Jökull - 01.01.2017, Blaðsíða 9
Zeinab Jeddi et al.
et al., 2003; Sturkell et al., 2008). Árnadóttir et
al. (2009) suggested that the observed horizontal ve-
locities between 1993 and 2004 could be a conse-
quence of rheological structure and fast deglaciation
rates. Similarly, Spaans et al. (2015) explained the
long-wavelength uplift observed with satellite radar
interferometry (INSAR) and GPS measurements from
2001 to 2010 as the response to long-term shrinking
of the ice cap.
SEISMIC NETWORK AND DATA
After the 2010 eruption of Eyjafjallajökull, a tempo-
rary network of nine stations was deployed on and
around Katla volcano in May 2011 until August 2013
to monitor the volcano closely. In addition, the perma-
nent monitoring network around Mýrdalsjökull, run
by IMO, was expanded to 10 stations. Therefore, a to-
tal of 19 seismic stations were operating around Katla
between May 2011 and August 2013 (Figure 2). They
were equipped with broadband or intermediate-band
instruments. Batteries, wind generators and/or solar
panels were used as power sources. Details on the
instrumentation and data recovery are found in Jeddi
et al. (2016). Seismic data were acquired at a 100
Hz sampling rate in continuous mode, but technical
problems, often due to harsh weather conditions, es-
pecially during winter, caused long station downtime
in some cases.
A densified seismic network in a seismically ac-
tive area like Katla improves the detection level of
seismic events. Therefore, we ran an automatic de-
tection algorithm developed at IMO (Stefánsson et
al., 1993; Bödvarsson et al., 1999) using data from
the whole 19-station network. We identified four
main seismic clusters in the data. Three of them, i.e.
the caldera, the west, and south flank, were already
known and included also in the IMO catalog. The
fourth cluster, located on the east flank at the tip of
Sandfellsjökull (eastern Mýrdalsjökull, Figure 2), is
only represented by a few events in the IMO catalog
(only 6 events during 2011–2013). This is likely due
to the fact that the IMO catalog is based on the per-
manent stations only, while our catalog is based also
on the temporary ones. In particular, 2 out of the 3
(LOD, KKE, RJU) seismic stations closest to the east-
ern events are part of the temporary network (Figure
2). Those stations were crucial to improve the detec-
tion of the seismicity in this area.
CHARACTERISTICS OF THE EASTERN
EVENTS
Waveform characteristics
The eastern events are low magnitude (ML<1) events
with a frequency content in the range 4–25 Hz. Both
P- and S-wave arrivals are clear. They can, therefore,
be classified as VT events (McNutt, 2005). Based on
visual examination of the detected events and similar-
ity between waveforms, we identified two families of
events (Figure 3). Even though the frequency con-
tent of the two families is similar, their waveforms
are different, e.g., the amplitude ratio between P and
S phases is very different on different components.
Looking at vertical components of the waveforms, the
P-phase has much lower amplitude than the S in fam-
ily 1, while for family 2 the opposite is the case. This
suggests that the source mechanisms of the two fami-
lies are different.
Temporal evolution
Because the waveforms have similar appearance,
we were able to improve the event detection us-
ing a cross-correlation method. We set up a cross-
correlation scheme between a reference event and the
continuous recordings (see Lindblom et al. (2015)
for a detailed explanation of the cross-correlation
method). The event with highest signal to noise ra-
tio was chosen as a template for each family (family
1: 6th December, 2011 at ∼19:44:42; family 2: 5th
December, 2011 at ∼09:19:34). The waveforms were
filtered between 3 and 20 Hz and a window (∼2 s)
that included both P and S phases was cross-correlated
with the continuous data from July 2011 to August
2013. We ran this process using data from the best
available station at any given time. We used LOD
(where the waveforms have the highest signal to noise
ratio) between September 2011 and August 2013, and
KKE between July and September 2011.
With this process we were able to detect 301
events, 270 belonging to family 1 and 31 to family
4 JÖKULL No. 67, 2017