4.10 Earthquake Hydrology - Berkeley Seismological Laboratory

4.10 Earthquake HydrologyM. Manga and C.-Y. Wang, University of California Berkeley, Berkeley, CA, USA

ª 2007 Elsevier B.V. All rights reserved.

4.10.1 Introduction 2934.10.2 Hydrologic Response to Stress 2934.10.2.1 Effect of Static Stress 2944.10.2.1.1 Poroelastic flow and deformation 2944.10.2.1.2 Permanent deformation 2954.10.2.2 Effect of Dynamic Strain 2964.10.2.2.1 Poroelastic deformation and fluid flow 2964.10.2.2.2 Permanent deformation 2974.10.2.2.3 Liquefaction 2974.10.3 Observations and Their Explanations 2994.10.3.1 Wells 2994.10.3.1.1 Water-level oscillations 2994.10.3.1.2 Persistent and delayed postseismic changes in water level in the near and

intermediate field 3004.10.3.1.3 Near-field response of unconsolidated materials 3014.10.3.1.4 Far-field response 3054.10.3.1.5 Summary 3054.10.3.2 Streamflow 3054.10.3.3 Mud Volcanoes 3104.10.3.4 Geysers 3114.10.4 Feedback between Earthquakes and Hydrology 3124.10.5 Hydrologic Precursors 3134.10.6 Concluding Remarks 315References 315

4.10.1 Introduction

For thousands of years, changes in the amount, direction,and rate of water flow in streams and in fluid pressure inthe subsurface have been documented following largeearthquakes. Examples include the formation of newsprings, disappearance of previously active springs,increased discharge in streams by an order of magnitudethat persists for weeks or even months, fluctuations inwater levels in wells by several meters, and permanentchanges in water levels in wells coincident with earth-quakes located thousands of kilometers away. Suchobservations, although often simply assigned to the cate-gory of curious phenomena, provide unique insight intohydrogeologic processes and, in particular, their corre-lation with tectonic processes at spatial and temporalscales that otherwise could not be studied.

Hydrologic responses are not unexpected: earth-quakes cause strain, and strain changes fluid pressureand alters hydrogeologic properties such as

permeability, which controls the rate of fluid flow.What is more surprising about many hydrologicresponses, however, is their large amplitude and thegreat distances over which they occur.

This review is divided into two main parts. Firstwe describe the processes by which fluid pressureand hydrologic properties can change following anearthquake. Next we review hydrologic phenomenaattributed to earthquakes and summarize our currentstate of understanding about the processes that causethese phenomena. We then conclude with a briefdiscussion of hydrologic precursors.

4.10.2 Hydrologic Response toStress

Earthquakes change the static stress (i.e., the offset ofthe fault generates a static change in stress in thecrust) and also cause dynamic stresses (from the

293

seismic waves). Both stresses increase as the seismicmoment of the earthquake, but they decay very dif-ferently with distance r. Static stresses decrease as1/r3. By comparison, dynamic stresses, which areproportional to the seismic wave amplitude, decreasemore gradually. A standard empirical relationshipbetween surface-wave amplitude and magnitude hasdynamic stress decreasing as 1/r1.66 (e.g., Lay andWallace, 1995). Other factors such as directivity,radiation pattern, and crustal structure will also influ-ence the amplitude of shaking. However, thesignificant difference in the dependence on distanceis a robust feature distinguishing static and dynamicstresses.

Table 1, reproduced from Manga and Brodsky(2006), lists and compares the magnitude of earth-quake stresses with other external stresses ofhydrologic interest. Both static and dynamic stressesmay be significant in the near and intermediate field(within up to a few fault lengths), but only dynamicstresses are larger than tidal stress in the far-field(many fault lengths away).

4.10.2.1 Effect of Static Stress

4.10.2.1.1 Poroelastic flow anddeformationIn response to stresses, porous solids deform, thepressure of fluids within pores changes, and porefluids can flow. The basic theory of poroelasticity,first developed by Biot (1941, 1956a, 1956b),describes the coupling between changes in stress !,strain ", pore pressure p, and the material propertiesthat relate these three variables. A complete discus-sion and derivation of the theory, even its simplestforms, is beyond the scope of this review, but severalkey relationships are useful for understanding thehydrological response to earthquakes and interpret-ing observations. Wang’s (2000) text on linear

poroelasticity, and review papers by Rice andCleary (1976), Roeloffs (1996), and Neuzil (2003),provide thorough and pedagogic reviews.

Consider an isothermal, linear, isotropic poroelas-tic material, subjected to a stress, !ij . The materialwill respond by deforming with the total strain "ij ,given by

"ij ¼1

2G!ij –

#

1þ #!kk$ij

! "þ %

3Kp$ij ½1$

where G is the shear modulus, K is the bulk modulus,# is the Poisson’s ratio, and % is the Biot–Williscoefficient, which is the ratio of the increment influid content to the volumetric strain at constantpore pressure. The increment in fluid content,denoted , is the change in fluid mass per unitvolume divided by the density of the fluid at thereference state and is given by

¼ %

3K!kk þ

%

KBp ½2$

B is called Skempton’s coefficient and has avalue between 0 and 1, with ‘hard rocks’ such assandstone and granite having values between about0.5 and 0.9 and unconsolidated materials having avalue close to 1. The variable B is related to theporosity & and to the compressibilities of the porefluid, solid grains, and saturated rock. The sign con-vention is chosen so that negative !kk indicatescompression.

First we consider the so-called ‘undrained’ limit inwhich there is no flow of pore fluid, that is, ¼ 0 (theopposite extreme, termed ‘drained’, corresponds to thelimit in which there is no change in fluid pressure, thatis, p ¼ 0). The change in pore pressure p caused by achange in mean stress !kk can be found from eqn [2]:

p ¼ –B

3!kk ½3$

Table 1 Amplitude and period of stress changes of earthquakes compared with other sources of stress

Stress (MPa) Period

Solid earth tides 10%3 12 hOcean tides 10%2 12 hHydrological loading 10%3 to 10%1 Days to yearsGlacier loading 101 to 102 103 years

102 km 103 km 104 kmStatic stress changes, M8 10%1 10%4 10%7 NADynamic stress changes, M8 3 0.06 0.001 20 s

(from Manga M, and Brodsky EE (2006) Seismic triggering of eruptions in the far field: Volcanoes and geysers. Annual Review of Earth andPlanetary Sciences 34: 263–291.)

294 Earthquake Hydrology

Equation [3] can be expressed in terms of the volu-metric strain "kk using constitutive relationship [1]and physical relationship [2]:

p ¼ –B

3!kk ¼ – BKu"kk ¼ –

2GB

3

1þ #u1 – 2#u

"kk ½4$

where the subscript u indicates a value underundrained conditions. The variable G should havethe same value in drained and undrained conditionsfor a linear poroelastic material.

If there are gradients of fluid pressure, there willbe a flux of fluid q (volume/area per unit time)governed by the Darcy’s equation

q ¼ –k

'rp ½5$

where the pressure p is the excess pressure, that is, thepressure difference from hydrostatic, k is permeabil-ity, and ' is fluid viscosity; p, as defined, is equivalentto the hydraulic head multiplied by (g, where ( is thedensity of the fluid and g the gravitational accelera-tion. Darcy’s equation, coupled with the continuityequation governing conservation of mass, if the per-meability k and ' are constant, leads to pore pressurebeing governed by an inhomogeneous diffusionequation

%

KB

B

3

q!kkqt

þ qpqt

! "¼ k

'r2p ½6$

The term involving the time derivative of stresscouples the mechanical deformation and the fluidflow. The time derivative of stress will not appearin the case of uniaxial strain and constant verticalstress, both reasonable approximations for near-sur-face aquifers (but not rigorously correct if there istwo- or three-dimensional flow), and can beneglected for very compressible fluids such as gases(Wang, 2000). In these cases, the evolution of fluidpressure resulting from the flow is uncoupled fromthat for the stress and can be solved independently.Then eqn [6] reduces to the form commonly used tomodel groundwater flow:

Sqpqt

¼ k(g

'r2p or

qpqt

¼ Dr2p ½7$

where S is standard hydrogeologic specific storageand is equal to volume of water released per unitchange in head while maintaining no lateral strainand constant vertical stress, and D is called thehydraulic diffusivity.

Assuming that eqn [7] is a good approximation,the timescale ) for changes in pore pressure to diffuse

or relax can be calculated by ) & L2/D, where L is thelength scale over which the pressure diffuses. Typicalvalues of D are 10%10–104m2 s%1, with small valuescharacterizing low-permeability rocks (e.g., shale),and large values characterizing coarse, unconsoli-dated sediments or highly permeable rocks (e.g.,gravel and karst, respectively). A value of 100m2 s%1

would be typical of sandstone aquifers. For a 100mthick aquifer, then, pore pressure changes will diffuseon timescales that range from minutes to many dec-ades, and with a timescale of hours for a sandstoneaquifer.

On examining hydrological observations, we seethat many of the most dramatic hydrologic phe-nomena mentioned in Section 4.10.1, and inparticular those that involve water discharge atthe Earth’s surface, cannot be explained by tradi-tional linear poroelasticity theory. The reason isquite straightforward. Consider the case in whicha stress change !kk causes an undrained pressurechange p given by eqn [3]. For a (very large) staticstress change of 1MPa (see Table 1), eqn [3]implies pressure changes of 0.1–0.3MPa. Thiswould cause a 10–30 m change in water level ina well. To bring the pore pressure back to itsoriginal value requires the removal of an incrementof fluid content

¼ 3ð#u – #Þ2GBð1þ #Þð1þ #uÞ

!kk ½8$

Using values for Berea sandstone (Wang, 2000)of #¼ 0.17, #u¼ 0.33, G¼ 5.6 GPa, and B¼ 0.75,we obtain ¼ 3.5) 10%5. For a 100 m thick aqui-fer with porosity &¼ 0.4, this is equivalent to about1mm of water per unit area discharged at thesurface–an amount that is small compared to thatneeded to explain the excess discharge in streamfollowing large earthquakes (see table 1 in Manga(2001)).

Equations [1]–[3] apply for linear, isotropic por-ous materials. In this limit, shear stresses cause nochange in pore pressure. In natural rocks, however,shear stresses can induce changes in pore pressurebecause of anisotropy or the nonlinear elastic beha-vior of rocks (e.g., Hamiel et al., 2005).

4.10.2.1.2 Permanent deformationAt shear strains from 10%4 to 10%3, brittle rocks,sediments, and soils show permanent deformation.At even greater shear strains (&10%2), failure occurs.The changes in porosity of sediments and soils

Earthquake Hydrology 295

during permanent deformation may be many ordersof magnitude greater than that during elastic defor-mation. Thus, under undrained conditions,permanent deformation can be associated with sig-nificant changes in pore pressure and groundwaterflow.

During permanent deformation, the cohesionamong the grains of sediments and soils is disruptedand the grains tend to move to reach a new state ofequilibrium. The basic characteristics of deformation,however, depend on the consolidation state of sedi-ments and soils. For loose deposits, shear deformationcauses grains to move into pre-existing pores. Thisreduces the original porosity, and thus the volume ofsediments (soils) decreases–a process commonlyknown as consolidation. Shear deformation of densedeposits, in contrast, will cause sediment grains toroll over each other and, in so doing, create newporosity; thus, the volume of sediments (soils)increases – a process commonly known as dilatancy.In both cases, a critical state is eventually reached inundrained deformation, beyond which porosity nolonger changes with continued shearing.

The deformation behavior of brittle rocks is dis-tinctly different from that of sediments and soils,leading to different predictions for the hydrologicand seismic consequences under applied stresses.Laboratory measurements show that beyond the elas-tic limit, shearing of consolidated rocks causesmicrocracks to open and the volume of rock toincrease – that is, they become dilatant (e.g., Braceet al., 1966). At still higher deviatoric stresses, micro-cracks may coalesce and localize into a shear zone,leading to eventual rupture (Lockner and Beeler,2002). Repeated rupturing of the brittle crust along afault zone can produce pulverized gouge material withsignificant porosity. Geologic investigations ofexhumed fault zones in California (e.g., Chester et al.,1993, 2004) and in Japan (Wibberley and Shimamoto,2003; Forster et al., 2003), and studies of drilled coresfrom fault zones (e.g., Ohtani et al., 2001) have demon-strated that some fault zones are composed of fluid-saturated, porous fault gouge several hundred metersthick. Attempts to image the structure of fault zones byusing conventional seismic reflection and refractionmethods (e.g., Thurber et al., 1997) suffered from nothaving high enough station and source densities toresolve deep fault-zone structures on the scale ofhundreds of meters (Mizuno, 2003). More recent seis-mic imaging of the fault-zone structures relies on theuse of fault-zone trapped waves (e.g., Li et al., 2000,2006; Mizuno, 2003) and high-frequency body waves

(Li and Zhu, 2006). These studies have delineated a100–250m wide gouge zone along active faults inCalifornia (Li et al., 2000; Li and Zhu, 2006) and inJapan (e.g., Mizuno, 2003), in which the seismic velo-cities are reduced by 30–50% from the wall-rockvelocities. Direct drilling of the San Andreas FaultZone to a depth of &3 km near Parkfield, CA, alsorevealed a process zone 250 m in width, in which theseismic velocities are 20–30% below the wall-rockvelocities (Hickman et al., 2005).

Using high-resolution electromagnetic imaging,Unsworth et al., (2000) showed that the San AndreasFault Zone contains a wedge of low-resistivity mate-rial extending to several kilometers in depth; theyfurther suggested a range of saturated porosity toaccount for the observed electrical resistivity. Usingexperimentally measured seismic velocity and elec-trical conductivity of fault gouge at elevatedpressures to interpret the then-available seismicvelocity and electrical resistivity of the San AndreasFault Zone in central California, Wang (1984) sug-gested that the fault zone may consist of saturatedfault gouge. Inverting the gravity anomaly across theSan Andreas Fault Zone in central California andconstraining the inversion by using existing seismicreflection profiles, Wang et al., (1986) further showedthat the fault zone may be characterized by relativelylow density with a corresponding porosity greaterthan 10% extending to seismogenic depths.

