2. Earthquakes-induced hydrologic phenomena
Earthquakes often cause large changes in spring and stream discharge, water levels in wells, and geyser behavior. Coseismic and postseismic hydrologic changes are abundant and well documented. Many preseismic or precursory hydrologic changes have also been reported, but in most cases the documentation has not been sufficient to convince the scientific community or to rule out explanations unrelated to earthquakes. The suggestion of precursory hydrologic changes is tantalizing, because the ultimate goal of most earthquake research is to reduce the hazard posed by major events, and some warning of impending activity would be very useful in that regard. Regardless of whether the various reported precursory changes are in fact “real,” their physical basis is poorly understood.
-6-
It is also worth nothing the relatively well-documented hydrogeochemical anomalies associated with the more recent, disastrous M 7.2 earthquake at Kobe, Japan. The chloride and sulfate contents of wells about 20 km east of the Kobe epicenter increased steadily from August 1994 to just before the earthquake on 17 January, 1995 (Tsunogai and Wakita, 1995). The radon content in a well about 30 km northeast of the epicenter peaked dramatically about 9 days before the eruption (Igarashi and others, 1995). Gas rations in mineral springs 220-230 km east of Kobe changed markedly in September 1994 and again on January, 1995, coincident with the earthquake (Sugisaki and others, 1996). The hydrology of the Kobe area is well known and carefully monitored in part because the groundwater there is regarded as highly suitable for drinking and for brewing sake. In fact, the preseismic changes in chloride and sulfate content were reconstructed by analyzing dated, bottled mineral water distributed in the domestic market.
References
Groundwater in Geologic Processes. Cambridge University Press. S.E. Ingebristen, W.E. Sanford. 1998. - 341 p
3. Streamflow and springs
Many of the quantitative data on streamflow response to seismicity are derived from the U.S. Geological Survey’s extensive hydrologic monitoring network n the tectonically active western United States. There are many examples of postseismic increases in stream and spring discharge, some instances of springs going dry, and very few examples (e.g., Waller, 1966) of decreased stream discharge. The hydrologic disturbances associated with fairly recent moderate-to-large earthquakes in California are particularly well documented (Sorey and Clark, 1981; Roeloffs and onthers, 1995; Quilty and others, 1995).
-7-
In some instances the streamflow data are actually sufficient to permit estimates of the total “excess” streamflow derived from a particular seismic event. The M 7.5 Hebgen Lake earthquake (17 August, 1959) apparently produced about 0.5 km3 of water, the M 7.3 Borah Peak earthquake (28 October, 1983) about 0.3 km3 of water, and the M 7.1 Loma Prieta earthquake (17 October, 1989) only about 0.01 km3 of water (Muir-Wood and King, 1993; Rojstaczer and others, 1995). These excess flows were estimated by comparing the actual post earthquake stream hydrographs to synthetic hydrographs that approximate what the flow would have been in the absence of an earthquake.
As nearly as can be determined, the streamflow increases are often essentially coseismic. The U.S. Geological Survey gaging stations generally record streamflow every 15 minutes, and flow increases tend to be observed within 15 minutes at those stations that are not temporarily disabled by the ground motion.
Streamflow may continue to increase for a few days and then gradually decay toward the prequake baseline condition over several months.
Selected examples of stream- and spring-discharge increases associated with California earthquakes are shown in Figure 13.1. After the 1989 Loma Prieta event, many streams within about 50 km of the epicenter showed 4- to 24-fold increases in discharge (e.g., Figure 13.1a). There was no clear relation between the magnitude of the increase and proximity to the epicenter. The Loma Prieta event occured during a prolonged California drought, so that some water managers viewed the unexpected “earthquake water” as a boon. However, the excess streamflow decayed fairly rapidly, and at Big Tress (Figure 13.1a), for example, the total earthquake-induced streamflow proved to be only about 20% of the total annual flow (Rojstaczer and Wolf, 1994).
-8-
Figure 13.1 (a) Stream discharge in response to the 17 October, 1989 Loma Prieta earthquake at Pescadero (~50 km from epicenter), Big Trees (~20 km from epicenter), and San Lorenzo Park (~30 km from epicenter and ~20 km upstream from Big Trees). Streamflow values are daily means and precipitation values are daily totals. From Rojstaczer and Wolf (1994). (b) Water stage at the weir on Little Hot Creek in Long Valley, California, and record of M > 4 earthquakes recorded at Mammoth Lakes on 24-28 May, 1980. From Sorey and Clark (1981).
