Subpart 5A: SAR interferometry study of the South Iceland seismic zone

The objective was to measure ongoing crustal deformation in the South Iceland seismic zone (SISZ) and relate it to distribution of faults and seismicity there. We have met that objective at the western edge of the SISZ, around the Hengill volcanic center, and published the results in an international, peer-reviewed scientific journal (Feigl et al. 2000).

We have analyzed synthetic aperture radar (SAR) images acquired by the ERS-1 and ERS-2 satellites between July 1993 and September 1998 using interferometry. In spite of our careful image selection, correlation is poor in the relatively flat and wet lowlands of southern Iceland, which unfortunately includes most of the faults in the SISZ. On the other hand, coherence remains good, even after 4 or 5 years, in the mountainous areas around Hengill.

The predominant signature in all the interferograms spanning at least 1 year, is a concentric fringe pattern centered just south of the Hrómundartindur volcanic center (Figure 26). This we interpret as mostly vertical uplift caused by increasing pressure in an underlying magma source. The volume source that best fits the observed interferograms lies at 7$\pm$1 km depth and remains in the same horizontal position to within 2 km. It produces 19$\pm$2 mm/year of uplift. This deformation accumulates as elastic strain energy at a rate 2.8 times the rate of seismic moment release.

Under our interpretation, magma is injected at 7 km depth, just below the seismogenic zone formed by colder, brittle rock. There, the inflation induces stresses that exceed the Coulomb failure criterion, triggering earthquakes. Accumulated over 5 years, the deformation increases the Coulomb failure stress by >0.6 bar in an area that includes some 84% of the earthquakes recorded between 1993 and 1998 (Figure 27 and Figure 28).

Our model suggests that magmatic inflation can trigger earthquakes, with stress rising slowly to failure and then dropping instantaneously in an earthquake. Thus a plot of stress as a function of time on a given fault forms a sawtooth pattern. Prior to an earthquake, on the leading edge of the sawtooth, the stress increases at a rate of the order of $\sim 1$ bar/year. After accumulating for a time interval $\Delta t$ years, the stress then decreases abruptly in an earthquake with stress drop $\delta \tau$. For the magnitude 5.2 earthquake of June 4, 1998, we take a mean stress drop of the order of $\tau \sim 20$ bars, assuming that $\mu =$33 GPa and $\delta \tau = \mu U ( L W )^{1/2}$. If this rupture returns the state of its stress to its initial level, then the accumulation interval is of the order of $\Delta t \sim 20$ years. If this process is cyclical, then this interval is the recurrence time of a characteristic earthquake. It suggests that inflation of a magma chamber can furnish the primary driving force to actually break rocks on a fault in an earthquake.

  

Figure 26: Observed interferograms of the Hengill area for four different time intervals. The time interval appears in the upper right corner of each panel, the orbit numbers appear at lower left, and the altitude of ambiguity h_a appears at lower right. One fringe represents 28 mm of range change. Two concentric fringes are visible in the 4-year interferogram (d) indicating at least 6 cm of uplift between August 1993 and August 1997. The white box in (d) includes the discontinuity enlarged in (b) and (c).

  

Figure: (a) Structural map showing V_P/V_S wave speed ratio at 4 km depth (Miller et al. 1998); our best fitting volume source (crossed square); contours of P-wave speed velocity anomaly $\delta V_P = +1 \%$ at 4 km depth (dashed white lines); contours of Miller et al. (1998); limits of volcanic system (solid black lines); outlines of three central volcanos (dashed black lines). Circles and dots show earthquake hypocenters for events between 1993 and 1998 (Rögnvaldsson et al. 1998a; Rögnvaldsson et al. 1998b). The largest circle is the M_w=5.2 event of June 4, 1998. Events with magnitude smaller than 2.5 plot as dots. (b-e) Enlargement of two interferograms that span August 1995 [b) and c)], and two which do not [d) and e)]. A discontinuity is clearly visible in b) and c) and is interpreted as a fault rupturing. We cannot discern a discontinuity in Figures d) and e).

  

Figure: (a) Coulomb failure stress change on optimally oriented vertical faults shown in a horizontal slice at 4.3 km depth. White color denotes where the stress increase is >1 bar but <10 bar. Source parameters include: latitude $64.032^\circ$N, longitude   $21.213^\circ$W, depth  7.0 km, tensile  opening 12.1 m on each of three dykes with length L=1 km and width W=1 km; the apparent coefficient of friction $\mu '_f$=0.4, the azimuth $\phi$=$70^\circ$ (measured clockwise from north) of the most compressive principal value of the stress tensor, and its magnitude $\vert\sigma _1\vert$=1 bar, and the two Lame' constants $\mu$=$\lambda$=33 GPa. Hypocenters of the earthquakes (circles and dots) are as in Figure 27. (b) Coulomb failure stress change on optimally oriented vertical faults shown in a vertical cross section passing E-W through the center of the modelled volcanic point source (crossed square). Model parameters and plotting conventions as in Figure 27a.