That fault zones may be porous and saturated atseismogenic depths has important implications on ahost of issues related to the dynamics of faulting (e.g.,Sleep and Blanpied, 1992; Sibson, 1973; Lachenbruch,1980; Mase and Smith, 1987; Brodsky and Kanamori,2001; Andrew, 2003; Wibberley and Shimamoto, 2005)and the frictional strength (or weakness) of faults (e.g.,Lachenbruch and Sass, 1977; Mount and Suppe, 1987;Zoback et al., 1987; Rice, 1992). These topics, whilebeyond the objective of the present chapter, highlightthe importance of understanding better the interactionbetween hydrology and earthquakes.

4.10.2.2 Effect of Dynamic Strain

4.10.2.2.1 Poroelastic deformation andfluid flowSeismic waves cause spatial variations in strain andhence spatial variations in pore pressure. The result-ing pore pressure gradients cause fluid flow. Seismicwave speeds in saturated rocks will thus differ fromthose in unsaturated rocks and will be frequency


dependent. As the fluid moves, energy is also dissi-pated by the fluid flow, which in turn is governed bythe permeability of the porous material, the geometryof the pores, and the viscosity of the fluid. Thiscontributes to seismic attenuation. Wave speed andattenuation can be calculated theoretically as a func-tion of frequency for porous materials with differentpore structures by solving the poroelastic equationscoupled with Darcy’s equation for fluid flow at themacroscopic scale. As frequency increases, pore scale(or ‘squirt’) flows become more important (e.g.,Mavko et al., 1998). These ideas were first quantifiedin a series of papers by Biot (e.g., Biot, 1941, 1956a,1956b, 1962) and extended by many authors since, forexample, by accounting for squirt flows (e.g., Dvorkinet al., 1994). Additional important extensions to actualgeological settings include accounting for heteroge-neity with dual porosity models (e.g., Berryman andWang, 2000) or fractal geometries (e.g., Masson andPride, 2006) and partial or multiphase saturation (e.g.,Pride et al., 2004).

Changes in pore pressure and stress also changefracture and pore geometry, which in turn alterswave speeds and anisotropy. The frequency depen-dence of attenuation and wave speeds also dependson hydrogeologic properties such permeability andthe nature of permeability heterogeneity. Therelationship between seismic properties and hydro-geologic properties and fluid pressure involves somany parameters that inferring properties of the sub-surface from seismic measurements is challenging.However, measurements of seismic attenuation,seismic anisotropy, and wave speeds can be used asa noninvasive probe to monitor changes in the sub-surface (e.g., Liu et al., 2004).

4.10.2.2.2 Permanent deformationThe dynamic deformation of sediments and soils, incontrast to the static deformation, is affected by iner-tial forces that depend on loading rates and numberof loading cycles. An often-used measure of the totalamount of seismic energy imparted to an engineeredstructure is the Arias intensity, which is proportionalto the square of acceleration (Arias, 1970). Therefore,even if the magnitude of stresses is small, inertialforces may significantly affect deformation at highfrequencies. For this reason, it is necessary in earth-quake engineering to examine the dynamicdeformation of sediments and soils for dynamicstrains as small as 10%6. Permanent volumetric defor-mation, however, does not occur until the magnitudeof shear strain reaches some threshold. Dobry (1989)

summarized undrained experimental results in whichexcess pore pressure developed during the applica-tion of 10 cycles of sinusoidal loads. He showed that,for different sands with a wide range of dry densitiessubjected to a wide range of effective confining pres-sures, pore pressure did not increase until the cyclicshear strain amplitude was increased to 10%4. Theincrease in pore pressure indicates a decrease in porevolume, and hence consolidation. It is interesting tonote that the strain threshold is comparable in mag-nitude to that for permanent deformation to occur instatic deformation (Section 4.10.2.1).

The number of loading cycles during dynamicloading is another important factor in determiningthe mechanical responses of sediments and soils.Results of laboratory experiments show that duringcyclic loading at constant strain amplitude, porepressure in sediments and soils increases with thenumber of loading cycles. At decreasing strain ampli-tudes, a greater number of loading cycles is requiredto build up excess pore pressure to a given magnitude(Ishihara, 1996). During large earthquakes, about10–20 cycles of major ground shaking occur, thoughthe amplitude of each cycle will vary.

4.10.2.2.3 LiquefactionWhen pore pressures approach the overburden pres-sure, soils lose their rigidity and become fluid-like – aphenomenon widely known as liquefaction (Seed andLee, 1966; Terzaghi et al., 1996), which is a majorsource of seismic hazard for engineered structures.Liquefied sediments and soils in the subsurface areassociated with substantial pore pressure, as indicatedby the height of the smear on the wall in Figure 1,left by ejected sands during the 1999 M 7.5 Chi-Chiearthquake in Taiwan. Most liquefaction occurs nearthe ruptured fault, but liquefaction can also occur atlarge distances from the earthquake epicenter; forexample, ejection of fluidized sediments wasobserved during the 1964 M 9.2 Alaska earthquakeat distances more than 400 km from the epicenter(Waller, 1966).

Field and laboratory studies show that the occur-rence of liquefaction depends on many factors, suchas earthquake magnitude, shaking duration, peakground motion, depth to the groundwater table,basin structures, site effects, and liquefaction suscept-ibility of sediments (e.g., Youd, 2003). Thus, theoccurrence of liquefaction is difficult to predict on atheoretical basis, and empirical approaches are, as arule, adopted in assessing the liquefaction potential ofan area. Various ground penetration tests (e.g.,


Bardet, 2003) are routinely applied at sites of engi-neering importance and in some metropolitan areas(e.g., Los Angeles and Memphis in the United States).On large scales, empirical relations between the mag-nitude of ground shaking and liquefaction (e.g., Wanget al., 2003) may be combined with numerical simula-tions of ground shaking during an earthquake (e.g.,Stidham et al., 1999) to evaluate the regional liquefac-tion potential. Such applications, however, requireextensive information about sediment propertiesand subsurface structures of the area of interest. Inareas where such data are not available, a simplerapproach may be applied to set some limits to theexpected extent of liquefaction during potentialearthquakes. Field observations show that, for earth-quakes of a given magnitude M, the occurrence ofliquefaction is confined within a particular distancefrom the earthquake epicenter, that is, the liquefac-tion limit Rmax beyond which liquefaction may not beexpected. Figure 2 shows a compilation of data forthe occurrence of liquefaction (Kuribayashi andTatsuoka, 1975; Ambraseys, 1988; Papadopoulosand Lefkopulos, 1993; Galli, 2000; Wang et al.,2005a) and an empirical bound on the relationshipbetween Rmax andM. The liquefied sites at the lique-faction limit are likely to be those with the optimalconditions for liquefaction, that is, saturated soilswith high liquefaction susceptibility. Liquefaction atcloser distances may include less optimal conditionsbut are exposed to greater seismic input.

Because liquefaction damages engineered struc-tures, there has been a concerted research effort,both in the laboratory and in the field, to understand

the processes of liquefaction and to predict its occur-rences (for an overview of these efforts, see NationalResearch Council (1985), Bardet (2003), and Youd(2003)). In laboratory studies, sediment and soil sam-ples are often obtained from the field and subjected tocyclic loading. At constant amplitude of deviatoricstress (stress minus the mean normal stress), defor-mation may be stable up to some number of cycles ofshearing, but sudden onset of large shear deformation(onset of liquefaction) may occur if the cycles ofshearing exceed some threshold number (e.g.,National Research Council, 1985). This onset ofliquefaction marks the moment when the stress pathbetween the deviatoric stress and the effective stress(defined later in eqn [15], which decreases withincreasing pore pressure) eventually intersects thefailure envelope of sediments and soils.

These laboratory studies have helped to advanceour understanding of the processes of liquefactionand to identify the major physical factors that affectliquefaction. However, it is difficult to duplicate nat-ural conditions in the laboratory or to obtainundisturbed samples from the field. Thus laboratorystudies may have limited practical application. In thefield, various penetration tests, such as the ‘standard’penetration test, are employed to assess the liquefac-tion potential of specific sites. Despite their wide

Figure 1 Smear left on a building wall created by ejectedsand during the 1999 M 7.5 Chi-Chi earthquake in Taiwan.Motorcycle at lower left shows scale. Reproduced from SuT-C, Chiang K-W, Lin S-J, Wang F-G, and Duann S-W(2000) Field reconnaissance and preliminary assessment ofliquefaction in Yuan-Lin area. Sino-Geotechnics 77: 29–38.

Magnitude

Dis

tanc

e (k

m)

4 5 6 7 8 9 10

10

100

1000

Figure 2 Relationship between earthquake momentmagnitude and maximum distance between the epicenterand occurrences of liquefaction. The solid line is anempirical upper bound for the maximum distance overwhich shallow liquefaction has been observed (Wang et al.,2005a): M¼%5.0þ2.26 logRmax, where Rmax is in meters.Data shown based on the compilation inWang et al. (2005a).


application, these tests are empirical. Thus, muchprogress can still be made toward a better physicalunderstanding of the liquefaction process.

4.10.3 Observations and TheirExplanations

Here we review hydrologic phenomena that accom-pany earthquakes. We will see that the magnitude ofhydrological changes caused by earthquakes arelikely to be of little human consequence, thoughdepletion of groundwater induced by seismic activityhas been invoked to explain the abandonment ofregions of Crete, Greece, during the Late Minoanperiod (Gorokhovich, 2005).

We categorize observations by the type of obser-vation (wells, streamflow, mud volcanoes, andgeysers). It will also be helpful when discussingresponses to distinguish hydrologic responses by thespatial relationship between the observation and theearthquake. We will use the terms near-field, inter-mediate-field, and far-field, for distances within onefault length, up to several fault lengths, and manyfault lengths, respectively. In the near-field, changesin the properties of the fault zone itself may beresponsible for hydrologic responses. Examplesinclude the formation of new springs along rupturedfaults (e.g., Lawson, 1908) or changes in groundwaterflow paths because of changes in the permeability ofor near the fault zone (e.g., Gudmundsson, 2000). Inboth the near- and intermediate-fields, dynamic andstatic strains are large enough that they might cause ameasurable hydrologic response. In the far-field,however, only dynamic strains are sufficiently largeto cause responses. In this chapter, we focus onhydrologic responses outside the fault zone.

Hydrologic changes following earthquakes mod-ify pore pressures, which in turn can influence localseismicity. The topic of triggered seimicity is cov-ered in the chapter by Hill and Prejean, and wesimply note here that the mechanisms responsiblefor the observed hydrological phenomena discussednext may be connected to those responsible for dis-tant, triggered seismicity.

4.10.3.1 Wells

The water level in wells measures the fluid pressureat depths the well is open to the surrounding forma-tions. Several types of coseismic and postseismicresponses are observed: water-level oscillations,

coseismic changes in water level, and delayedchanges in water level. Figure 3 shows a particularlydramatic example of a response in a well in whichwater erupted to a height of 60m above the surfacefollowing the 2004 M 9.2 Sumatra earthquake3200 km away.

4.10.3.1.1 Water-level oscillationsWater wells can act like seismometers by amplifyingground motions, in particular long-period Rayleighwaves. Water-level fluctuations as large as 6m (peak-to-peak amplitude) were recorded in Florida, thou-sands of kilometers away from the epicenter, duringthe 1964 Alaska earthquake (Cooper et al., 1965).Hydroseismograms have been recorded since theearly days of seismometer use (Blanchard andByerly, 1935).

Figure 4 (from Brodsky et al., 2003) shows waterlevel in a well in Grants Pass, OR, following the 2002M 7.9 Denali earthquake 3100 km away. Also shownis the vertical component of ground velocity mea-sured on a broadband seismometer located adjacentto the well. Large water-level fluctuations, such asthose shown in Figure 4, are unusual and occur onlyfor the right combination of geometric (water depthin the well, well radius, aquifer thickness) and hydro-logic properties (transmissivity). The frequency-dependent response of the water level in a well todynamic strain in an aquifer can be determined the-oretically by solving the coupled equations forpressure change in the aquifer, flow toward/awayfrom the well, and flow within the well (e.g.,

Figure 3 Well in China responding to the 2004 M 9.2Sumatra earthquake 3200km away. The picture was takenby Hou Banghua, Earthquake Office of Meizhou County,Guangdong, 2 days after the Sumatra earthquake. Thefountain was 50–60m high when it was first sighted, 1 dayafter the earthquake.


Cooper et al., 1965; Liu et al., 1989). In general, hightransmissivity favors large amplitudes.

Kono and Yanagidini (2006) found that in closed(as opposed to open) borehole wells, pore pressurevariations are consistent with an undrained poroelas-tic response (see Section 4.10.2.1) at seismicfrequencies. In addition, they found no pore pressureresponse to shear deformation.

4.10.3.1.2 Persistent and delayedpostseismic changes in water levelin the near and intermediate fieldPersistent changes in water level in wells are probablythe best-documented hydrologic responses to earth-quakes. Significant advances in quantitative analysis ofwater-level changes during earthquakes have beenmade during the last decade (e.g., Roeloffs, 1998;King et al., 1999; Roeloffs et al., 2003; Brodsky et al.,2003; Matsumoto and Roeloffs, 2003; Montgomeryand Manga, 2003; Wang et al., 2003; Sato et al., 2004;Kitagawa et al., 2006; Sil and Freymueller, 2006).

Sustained changes in water level in the near- and,possibly, intermediate-field can in some cases beexplained by the coseismic static strain created bythe earthquake. In this case, the water level will risein zones of contraction, and fall in regions of dilation.

Figure 5 (from Jonsson et al., 2003) shows a pat-tern of water-level change that mimics the pattern ofcoseismic volumetric strain after a strike-slip event inIceland. An analogous correlation of volumetric

strain and water-level changes was found after the2003 M 8.0 Tokachi-oki thrust event in Japan (Akitaand Matsumoto, 2004). Similar conclusions arereported by others (e.g., Wakita, 1975; Quilty andRoeloffs, 1997). This pattern of water-level changeis not, however, universal. Following the 1999 M 7.5Chi-Chi earthquake in Taiwan, the water level rosein most near-field wells where the coseismic strainshould have caused aquifer dilation (Koizumi et al.,2004). Following the 2004 M 9 Sumatra earthquake,more than 5000 km away in Japan, Kitagawa et al.,(2006) found that only about half the monitoringwells equipped with strain instruments recordedwater-level changes consistent with the measuredcoseismic strain, implying that dynamic strains canalso change water levels in the intermediate field.

