Tasks 2, 3, and 4: Cross correlation of logs of the same type from different campaigns and earlier loggings; comparison of changes in logs of different type; comparison of changes in logs with changes in seismicity, etc.

Due to a delay in the year 2000 logging campaign, we had the chance to take at least some data after the two June 2000 earthquakes in the SISZ. Comparing the data of July 1996 to October 1999 with those of August 2000 gives the opportunity to see which parameters have changed across the time from before the event to 6 weeks after.

And there were remarkable changes at the borehole. Our coworker Steinar Þór Guðlaugsson reports: "Water started to flow from well LL-03 on July 20-22 after a rapid rise in the water table from a depth of several meters since the earthquakes. This was probably caused by changes in crustal stress resulting from the earthquakes, although reinjection into the Laugaland geothermal field, which was initiated recently, may also have played some role. On August 3, the flow-rate was 0.5 l/s." The effects seen in the logs will be described below.

Repeated logs in LL-03: borehole temperature
The longest time series of logs is available in the temperature logs (BHT). There are data from the time of the drilling of the well (1977), later from 1980 and 1992, and finally from 1999 and 2000 taken during the work in the PRENLAB projects. The temperature logs are plotted together in Figures 13 and  14. Following a heating phase after drilling, the temperatures went down. The most recent of the logs predating the earthquakes show a regime of inflow into the well at several feed-points and downward flow of this fluid to a depth of 830 m where it re-enters the formation. The new log of August 2000 can be interpreted in two different ways: (1) The fracture at 830 m now feeds water into the well in addition to the other feed-points and above this depth water everywhere flows upwards; (2) Water now flows upwards above the feed-point at 120 m, whereas below this depth water continues to flow down the well to a depth of 830 m as before. In either case, it is evident that water entering the well at feed-points in the 200-250 m depth range is warmer than before. Temperature differences between the feed-points in this depth range also seem to have increased. At present it is unclear whether this is caused by changes in the fracture network feeding this interval or by the reinjection.

In conclusion, a comparison of the August 2000 temperature log with the earlier ones does not reveal any new water-conducting fractures (feed-points), but does show changes in the temperature and flow rate of the water flowing through previously existing fractures. These changes may probably be explained by a combination of stress change (general increase in formation pressure) associated with the earthquakes and reinjection, but do not rule out movements on some of the fractures feeding the well.

  

Figure 13: Borehole temperature logs since the drilling of well LL-03. Numbers following "T" give date of log.

  

Figure 14: Borehole temperature logs since the drilling of well LL-03. Numbers following "T" give date of log.

Repeated logs in LL-03: latero log and "normal" resistivity tools, self potential
The latero log (the one included in the GFZ DIL tool is of LL3 type) and the "short normal" resistivity tool of OS (16" electrode spacing) are designed for shallow penetration into the formations. Whereas the "induction log medium" and "induction log deep" of the DIL tool and the "long normal" resistivity tool (64") are dedicated for investigations deeper into the rock surrounding the borehole.

The latero log resistivity shows good repeatability (see red curves in Figure 15). So do the 16" and 64" tools, as will be discussed later. Figure 16 gives a comparison of LL3 logs from 2 different campaigns with logs of the short and long normal resistivity tools. The LL3 curves correlate very well besides a small depth shift. They also correlate qualitatively well with the other two logs, but there seems to be a calibration problem in the LL3, leading to much lower values than the other tools. We did not care about this, as only changes in the logs were of interest here. The match between the normal resistivity tools is expected to be moderate as their penetration depths are different. There was no indication of changes in LL3 logs before June 2000.

  

Figure 15: Data example showing from left to right: gamma-ray log in API units, compensated travel time in microseconds per foot and the sonic amplitude, and three repeated measurements of latero log (LL3), induction log of medium penetration (ILM) and induction log of deep penetration (ILD). Month and year of campaign are given.

  

Figure 16: Data example showing latero log data (LL3) from July 1996 and December 1997 together with short (SN) and long normal (LN) resistivity data from October 1999.

