–
800 m depth north of and slightly deeper beneath Bæj-
arfjall. Low resistivity caps overlying high resistivity
cores are characteristic structures in high-temperature
geothermal systems in Iceland correlating well with
the dominant alteration minerals of low and high tem-
perature systems, respectively (Árnason et al., 2008).
Thus, the TEM/MT results appear to coincide well
with the hydrothermal model based on well informa-
tion. Two distinct low-resistivity anomalies are found
in the southeastern part of Bæjarfjall at 5–6 km depth
b.s.l. and northeast of Bæjarfjall (beneath Ketilfjall)
at 2.5–12 km depth b.s.l. and are interpreted as up-
flow zones of heat into the geothermal systems above
(Karlsdóttir et al., 2012).
Petrological analyses of erupted lava (Borgar-
hraun flow) suggest that crystallization of basalts oc-
curred over a range of depths beneath Þeistareykir,
where primitive basalts started crystallizing at the
base of the crust (20–30 km depth) or uppermost man-
tle and more evolved basalts formed in the shallow-
mid crust (Maclennan 2008; Winpenny and Maclen-
nan, 2014).
SEISMIC DATA PROCESSING
AND RESULTS
A temporary seismic array of up to 31 three-
component seismometers recorded the seismicity
within the northern NVZ in the period August 2009 to
July 2012. The array was operated by the University
of Cambridge with additional stations from the IMO
and covers the Krafla and Þeistareykir geothermal sys-
tems (Figures 1a and 2c). A coalescence microseis-
mic mapping algorithm (Drew et al., 2013) was used
for automated event detection, where a short- to long-
term average onset function is continuously computed
from the horizontal and vertical amplitude compo-
nents. The magnitudes are then back-migrated into
the subsurface forming ‘coalescence’ functions with
peak values indicating likely hypocenter locations and
origin times. The Krafla starting model (Schuler et al.,
2015) was used to estimate the travel times and ini-
tial hypocenter localizations. Of the 5338 identified
earthquakes within the Þeistareykir area, we removed
events that had an average signal-to-noise ratio < 2.5
at a minimum of five stations. P- and S-wave arrivals
times of 199 events were then manually repicked and
picking errors (0.01 s, 0.02 s, 0.05 s, 0.1 s, or 0.2 s)
assigned to each by visually inspecting the change in
amplitude exceeding the background noise as well as
change of signal frequency.
JÖKULL No. 65, 2015 53