Both the pattern and magnitude provide insightinto the origin of water-level changes. The magni-tude of expected water-level changes caused bycoseismic strain can be determined from models ofcoseismic strain and calibrated models for well sensi-tivity that are calibrated on the basis of their responseto barometric and tidal strains (e.g., Roeloffs, 1996).In some instances, recorded changes are similar tothose predicted (e.g., Igarashi and Wakita, 1991), butchanges can also be much larger than predicted, evenin the near-field (e.g., Igarashi and Wakita, 1995).

The time evolution of water-level changes pro-vides additional constraints on the location andmagnitude of pore pressure changes. Figure 6 (from

−4

–2

0

2

4

Time from earthquake origin time (s)

R

Wat

er le

vel (

m)

Gro

und

velo

city

(m

m s

–1)

400 500 600 700 800 900 1000 1100 1200 1300

400 500 600 700 800 900 1000 1100 1200 1300

9

9.5

10

10.5∆h

(a)

(b)

Figure 4 (a) Hydroseismograph recorded at 1 s intervals and (b) vertical component of ground velocity measured on anbroadband seismometer at a well in Grants Pass, OR, following the 2002M 7.9 Denali earthquake located 3100 km away. !hshows the 12cm permanent change in water level that followed the passage of the seismic waves. Figure based on figure 7 inBrodsky EE, Roeloffs E, Woodcock D, Gall I, and Manga M. (2003) A mechanism for sustained ground water pressurechanges induced by distant earthquakes. Journal of Geophysical Research 108: doi:10.1029/2002JB002321.


Roeloffs, 1998) shows water level for a period of 10days following the 1992 M 7.3 Landers earthquake433 km from the well. The observed water-levelchanges, shown with data points, can be modeled bya coseismic, localized pore pressure change at somedistance from the well, as shown by the solid curve.The origin, however, of the hypothesized pressurechanges cannot be determined from measurements atthe well (Roeloffs, 1998). Other processes that can

cause delayed or gradual changes in water level arelocalized changes in porosity and/or permeability(e.g., Gavrilenko et al., 2000).

4.10.3.1.3 Near-field response ofunconsolidated materialsAt shallower depths and in unconsolidated materials,changes in water level in wells often exhibit largeramplitudes and different signs than predicted bythe coseismic elastic strain. The 1999 Chi-Chiearthquake in Taiwan provided an excellent oppor-tunity to examine models for postseismic water-levelchanges because of the very high density of moni-tored wells. We discuss these observations in muchmore detail than other types of well responsesbecause the near-field hydrologic response at shallowdepths and in unconsolidated materials is much lesswell studied.

As a fortuitous coincidence, a network of 70evenly distributed hydrological stations (Figure 7),with a total of 188 wells, were installed in 1992 on alarge alluvial fan (the Choshui River fan) near theChi-Chi epicenter. The network was installed for thepurpose of monitoring the groundwater resourcesover the Choshui River fan. At each station, one tofive wells were drilled to depths ranging from 24 to306 m to monitor groundwater level in individualaquifers. Each well was instrumented with a digitalpiezometer that automatically records the ground-water level hourly, with a precision of 1mm (Hsu

0

0.1

0.2

0.3

0.4

0.5

Time (days)

h/h 0

Scaled BV response toLanders earthquake

Calculated response topressure step at time ofearthquake

0 2 4 6 8 10

Figure 6 Water level as a function of time after the 1992M7.2 Landers, CA, earthquake. The epicenter is 433 km awayfrom the well (named BV). The smooth curve shows thepredicted evolution of water level for a coseismic, localizedpressure change at some distance from the well. Water levelis normalized by the magnitude of the head change wherepore pressure changed. Reproduced from Roeloffs EA(1998) Persistent water level changes in a well nearParkfield, California, due to local and distant earthquakes.Journal of Geophysical Research 103: 869–889.

21 Junecoseismic

10 km

0.5

0

–0.5

∆p/B = –∆σkk /3 (MPa)64° N

20° W

Figure 5 Predicted (color map) and observed (circles) coseismic water-level changes following a 2000 M 6.5 strike-slipearthquake in Iceland. Black and white circles indicate water-level increases and decreases, respectively. The white lineshows the mapped surface rupture. Reproduced form Jonsson S, Segall P, Pedersen R, and Bjornsson G., (2003) Post-earthquake ground movements correlated to pore-pressure transients. Nature 424: 179–183.


et al., 1999). The dense network of hydrological sta-tions and its close proximity to a large earthquakemade the groundwater records during the Chi-Chiearthquake the most comprehensive and systematicobtained so far. In addition, a dense network ofbroadband strong-motion seismographic stationscaptured the ground motion in the area, and detailedstudies of liquefaction occurrences were made.Finally, the isotopic composition of the groundwaterwas measured before and after the earthquake (Wanget al., 2005c).

The coseismic changes of groundwater level dur-ing this earthquake have been reported in severalpapers (Hsu et al., 1999; Chia et al., 2001; Wang et al.,2001; Lee et al., 2002). Near the ruptured fault, whereconsolidated sedimentary rocks occur near the sur-face, the groundwater level showed a coseismic,stepwise decrease (Figures 8(a) and 8(b)). Awayfrom the ruptured fault, in the fan of unconsolidatedsediments, confined aquifers showed a distinctly dif-ferent pattern of coseismic changes than in theuppermost aquifer. In the confined aquifers, ground-water levels showed a coseismic stepwise rise(Figures 8(c) and 8(d)). The magnitude of this riseincreases with distance from the ruptured fault,reaching more than 5 m at distances of 20–30 km

away from the fault before decreasing at greaterdistances (Figure 7(a); Wang et al., 2001). Inthe uppermost aquifer, the coseismic changes in thegroundwater level were much smaller in magnitudewith an irregular pattern (Figure 7(b)). Liquefactionsites on the Choshui River fan are clearly correlatedwith significant coseismic rise of the groundwaterlevel in the uppermost aquifer, but no such correla-tion exists for the confined aquifers (Figure 7).This observation strongly indicates that liquefactionoccurred in the uppermost aquifer but not inthe confined aquifers. In the basins nearest the rup-tured fault (east of the Choshui River fan), acomparison between liquefaction sites and the coseis-mic change in the groundwater level cannot bemade, because no wells were installed and thus nogroundwater-level data were available in thesebasins.

Lee et al. (2002) claimed that the spatial distribu-tion of the groundwater-level change during theChi-Chi earthquake may be accounted for by aporoelastic model. The widespread occurrence ofliquefaction, however, provides clear evidence thatat least part of the sediment deformation was notelastic. Furthermore, for confined aquifers that havea typical thickness of 1–10m, the required changes in

24.13

(a) (b)

N N

0 10 20km

23.77

23.41

24.13

23.77

23.41

120.11 120.51 120.90 120.11 120.51 120.90

0 10 20km

0.5

00.5

0.5

0

0

0 1

3

5

3

5

5

Figure 7 Spatial distribution of coseismic changes in groundwater level during the Chi-Chi earthquake. (a) Contours (m)showing coseismic groundwater-level changes in a confined aquifer (Aquifer II) in the Choshui River fan. (b) Contours (m)showing coseismic groundwater-level changes in the uppermost aquifer (Aquifer II) in the Choshui River fan. Groundwatermonitoring stations are shown in open circles, liquefaction sites in open diamonds, epicenter of Chi-Chi earthquake in star,and the ruptured fault in discontinuous red traces. Reproduced from Wang C-Y Wang C-H, and Kuo C-Y (2004a) Temporalchange in groundwater level following the 1999 MW¼7.5 Chi-Chi earthquake, Taiwan. Geofluids 4: 210–220.


porosity would have to be from 10%3 to 10%2 toaccount for the range of coseismic groundwater-level changes (Wang et al., 2001), beyond the magni-tude of strains calculated from an elastic model (Leeet al., 2002). Finally, the sign of the changes in waterlevel are in general the opposite of those expectedfrom a poroelastic model (Koizumi et al., 2004).

As noted in Section 4.10.2.2, when loose sedimentsand soils are sheared at stresses between some upperand lower thresholds, grains tend to move together,leading to a decrease in porosity. If the deformation isundrained, as expected during seismic shaking, porepressure would increase when the strain exceedsabout 10%4. This may explain the stepwise rise inthe groundwater level in the fan (Figures 8(c) and8(d)). On the other hand, when consolidated sedi-ments or sedimentary rocks are cyclically shearedbeyond some critical threshold, microfracture andfracture may occur, leading to increased porosityand decreased pore pressure. This may explain thecoseismic decline in the groundwater level in sedi-mentary rocks close to the ruptured fault (Figures8(a) and 8(b)).

It is possible to make an order-of-magnitude esti-mate of the change in porosity required to account

for the coseismic water-level change. Becausethe coseismic water-level change does not involvefluid flow, the mass of pore water must be conserved,that is,

d &(ð Þdt

¼ 0 ½9$

where & again denotes the porosity. Equation (9) canbe rewritten as

S

b

dh

dtþ d&

dt¼ 0 ½10$

where S, as before, is the aquifer storativity, h thehydraulic head, and b the aquifer thickness. Thesecond term in eqn [10] is the rate of change ofsediment porosity at a given location resulting fromground shaking, which we assume is proportional tothe intensity of ground shaking. Given that the inten-sity of ground shaking during the Chi-Chiearthquake decayed approximately exponentiallywith time (Ma et al., 1999), the rate of change insediment porosity at a given location under earth-quake shaking may be expressed as

d&

dt¼ Ae –*t ½11$

120

116

112

–1.8

–2.4

–3.0

–3.6

–4.2

–4.8

180

176

172

168

30

25

20

15

10

5

Chukou l

(a) (b)

(c) (d)

Liyu II

Hsikang IVYuanlin I

Time (days)

h (m

)h

(m)

Time (days)

0 20 40 60 80 100 120 0 20 40 60 80 100 120

Figure 8 Four types of groundwater level responses during the Chi-Chi earthquake: (a) Groundwater level dropped duringthe earthquake and gradually rose afterward; (b) groundwater level dropped during the earthquake and continued to dropgradually afterward; (c) groundwater level rose during the earthquake and gradually declined afterwards; (d) groundwaterlevel rose during the earthquake and remained at the new level for some time before slowly declining. The differentgroundwater-level responses may be due to differences in the proximity to the ruptured fault and different hydrogeologicalconditions as discussed in Wang et al. (2001). Figure created by Chung-Ho Wang.


where * is of the order of 10%2–10%1 s%1, and A is aconstant to be determined. Substituting eqn [11] intoeqn [10] and integrating with respect to time, we getthe coseismic change in the hydraulic head:

!hðtÞ ¼ bA

S*e –*t – 1# $

½12$

Pumping tests in the Choshui River fan yieldS& 10%3 (Lee and Wu, 1996). Given that groundshaking on the Choshui River fan during the Chi-Chi earthquake lasted between 20 and 100 s (Leeet al., 2002a), for a 10m groundwater-level changein a 1m thick confined aquifer, A¼%10%4 to%10%3 s%1, and for the same level of change in a10m aquifer, A¼%10%5 to %10%4 s%1. Integratingeqn [10], we also obtain a coseismic change in por-osity between 10%3 and 10%2.

The uppermost aquifer on the Choshui River fan iseither unconfined or poorly confined, resulting in lessregular changes in pore pressure and groundwaterlevel. Thus, qualitatively, the processes of consolida-tion and dilatation of sediments and sedimentaryrocks, respectively, appear to provide a self-consistentinterpretation of the spatial pattern of the coseismicgroundwater-level changes on the Choshui River fanduring the Chi-Chi earthquake (Wang et al., 2001).

The availability of both well records and strong-motion records at the time of the Chi-Chi earthquake

allows quantitative examination of the relationshipbetween ground motion, groundwater-level change,and the distribution of liquefaction sites.Interpolating the strong-motion parameters at thewell sites, Wang et al. (2003) showed that there isonly a weak correlation between the horizontal com-ponent of the peak ground acceleration (PGA) andthe occurrence of water-level changes and liquefac-tion (Figure 9(a)). Wong (2005) confirmed thisfinding by using a more extensive data set. Thisresult is unexpected, because PGA is frequentlyused to predict the occurrence of liquefaction.Instead, Wong (2005) showed that there is a bettercorrelation between the horizontal peak ground velo-city (PGV) and the sites of liquefaction. BecausePGV depends more on low-frequency componentsof ground motion than does PGA, this indicates thatlow-frequency ground motions are better correlatedwith elevated pore pressure and groundwater levelthan are high-frequency motions. This conclusionsupports the finding that spectral accelerations andvelocities at frequencies of about 1Hz and lowerwere strongly correlated with the distribution ofliquefaction sites (Figure 9(b)), whereas thoseabove about 1Hz were not (Wang et al., 2003;Wong, 2005). A similar conclusion was reached bycomparing the hydrologic changes with various mea-sures of earthquake intensity computed from

24.13

(b)

N

0 10 20km

23.77

23.41

120.11 120.51 120.90

6

8

10

6

9

812

10 16

18

2016

1614

14

12

8

24

12

4

16 18

20

24.13

(a)

N

0 10 20km

23.77

23.41

120.11 120.51 120.90

1

1.5 2.5

3.5

4

53

2

2

3

5

6

7

67

43.

5

Figure 9 (a) Contours of the horizontal PGA during the Chi-Chi earthquake together with liquefaction sites in opendiamonds. Note the weak correlation between PGA and liquefaction sites. (b) Contours of the spectral acceleration at 1Hz.Note the fairly strong correlation between this and liquefaction sites.


seismograms passed through low-pass and high-passfilters (Wong, 2005). Although there is extensiveevidence that low-frequency ground motions arebetter correlated with elevated pore pressure andgroundwater level than high-frequency motions, itis unclear at this time whether the low-frequencyground motion was the cause for the coseismicgroundwater-level change and liquefaction, orwhether the hydrological changes, including lique-faction, caused the different spectral composition ofthe observed seismograms.

4.10.3.1.4 Far-field responseIn the far-field (many fault lengths), the static stressdue to the earthquake is nearly zero. Thus, sustainedchanges in water level at such distances (e.g., Figure 4)must be caused by the interactions between the aquiferand seismic waves. Such changes are not common, butconfined to a small number of unusually sensitive sitesusually located near active faults (King et al., 2006).The mechanism for these interactions, however, is amatter of debate (Roeloffs, 1998; Matsumoto andRoeloffs, 2003; Brodsky et al., 2003).

