Shear-wave splitting (seismic bi-refringence) is widely observed in all crustal rocks worldwide, whenever there are small earthquakes within the shear-wave window of seismic stations [12,13]. Caused by stress-aligned fluid-saturated grain-boundary cracks and pores, the microcracks are highly compliant to changes in stress, and changes in shear-wave splitting have been seen before four earthquakes worldwide [12,13]. These changes are now seen routinely before earthquakes in the transform zone of the Mid-Atlantic Ridge in SW Iceland .
Figure 9 shows changes in shear-wave splitting from 1st January 1997 to 1st December 1999 at Stations BJA and KRI in SW Iceland. The time-delays between the split shear-waves, normalized to ms/km, are shown in two bands. The middle diagrams show time-delays in Band-1 , which are sensitive to changes in crack aspect-ratios and it has been demonstrated  that the immediate effect of increasing stress is to modify crack aspect-ratios. The upper diagrams show time-delays in Band-2 which are sensitive to crack density. The lower diagrams show the magnitudes of earthquakes within 20km of the stations.
Both bands of time-delays have nine-point moving averages. The time-delays in Band-2 show no identifiable correlation with the earthquakes. The time-delays in Band-1 have least-squares lines added to data where the averages show increases. These least-squares lines begin just before minima of the moving average and end where there is a larger earthquake nearby marked below the diagram. The first four increases at Station BJA show regular behaviour. It was recognized towards the end of October 1998 that another increase had begun. This allowed the successful stress-forecast of the time and magnitude of a M=5 earthquake on 13th November 1998 [15,16] (see the exchange of emails between the University of Edinburgh and the Icelandic Meteorological Office (IMO) listed in Table 1.).
Note that changes in shear-wave splitting do not provide information about the location of the forecast event, but IMO recognised continuing seismicity from a previous event and suggested correctly that the forecast event would occur on a fault close to Station BJA .
Note changes in Band-1 monitor the effect of changes of stress in microcrack geometry. Stress changes before volcanic events also effect shear-wave splitting. The overall decline in time-delays in both bands at both stations follows increases first identified before the Vatnajokull eruption, and is believed to represent the adjustment of the Mid-Atlantic Ridge to the effect of the eruption.
In 1999, changes in shear-wave splitting at BJA do not show the previous regular behaviour. This is believed to be due to volcanic activity at the volcano Katla which erupted in July 1999.
Time delays in Band-1 at Station KRI, show an exceptionally large scatter but there is consistent increase since December, 1998. At one year, this is the longest duration of increase yet observed in Iceland. Assuming that the impending event is close to Station KRI, and assuming linearity, the increase will reach the level of fracture-criticality (11 to 14 ms/km at BJA) between February and July 2000. Note however that linearity cannot be assumed, neither can it be assumed that the increase is due to an impending earthquake, and the size of a volcanic event is probably not directly related to the duration or rate of increasing time-delays. All we can suggest is that a big event (M>=6 earthquake or volcanic eruption) is approaching. Since no clear evidence of an increase can be seen at BJA or SAU this suggests that the event will be probably be at the west end the Gullbringu Peninsula.
Note: Stress-forecasts using small earthquakes as a source of shear-waves can only made in areas of exceptionally persistent small earthquakes. To stress-forecast earthquakes elsewhere would require the development of stress-monitoring sites (SMSs) where shear-wave splitting is monitored in cross-well seismics. This was first suggested in 1998 . The recent commercialization of an effective shear-wave source now allows much cheaper borehole geometries , and the EC has recently funded the development of a SMS near Husavik on the Flatey- Husavik Fault in Northern Iceland .