The resistivity logs from before and after the June 2000 events are plotted together in Figure 17. Two kinds of adjustments were made to the final logs before they were plotted: Firstly, a correction was applied to the logs based on a calibration carried out at the surface with a set of known resistances. Secondly, the origin of the August 2000 neutron-neutron log was taken as a common depth reference and the logs shifted to this datum.

The new logs agree remarkably well with those obtained in 1999. The main change observed is a decrease in the 16" resistivity log above 110 m. This can probably be explained by the increased temperature (and - less likely - salinity) in this depth range caused by the newly established upflow. Too much should not be read from the differences between the 16" logs in the 400-700 Ohmm range, because of the large correction that was applied to the 1999 log as a result of the surface calibration. The differences may simply reflect the inaccuracy of the correction. Keeping this in mind, the only significant difference between the 16" resistivity logs that may require other explanations seems to be a decrease in resistivity in the 465-470 m depth interval.

The main change in the 64" resistivity log is a decrease in resistivity from 150-155 m. This may or may not be associated with the upflow. We also note a slight decrease in the 465-470 m range which correlates spatially with the main anomaly in the 16" log. At the same depth, the spontaneous self potential log (SP) of October 1999 shows an anomaly too (cf. Figure 18), which could indicate a crack, existing already at that time (there is no SP log of August 2000). A possible interpretation of the 3 correlating signals might be a crack that has widened but does not feed-in water with a temperature different from the local temperature in the borehole.

The long normal tool shows variations between October 1999 and August 2000 also at depths other than 465 m and at resistivities lower than 400 Ohmm. They usually display lower values in August 2000, which could mean that water filled cracks have opened. However, none of the variations is also seen in the "short normal" log, which leads to doubts on real changes in rock physical parameters, as new cracks could be formed much more easily near the borehole wall, an area sampled by the 16" tool. See also the discussion on the neutron-neutron log.

  

Figure 17: Data example comparing short (SN) and long normal (LN) resistivity data from October 1999 and August 2000.

  

Figure 18: Data example comparing short (SN) and long normal (LN) resistivity data from October 1999 and August 2000.

Repeated logs in LL-03: dual-induction log (deep and medium penetration)

The repeated measurements with the dual-induction log (deep and medium penetration) show variations in resistivity even between different logs of one day, therefore these values cannot be used for an analysis of changes. More details can be found in the first annual PRENLAB-2 report.

Repeated logs in LL-03: sonic logs, P-wave travel times

The difference in the sonic velocity of logs measured at one day is in average $3.0\%$ of the average value (Figure 19). At some depth intervals greater differences between repeated logs are caused by inaccurate depth matching, e.g. at 740 m in Figure 19. Depth intervals where the amplitude of the measured signal is very low showed significant variation in the (automatically) picked travel times, thus cannot be considered for an analysis of changes. Greater variations between logs of different logging campaigns could not be found.

  

Figure 19: Data example showing from left to right panel: an average curve of sonic amplitude, a superposition of the average compensated travel times measured in October 1996 and in December 1997 (each averaged over four runs performed immediately one after the other), the difference between these two average travel time curves (absolute value, green curve) and the standard deviation (absolute value) derived from the four runs performed in December 1997. Compensated travel times are given in microseconds per foot. The difference between the average travel times exceeds the standard deviation only in low amplitude intervals (not shown here) or at depths with inaccurate depth-matching (for example around 740 m).

Repeated logs in LL-03: neutron-neutron logs

Neutron-neutron logs were run before and after the June 2000 earthquakes. In the environment of borehole LL-03 at Nefsholt they are expected to reflect mainly the water content of the rock. They are plotted together in Figures 20 and 21. The one of April 1999 was measured with a log speed of 40 m/min. Those of October 1999 and August 2000, both were measured with a logging speed of 6 m/min. The April 1999 and the August 2000 logs were measured at 0.5 m depth interval. The 1999 log was measured at a depth interval of 0.1 m. The depth scale was shifted as described above.