A permanent change in some aspect of pore struc-ture is required, but the process by which smallperiodic strains change permeability or storage prop-erties is unclear. Hydrogeochemical (e.g., Wang et al.,2004d) and temperature (e.g., Mogi et al., 1989)changes that accompany water-level changes confirmthat pore-space connectivities change. One mechan-ism that has been proposed (to explain the particularpostseismic change in water level shown in Figure 4)is that oscillatory flow back and forth in fracturescaused by cyclic strain removes ‘barriers’ of frac-ture-blocking deposits, which then increasespermeability and affects the final distribution ofpore pressure (Brodsky et al., 2003). Elkhoury et al.(2006) confirmed that distant earthquakes can indeedchange permeability. Other proposed, but also unver-ified, mechanisms include pore pressure increasescaused by a mechanism ‘akin to liquefaction’(Roeloffs, 1998), shaking-induced dilatancy (Bowerand Heaton, 1978), or increasing pore pressurethrough seismically induced growth of bubbles (e.g.,Linde et al., 1994). A feature of at least some far-fieldresponses is that the magnitude of water-levelchanges scales with the amplitude of seismic waves(Sil and Freymueller, 2006).

One common feature of wells that exhibit far-fieldpostseismic responses is that the sign of the water-levelchanges always appears to be the same (Matsumoto,1992; Roeloffs, 1998; Sil and Freymueller, 2006). Any

proposed mechanism should explain this latter feature,and the barrier explanation can only do so if the loca-tion of barriers is always located either upgradient ordowngradient of the well. Far-field shaking-induceddilatancy (e.g., Bower and Heaton, 1978) or a processanalogous to liquefaction (Roeloffs, 1998) wouldalways cause the response to have the same sign, butnone of them can be universal mechanisms because thesign of water-level changes varies from well to well.

Observations from wells that have responded tomultiple earthquakes indicate that wells may have anenhanced sensitivity over a specific range of frequen-cies (Roeloffs, 1998). Interestingly, the threshold fortriggered earthquakes also appears to have a fre-quency dependence (Brodsky and Prejean, 2005),though the frequency sensitivities for the two studiesjust cited are different. Observations of, and modelsthat account for, the frequency-dependence responsemay be useful for distinguishing between processes.

4.10.3.1.5 SummaryTo generalize, on the basis of the reported changes inwater level in the near- and intermediate-field, itappears that deep wells in sound rock display aporoelastic response that mimicks coseismic strain(both in sign and magnitude), whereas shallow wellsand wells in poorly consolidated materials tend toshow larger changes and increases in water level.

4.10.3.2 Streamflow

One of the most spectacular surface hydrologicresponses to earthquakes is large changes in stream-flow. Figure 10 shows two typical examples. Theincreased discharge is persistent, at least until rainfallobscures the changes caused by the earthquake. Thepeak discharge can occur from within a day to asmuch as several weeks after the earthquake.Increased streamflow occurs in the near- and inter-mediate-field (Manga, 2001). The total excessdischarge (i.e., the total volume of water released inexcess of that expected in the absence of the earth-quake) can be large: 0.7 km3 for the M 7.5 Chi-Chiearthquake (Wang et al., 2004b), and 0.5 km3 after theM 7.5 Hebgen Lake earthquake (Muir-Wood andKing, 1993). Because streamflow responds to preci-pitation in the drainage basin, earthquake-inducedchanges are best recorded and studied during thedry season or during dry periods when there is littleor no precipitation.

Explanations for changes in streamflow can bedivided into five categories: expulsion of deep crustal


fluids resulting from coseismic elastic strain (e.g.,Muir-Wood and King, 1993), changes in near-surfacepermeability (Briggs, 1991; Rojstaczer and Wolf,1992; Rojstaczer et al., 1995; Tokunaga, 1999; Satoet al., 2000), consolidation or even liquefaction ofnear-surface deposits (Manga, 2001; Manga et al.,2003; Montgomery et al., 2003), rupturing of subsur-face reservoirs (Wang et al., 2004c), and the release ofwater trapped in fault zones. The differencesbetween these different explanations are nontrivialbecause of their implications for the magnitude ofcrustal permeability, its evolution, and the nature ofgroundwater flow paths. We thus summarize thebasis and some problems with each explanation andthen consider the streamflow response to the 1999Chi-Chi earthquake as an illustrative example:

1. Coseismic elastic strain. Muir-Wood and King(1993) applied the coseismic elastic strain model,proposed by Wakita (1975) to explain coseismicgroundwater-level changes, to explain the increasedstream discharge after some large earthquakes,including the 1959 M 7.5 Hebgen Lake earthquakeand the 1983 M 7.3 Borah Peak earthquake. Muir-Wood and King (1993) suggested that saturatedmicrocracks in rocks opened and closed in responseto the stress change during an earthquake, causing

pore volume either to increase or decrease, resultingin a decrease or increase in the groundwater dis-charge into streams.

Following the 1989 Loma Prieta earthquake, how-ever, chemical analysis of the stream water showedthat the extra water that appeared in the streamsfollowing the earthquake had a shallow, rather thandeep, origin (Rojstaczer and Wolf, 1992; Rojstaczeret al., 1995). Furthermore, Rojstaczer and colleaguesshowed that to account for the extra water in theincreased streamflow by coseismic elastic strain, avery large volume of the crust must be involved inthe expulsion of groundwater, which in turn wouldrequire an unreasonably high permeability for thecrust. In addition, one important constraint on modelsis that streamflow generally increases, and the fewstreams with decreased discharge show very smallchanges.

Manga et al. (2003) reported that streamflowincreased at Sespe Creek, California, after several earth-quakes, irrespective of whether the coseismic strain inthe basin was contraction or dilatation. For this reason,the dynamic strain caused by the earthquake must beresponsible for the observed changes in discharge.

2. Enhanced permeability. Briggs (1991) andRojstaczer and Wolf (1992; also Rojstaczer et al.,1995) proposed a model of enhanced permeability ofthe shallow crust resulting from seismically inducedfractures and microfractures to explain the increasedstream discharge in the nearby basins and the changesin the ionic concentration of stream water followingthe 1989 Loma Prieta earthquake in California. Ashallow origin of the excess water is supported by adecrease in water temperature after dischargeincreased. Similar models were applied to the 1995Kobe earthquake in Japan to explain the observedhydrological changes (Tokunaga, 1999; Sato et al.,2000). Permeability enhancement was also invokedto explain the increased electrical conductivity ofwater discharged after an earthquake (Chamoilleet al., 2005). Elkhoury et al., (2006) found that earth-quakes in southern California caused phase shifts inthe water-level response to tidal strain, and inter-preted these to be due to increased permeability byseismic waves. The increase in permeability may betemporary because biogeochemical processes may actto reseal or block any newly created fresh fractures.

Manga (2001) found, however, that the rate of base-flow recession (i.e., the slow decrease in dischargeduring periods without precipitation) was unchangedby earthquakes that caused increases in streamflow.Assuming that the groundwater discharge Q to the

0.1

0.01

0.05

EQ

(a)

(b)

San Lorenzo River, CA

1

3

0.6

1991.6 1991.7 1991.8Date

Dis

char

ge (

m3 s

–1)

EQ

Mattole River, CA

Aug. 1989 Sep. 1989

Figure 10 Hydrographs of (a) San Lorenzo River, CA, and(b) Mattole River, CA, showing postseismic response to the1989 M 7.3 Loma Prieta earthquake and 1991 M 6Honeydew earthquake, respectively. The vertical linelabeled EQ indicates the time of earthquake. Small arrowsshow the time of rainfall events. Figure made with USGeological Survey stream gauge data.


stream is governed by Darcy’s equation [5], baseflowrecession (a long time after recharge) is given by

dQ

dt_ exp½ – aDt $ ½13$

where a is a constant that characterizes the geometryof the drainage basin, D is the hydraulic diffusivity,and t is time. Although this is a great simplification ofthe complex subsurface flow paths of water tostreams, such models in general provide an excellentempirical fit to discharge records. The slope of thehydrograph in Figure 11(a) is proportional to aD,and Figure 11(b) shows that the recession constantdoes not change following the earthquake. Given thatthe basin geometry and flow pathways are unlikely tohave been reorganized by the earthquake, anunchanged baseflow recession implies that aquiferpermeability did not change after the earthquake,even over the time period of increased discharge.This conclusion was substantiated by later studiesin other areas (Montgomery et al., 2003) and duringother earthquakes (Manga et al., 2003).

3. Coseimic consolidation and liquefaction.Consolidation of loose materials is one way to

increase pore pressures. Indeed, liquefactionoccurred in many of the areas where streamflowincreased (Montgomery and Manga, 2003).Consequently, Manga (2001; also Manga et al., 2003)suggested that coseismic liquefaction of loosesediments on floodplains may provide the water forthe increases in stream discharge followingearthquakes.

Figure 12 shows the relationship betweenearthquake magnitude and distance between the epi-center and the center of the gauged basin for streamsthat responded to earthquakes. Also shown for refer-ence is the maximum distance over whichliquefaction has been reported (from Figure 2).That the limit for both responses is similar does notimply that liquefaction causes streamflow to increase.Instead, the correlation simply means that dynamicstrains sufficient to induce liquefaction underoptimal conditions are also sufficient for streamflowto change.

Wang et al., (2004b), however, used the extensivenetwork of stream gauges in Taiwan to identify thesource of excess discharge in streams. Excess dis-charge did not originate on the unconsolidated

Date

0

0.01

0.02

0.03

Recession of avg daily hydrographfrom (a)

100

10

1

0.1

San Lorenzo River near Boulder Creek, CA

(0.98 ± 0.07) × 10–2 day–1; R 2 = 0.90

(1.07 ± 0.04) × 10–2 day–1; R 2 = 0.95

Date

Dis

char

ge (

cfs)

Rec

essi

on c

onst

ant

(a)

(b)

1989.6 1989.8 1990

1970 1975 1980 1985 1990

Figure 11 (a) Hydrograph of the San Lorenzo River, CA, showing postseismic response to the 1989 M 7.3 Loma Prietaearthquake. The vertical line labeled EQ indicates the time of the earthquake. The postseismic period of baseflow recession isshown by the bold sloping line. (b) The baseflow recession constant for periods of baseflow before and after the earthquakeshows that even though discharge increased by an order of magnitude there was no significant change in baseflow recession.Figure made with US Geological Survey stream gauge data. Modified fromManga M (2001) Origin of postseismic streamflowchanges inferred from baseflow recession and magnitude-distance relations. Geophysical Research Letters 28: 2133–2136.


alluvial fan where there was, in fact, widespreadliquefaction. Rather, the excess discharge originatedin the mountains where there is little alluvial mate-rial and groundwater originates in bedrock. Inaddition, after the 2001 M 6.8 Nisqually, WA, earth-quake, Montgomery et al., (2003) did not find asignificant field association between liquefactionoccurrence and streamflow changes.

4. Ruptured subsurface reservoirs. Following the M6.5 San Simeon, CA, earthquake on 22 December2003, a different type of streamflow increase wasrevealed in the central Coast Ranges, occurringwithin 15min of the earthquake and lasting aboutan hour. At the same time, several new hot springsappeared across a dry riverbed, a short distanceupstream of the stream gauge that registered theabrupt increase in streamflow. These observationssuggest that the occurrence of streamflow increaseand the appearance of hot springs were causallyrelated and may be the result of earthquake-inducedrupture of the seal of a hydrothermal reservoir.Recession analysis of the stream gauge data showD/L2&1000 s%1, where L is the depth of the reservoir,implying a small L or large D or both. Assuming aone-dimensional model for flow between theruptured seal and the stream, Wang et al., (2004c)estimated that the extra discharge after this

earthquake was &103 m3. Abrupt changes in hot-spring discharge are known to occur in theLong Valley, California hydrothermal area (Soreyand Clark, 1981), suggesting that this type ofhydrologic response may be characteristic of hydro-thermal areas.

Rupturing permeability barriers may occur atdepths of many kilometers. Husen and Kissling(2001) suggest that postseismic changes in the ratioof P-wave and S-wave velocities above the subductingNazca Plate reflect fluid migration into the overlyingplate following the rupture of permeability barriers.

5. Fault valves. In convergent tectonic regions,large volumes of pore water may be locked in sub-ducted sediments (Townend, 1997). Sealing may beenacted partly by the presence of low-permeabilitymud and partly by the prevailing compressionalstresses in such tectonic settings (Sibson andRowland, 2003). Earthquakes may rupture the sealsand allow pressurized pore water to erupt to thesurface and recharge streams. This process mayexplain time variations in submarine fluid dischargeat convergent margins (Carson and Screaton, 1998).Episodes of high discharge are correlated with seis-mic activity having features similar to tremor but arenot correlated with large regional earthquakes(Brown et al., 2005).

To illustrate some approaches that are used to testthese explanations, we again consider the recordsfrom the 1999 M 7.5 Chi-Chi earthquake in centralTaiwan. In central Taiwan, 17 stream gauges arelocated on three streams systems (Figure 13(a))that showed different responses to the 1999 Chi-Chiearthquake (Wang et al., 2004b). In addition, the localrainy season was over well before the earthquake.These records have made it possible to test the dif-ferent hypotheses listed above. Among the threestreams, two (Choshui Stream and the WushiStream) have extensive tributaries in the mountains(Figure 13(a)), but the third (Peikang Stream) origi-nates on the sloping side of the Choshui fan, with notributaries in the mountains. All the streams in themountains showed large postseismic streamflowincreases. On the alluvial fan, both the ChoshuiStream and theWushi Stream, which have tributariesin the mountains, showed large increases in stream-flow. The amount of increase in streamflow in theproximal area of the Choshui alluvial fan was aboutthe same as that in the distal area of the fan, indicat-ing that most of the excess water originated in themountains and that the contribution from the

Dis

tanc

e (k

m)

Magnitude

8 10

10

100

1000

4 5 6 7 9

Figure 12 Relationship between earthquake magnitudeand distance from the epicenter of streams that exhibitedclear and persistent increases in discharge. The solid line isan empirical upper bound for the maximum distance overwhich shallow liquefaction has been observed (Wang et al.,2005a): M ¼ –5.0 þ 2.26 log Rmax, where Rmax is in meters.Data shown based on the compilation in Montgomery andManga (2003) with additional data from Wang et al. (2005a).


sediments in the fan was insignificant. By contrast,the Peikang Stream system, which does not havetributaries in the mountainous area, did not showany noticeable postseismic increase in streamflow.Thus it may be concluded that the excess water inthe streams after the Chi-Chi earthquake requires anearthquake-induced discharge from the mountains.Any coseismic consolidation and liquefaction of sedi-ments on the floodplain (alluvial fan) were notsubstantial enough to cause noticeable postseismicincrease in streamflow.