-9-
A series of earthquakes as large as M 6.3 on 25-28 May, 1980, caused temporary increases in the discharge (Figure 13.1b) and turbidity of hot springs in the Long Valley caldera of east-central California. These earthquakes had other obvious effects on the hydrothermal system, including emptying and refilling of boiling pools and temporary increases in fumarolic activity (Sorey and Clark, 1981).
Two mechanisms have been invoked to explain earthquake-related changes in spring and stream discharge: near-surface permeability enhancements (e.g., Rojstaczer and Wolf, 1992) and expulsion of water from depth due to elastic compression (e.g., Muir-Wood and King, 1993).
4. Fluid pressures at seismogenic depths
Because there is no direct information on the fluid-pressure regimes associated with natural (i.e., noninduced) seismicity, hypotheses about fluid pressures at seismogenic depths are necessarily inference- or model-based.
Hubbert and Rubey (1959) introduced the concept of fault movement due to Mohr-Coulomb failure of a fluid-saturated medium. They focused specifically on overthrust faulting and aimed at resolving the mechanical problem of moving very long fault blocks over horizontal surfaces. Without consideration of pore-fluid pressures, such movement appears to be precluded by the limited strength of the overthrust block and/or the large frictional resistance to low-angle sliding.
-10-
Much more recently, a consensus has developed that the mechanical behavior of the San Andreas fault (and presumably other transform or strike-slip faults) is profoundly influenced by the presence and distribution of high pore-fluid pressures. This realization has developed through detailed investigation of the so-called stress-heat flow paradox of the San Andreas fault. If the San Andreas had the relatively high frictional strength typical of the upper crust, its movement should generate significant frictional heat. In fact, there is no detectable heat flow anomaly associated with the fault. A number of models for the fluid-pressure distribution in and around the fault zone have been proposed, and we choose three of these to illustrate the range of possibilities. Irwin and Barnes (1975) suggested that the San Andreas operates in an environment where ambient fluid pressures are maintained at elevated levels by generation of metamorphic fluids. Byerlee (1990) suggested that the high fluid pressures are local to the fault zone, rather than ambient, and result from the sealing of locally derived fluids within the fault zone. Rice (1992) suggested continual upwelling of overpressured fluids within the fault zone, with associated lateral leakage of such fluids away from the fault zone.
An important caveat to the discussion in this lecture is that, except in geothermal areas (Fournier, 1991), there is as yet little convincing evidence for greatly elevated pore pressures at midcrustal depth in crystalline rocks, whereas there is good evidence for hydrostatic pore pressures at depths as great as 12 kilometers (Zoback and Zoback, 1997).
-11-
The stress-heat flow paradox of the San Andreas fault. The issue of heat transfer via topographically (gravity) driven regional groundwater flow is also pertinent to so-called stress-heat flow paradox of the San Andreas fault. The San Andreas is a major right-lateral strike-slip fault that defines the boundary between the Pacific and North American plates in California and northwestern Mexico. As first pointed out by Brune and others (1969), displacement rates along the San Andreas are large enough that frictional heat generation should cause a pronounced, local heat flow anomaly unless the fault is “weak,” that is, unless it has low frictional resistance. Because no such heat flow anomaly had yet been observed, Brune and others (1969) concluded that the fault is indeed weak. A subsequent compilation of about 100 high-quality heat flow data by Lachenbruch and Sass (1980) confirmed that there is no local conductive heat flow anomaly associated with the fault (Figure 13.2). They agreed with Brune and others (1969) that this implies a frictional resistance of less than 10 MPa. If this rather low value is correct, then the stress drops caused by earthquakes, typically in the range of 1−10 MPa (Kanamori and Anderson, 1975), may represent nearly complete relief of shear stress along the fault.
The inference of low frictional resistance along the San Andreas system is inconsistent with most in situ stress measurements in the upper few kilometers of the crust and with the results of laboratory rock strength studies. The in situ measurements suggest that in intraplate regions the maximum horizontal shear stress generally increases with depth at a rate of 7−8 MPa/km, so that the shear stress at seismogenic depths would be projected as 50−100 MPa. Attempts to extend laboratory studies (e.g., Byerlee, 1968) to midcrudtal conditions predict even somewhat larger shear-stress values of 100 −200 MPa at 10−15-km depth (Brace and Kohlstedt, 1980). This apparent conflict between the stress and heat flow observations become known as the stress-heat flow paradox of the San Andreas fault (e.g., Zoback and Lachenbruch, 1992).