Small changes can be found as well as between logs of April and October 1999 as between those of October 1999 and August 2000. The differences in logs of 1999 are larger than between those of October 1999 and August 2000. This might be due to the comparatively high logging speed in April 1999. At the depth of 465 m the neutron logs do not show a difference, as would be expected if a crack would have widened there. Generally, the match between the logs is remarkably good - so good that all of the observed differences can probably be attributed to statistical fluctuations.

  

Figure 20: Data example comparing short (SN) and long normal (LN) resistivity data from April 1999 and October 1999.

  

Figure 21: Data example comparing short (SN) and long normal (LN) resistivity data from October 1999 and August 2000.

Borehole-televiewer measurements

Nefsholt:

In drillhole LL-03 (63.92$^\circ$N, 20.41$^\circ$W), the orientation of the borehole breakouts found is about 120.5$^\circ$ to 126$^\circ$, implying a direction of the maximum horizontal principal stress of N30.5$^\circ$E to N36$^\circ$E. The circular standard deviation (1$\sigma$) is of the order of 10$^\circ$. The stress direction is in agreement with the expected stress direction from large-scale tectonics (left-lateral strike-slip). The length of picked breakouts sums up to approximately 5.5 m. They are supposed to have formed while drilling, i.e. in 1977.

Þykkvibær:

Borehole breakouts have been found in the depth intervals 925 m to 927 m and 937 m to 941 m in THB-13 (63.77$^\circ$N, 20.67$^\circ$W). The breakouts in these two depth intervals sum up to approximately 3.5 m. Data quality is rather poor due to weak reflection amplitudes. Figure 22 shows a data example. The average breakout azimuth is between N105$^\circ$E (upper depth-interval) and N121$^\circ$E (lower depth-interval). Statistical analysis over the whole depth range gives a breakout-orientation of N111$^\circ$E with a circular standard deviation (1$\sigma$) of about 10$^\circ$. This would mean that the larger principal horizontal stress is in average oriented N21$^\circ$E. This is in agreement with the stress directions expected from the overall tectonics as well as with the results from Nefsholt. The breakouts in THB-13 have most likely formed during drilling in 1997, i.e. just before they were measured.

Böðmóðsstaðir

No breakouts have been observed in borehole BS-11 (64.20$^\circ$N, 20.55$^\circ$W), but there are vertical fractures visible between 713 m and 934 m depth. The length of the vertical fractures sums up to 45 m. Vertical fractures are expected to occur in the direction of the maximum horizontal principal stress because of tensile failure of the borehole wall. They are supposed to be not of natural origin but drilling induced; so they stem out of 1992. These fractures occur at an azimuth of N45$^\circ$E to N90$^\circ$E. A data example is presented in Figure 23.

Breakout orientations and fracture statistics from the measurements in the SISZ are shown in Table 5 and Figure 24 as an overview. In Subpart 7B stress orientations were calculated for the SISZ using the co-seismic stress release of large earthquakes in the SISZ since 1706 and taking into account the stress build-up by plate motion. For comparison, we give the values for the boreholes where measurements were performed in Figure 25. Both results show stress orientations similar to each other. In general, the modelled values are more oriented to E-W than the measured ones.

 

Table 5: Stress orientations found from borehole-televiewer logs.

Well:	Logged	Interval with	Total length	Orientation	circ.
	interval:	BOs / vert.	of BOs/ vert.	of $\sigma_H$:	std.
		fractures:	fractures:		devi.:
BS-11	703-1090 m	713-934 m	45.0 m fract.	N45$^\circ$E-N90$^\circ$E	--
LL-03	80-1100 m	780-983 m	5.0 m BOs	N30$^\circ$E	12$^\circ$
THB-13	466-1225 m	925-941 m	3.5 m BOs	N21$^\circ$E	10$^\circ$

 

  

Figure 22: Example of the borehole breakouts found in well THB-13 with two cross sections. The two panels show the amplitude of the reflected signal (left) and the radius calculated from the travel time (right), both unwrapped from north over east, south, west to north. Vertical axis: depth in meters. Breakouts appear as vertical bands of low reflection amplitudes. Due to poor reflection amplitudes, the values for the radius are often missing in these parts, resulting in black bands. In the two cross sections, the black lines indicate the range in azimuth of the picked breakouts.