Figure 13(b) shows the discharge of a stream inthe mountains before and after the Chi-Chi earth-quake, together with the precipitation record from anearby station. In addition to a coseismic increase instreamflow and a slow postseismic return to normal,the figure shows that the baseflow recession ratesboth before and after the earthquake are identical.This observation is of interest because it suggests thatthe Chi-Chi earthquake did not cause significantchanges in the permeability of the aquifer that fedthe stream, as Manga (2001) first noticed in his studyof the coseismic responses of streams in the UnitedStates. However, the excess water in streams in cen-tral Taiwan after the Chi-Chi earthquake requiresextra discharge from the mountains that can appearonly through enhanced permeability during theearthquake. This apparent contradiction may beresolved by recognizing that the earthquake-inducedchange in permeability may be anisotropic. The foot-hills of the Taiwan mountains consist of alternatingbeds of sandstone and shale. Thus, before theearthquake, the vertical permeability would be con-trolled by the impervious shales and groundwaterflow would be mostly subhorizontal. During theearthquake, subvertical fractures breached theimpervious shales and enhanced vertical permeabil-ity, allowing rapid downward draining of water torecharge underlying aquifers. The horizontal con-ductivity of the aquifers, however, was essentiallyunaffected. Field surveys after the Chi-Chi earth-quake revealed numerous subvertical tensile cracksin the hanging wall of the thrust fault (Angelier et al.,2000; Lee et al., 2000, 2002b). Many wells in thefoothills above the thrust fault showed a significantdrop in water level (Lin, 2000; Yan, 2001; Chia et al.,2001; Wang et al., 2001), and a sudden downpouroccurred in a tunnel beneath the foothills right afterthe earthquake. These reports are consistent with themodel of enhanced vertical permeability during theChi-Chi earthquake.

(a)

(b)

(c)

(d)

24.22° N

23.86° N

23.50° N

120.12° E 120.90° E

B

N

0 km 20

Choshui S

Peikang S.

Wushi S025 028

038

058 057 064063

049

024029

014009

042032

037

54328/1 9/1 10/1 11/1 12/1

H032

0

100D

aily precipitation (mm

)

In q

In q

x = L x = Ox = L′

4

3

2

0 10 3020t (day)

40 50

Data (H032)Model

A

Figure 13 (a) Map showing the three river systems incentral Taiwan and the locations of the stream gauges (intriangles). Note that both the Wushi River and the ChoshuiRiver have extensive tributaries in the mountains, but thePeikang River does not. Hydrological monitoring wells aremarked by bull’s-eyes. (b) Logarithm of discharge of astream as recorded by a gauge in central Taiwan as afunction of time before and after the Chi-Chi earthquake,together with precipitation record at a nearby station. Notethat the recession curve before and after the earthquake(indicated by arrow) has the same slope. (c) Conceptualmodel showing increased recharge of aquifer bygroundwater from mountains due to enhanced verticalpermeability during and after the earthquake. (d)Comparison of excess discharge as a function of time withmodel result (thin curve). Excess discharge at the time ofearthquake (t¼ 0) is taken to be zero and therefore does notappear on the diagram.


A one dimensional leaky aquifer model(Figure 13(c)) may be used to quantify this concep-tual model. Flow is governed by eqn [7], and therecharge to the aquifer resulting from enhanced ver-tical permeability is treated as a source. Even thoughthis model is highly simplified, several studies (e.g.,Roeloffs, 1998; Manga, 2001; Manga et al., 2003;Brodsky et al., 2003) have demonstrated that it isuseful for characterizing the first-order response ofhydrological systems to earthquakes.

We treat the recharge of the leaky aquifer ascoseismic with constant recharge for 0 < x* L9

(length of aquifer beneath the mountain withenhanced permeability) and no recharge for L9 > x> L (length of aquifer beneath the floodplain). If wefurther assume a boundary condition h ¼ 0 at x ¼ Land no-flow condition at x ¼ 0, the excess dischargeto the stream at time t is given by

Qex ¼2DVQ0

L2ðL9=LÞX1

n¼1

ð – 1Þn – 1sin ð2r – 1Þ+L92L

! "

) exp –ð2r – 1Þ2+2D

4L2t

! "½14$

where L is the total length of the aquifer and V is thevolume of the aquifer between x ¼ 0 and x ¼ L9(Figure 13(a)); thus, VQ0 is the total excess waterrecharging the aquifer. For large times, eqn [14]results in an exponential decrease in discharge, theso-called baseflow recession given by eqn [13].Figure 13(d) compares the data for stream gauge032 (adjusted to Qex¼ 0 just before earthquake)with the modeled excess discharge Qex determinedby eqn [14]. An excellent fit is obtained with D/L2 ¼2.4) 10%7 s%1(from recession analysis) and L9/L ¼0.8. The latter value is consistent with the fact thatgauge 032 is located in the mountains, where thefloodplain is small in comparison with the total aqui-fer length. The amount of excess water VQ0 may bedetermined from the value of DVQ0/L

2 from modelfitting. By summing the excess discharges in theChoshui Stream and in the Wushi Stream, we obtaina total amount of 0.7–0.8 km3 for the excess waterreleased from the mountains after the Chi-Chiearthquake.

For the example shown in Figure 13, the peakpostseismic discharge occurs within a day of theearthquake. Figure 14 shows an example in whichthe peak discharge is reached 9–10 days later, thoughdischarge begins to increase coseismically. Thisexample is for Sespe Creek, CA, responding to the1952 M 7.5 Kern County earthquake located 63 km

away from the center of the drainage basin. Again, themodel for excess discharge given by eqn [14], shownin the solid curve in Figure 14, fits the observedpostseismic discharge very well (baseflow describedby eqn [13] has been added back to the calculatedexcess discharge).

The different hypotheses listed in this sectionimply different crustal processes and differentwater–rock interactions during an earthquake cycle.In most instances, the hydrological models are under-constrained. A reasonable approach is to test thedifferent hypotheses against cases (such as Chi-Chi)in which abundant and accurate data are available.We note that a single explanation need not apply forall cases of increased streamflow, so that identifyingwhen and where different mechanisms dominate isimportant.

4.10.3.3 Mud Volcanoes

Mud volcanoes also appear to erupt in response toearthquakes (e.g., Panahi, 2005) and thus provideanother probe of earthquake–hydrology interactions.Unfortunately though, the number and quality ofreports prevent a rigorous statistical analysis of thecorrelation (Manga and Brodsky, 2006). Mud volca-noes range from small, centimeter-sized structures tolarge edifices up to few hundred meters high andseveral kilometers across. Mud volcanoes erupt pre-dominantly water and fine sediment. They occur inregions where high sedimentation rates and fine-grained materials allow high pore pressures todevelop (Kopf, 2002). Large mud volcanoes eruptfrom depths of more than several hundred meters.

A necessary condition to create mud volcanoes isthe liquefaction or fluidization of erupted materials,

Date1952.5 1952.7

0.1

1

Dis

char

ge (

m3 s

–1)

Figure 14 Response of Sespe Creek, CA to the 1952 M7.5 Kern County earthquake. Daily dischargemeasurements collected and provided by the USGeological Survey are shown with circles. Curve is solutionto eqn [14] for the excess flow with L’/L ¼ 0.4 added to thebaseflow, eqn [13], to recover the entire hydrograph.Vertical line shows the time of the earthquake. There was noprecipitation during the time interval shown in this graph.


so that the erupted materials lose strength and canbehave in a liquid-like manner (Pralle et al., 2003).Unconsolidated sediment can be liquefied and flui-dized through mineral dehydration, gas expansion,tectonic stresses, inflow of externally derived fluids,and even ocean waves (Maltman and Bolton, 2003).

Figure 15 shows that the occurrence of triggeredmud volcanism falls within the liquefaction limit forshallow (upper few to tens of meters) liquefaction.Reports of mud volcanoes erupting within a day oflarge, distant earthquakes are compiled in Manga andBrodsky (2006) and are listed in the caption ofFigure 15.

Liquefaction is usually thought to be a shallowphenomenon confined to the upper few to tens ofmeters. This is because the increase in fluid pressureneeded to reach lithostatic pressure usually increaseswith depth. However, sedimentary basins with highsedimentation rates, fine sediments (with low perme-ability), and lateral compression often have high fluidpressures. Indeed, it is these settings in which mudvolcanoes seem to occur (Milkov, 2000). Seismicallytriggered liquefaction may not necessarily be only ashallow phenomenon.

4.10.3.4 Geysers

Geysers change eruption frequency following distantearthquakes that generate coseismic static strainssmaller than 10%7 or dynamic strains less than 10%6

(e.g., Hutchinson, 1985; Silver and Vallette-Silver,1992). Geysers, despite requiring very special ther-mal conditions and hydrogeological properties tooccur (Steinberg et al., 1982a), thus provide anotheropportunity to understand how earthquakes caninfluence hydrogeological processes and properties.

Figure 16 shows the response of Daisy geyser inYellowstone National Park to the 2002 M 7.9 Denaliearthquake located more than 3000 km away (Husenet al., 2004). The eruption interval decreases by abouta factor of 2 and then slowly increases to the pre-earthquake frequency over a period of a few months.Among the many geysers at Yellowstone, the erup-tion frequency of some increased, whereas for othersit decreased.

Although geysers often respond to earthquakes,their response to barometric pressure changes (e.g.,White, 1967), solid Earth tides (e.g., Rinehart, 1972),and hypothetical preseismic strains (Silver andVallette-Silver, 1992) is weaker. The apparent

Dis

tanc

e (k

m)

Magnitude4 6 8 10

10

100

1000

5 7 9

Figure 15 Relationship between earthquake magnitudeand distance from the epicenter of large mud volcanoes thatoriginate from depths greater than hundreds of meters anderupt within 2 days of the earthquake; shown are AndamanIslands (responding to M 9.2 Sumatra earthquake) andNiikappu Mud Volcano, Hokkaido, Japan (responding toevents withM 7.1, 7.9, 7.9, 8.2, and 8.2, in 1982, 1968, 2003,1952, and 1994, respectively). The solid line is an empiricalupper bound for the maximum distance over which shallowliquefaction has been observed (Wang et al., 2005): M ¼%5.0þ 2.26 log Rmax, whereRmax is in meters Niikappu dataprovided by Nakamukae Makoto.

1:00

2:00

3:00

Eru

ptio

n in

terv

al (

h : m

in)

DFE

02:30:12(02:39:47)

01:57:41(01:36:46)

1Sep.

15Sep.

29Sep.

13Oct.

27Oct.

10Nov.

24Nov.

8Dec.

22Dec.

Date (2002)

Figure 16 Eruption interval (time between two successiveeruptions) in h :min format as a function of times Daisygeyser, Yellowstone National Park. Thin grey line shows rawdata, that is, the actual measured eruption intervals. Theblack line is smoothed data obtained by averaging overseveral measurements. Median eruption intervals are alsoshown in h :min : s format. Median eruption intervals arecomputed over a several-week period; intervals inparentheses are based on a few days. Time of the 2002 M7.9 Denali earthquake is shown be the vertical line. Figureprovided by Stephan Husen and based on data presented inHusen S, Taylor R, Smith RB, and Healser (2004) Changes ingeyser eruption behavior and remotely triggered seismicityin Yellowstone. National park, produced by the 2002 M 7.9Denali fault earthquake, Alaska. Geology 32: 537–540.


correlation of eruption interval and nonseismicstrains is, in fact, controversial. Rojstaczer et al.(2003) reanalyzed the eruption records atYellowstone and found that geysers were insensitiveto earth tides and diurnal barometric pressurechanges, and thus concluded that geysers were notsensitive to strains less than about 10%8%10%7. One ofthese records was from the Daisy geyser (Figure 16).Given that Daisy geyser responded to earthquakesthat caused equally small static strain, we can reason-ably conclude that it is the dynamic strain, ratherthan static strain, that causes the observed responses(e.g., Husen et al., 2004).

The mechanism by which the eruption intervalchanges is not known. Steinberg et al. (1982b) notethat one mechanism for initiating a geyser eruption issuperheating the water. Using a lab model of a geyser,Steinberg et al. (1982c) show that vibrations canreduce the degree of superheating and can thusincrease the eruption frequency. It is worth notingthat while earthquakes can decrease the intervalbetween eruptions, sometimes the frequency of erup-tions decreases. Ingebritsen and Rojstaczer (1993)propose instead that much of the observed temporalvariability in eruption frequency can be explained bychanges in matrix permeability, which in turn gov-erns the recharge of the conduit. Geyser frequency isalso sensitive to the (often complex) conduit geome-try (Hutchinson et al., 1997) and hence permeabilityof the geyser conduit. However, because the conduitis already much more permeable, small strains withinthe conduit itself are unlikely to have a significanteffect on geyser eruptions (Ingebritsen et al., 2006). Asa consequence, the observed temporal variability ineruption frequency is probably caused by changes inthe permeability of the matrix surrounding the maingeyser conduit–it is through this matrix that theconduit is recharged between eruptions.

The high sensitivity and long-lasting response ofgeysers to small seismic strains requires reopening,unblocking, or creating fractures to induce largeenough permeability changes to influence eruptionfrequency. Geysers thus provide evidence thatdynamic strains are able to create permanent changesin permeability from small dynamic strains and atgreat distances from the earthquake.

In summary, the response of geysers to distantearthquakes is most easily explained by changes inpermeability, and the sensitivity of geysers to earth-quakes compared with earth tides and changes inbarometric pressure indicates that dynamic strainscause the response.

4.10.4 Feedback betweenEarthquakes and Hydrology

The presence of water in the subsurface, and changesin the amount of water on the surface or within thesubsurface, can influence the occurrence of earth-quakes, as summarized in Figure 17. There arethree basic ways water can influence seismicity.

First, changes in loading of the surface canincrease deviatoric stresses. The most clear examplesfollow the impoundment of water by reservoirs (e.g.,Simpson et al., 1988; Gupta, 1992). Natural examplesare more difficult to identify, probably in partbecause the magnitude of the surface load is muchsmaller; proposed examples include snow loading(Heki, 2003) and loading from ocean tides (Cochranet al., 2004), or other seasonal processes (Wolf et al.,1997).