-12-
Figure 13.2 Heat flow as a function of distance from main trace of the San Andreas fault. Dots are measured values; cures represent the conductive heat flow anomaly expected for an average fault friction of 50 MPa and a slip velocity of 3 cm per year at various times. The uppermost curve is the infinite-time (steady-state) solution. From Lachenbruch and Sass (1980).
Lachenbruch and Sass (1980) considered and rejected the possibility that groundwater flow is masking frictional heating effects by removing significant quantities of heat from the fault zone. Thermal springs are scarce along the length of the San Andreas, and the total hydrothermal heat discharge is negligible relative to the heat represented by the “missing” conductive heat flow anomaly. Although they could not categorically rule out substantial undetected discharge at temperatures slightly above ambient, they saw no obvious driving mechanism for such flow.
-13-
Williams and Narasimhan (1989) later recognized the possibility of topographically driven groundwater flow away from the San Andreas fault. They demonstrated that cross-fault topographic profiles throughout central and southern California show significant relief associated with the fault zone (Figure 10.7). The fault one is generally elevated and might represent a regional recharge area in which heat flow is depressed. Williams and Narasimhan (1989) found that topographically driven flow could completely suppress a fault-centered conductive heat flow anomaly if the average upper crustal permeability is at least 10-16 m2 (Figure 13.3).
In the late 1980s a 3.5-km-deep research borehole was drilled in Cajon Pass, southern California, in order to better determine the thermal, mechanical, and hydrologic conditions along the San Andreas fault. A relatively high conductive heat flow value of about 90 mW/m2 had been measured in a nearby 1.7-km-deep oil well, tending to support the idea of a “strong” fault. However, heat flow in the Cajon Pass borehole itself showed a systematic decrease from about 100 mW/m2 at less than 400-m depth to 75 mW/m2 in the lowermost 300 m (Sass and others, 1992). The shape of the Cajon Pass heat flow profile is consistent with uplift and erosion at an estimated rate of about 1 mm/yr over the past 1 Ma. The elevated heat flow at shallow depths is likely a transient feature related to this rapid uplift. The deeper heat-flow value more likely represents the thermal steady state and is consistent with the average value obtained from relatively shallow (<300 m) drillholes along the length of the San Andreas (Figure 13.3).
-14-
Figure 13.3 (a) Three averaged topographic profiles across the San Andreas fault (0 km on the horizontal axis). Averaging was done for 45-km (San Francisco), 70-km (Cholame Hills), and 60-km (Palmdale) segments of the fault. (b) Effect of topographically driven groundwater flow on near-surface heat flow. For a permeability−depth relation k = 10 –(0.2z+15) m2, where z is depth in kilometers, the conductive heat flow anomaly caused by frictional heating is totally dissipated. From Williams and Narasimhan (1989).
In situ stress measurements in the Cajon Pass borehole revealed that the frictional strength of the crust near the fault is actually quite high, consistent with laboratory values. However, surprisingly, there was no evidence of right-lateral shear on planes parallel to the San Andreas fault.
-15-
Instead, the stresses measured in the borehole are consistent with the active normal faults in the vicinity and with left-lateral displacement on a fault that locally runs parallel to the San Andreas. This suggests that the San Andreas fault itself moves at low shear stresses and in fact is extremely weak relative to the surrounding crust (Zoback and Healy, 1992). The fault seems to operate in the context of roughly fault-normal compression (Zoback and others, 1987), with the surrounding crust failing under more typical “high-stress” frictional faulting conditions.
The permeability of core samples obtained from 2,100- to 3,500-m depth in the Cajon Pass borehole ranged from 10-22 to 10-19 m2 (Morrow and Byerlee,1992), much lower than the threshold value for heat dissipation via topographically driven flow determined by Williams and Narasimhan (1989). However, the cores are from crystalline rocks in which fluid movement occurs dominantly in major fracture zones. It is well known that core measurements on such rocks tend to underestimate the larger-scale in situ permeability by a factor of about 103, mainly because the core fail to sample the important fractures. Thus the permeability measurements do not exclude the possibility of effective advection of heat away from the fault.
At this point, arguments about the strength or weakness of the San Andreas ultimately depend upon the occurrence and movement of pore fluids. The current consensus is that the fault is weak; if so, the most likely explanation is that pore fluids within the fault zone are at near-lithostatic pressures, so that the effective stress is very low.
One possible explanation for elevated pore-fluid pressures is some sort of fluid source at depth. If the fault is strong (which now seems less likely), the associated frictional heating must be effectively obscured by groundwater movement.