Hydrological responses to earthquakes, such as increased stream discharge, changes in stream water chemistry and water level changes in monitored well sites are often observed. As discussed above, an increase in fracturing at the time of an earthquake will cause an increase in permeability. Such changes will therefore facilitate the movement of fluid in and around the fault zone at the time of an earthquake. As discussed in chapter 2, fluid content and pore fluid pressure will affect the seismic properties of a medium. Therefore, we discuss observations of the changes in fluid distribution at the time of an earthquake, and the potential influences on the seismic properties, with reference to chapter 2. An increase in saturation will cause an increase in VP, have little effect on
VS and increase the VP/VS ratio (see chapter 2, figures 2.1 and 2.2). An increase in
saturation will also increase S-wave attenuation. Maximum P-wave attenuation is seen at partial saturation (see chapter 2, figure 2.5).
6.2.1 Increased shallow permeability
After the 1989 Loma Prieta earthquake, an increase in stream flow was observed within 15 minutes of the earthquake. Ionic concentrations of the stream water also increased, although the overall proportions of the major ions remained the same. Elevated ionic concentrations of stream water were observed for almost 3 years after the 1989 Loma Prieta earthquake (Rojstaczer et al., 1995). Groundwater levels were lowered, in some areas by as much as 21 m, within weeks to months after the earthquake. Similarly, following the 1995 Kobe earthquake, a large amount of water was discharged, especially in mountainous areas and near active faults (Oshima et al., 1996; Tokunaga, 1999). The anomalous flow rate gradually decreased over 2 years (Sato et al., 2000). In some areas, the water table dropped more than 70 m within 90 days of the earthquake (GRG, 1996). The permeability has been estimated to have increased by a factor of 3.6 to 7 (GRG, 1996; Sato et al., 2000).
Such postseismic changes have been observed following other earthquakes (Briggs
and Troxell, 1955; Whitehead, 1985;Nur, 1974;Waller, 1966; Bell and Katzer, 1987).
Changes are observed to persist for months to years. These hydrological observations can be explained together by an increase in permeability at the time of the earthquake. Permeability enhancement at the time of the earthquake would initially cause an in- crease in flow rates. This in turn would lower the water table. The change in water chemistry can then be explained by an increase in the proportion of water that has spent a long time interacting with rock at depth. The lack of compositional changes in stream chemistry implies that water is expelled from relatively shallow depths. Af- ter the earthquake, the permeability would be expected to decrease gradually to pre-
earthquake levels, and the water table would recover.
Changes in both P and S wave velocities are observed after an earthquake. However, greater changes in P wave velocities are often observed (e.g. Schaff and Beroza (2004);
Li et al. (1998b, 2003, 2006)). A decrease in saturation is expected to decrease the P
wave velocity, but have little effect on the S wave velocity. The ratio of travel time changes (∆tP
∆tS) can therefore reveal the state of saturation of the crust. In a Poisson
solid, with isotropically oriented penny-shaped cracks, the ∆tP
∆tS ratio is 1.64 for dry
cracks and 0.17 for water-saturated cracks (with a Poisson’s ratio of 0.33) (Garbin and
Knopoff, 1975;Li et al., 2003). Applying this after the 2004 M6.0 Parkfield earthquake
implies partial fluid saturation postseismically (Li et al., 2006). This is also the case following the 1999 M7.1 Hector Mine earthquake (Li et al., 2003) and the 1992 M7.5 Landers earthquake (Li et al., 1998b). Furthermore, repeated seismic surveys (in 1994, 1996 and 1998) following the 1992 Landers earthquake show a decrease in the ratio of travel time changes. A value of 0.75 in the earlier 2 years decreases to 0.65 in the later 2 years, implying that cracks near the fault zone are becoming more fluid saturated with time after the earthquake (Li and Vidale, 2001). This observation fits well with the observed coseismic decrease in the water table, which recovers postseismically.
It is noted that following the 1989 Loma Prieta earthquake, the stream flow ob- servations recover more quickly than the seismic velocities. However, stream-flow is controlled by permeability (connectivity of pore space), whereas seismic velocity is controlled by porosity (density of pore space). These observations can therefore be reconciled. We would expect permeability to recover more quickly, as closure of con- nective cracks will reduce permeability much more than porosity (Rubinstein and
Beroza, 2004).
6.2.2 Poroelastic Flow
Large earthquakes coseismically cause pore pressures to increase in areas of compression and decrease in areas of dilatation ( Nur and Booker (1972); Bj¨ornsson et al. (2001) and see figure 6.1). These pore-pressure gradients induce groundwater flow (Roeloffs, 1996) which results in poroelastic rebound i.e. postseismic subsidence within (coseimic) compressional quadrants and uplift in (coseismic) extensional quadrants. Poroelastic rebound has been inferred from InSAR measurements following the 1992 M7.3 Landers earthquake (Peltzer et al., 1998; Fialko, 2004) the 2003 M6.6 Bam earthquake, Iran
(Fielding et al., 2009) and M6.0 earthquakes in Iceland (Jonsson et al., 2003).
Following two strike-slip earthquakes in Iceland, Jonsson et al. (2003) investigated coseismic and postseismic changes in water level in geothermal wells of depth < 1.5 km. Coseismic water level increase was observed in the compressional quadrants and coseismic water level decrease in the extensional quadrants of each strike slip earth- quake. Postseismically, the water level changes are in the opposite direction i.e. water
level decrease in the coseismic compressional quadrants and water level increase in the coseismic extensional quadrants. The water level changes recover within 1-2 months, similar to the surface deformation recorded in interferograms (Jonsson et al., 2003). In contrast, the duration of the aftershock sequence was projected to be 3.5 years. Longer pore-pressure transients may exist at the depths of the aftershocks (3-10 km), where the permeability is expected to be lower. Water level changes are seen at wells up to 25 km from the ruptures.
Following the 1992 Landers earthquake, poroelastic rebound was observed in InSAR data at distances of 15 km from a rupture and for 3.5 years after the earthquake
(Peltzer et al., 1998). Fialko(2004) infer, from the fit of models to the InSAR and GPS
data, that poroelastic rebound following the 1992 M7.3 Landers earthquake involves the whole upper crust (i.e. to a depth of 15 km).
The signal following the 2003 Bam, Iran earthquake has a much more limited spatial extent (500 m) and a relaxation time of 1.7 years. A small amount of broad-scale poroelastic rebound in the first two months after the earthquake cannot be ruled out
(Fielding et al., 2009).
The differences in spatial extent and recovery time in these examples are likely to be due to different levels of permeability. Longer relaxation times imply a lower permeability of the poroelastic rock volume (Fielding et al., 2009).
This mechanism would be expected to form a pattern in velocity and attenuation changes, matching the areas of extension and compression. In the compressional areas where the pore pressure increases, the effective pressure will decrease. This would cause a decrease in both P and S wave velocities and an increase in both P and S wave attenuation. Conversely, in extensional areas where the pore pressure decreases, the effective pressure increases, which would cause an increase in both P and S wave velocities and a decrease in P and S wave attenuation.