Second, changes in fluid pressure p reduce theeffective stress

!effij ¼ !ij –%p$ij ½15$

Pore pressurechange

Permeabilityincrease

Earthquake FlowStaticstrain

Dynamicstrain

Consolidation,liquefaction

Microfractures

Unclog fractures

–

–+ –

+

Near-field

Figure 17 Schematic illustration of the relationship between earthquakes and hydrologic responses; þ and–signs indicatethe sign of the response. Figure based on an illustration from David Mays.


where % is again the Biot–Willis coefficient. Theclassic example of seismicity induced by changes inpore pressure is caused by fluid injection (e.g.,Raleigh et al., 1976). Some cases of reservoir-inducedseismicity, in particular those in which the seismicitylags changes in water level, can be attributed tochanges in pore pressure p (e.g., Talwani and Acree,1985; do Nascimento et al., 2005). Changes in porepressure also diffuse, governed by eqn [6], so thetelltale signature of earthquakes triggered by changesin pore pressure would be a migration of the locus ofseismicity at a rate consistent with the (usuallyunknown) hydraulic diffusivity. Spatial migration ofreservoir-induced seismicity, seismicity induced byfluid injection (Tadokoro et al., 2000; Parotidis andShapiro, 2004), and some aftershock sequences (e.g.,Nur and Booker, 1972; Bosl and Nur, 2002; Hainzl,2004; Miller et al., 2004) have been attributed to porepressure diffusion. Pore pressure changes caused bynatural loading at the surface (e.g., groundwaterrecharge) have also been invoked to trigger seismi-city (e.g., Costain et al., 1987; Saar and Manga, 2003;Christiansen et al., 2005).

Third, fluid extraction, rather than fluid injection,can change stresses through poroelastic deformation(Segall, 1989). Earthquakes appear to have beeninduced by fluid extraction in gas fields (Segallet al., 1994) and oil fields (Gomberg and Wolf, 1999;Zoback and Zinke, 2002). A decrease in pore pressuremight be expected to stabilize faults by increasing theeffective stress. Poroelastic deformation, however,caused changes in fluid pressure, increases the mag-nitude of deviatoric stresses away from the region offluid extraction.

One general conclusion from all these studies ofhydrologically triggered seismicity, regardless of themechanism, is that very small stress changes, typi-cally 0.01–1Mpa, appear to trigger the earthquakes.Such small stress changes, however, are also similarto the stress changes thought to trigger earthquakesthrough nonhydrological means (e.g., Harris, 1998;Stein, 1999).

If earthquakes cause changes in hydrogeologicalproperties and changes in fluid pressure promoteseismicity, then hydrogeological and seismologicalprocesses are coupled. Triggered seismicity. (seeChapter 4.09) is one possible example of thiscoupling. Interaction is promoted if the state of stressis close to failure so that the small changes in stressassociated with either natural hydrologic processes orhydrological responses to earthquakes can in turninfluence seismicity. At least in tectonically active

areas, many faults do appear to be critically stressed(Zoback et al., 1987), consistent with earthquakesbeing triggered by small stress changes.

Rojstaczer and Ingebritsen (2005, personal com-munication) suggest that one manifestation of thisinteraction is the value of the mean large-scale per-meability of the crust – it should be of a size toaccommodate internal (fluid generation by meta-morphic and magmatic processes) and external(groundwater recharge and discharge) forcing.Diagenesis, in general, tends to seal fractures andfill pores, thus decreasing permeability. If ground-water recharge does not change with time, waterlevels and pore pressures will increase, promotingseismicity, which in turn generates new fracturesand increases permeability, A mean crustal perme-ability of 10%14 m2 (Manning and Ingebritsen, 1999)is consistent with the present mean rate of ground-water recharge and hydraulic head gradients drivingbasin-wide groundwater flow. Rojstaczer andIngebritsen propose that this balance is reached by afeedback between hydrological processes and seismi-city. A similar balance must occur in the lower crust,though the source of water is from metamorphicreactions rather than of meterological orgin. Themean permeability obtained from such a balance,however, is a temporal and spatial average. Instead,evidence of short-lived and locally high permeabil-ities are preserved in the spatial distribution ofmineral deposits in ancient fault systems (e.g.,Micklethwaite and Cox, 2004) and transient, loca-lized high temperatures in the lower crust(Camacho et al., 2005).

4.10.5 Hydrologic Precursors

As noted in Section 4.10.2.1, failure of brittle rocksunder deviatoric stress is usually preceded by perva-sive microcrack formation. In the subsurface, thisprocess (i.e., microfracturing) would greatly increasethe surface area of the affected rock in contact withgroundwater and thus allow the release of gases anddissolved ions from the rock into the groundwater,changing its chemical composition. Coalescing ofmicrocracks into larger fractures may connecthydraulically isolated aquifers, causing both changesin the hydraulic heads and mixing of groundwaterswith initially distinct hydrogeochemistry. These pro-cesses may further cause changes in the electricalconductivity of the rocks, and thus of the crust.Scenarios of this kind have led to the not


unreasonable expectation that hydrological, hydro-geochemical, and related geophysical precursors mayappear before the occurrence of large earthquakes.

During the 1960s and 1970s, several groupsreported precursory changes in the crustal seismicvelocity ratio (Vp/Vs) before some earthquakes in thethen Soviet Union (Savarensky, 1968; Semenov, 1969),in New York (Aggarwal et al., 1973), and in California(Whitcomb et al., 1973). Coupling the failure processesof brittle rocks, as noted previously, to a groundwaterflow model, Nur (1974) proposed a ‘dilatancy model’to explain the sequence of precursory changes in(Vp/Vs). Scholz et al., (1973) further develop the dila-tancy model as a common physical basis forprecursory phenomena. Later seismic experimentsdesigned to detect changes in both Vp and Vs, however,failed to reveal any precursory changes before manymajor and moderate earthquakes (e.g., McEvilly andJohnson, 1974). In 1988, in response to Bakun andLindh’s (1985) prediction that a large earthquakewould occur near Parkfield before 1993, Park installeda network of electrodes across the fault zone nearParkfield to detect any precursory changes in theelectrical resistivity of the crust (Park, 1997) – one ofthe major predictions of the dilatancy model (Scholzet al., 1973). The predicted Parkfield earthquake didnot occur; but an M 6.0 earthquake occurred on 28,September 2004 near Parkfield, providing an excellentopportunity to test the telluric method for detectingany precursory changes in crustal resistivity before alarge earthquake. Careful data processing was appliedto the time series of the dipole fields before the earth-quake, but no precursory changes were found (Park,2005). These failures, among others, have cast ques-tions about the general validity of the dilatancy model,and about earthquake precursors in general.

An intense search for hydrological and hydroche-mical precursors before earthquakes has also beenpursued. For example, a few days prior to the 1946M 8.3 Nankaido earthquake in Japan, water levels insome wells reportedly fell by more than 1m andsome wells went dry (Sato, 1978; Linde and Sacks,2002). Three days before the M 6.1 Kettleman Hill,CA, earthquake, Roeloffs and Quilty (1997) found agradual, anomalous rise in water level of 3 cm. Thisobservation was included in the IASPEI PreliminaryList of Significant Precursors (Wyss and Booth,1997). Following the 1995 M 7.2 Kobe earthquake,several papers reported precursory changes in theconcentrations of radon, chlorine, and sulfate ions ingroundwater (e.g., Tsunogai and Wakita, 1995;Igarashi et al., 1995) and in groundwater level (King

et al., 1995). Changes in radon concentration are themost commonly reported and discussed hydrogeo-chemical precursor (e.g., Wakita et al., 1988; Triqueet al., 1999) because the release of radon is especiallysensitive to crustal strains.

Definitive and consistent evidence for hydrologi-cal and hydrochemical precursors, however, hasremained elusive (Bakun et al., 2005). Difficultiesinclude that most reported changes were not cor-rected for the fluctuations in temperature,barometric pressure, earth tides, and other environ-mental factors, so that some changes taken to beearthquake related may in fact be ‘noise’ (e.g.,Hartman and Levy, 2005), that changes wererecorded at some sites but often were not recordedat other nearby sites (e.g., Biagi et al., 2001), and thatinstrument failures and personnel/program changesoften do not allow persistent and consistent monitor-ing over long periods of time (King et al., 2000) – anecessary condition for obtaining reliable precursorydata. Distinguishing a precursor from a response to aprevious earthquake, for example, the 1946 Nankaidoearthquake was preceded by the 1944 M 8.2Tonankai event, creates additional ambiguity.Notwithstanding these difficulties, progress hasbeen made in the past decade. For example, intensiveand continued observations of various kinds of pre-cursory hydrological and hydrochemical changeshave been made in Japan during the past half century(Wakita, 1996), and records are corrected to removethe noise introduced by fluctuations in temperature,barometric pressure, earth tides, and other factors(Igarashi and Wakita, 1995); both high- and low-pass filtering has been applied to the time series ofraw hydrochemical data in Kamchatka, Russia, toremove long- and short-period changes unrelated toearthquake processes (Kingsley et al., 2001); statisti-cal, rather than deterministic, procedures have beenintroduced (Maeda and Yoshida, 1990) to assess theconditional probability of future seismic events; mul-ticomponent, hydrochemistry analysis was applied togroundwater samples in Iceland before and after amajor earthquake to enhance the possibility ofdetecting possible precursors (Claesson et al., 2004);and, finally, relationships among various types ofhydrological and hydrochemical precursory signalswere sought to improve future earthquake predictionstrategies (Hartman and Levy, 2005). Thus, althoughwe may still be far from achieving a genuine under-standing of the underlying mechanisms of the variousearthquake-related anomalies, significant efforts areunderway.


Static strains of 10%8 can reasonably be expectedto produce changes in water level of about 1 cm foroptimal hydrogeologic properties. Even greater sen-sitivity to strain (though perhaps only dynamicstrains) is implied by some of the very distant hydro-logic responses seen at geysers and some very distantchanges in water in wells (e.g., Brodsky et al., 2003).The strong sensitivity of hydrologic processes andproperties to small strains is the primary basis forhope that hydrological and hydrogeochemical mon-itoring may detect any hypothetical pre-earthquakestrains. Recognizing precursors, and distinguishingbetween responses from previous earthquakes andprecursors to new earthquakes, is a different matterand may continue to be problematic (Hartman andLevy, 2005). Although it is unclear whether anyfuture documented hydrological precursors couldactually be used for earthquake prediction, theymay at least provide new insight into the physics ofearthquakes.

4.10.6 Concluding Remarks

Hydrologic responses to earthquakes provide con-straints on hydrogeologic processes in regions thatmight otherwise be inaccessible, for example, faultzones or the deep subsurface. Measured responsesmay also provide information at spatial and temporalscales that are difficult to study with more conven-tional hydrogeological measurements such as welltests. However, these novel features of hydrologicresponses also mean that it can be difficult, andperhaps even impossible in some cases, to obtainthe information needed to distinguish between com-peting models for hydrologic responses.

Explaining the hydrologic responses to earth-quakes should in principle be simple because theyreflect the strain caused by earthquakes. The greatvariety of hydrologic responses, however, highlightsthe complexity of deformation and structure of geo-logical materials and the interaction betweenprocesses. Over the last decade there has thus beena trend toward developing quantitative physicallybased models to explain hydrogeological responsesto earthquakes with an emphasis on explainingphenomena that cannot be explained by linear por-oelastic models alone.

We conclude by noting that advances in physicallybased models may not be sufficient for understandingthe interactions between earthquakes and water.Although expensive and time consuming, continued

monitoring of wells, springs, and streams, ideally athigh sampling rates and with complementary data sets(e.g., chemistry, temperature, pressure), provides thedata needed to test models and may also lead to thediscovery of new hydrological phenomena.

Acknowledgments

We thank Alex Wong for sharing the results of hisstudy of the Chi-Chi earthquake, Fu-Qiong Huangfor the information in Figure 3 and three reviewersfor comments. This study is supported by the SloanFoundation.

References

Aggarwal YP, Sykes LR, Armbruster J, and Sbar ML (1973)Premonitory changes in seismic velocities and prediction ofearthquakes. Nature 241: 101–104.

Akita F and Matsumoto N (2004) Hydrological responsesinduced by the Tokachi-oki earthquake in 2003 at hot springwell in Hokkaido, Japan. Geophysical Research Letters 31,doi:10.1029/2004GL020433.

Ambraseys NN (1988) Engineering seismology. EarthquakeEngineering and Structural Dynamics 17: 1–105.

Andrew DJ (2003) A fault constitutive relation accounting forthermal pressurization of pore fluid. Journal of GeophysicalResearch 107, doi:10.1029/2002JB001942.

Angelier J, Chu HT, Lee JC, et al. (2000) Geologic knowledgeand seismic risk mitigation: Insight from the Chi–Chi earth-quake: Taiwan. In: Lo C-H and Liao W-I (eds.) Proceeding ofInternational Workshop Annual Commemoration of Chi-ChiEarthquake, Science Aspect,, pp. 13–24. Taipei, Taiwan:National Center for Research on Earthquake Engineering.

Arias A (1970) A measure of earthquake intensity. In: Hansen RJ(ed.) Seismic Design for Nuclear Power Plant, pp. 438–483.Cambridge, MA: Massachusetts Institute of Technology.

Bakun W and Lindh A (1985) The Parkfield earthquake predic-tion experiment. Science 229: 619–624.

Bakun WH, Aagaard B, Dost B, et al. (2005) Implications forprediction and hazard assessment from the 2004 Parkfieldearthquake. Nature 437: 969–974.

Bardet JP (2003) Advances in analysis of soil liquefaction duringearthquakes. In: Lee WHK, Kanamori H, Jennings PC, andKisslinger C (eds.) International Handbook of Earthquake andEngineering Seismology, Part B, pp. 1175–1201. San Diego:Academic press.

Berryman JG and Wang HF (2000) Elastic wave propagationand attenuation in a double-porosity dual-permeabilitymedium. International Journal of Rock Mechanics andMining Sciences 37: 67–78.

Biagi PF, Ermini A, Kingsley SP, Khatkevich YM, and Gordeev EI(2001) Difficulties with interpreting changes in groundwatergas content as earthquake precursors in Kamchatka,Russia. Journal of Seismology 5: 487–497.