  

Figure 23: Example for drilling induced vertical fractures observed in borehole BS-11. The data are displayed as described in Figure 22. The fractures appear as narrow vertical stripes of low reflection amplitude, much narrower than breakouts. Values for the radius calculated from travel time are missing for these stripes due to low amplitudes.

  

Figure 24: Stress orientations found in boreholes in the South Iceland seismic zone. Long red lines indicate the orientation of the larger, short ones that of the lower principle horizontal stress.

  

Figure 25: Stress orientations calculated in Subpart 7B for borehole positions in and near the SISZ.

Summary:

The results can be summarized as follows:

$\bullet$

The measurements at LL-03 for changes in P-wave velocity with the sonic log, for porosity changes with the neutron-neutron log, and for resistivities with the latero log, 16", and 64" show good repeatability.

$\bullet$

The earthquakes in June 2000 happened two weeks before the end of PRENLAB-2 project. Therefore, the last logging campaign before the events took place 8 months before the events. No clear indication for changes before the events was found. This might be, because the time difference was too long or the signals were below the detection threshold of the tools used.

$\bullet$

Coseismic changes might be visible at about 465 m depth, where both the 16" and the 64" log show a decrease in resistivity. It would be interesting to check whether cracks can be found there and at the other feed-points (visible in the temperature logs) with a high resolution borehole televiewer.

$\bullet$

Non-vertical fractures found down to nearly 1100 m depth show the same dominant strike as those observed at the surface (see first annual PRENLAB-2 report). A dependence of fracture orientation upon depth or upon geographical latitude could not be found.

$\bullet$

The stress orientations found at all three locations are similar and agree with a left-lateral strike-slip regime. They are based on data that are assumed to originate from 1977, 1992 and 1997 and are not perpendicular to existing ruptures found in the SISZ and, therefore indicate, that the SISZ is not a weak fault, as postulated for the San Andreas fault. Instead, it appeared prepared for another earthquake. From a rock mechanics view the stress directions found at LL-03 and THB-13 correlate with N-S striking faults, as they are found in the SISZ. On the other hand the orientation of maximum horizontal principal stress found at BS-11 correlates with the model of an E-W striking strike-slip fault zone. A linear dependency of stress orientation on age, depth or geographical position (in-/outside the SISZ) of the boreholes could not be found. Stress orientations similar to ours were also found by Stefánsson et al. (1993) from fault plane solutions. This was confirmed by other groups in the framework of PRENLAB as by Bergerat and coworkers (Subproject 6), who found a mean orientation of $\sigma_H$ of N56$^\circ$E derived from 1916 fault plane solutions selected from 48669 earthquakes in the SISZ of the years 1995 to 1997 (Angelier et al. 2000). A NE-SW orientation of $\sigma_H$ was also the result of investigations of Crampin and coworkers on shear-wave splitting due to stress anisotropy at four of six seismic stations in the SISZ (Crampin et al. 1999) (Subproject 3).

Participants:
Besides the proposer and the subcontractor, the following scientists and technicians have supported the Subproject:

G. Axelsson (OS), H. Bäßler (Karlsruhe), C. Carnein (GFZ), E.T. Elíasson (OS), H.-J. Fischer (GFZ), P. Fleckenstein (GFZ) S.Þ. Guðlaugsson (OS), K. Henneberg (GFZ; now at PGS, Oslo), G. Hermannsson (OS), M. Hönig (GFZ), G. Kurz (GFZ; now at NLfB-GGA, Hannover), Fernando J. Lorén Blasco (GFZ), S. Mielitz (GFZ), H. Sigvaldason (OS), Ó. Sigurðsson (OS), and B. Steingrímsson (OS).