Evelyn Roeloffs

A mechanism for sustained groundwater pressure changes induced by distant earthquakes

[1] Large, sustained well water level changes (>10 cm) in response to distant (more than hundreds of kilometers) earthquakes have proven enigmatic for over 30 years. Here we use high sampling rates at a well near Grants Pass, Oregon, to perform the first simultaneous analysis of both the dynamic response of water level and sustained changes, or steps. We observe a factor of 40 increase in the ratio of water level amplitude to seismic wave ground velocity during a sudden coseismic step. On the basis of this observation we propose a new model for coseismic pore pressure steps in which a temporary barrier deposited by groundwater flow is entrained and removed by the more rapid flow induced by the seismic waves. In hydrothermal areas, this mechanism could lead to 4 � 10 � 2 MPa pressure changes and triggered seismicity. INDEX TERMS: 1829 Hydrology: Groundwater hydrology; 7209 Seismology: Earthquake dynamics and mechanics; 7212 Seismology: Earthquake ground motions and engineering; 7260 Seismology: Theory and modeling; 7294 Seismology: Instruments and techniques; KEYWORDS: earthquakes, triggering, time-dependent hydrology, fractures Citation: Brodsky, E. E., E. Roeloffs, D. Woodcock, I. Gall, and M. Manga, A mechanism for sustained groundwater pressure changes induced by distant earthquakes, J. Geophys. Res., 108(B8), 2390, doi:10.1029/2002JB002321, 2003.

Poroelastic Techniques in the Study of Earthquake-Related Hydrologic Phenomena

Publisher Summary This chapter presents several poroelastic techniques employed in the study of earthquake-related hydrologic phenomena. The well water levels can be used to measure tectonic strain over a limited frequency band that depends critically on the degree of aquifer confinement. Although chemistry plays an important role in earthquake hydrology, only mechanical and fluid-dynamic effects can be modeled using poroelasticity. The chapter focuses on hydrologic data related to earthquakes and their quantitative interpretation. Some examples of the earthquake-related hydrologic phenomena are provided. These examples include the 1964 Alaska earthquake, the 1975 Haicheng, China, earthquake, and the 1976 Tangshan, China, earthquake. This chapter describes how the water table drainage affects the time-varying aquifer response to strain imposed by earth tides, barometric pressure, or earthquakes. A graph illustrating the relationship among the depth below the water table, hydraulic diffusivity, and time scale of pore pressure dissipation to the water table is presented The question of whether the hydrologic phenomena precedes earthquakes is also addressed in the chapter.

Persistent water level changes in a well near Parkfield, California, due to local and distant earthquakes

Coseismic water level rises in the 30-m deep Bourdieu Valley (BV) well near Parkfield, California, have occurred in response to three local and five distant earthquakes. Coseismic changes in static strain cannot explain these water level rises because (1) the well is insensitive to strain at tidal periods; (2) for the distant earthquakes, the expected coseismic static strain is extremely small; and (3) the water level response is of the incorrect sign for the local earthquakes. These water level changes must therefore be caused by seismic waves, but unlike seismic water level oscillations, they are monotonic, persist for days or weeks, and seem to be caused by waves with periods of several seconds rather than long-period surface waves. Other investigators have reported a similar phenomenon in Japan. Certain wells consistently exhibit this type of coseismic water level change, which is always in the same direction, regardless of the earthquakes azimuth or focal mechanism, and approximately proportional to the inverse square of hypocentral distance. To date, the coseismic water level rises in the B V well have never exceeded the seasonal water level maximum, although their sizes are relatively well correlated with earthquake magnitude and distance. The frequency independence of the wells response to barometric pressure in the frequency band 0.1 to 0.7 cpd implies that the aquifer is fairly well confined. High aquifer compressibility, probably due to a gas phase in the pore space, is the most likely reason why the well does not respond to Earth tides. The phase and amplitude relationships between the seasonal water level and precipitation cycles constrain the horizontal hydraulic diffusivity to within a factor of 4.5, bounding hypothetical earthquake-induced changes in aquifer hydraulic properties. Moreover, changes of hydraulic conductivity and/or diffusivity throughout the aquifer would not be expected to change the water level in the same direction at every time of the year. The first 2.5 days of a typical coseismic water level rise could be caused by a small coseismic discharge decrease at a point several tens of meters from the well. Alternatively, the entire coseismic water level signal could represent diffusion of an abrupt coseismic pore pressure increase within several meters of the well, produced by a mechanism akin to that of liquefaction. The coseismic water level changes in the BV well resemble, and may share a mechanism with, coseismic water level, stream discharge, and groundwater temperature changes at other locations where preearthquake changes have also been reported. No preearthquake changes have been observed at the BV well site, however.