Biot MA (1941) General theory of three-dimensional consolida-tion. Journal of Applied Physics 12: 155–164.

Biot MA (1956a) Theory of propagation of elastic waves in fluid-saturated porous solid. I: Low-frequency range. Journal ofthe Acoustical society of America 28: 168–178.


Biot MA (1956b) Theory of propagation of elastic waves in fluid-saturated porous solid. I: Higher frequency range. Journal ofthe Acoustical Society of America 28: 179–191.

Biot MA (1962) Mechanics of deformation and acoustic propa-gation in porous media. Journal of Applied Physics33: 1482–1498.

Blanchard FB and Byerly P (1935) A study of a well gauge as aseismograph. Bulletin of the Seismological Society ofAmerica 25: 313–321.

Bosl WJ and Nur A (2002) Aftershocks and pore fluid diffusionfollowing the 1992 Landers earthquake. Journal ofGeophysical Research 107: 2366.

Bower DR and Heaton KC (1978) Response of an aquifer nearOttawa to tidal forcing and the Alaskan earthquake of 1964.Canadian Journal of Earth Sciences 15: 331–340.

Brace WF, Paulding J, and Scholz C (1966) Dilatancy in thefracture of crystalline rocks. Journal of the GeophysicalResearch 71: 3939–3953.

Briggs RO (1991) Effects of Loma Prieta earthquake on surfacewater in Waddell Valley. Water Resources Bulletin27: 991–999.

Brodsky EE and Kanamori H (2001) Elastohydrodynamic lubri-cation of faults. Journal of Geophysical Research 106:16357–16374.

Brodsky EE and Prejean SG (2005) New constraints onmechanisms of remotely triggered seismicity at Long ValleyCaldera. Journal of Geophysical Research 110, dio:10.1029/2004JB003211.

Brodsky EE, Roeloffs E, Woodcock D, Gall I, and Manga M(2003) A mechanism for sustained ground water pressurechanges induced by distant earthquakes. Journal ofGeophysical Research 108, doi:10.1029/2002JB002321.

Brown KM, Tryon MD, DeShon HR, Schwartz S, andDorman LM (2005) Correlated transient fluid pulsing andseismic tremor in the Costa Rica subduction zone. Earth andPlanetary Science Letters 238: 189–203.

Camacho A, Lee JKW, Hensen BJ, and Braun J (2005) Short-lived orogenic cycles and the eclogitization of cold crust byspasmodic hot fluids. Nature 435: 1191–1196.

Carson B and Screaton EJ (1998) Fluid flow in accretionaryprisms: Evidence for focused, time-variable discharge.Reviews of Geophysics 36: 329–352.

Charmoille A, Fabbri O, Mudry J, Guglielmi Y, and Bertrand C(2005) Post-seismic permeability change in a shallow frac-tured aquifer following a M-L 5.1 earthquake (Fourbannekarst aquifer, Jura outermost thrust unit, eastern France).Geophysical Research Letters 32: L18406 (doi:10.1029/2005GL023859).

Chester FM, Chester JS, Kirschner DL, Schulz SE, andEvans JP (2004) Structure of large-displacement strike-slip fault zones in the brittle continental crust.In: Karner GD, Taylor B, Driscoll NW, and Kohlstedt DL(eds.) Rheology and Deformation in the Lithosphere atContinental Margins, pp. 223–260. New York: ColumbiaUniversity Press.

Chester FM, Evens JP, and Biegel RL (1993) Internal structureand weakening mechanisms of the San Andreas fault.Journal of Geophysicall Research 98: 771–786.

Chia YP, Wang YS, Wu HP, Chiu JJ, and Liu CW (2001)Changes of groundwater level due to the 1999 Chi-Chiearthquake in the Choshui River fan in Taiwan. Bulletin of theSeismological Society of America 91: 1062–1068.

Christiansen LB, Hurwitz S, Saar MO, Ingebritsen SE, andHsieh PA (2005) Seasonal seismicity at western UnitedStates volcanic centers. Earth and Planetary Science Letters240: 307–321.

Claesson L, Skelton A, Graham C, et al. (2004)Hydrogeochemical changes before and after a major earth-quake. Geology 32: 641–644.

Cochran ES, Vidale JE, and Tanaka S (2004) Earth tides cantrigger shallow thrust fault earthquakes. Science306: 1164–1166.

Cooper HH, Bredehoeft JD, Papadopulos IS, and Bennett RR(1965) The response of well-aquifer systems to seismicwaves. Journal of Geophysical Research 70: 3915–3926.

Costain JK, Bollinger GA, and Speer JA (1987) Hydroseismicity:A hypothesis for the role of water in the generation of intro-plate seismicity. Seismological Research Letters 58: 41–64.

do Nascimento AF, Lunn RJ, and Cowie PA (2005) Numericalmodelling of pore-pressure diffusion in a reservoir-inducedseismicity site in northeast Brazil. Geophysical JournalInternational 160: 249–262.

Dobry R (1989) Some basic aspects of soil liquefaction duringearthquakes. Annals of the New York Academy of Sciences558: 172–182.

Dvorkin J, Nolen-Hoeksema R, and Nur A (1994) The squirt-flowmechanism: Macroscopic description. Geophysics59: 428–438.

Elkhoury JE, Brodsky EE, and Agnew DC (2006) Seismic wavesincrease permeability. Nature 441: 1135–1138.

Forster CB, Evens JP, Tanaka H, Jeffreys R, and Noraha T(2003) Hydrologic properties and structure of theMozumi Fault, central Japan. Journal of GeophysicalResearch 30, doi:10.1029/2002GL014904.

Galli P (2000) New empirical relationships between magnitudeand distance for liquefaction. Tectonophysics 324: 169–187.

Gavrilenko P, MeliKadze G, Chelidze T, Gibert D, andKumsiashvili (2000) Permanent water level drop associatedwith the Spitak earthquake: Observations at Lisi borehole(Republic of Georgia) and modeling. Geophysical JournalInternational 143: 83–98.

Gomberg J and Wolf L (1999) A possible cause for an improb-able earthquake: The 1997 Mw 4.9 southern Alabamaearthquake and hydrocarbon recovery. Geology27: 367–370.

GorokhovichY (2005) Abandonment ofMinoanpalacesonCrete inrelation to the earthquake induced changes in groundwatersupply. Journal of Archeological Science 32: 217–222.

Gudmundsson A (2000) Active fault zones and groundwaterflow. Geophysical Research Letters 27: 2993–2996.

Gupta HK (1992) Reservoir-Induced Earthquakes, 264 pp. NewYork: Elsevier.

Hainzl S (2004) Seismicity patterns of earthquake swarms dueto fluid intrusion and stress triggering. Geophysical JournalInternational 159, doi: 10.1111/j.1365-246X.2004.02463.x.

Hamiel Y, Lyakhovsky V, and Agnon A (2005) Rock dilation,nonlinear deformation, and pore pressure changeunder shear. Earth and Planetary Science Letters237: 577–589.

Harris RA (1998) Introduction to the special section: Stresstriggers, stress shadows, and the implication for seismichazard. Journal of Geophysical Research 103: 24247–24258.

Hartman J and Levy JK (2005) Hydrogeological and gasgeo-chemical earthquake precursors–a review for application.Natural Hazards 34: 279–304.

Heki K (2003) Snow load and seasonal variation of earthquakeoccurrence in Japan. Earth and Planetary Science Letters207: 159–164.

Hickman SH, Zoback MD, and Ellsworth WL (2005) Structureand composition of the San Andreas fault zone at Parkfield:Intial results from SAFOD, phase 1 and 2. EOS Transaction,American Geophysical Union 83: 237.

Hsu SK, Chin CV, Cheng LH, and Lin WS 1999 The change ofgroundwater level of earthquake 921 in central Taiwan. In:Proceedings of Conference on Sustainable Developmentand Geosciences Prospect, National Taiwan University,Department of Geololgy, pp. 7–35 (in Chinese). Taipei,Taiwan: Department of Geology, National Taiwan University.


Husen S and Kissling E (2001) Postseismic fluid flow after thelarge subduction earthquake of Antofagasta, Chile. Geology29: 847–850.

Husen S, Taylor R, Smith RB, and Healser H (2004) Changesin geyser eruption behavior and remotely triggered seis-micity in Yellowstone. National park produced by the 2002M 7.9 Denali fault earthquake, Alaska. Geology32: 537–540.

Hutchinson RA (1985) Hydrothermal changes in the upper gey-ser basin, Yellowstone National Park, after the 1983 BorahPeak, Idaho, earthquake. US Geological Survey Open FileReport 85-290, 612–624.

Hutchinson RA, Westphal JA, and Kieffer SW (1997) Insitu observations of Old Faithful Geyser. Geology25: 875–878.

Igarashi G, Saeki S, Takahata N, et al. (1995) Ground-waterradon anomaly before the Kobe earthquake in Japan.Science 269: 60–61.

Igarashi G and Wakita H (1991) Tidal responses and earth-quake-related changes in the water level of deep wells.Journal of Geophysical Research 96: 4269–4278.

Igarashi G and Wakita H (1995) Geochemical and hydrologicalobservations for earthquake prediction in Japan. Journal ofPhysics of the Earth 43: 585–598.

Ingebritsen SE and Rojstaczer SA (1993) Controls on geyserperiodicity. Science 262: 889–892.

Ingebritsen SE, Sanford WE, and Neuzil CE (2006)Groundwaterin Geologic Processes, 2nd edn. New York: CambridgeUniversity Press.

Ishihara K (1996) Soil Behavior in Earthquake Geotechnics,350p. Oxford: Clarendon Press.

Jonsson S, Segall P, Pedersen R, and Bjornsson G (2003) Post-earthquake ground movements correlated to pore-pressuretransients. Nature 424: 179–183.

King C-Y, Azuma S, Igarashi G, Ohno M, Saito H, and Wakita H(1999) Earthquake-related water-level changes at 16 closelyclustered wells in Tono, central Japan. Journal ofGeophysical Research 104: 13073–13082.

King C-Y, Azuma S, Ohno M, et al. (2000) In search of earth-quake precursors in the water-level data of 16 closelyclustered wells at Tono, Japan. Geophysical JournalInternational 143: 469–477.

King C-Y, Koizumi N, and Kitagawa Y (1995) Hydrogeochemicalanomalies and the 1995 Kobe earthquake. Science269: 38–39.

King CY, Zhang W, and Zhang ZC (2006) Earthquake-inducedgroundwater and gas changes. Pure and AppliedGeophysics 163: 633–645.

Kingsley SP, Biagi PF, Piccolo R, et al. (2001)Hydrogeochemical precursors of strong earthquakes: Arealistic possibility in Kamachatka. Physics and Chemistry ofthe Earth (C) 26: 769–774.

Kitagawa Y, Koizumi N, Takahashi M, Matsumoto N, and Sato T(2006) Changes in groundwater levels or pressures asso-ciated with the 2004 earthquake off the west coast ofnorthern Sumatra (M9.0). Earth, Planets and Space58: 173–179.

Koizumi N, Lai W-C, Kitagawa Y, and Matsumoto N (2004)Comment on ‘‘Coseismic hydrological changes associatedwith dislocation of the September 21, 1999 Chichi earth-quake, Taiwan’’ by Lee M et al. Geophysical ResearchLetters 31, doi:10:1029/2004GL019897.

Kono Y and Yanagidani T (2006) Broadband hydroseismogramsobserved by closed borehole wells in the Makioka mine,central Japan: Response of pore pressure to seismic wavesfrom 0.05 to 2Hz. Journal of Geophysical Research111: B03410 (doi:10.1029/2005JB003656).

Kopf AJ (2002) Significance of mud volcanism. Reviewsof Geophysics 40, doi:10.1029/2000RG000093.

Kuribayashi E and Tatsuoka F (1975) Brief review of liquefactionduring earthquakes in Japan. Soil and Foundations15: 81–92.

Lachenbruch A (1980) Frictional heating, fluid pressure, and theresistance to fault motion. Journal of Geophysical Research85: 6097–6112.

Lachenbruch AH and Sass JH (1977) Heat flow in the UnitedStates and the thermal regime of the crust. In: Heacock JH(ed.) The Earth’s Crust, Geophys. Monogr, Ser., 20,pp. 625–675. Washington, DC: American GeophysicalUnion.

Lawson AC (1908) The California Earthquake of April 18,1906. Report of the state earthquake investigationcommission, vol. 1, Carnegie Institution of Washington,Washington, DC.

Lay T and Wallace TC (1995) Modern Global Seismology.New York: Academic Press.

Lee CS and Wu CL (1996) Pumping tests of the Choshuichialluvial fan. In:Conference on Groundwater and Hydrology ofthe Choshui River Alluvial Fan, Taiwan, Water ResourcesBureau, pp.165–179 (in Chinese). Taipei, Taiwan: WaterResources Bureau, MOEA.

Lee CT, Kelson KI, and Kang KH (2000) Hanging wall deforma-tion and its effect on buildings and structures as learned fromthe Chelungpu faulting in the 1999 Chi-Chi Taiwan earth-quake. In: Lo C-H and Liao W-I (eds.) Proceedings ofInternational Workshop Annual Commemoration of Chi-ChiEarthquake, Science Aspect, pp. 93–104. Taipei, Taiwan:National Center for Research on Earthquake Engineering.

Lee J-C, Chu H-T, Angelier J et al. (2002a) Geometry andstructure of northern rupture surface ruptures of the 1999Mw

¼ 7.6 Chi-Chi Taiwan earthquake: Influence from inheritedfold belt structures. Journal of Structural Geology24: 173–192.

Lee M, Liu T-K, Ma K-F, and Chang Y-M (2002b) Coseismichydrological changes associated with dislocation of theSeptember 21, 1999 Chichi earthquake, Taiwan. 29,doi:10.1029/2002GL015116.

Li H and Zhu L (2006) High-resolution structures of the Landersfault zone inferred from aftershock waveform data.Geophysical Journal International (in press).

Li Y-G, Vidale JE, Aki K, and Xu F (2000) Depth-dependentstructure of the Lander Fault zonefrom trapped waves gen-erated by aftershocks. Journal of Geophysical Research105: 6237–6254.

Li YG, Vidale JE, and Malin PE (2006) Parkfield fault-zoneguided waves: High-resolution delineation of the low-velo-city damage zone on the San Andreas at depth near SAFODsite. The Proceeding of ICDP-IODP Fault-Zone Drilling,pp. 1–4. Miyazaki, Japan.

