Jökull - 01.01.2010, Qupperneq 99

Geothermal noise at Ölkelduháls, SW Iceland
The simple structure that is clear in the correlo-
grams in Figure 9 is that along line A the energy is
predominantly at positive time shifts. This indicates a
source to the west of station LA0. The first few sta-
tions along the B line have energy at negative time
shifts indicating a source to the west of LB2. Further
west the energy shifts to positive time shifts indicating
a source to the east of the westernmost sites. In order
to combine the information in all cross-correlations
available from the data we compute the complex en-
velopes of the correlograms and use them as a smooth
measure of energy. We then fill a grid covering hypo-
thetical sources (in the surface) in the vicinity of the
profile with hypothetical slowness or velocity (hori-
zontal) with the stacked correlation envelopes shifted
in time to the appropriate time shift for the particular
geometry and velocity. This procedure searches for
those source locations, combined with velocity, that
best predict the energy in the correlograms of Figure 9
and the 72 other cross correlations that are computable
from the data. The results are shown in Figure 10 for
four different velocities in increasing order from 0.5,
1.0, 1.5 and 2.0 km/s for frames a) to d) respectively.
The maximum correlation stack is about 30, i.e. one
third of the maximum possible value for 90 station
pairs. The minimum stacked correlation is about half
the maximum, i.e. 15. This search procedure does not
find a perfect model, or a perfect distribution of mod-
els, but a crude map of which source locations (and
velocity) are more likely than others. Quantitative cal-
ibration of the stacked correlation in terms of proba-
bility requires statistical justification, which we have
not performed, but we can at least state qualitatively
that those source locations that give a high stack are
in general more likely than others.
We achieve the highest stacked correlation at a
low velocity of 0.5 km/s (Figure 10a). The distribu-
tion of the most likely sources is found 3 to 5 km NW
of Ölkelduháls. This is inconsistent with the ampli-
tude data which would be predicted to vary little along
the entire profile which is of comparable length (4–5
km). At a somewhat higher velocity of 1 km/s (Fig-
ure 10b) the region of high stacked correlation is not
significantly reduced in amplitude, but more confined
in space compared to Figure 10a. The most likely
source locations are found in the immediate vicinity
of Ölkelduháls and slightly to the north. A secondary
localized peak is also found in the vicinity of stations
LA3 and LA4 that we can speculate to be related to
the anomaly A in Figure 7. No localized peak is found
near the SW end of the profile. Figure 10c shows the
stacked correlation for a velocity of 1.5 km/s. In this
case the highest stacked correlation has dropped sig-
nificantly or by about 15%. The best solutions are
localized precisely at Ölkelduháls and then at a lower
level of fit along a NW trending ridge centered on Öl-
kelduháls. With a velocity as high as 2.0 km/s (Fig-
ure 10d) the fit has deteriorated further by 15% and
the best-fit sources are confined to an elongated NW
trending ridge including Ölkelduháls.
On the basis of these calculations and in combi-
nation with the amplitude relations shown in Figure 7
we conclude that the most likely source of the geother-
mal noise measured along our profile is in large part
confined to the Ölkelduháls geothermal activity. We
regard this as a strong argument for association of this
noise in the 3–7 Hz range with geothermal activity.
The results suggest a velocity close to 1 km/s, which
we must interpret as group velocity due to use of the
complex envelope of correlograms. Combined with
snap shots of particle motion and polarization anal-
yses this suggests primarily a surface-wave content
in the noise. The wavelength is then about 200 m
and intra-station distance exceeds a quarter of a wave
length. Array methods are, therefore, poorly suited
for our observational geometry (in addition to the fact
the the profile is roughly linear and thus strongly di-
rectional in its sensitivity). Many questions still re-
main about the detailed characterization of the noise
and its sources. The crude tools available to us with
the limited data from a small linear array will proba-
bly not answer those. But, we believe we now know
what progress does require.
CONCLUSIONS
We have argued, based on amplitude and timing of
energy packets, that the noise in the range between 3
and 7 Hz along the 4–5 km long profile at Ölkelduháls
is generated by the geothermal activity in the region.
The wave field of the noise appears to be dominantly
JÖKULL No. 60 99