Implications for prediction and hazard assessment from the 2004 Parkfield earthquake

Obtaining high-quality measurements close to a large earthquake is not easy: one has to be in the right place at the right time with the right instruments. Such a convergence happened, for the first time, when the 28 September 2004 Parkfield, California, earthquake occurred on the San Andreas fault in the middle of a dense network of instruments designed to record it. The resulting data reveal aspects of the earthquake process never before seen. Here we show what these data, when combined with data from earlier Parkfield earthquakes, tell us about earthquake physics and earthquake prediction. The 2004 Parkfield earthquake, with its lack of obvious precursors, demonstrates that reliable short-term earthquake prediction still is not achievable. To reduce the societal impact of earthquakes now, we should focus on developing the next generation of models that can provide better predictions of the strength and location of damaging ground shaking.

Hydrologic precursors to earthquakes: A review

AbstractsThis review summarizes reports of anomalous flow rates or pressures of groundwater, oil, or gas that have been interpreted as earthquake precursors. Both increases and decreases of pressure and flow rate have been observed, at distances up to several hundred kilometers from the earthquake epicenter, with precursor times ranging from less than one day to more than one year. Although information that might rule out nontectonic causes does not appear in many published accounts of hydrologic anomalies, several recent studies have critically evaluated the possible influences of barometric pressure, rainfall, and groundwater or oil exploitation. Anomalies preceding the 1976 Tangshan, China, and the 1978 Izu-Oshima-Kinkai, Japan, earthquakes are especially well-documented and worthy of further examination.Among hydrologic precursors, pressure head changes in confined subsurface reservoirs are those most amenable to quantitative interpretation in terms of crustal strain. The response of pressure head to earth tides determines coefficients of proportionality between pressure head and crustal strain. The same coefficients of proportionality should govern the fluid pressure response to any crustal strain field in which fluid flow in the reservoir is unimportant. Water level changes in response to independently recorded tectonic events, such as earthquakes and aseismic fault creep, provide evidence that a calibration based on response to earth tides may be applied to crustal strains of tectonic origin.Several models of earthquake generation predict accelerating stable slip on part of the future rupture plane. If precursory slip has moment less than or equal to that of the impending earthquake, then the coseismic volume strain is an upper bound for precursory volume strain. Although crustal strain can be only crudely estimated from most reported pressure head anomalies, the sizes of many anomalies within 150 kilometers of earthquake epicenters appear consistent with this upper bound. In contrast, water level anomalies at greater epicentral distances appear to be larger than this bound by several orders of magnitude.It is clear that water level monitoring can yield information about the earthquake generation process, but progress higes on better documentation of the data.

Water-level changes induced by local and distant earthquakes at Long Valley caldera, California

Distant as well as local earthquakes have induced groundwater-level changes persisting for days to weeks at Long Valley caldera, California. Four wells open to formations as deep as 300 m have responded to 16 earthquakes, and responses to two earthquakes in the 3-km-deep Long Valley Exploratory Well (LVEW) show that these changes are not limited to weathered or unconsolidated near-surface rocks. All five wells exhibit water-level variations in response to earth tides, indicating they can be used as low-resolution strainmeters. Earthquakes induce gradual water-level changes that increase in amplitude for as long as 30 days, then return more slowly to pre-earthquake levels. The gradual water-level changes are always drops at wells LKT, LVEW, and CH-10B, and always rises at well CW-3. At a dilatometer just outside the caldera, earthquake-induced strain responses consist of either a step followed by a contractional strain-rate increase, or a transient contractional signal that reaches a maximum in about seven days and then returns toward the pre-earthquake value. The sizes of the gradual water-level changes generally increase with earthquake magnitude and decrease with hypocentral distance. Local earthquakes in Long Valley produce coseismic water-level steps; otherwise the responses to local earthquakes and distant earthquakes are indistinguishable. In particular, water-level and strain changes in Long Valley following the 1992 M7.3 Landers earthquake, 450 km distant, closely resemble those initiated by a M4.9 local earthquake on November 22, 1997, during a seismic swarm with features indicative of fluid involvement. At the LKT well, many of the response time histories are identical for 20 days after each earthquake, and can be matched by a theoretical solution giving the pore pressure as a function of time due to diffusion of a nearby, instantaneous, pressure drop. Such pressure drops could be produced by accelerated inflation of the resurgent dome by amounts too small to be detected by the two-color electronic distance-measuring network. Opening-mode displacement in the south moat, inferred to have followed a M4.9 earthquake on November 22, 1997, could also create extensional strain on the dome and lead to water-level changes similar to those following dome inflation. Contractional strain that could account for earthquake-induced water-level rises at the CW-3 well is inconsistent with geodetic observations. We instead attribute these water-level rises to diffusion of elevated fluid pressure localized in the south moat thermal aquifer. For hydraulic diffusivities appropriate to the upper few hundred meters at Long Valley, an influx of material at temperatures of 300°C can thermally generate pressure of 6 m of water or more, an order of magnitude larger than needed to account for the CW-3 water-level rises. If magma or hot aqueous fluid rises to within 1 km of the surface in the eastern part of the south moat, then hydraulic diffusivities are high enough to allow fluid pressure to propagate to CW-3 on the time scale observed. The data indicate that seismic waves from large distant earthquakes can stimulate upward movement of fluid in the hydrothermal system at Long Valley.

