Temporal variations in time-delays

Polarization directions and time-delays were measured during March 1998 - June 2000. The suitable events recorded within the shear-wave window (station-to-epicenter distance less than hypocentral depth) ensures that the shear-waves are not distorted by surface conversions. This constrains the number of events that can be used for shear-wave splitting analysis. There is also need for sufficient activity to show temporal variations. Other constraints include restrictions in focal depths, location errors, and deviations from the mean polarization direction. In the PRENLAB-2 period, these criteria were fulfilled mainly for station BJA, SAU and KRI. Variations in time-delays for the above stations are shown in Figures 10a, 10b and 10c. At each station the mean polarization is calculated and time-delay measurements with polarizations within a standard deviation of this direction are selected. There were 4230 observations at BJA, 5145 at KRI and 44 at SAU. The time-delay measurements are normalized over straight-line path distance and separated into two bands defined by incidence to the vertical plane of symmetry parallel to the mean polarization direction (the average strike of aligned near-vertical cracks). Band-1 ($15^\circ$-$45^\circ$) is more sensitive to changes in crack aspect-ratio, the result of gradually increasing stress. Band-2 ($0^\circ$-$15^\circ$) is more sensitive to crack density, which does not vary consistently with increasing stress.

Figures 10a, 10b, and 10c show data from March 1998 to June 2000. The middle cartoons show nine-point moving averages through the time-delays in Band-1. The least-squares lines begin before a minima in the nine-point moving average and end when a large earthquake occurs. This is followed by an abrupt decrease in time-delays. The upper cartoons show nine-point moving averages through the time-delays in Band-2. No consistent pattern or relationship to earthquakes can generally be seen in Band-2. The lower cartoons show the magnitudes of all M$\ge$2 earthquakes within 20 km of the station.

Since 1996, station BJA displayed a relatively simple relationship between the magnitude of an earthquake and the duration (and slope) in the increase of time-delays before the event: the magnitude was proportional to the duration of the increase, and inversely proportional to the slope (Volti and Crampin 2000). In Figure 4a, Band-1 of BJA, variations before two large earthquakes (M=5.1 and 5) are indicated. They both show similar patterns of increase and reach maximum time-delays up to 14 ms/km after an increase of about four months. The second event was stress-forecast (see Subsection 3.3.1.2).

(Note that earthquake magnitudes, written as M, are local magnitudes approximately equivalent to body-wave magnitudes, mb, where in this magnitude range M=5 is approximately equal to surface wave magnitudes, Ms=6).

Three months later, there was a M=4 earthquake, with longer duration and smaller slope from that expected from previous events. The time-delays do not drop immediately afterwards, but show a slight increase. Two months later, July 1999, a volcanic event in Katla took place, after which time-delays started to drop gradually. Also, in contrast with the previous earthquakes, a similar pattern exists in Band-2. In February 2000, there was a further eruption at Hekla. It appears that magmatic activity complicates the behaviour of the stress-field, so the simple behaviour before earthquakes in 1997 and 1998 is no longer apparent in 1999 and 2000 (Volti and Crampin 2000).

In Figure 10b, the middle cartoon shows time-delays variations in Band-1 at station SAU. The nine-point moving averages in Band-1 show several broad maxima, one during 1998, three during 1999 and two in 2000. The first can be associated with the M=5.1 event (which is 43 km from SAU). Although during 1999, the peak in March does not seem to correlate with any nearby activity, the other, in July, may be correlated with the event in Katla (75 km SW of SAU). There are two earthquakes of M=4.2 and M=3.8 in September, 1999, but the time-delays do not show any increase. Again, the complications are believed to be due to the combined effects of the build-up of stress before earthquakes and the movement of magma before eruptions.

Recently, (June 17 and 21, 2000) three large earthquakes of M=5.6, M=5 and M=5.3 (Ms$\cong$6.6, M=6, and M=6.3) occurred near SAU. These are the largest earthquakes in Iceland since 1963. Time-delays increased before the events for about four months, with the rate of increase being almost consistent with a M=5.6 earthquake, but the duration was too short. The complication is that there is a period of about seven weeks without source earthquakes to monitor the rockmass. As a consequence the start of the increase in time-delays in Band-1 was not recognized, and the events were consequently not stress-forecast (Volti and Crampin 2000). Such irregularities indicate the need for controlled source seismology, as is proposed in the SMSITES project, even in highly seismic SW-Iceland.

Figure 10c shows variations in station KRI. The there are two increases in time-delays. The first is associated with the M=5 event, November 13, 1998, which was stress-forecast. The second increase is gradual but steady increase in time-delays in Band-1 since December 1998 which may indicate the build-up of stress, and subsequently a future large earthquake.

 

Figure: Shear-wave splitting at BJA (a), SAU (b) and KRI (c) from March 1, 1998 to June 31, 2000. The upper and middle cartoons show variaton of normalized time-delays with time, for ray paths in bands with incidence $0^\circ$ to $15^\circ$ to the crack face (Band-2) and with incidence $15^\circ$ to $45^\circ$ to the crack face (Band-1). Band-2 is sensitive to crack density whereas Band-1 is sensitive to aspect ratio. There are nine-point moving averages through the time-delays. Lines are least square fits to data before a major earthquake. The lower cartoons show the magnitudes of $M\geq 2$earthquakes within 20 km of each station.