The Earthquake Prediction Experiment at Parkfield, California

Since 1985, a focused earthquake prediction experiment has been in progress along the San Andreas fault near the town of Parkfield in central California. Parkfield has experienced six moderate earthquakes since 1857 at average intervals of 22 years, the most recent a magnitude 6 event in 1966. The probability of another moderate earthquake soon appears high, but studies assigning it a 95% chance of occurring before 1993 now appear to have been over-simplified. The identification of a Parkfield fault “segment” was initially based on geometric features in the surface trace of the San Andreas fault, but more recent microearthquake studies have demonstrated that those features do not extend to seismogenic depths. On the other hand, geodetic measurements are consistent with the existence of a “locked” patch on the fault beneath Parkfield that has presently accumulated a slip deficit equal to the slip in the 1966 earthquake. A magnitude 4.7 earthquake in October 1992 brought the Parkfield experiment to its highest level of alert, with a 72-hour public warning that there was a 37% chance of a magnitude 6 event. However, this warning proved to be a false alarm. Most data collected at Parkfield indicate that strain is accumulating at a constant rate on this part of the San Andreas fault, but some interesting departures from this behavior have been recorded. Here we outline the scientific arguments bearing on when the next Parkfield earthquake is likely to occur and summarize geophysical observations to date.

Case 21 water level and strain changes preceding and following the August 4, 1985 Kettleman Hills, California, earthquake

Two of the four wells monitored near Parkfield, California, during 1985 showed water level rises beginning three days before theMw 6.1 Kettleman Hills earthquake. In one of these wells, the 3.0 cm rise was nearly unique in five years of water level data. However, in the other well, which showed a 3.8 cm rise, many other changes of comparable size have been observed. Both wells that did not display pre-earthquake rises tap partially confined aquifers that cannot sustain pressure changes due to tectonic strain having periods longer than several days. We evaluate the effect of partial aquifer confinement on the ability of these four wells to display water level changes in response to aquifer strain. Although the vertical hydraulic diffusivities cannot be determined uniquely, we can find a value of diffusivity for each site that is consistent with the sites tidal and barometric responses as well as with the rate of partial recovery of the coseismic water level drops. Furthermore, the diffusivity for one well is high enough to explain why the preseismic rise could not have been detected there. For the fourth well, the diffusivity is high enough to have reduced the size of the preseismic signal as much as 50%, although it should still have been detectable. Imperfect confinement cannot explain the persistent water level changes in the two partially confined aquifers, but it does show that they were not due to volume strain. The pre-earthquake water level rises may have been precursors to the Kettleman Hills earthquake. If so, they probably were not caused by accelerating slip over the part of the fault plane that ruptured in that earthquake because they are of opposite sign to the observed coseismic water level drops.

Removal of barometric pressure response from water level data

The removal from water level data fluctuations due to barometric forcing at subtidal frequencies is studied using transfer functions. In a poorly confined well-aquifer system these transfer functions are, in general, dependent upon air diffusivity through the unsaturated zone and diffusivity from the water table into the unsaturated zone. The stability in time of estimated transfer functions from selected water well sites near Parkfield, California, is examined, and an average transfer function is computed for each site that demonstrates reasonable stability over time. Two of these transfer functions that show marked frequency dependence in the subtidal frequency band are used to filter water level data in this frequency range. A comparison is made with residuals after subtracting the best fitting multiple of band-passed barometric pressure from the water level data, and transfer function filtering is shown to be superior. Comparing the estimated transfer functions with previously derived theoretical ones shows that in one case the well-aquifer system can be reasonably well modeled by assuming that the effects of air and water diffusivity are nonnegligible, while the remaining frequency-dependent transfer function can be better modeled by assuming minimal air diffusivity through the unsaturated zone.

Creep rate changes at Parkfield, California 1966–1999: Seasonal, precipitation induced, and tectonic

Measured fault creep at Parkfield contains information about the slip rate on the San Andreas fault but also is affected by rainfall and other seasonal effects. Seven creep meters record a repeatable seasonal variation that can be removed from the data. Four of the creep meters exhibit surges in creep rate during the rainy season. For these instruments a function of rainfall is developed that crosses a threshold within a few days of the onset of accelerated creep. In addition to the seasonal variations and surges, most Parkfield creep meters record higher creep rates more frequently during the wet season than during the dry season. Characterizing this behavior makes it possible to distinguish between rainfall-influenced creep and creep rate changes of tectonic origin. Accelerated creep beginning in January 1983 was probably induced by rainfall, rather than by tectonic sources related to the Coalinga earthquake in May 1983. Four creep meters in the central part of the Parkfield fault segment are recording higher creep rates since 1993. A strong case can be made that the increased creep rate at CRR1 is tectonic, but for the other three creep meters an influence of rainfall is difficult to rule out. Strain rate changes recorded on three-component borehole strain meters suggest that the rate of subsurface creep to the northwest of these instruments is even greater. The 1993 creep rate increase is more abrupt and farther south than the expected signal from accelerating slip near the hypocenter of the 1966 Parkfield earthquake but is in the same location as rapid fault slip reported prior to that event.