Lin WY (2000) Unstable groundwater supply after the bigearthquake. United Daily Newspaper, Taiwan (in Chinese).

Linde AT and Sacks IS (2002) Slow earthquakes and greatearthquakes along the Nankai trough. Earth and PlanetaryScience Letters 203: 265–75.

Linde AT, Sacks IS, Johnston MJS, Hill DP, and Bilham RG(1994) Increased pressure from rising bubles as a mechan-ism for remotely triggered seismicity. Nature 371: 408–410.

Liu E, Vlatsos S, Li S-Y, Main IG, and Schoenberg M (2004)Modeling seismic wave propagation during fluid injectionin a fractured network: Effects of pore fluid pressureon time-lapse seismic signatures. Leading Edge23: 778–783.

Liu LB, Roeloffs E, and Zheng XY (1989) Seismically inducedwater level oscillations in the Wali well, Beijing, China.Journal of Geophysical Research 94: 9453–9462.

Lockner DA and Beeler NM (2002) Rock failure and earth-quakes. In: Lee WHK, Kanamori H, Jennings PC, andKisslinger C (eds.) International Handbook of Earthquake and


Engineering Seismology, pp. 505–537. San Diego: AcademicPress.

Ma KF, Lee CT, and Tsai YB (1999) The Chi-Chi, Taiwan earth-quake: Large surface displacement on an inland thrust fault.EOS Transaction, American Geophysical Union 80: 605–611.

Maeda K and Yoshida A (1990) A probabilistic estimation ofearthquake occurrence on the basis of the appearance timesof multiple precursory phenomena. Journal of Physics of theEarth 38: 431–444.

Maltman AJ and Bolton A (2003) How sediments becomemobilized. In: Van Rensberger P, Hillis RR, Maltman AJ, andMorley CK (eds.) Subsurface Sediment Mobilization,Geological Society of London Special Publications 216,pp. 9–20. London: Geological Society of London.

Manga M (2001) Origin of postseismic streamflow changesinferred from baseflow recession and magnitude-distancerelations. Geophysical Research Letters 28: 2133–2136.

MangaM and Brodsky EE (2006) Seismic triggering of eruptionsin the far field: Volcanoes and geysers. Annual Review ofEarth and Planetary Sciences 34: 263–291.

Manga M, Brodsky EE, and Boone M (2003) Response ofstreamflow to multiple earthquakes and implications for theorigin of postseismic discharge changes. GeophysicalResearch Letters 30, doi:10.1029/2002GL016618.

Manning CE and Ingebritsen SE (1999) Permeability of thecontinental crust: Constraints from heat flow models andmetamorphic systems. Reviews of Geophysics 37, 127–150.

Mase CW and Smith L (1987) Effect of frictional heating on thethermal, hydrological, and mechanical response of a fault.Journal of Geophysical Research 92: 6249–6272.

Masson Y and Pride SR (2006) Computation of seismicattenuation and dispersion due to microscopic heterogene-ity in porous materials. Journal of Geophysical Research (inpress).

Matsumoto N (1992) Regression analysis for anomalouschanges of ground water level due to earthquakes.Geophysical Research Letters 19: 1192–1196.

Matsumoto N and Roeloffs EA (2003) Hydrological response toearthquakes in the Haibara well, central Japan. II: Possiblemechanism inferred from time-varying hydraulic properties.Geophysical Journal International 155: 899–913.

Mavko G, Mukerji T, and Dvorkin J (1998) The Rock PhysicsHandbook: Tools for Seismic Data Analysis in Porous Media.New York: Cambridge University Press.

McEvilly T and Johnson L (1974) Stability of P and S velocitiesfrom central California quarry blasts. Bulletin of theSeismological Society of America 64: 343–353.

Micklethwaite S and Cox SF (2004) Fault-segment rupture,aftershock-zone fluid flow, and mineralization. Geology32: 813–816.

Milkov AV (2000) Worldwide distribution of submarine mudvolcanoes and associated gas hydrates. Marine Geology167: 29–42.

Miller SA, Collettini C, Chiaraluce L, Cocco M, Barchi M, andKaus BJP (2004) Aftershocks driven by a high-pressure CO2

source at depth. Nature 427: 724–727.Mizuno T (2003) The subsurface observation of fault-zone

trapped waves: Applications to investigations of the deepstructure of active faults. Bulletin of the Earthquake ResearchInstitute 78: 91–106.

Mogi K, Mochizuki H, and Kurokawa Y (1989) Temperaturechanges in an artesian spring at Usami in the lzu Peninsula(Jaoan) and their relation to earthquakes. Tectonophysics159: 95–108.

Montgomery DR, Greenberg HM, and Smith DT (2003)Streamflow response to the Nisqually earthquake. Earth andPlanetary Science Letters 209: 19–28.

Montgomery DR and Manga M (2003) Streamflow and waterwell responses to earthquakes. Science 300: 2047–2049.

Mount VS and Suppe J (1987) State of stress near the SanAndreas Fault – Implications for wrench tectonics. Geology15: 1143–1146.

Muir-Wood R and King GCP (1993) Hydrological signatures ofearthquake strain. Journal of Geophysical Research98: 22035–22068.

National Research Council (1985) Liquefaction of Soils duringEarthquakes, 204 pp. Washington, DC: National AcademyPress.

Neuzil CE (2003) Hydromechanical coupling in geologic pro-cesses. Hydrogeological Jounal 11: 41–83.

Nur A (1974) Matsushiro, Japan, earthquake swarm:Confirmation of the dilatancy-fluid diffusion model. Geology2: 217–221.

Nur A and Booker JR (1972) Aftershocks caused by pore fluidflow? Science 175: 885–887.

Ohtani T, Tanaka H, Fujimoto K, Higuchi T, Tomida N, and Ito H(2001) Internal structure of the Nojima fault zone fromHirabayshi GSJ drill core. Island Arc 10: 392–400.

Panahi BM (2005) Mud volcanism, geodynamics and seismicityof Azerbaijan and the Caspian sea region. In: Martinelli G andPanahi B (eds.) Mud Volcanoes, Geodynamics andSeismicity, pp. 89–104. Dordrecht: Springer.

Papadopoulos GA and Lefkopulos G (1993) Magnitude-dis-tance relations for liquefaction in soil from earthquakes.Bulletin of the Seismological Society of America83: 925–938.

Park SK (1997) Monitoring resistivity changes in Parkfield,California: 1988–1995. Journal of Geophysical Research102: 24545–24559.

Park SK (2005) The agony and ecstasy of earthquake pre-diction (talk presented at US Geological Survey, MenloPark, CA).

Parotidis M and Shapiro SA (2004) A statistical model for theseismicity rate of fluid-injection-induced earthquakes.Geophysical Research Letters 31, doi:10.1029/2004GL020421.

Pralle N, Kulzer M, and Gudehus G (2003) Experimental evi-dence on the role of gas in sediment liquefaction and mudvolcanism. In: Van Rensberger P, Hillis RR, Maltman AJ, andMorley CK (eds.) Subsurface Sediment Mobilizationpp. 159–172. Geological Society of London SpecialPublications 216.

Pride SR, Berryman JG, and Harris JM (2004) Seismicattenuation due to wave-induced flow. Journal ofGeophysical Research 109: B01201, (doi:10.1029/2003JB002639).

Quilty EG and Roeloffs EA (1997) Water level changes inresponse to the December 20, 1994 M4.7 earthquake nearParkfield, California. Bulletin of the Seismological Society ofAmerica 87: 1018–1040.

Raleigh CB, Healy JH, and Bredehoeft (1976) An experiment inearthquake control at Rangely, Colorado. Science191: 1230–1237.

Rice JR (1992) Fault stress state, pore pressure distributions,and the weakness of the san Andreas fault. In: Evans B andWong T-F (eds.) Fault Mechanics and Transport Properties ofRocks, pp. 475–504. San Diego: Academic Press.

Rice JR and Cleary MP (1976) Some basic stress diffusionsolutions for fluid saturated elastic porous media with com-pressible constituents. Reviews of Geophysics and SpacePhysics 14: 227–241.

Rinehart JS (1972) Fluctuations in geyser activity caused byvariations in earth tidal forces, barometric pressure, andtectonic stresses. Journal of Geophysical Research77: 342–350.

Roeloffs EA (1996) Poroelastic techniques in the study ofearthquake-related hydrologic phenomena. Advances inGeophysics 37: 135–195.


Roeloffs EA (1998) Persistent water level changes in a well nearParkfield, California, due to local and distant earthquakes.Journal of Geophysical Research 103: 869–889.

Roeloffs EA (2006) Evidence for aseismic deformation-ratechanges prior to earthquakes. Annual Review of Earth andPlanetary Sciences 34: 591–627.

Roeloffs EA and Quilty E (1997) Water level and strain changespreceding and following the August 4, 1985 Kettleman Hills,California, earthquake. Pure and Applied Geophysics149: 21–60.

Roeloffs EA, Sneed M, Galloway DL, et al. (2003) Water-levelchanges induced by local and distant earthquakes at LongValley caldera, California. Journal of Volcanology andGeothermal Research 127: 269–303.

Rojstaczer S and Ingebritsen SE 2005 Crustal permeabilityconstrained by internal and external forcing (in review).

Rojstaczer S and Wolf S (1992) Permeability changes asso-ciated with large earthquakes: An example from LomaPrieta, California. Geology 20: 211–214.

Rojstaczer S, Wolf S, and Michel R (1995) Permeabilityenhancement in the shallow crust as a cause of earth-quake-induced hydrological changes. Nature373: 237–239.

Rojstaczer S, Galloway DL, Ingebritsen SE, and Rubin DM(2003) Variability in geyser eruptive timing and its causes:Yellowstone National Park. Geophysical Research Letters30: (art No. 1953), doi: 10.1029/2003GL017853.

Rudnicki JW, Yin J, and Roeloffs EA (1993) Analysis of waterlevel changes induced by fault creep at Parkfield, California.Journal of Geophysical Research 98: 8143–8152.

Saar MO and Manga M (2003) Seismicity induced by seasonalgroundwater recharge at Mt. Hood, Oregon. Earth andPlanetary Science Letters 214: 605–618.

Sato H (1978) Precursory land tilt prior to the Tonankai earth-quake of 1944. In some precursors prior to recent greatearthquakes along the Nankai trough. Journal of Physics ofthe Earth 25: s11521.

Sato T, Sakai R, Furuya K, and Kodama T (2000) Coseismicspring flow changes associated with the 1995 Kobe earth-quake. Geophysical Research Letters 27: 1219–1222.

Sato T, Matsumoto N, Kitagawa Y, et al. (2004) Changes inwater level associated with the 2003 Tokachi-oki earth-quake. Earth, Planets and Space 56: 395–400.

Savarensky EF (1968) On the prediction of earthquakes.Tectonophysics 6: 17–27.

Scholz CH, Sykes LR, and Aggarwal YP (1973) Earthquakeprediction – Physical basis. Science 181: 803–810.

Seed H and Lee KL (1966) Liquefaction of saturated sandsduring cyclic loading. Journal of the Soil Mechanics andFoundation Division 92: 105–134.

Segall P (1989) Earthquakes triggered by fluid extraction.Geology 17: 942–946.

Segall P, Grasso JR, and Mossop A (1994) Poroelastic stressingand induced seismicity near the Lacq gas field, southwes-tern France. Journal of Geophysical Research99: 15423–15438.

Semenov AM (1969) Variations in the travel-time of transverseand longitudinal waves before violent earthquakes. Physicsof the Solid Earth 3: 245–248.

Sibson RH (1973) Interactions between temperature and pore-fluid pressure during earthquake faulting and a mechanismfor partial or total stress relief. Nature 243: 66–68.

Sibson RH and Rowland JV (2003) Stress, fluid pressure andstructural permeability in seismogenic crust, North Island,New Zealand. Geophysical Journal International154: 584–594.

Sil S and Freymueller JT (2006) Well water level changes inFairbanks, Alaska, due to the great Sumatra-Andamanearthquake. Earth Planets Space 58: 181–184.

Silver PG and Valette-Silver NJ (1992) Detection of hydrother-mal precursors to large northern California earthquakes.Science 257: 1363–1368.

Simpson DW, Leith WS, and Scholz CH (1988) Two types ofreservoir-induced seismicity. Bulletin of the seismologicalSociety of America 78: 2025–2050.

Sleep NH and Blanpied ML (1992) Creep, compaction and theweak rheology of major faults. Nature 359: 687–692.

Sorey ML and Clark MD (1981) Changes in the discharge char-acteristics of thermal springs and fumaroles in the LongValley caldera, California, resulting from earthquakes on May25–27, 1980. US Geological Survey Open-File Report81–203.

Stein RS (1999) The role of stress transfer in earthquakeoccurence. Nature 402: 605–609.

Steinberg GS, Merzhanov AG, and Steinberg AS (1982a)Geyser process: Its theory, modeling and field experiment.Part 1: Theory of the geyser process. Modern Geology8: 67–70.

Steinberg GS, Merzhanov AG, and Steinberg AS (1982b)Geyser process: Its theory, modeling and field experiment.Part 3: On metastability of water in geysers.ModemGeology8: 75–78.

Steinberg GS, Merzhanov AG, and Steinberg AS (1982c) Geyserprocess: Its theory, modeling and field experiment. Part 4:On seismic influence on geyser regime. Modern Geology8: 79–86.

Stidham C, Antolik M, Dreger D, Larsen S, and Romanowicz B(1999) Three-dimensional structure influences on the strong-motion wavefield of the 1989 Loma Prieta earthquake.Bulletin of the Seismological Society of America89: 1184–1202.

Su T-C, Chiang K-W, Lin S-J, Wang F-G, and Duann S-W (2000)Field reconnaissance and preliminary assessment of lique-faction in Yuan-Lin area. Sino-Geotechnics 77: 29–38.

Tadokoro K, Ando M, and Nishigami K (2000) Induced earth-quakes accompanying the water injection experiment at theNojima fault zone, Japan: Seismicity and its migration.Journal of Geophynical Research 105: 6089–6104.

Talwani P and Acree S (1985) Pore pressure diffusion and themechanism of reservoir-induced seismicity. Pure andApplied Geophysics 122: 947–965.

Terzaghi K, Peck RB, and Mesri G (1996) Soil Mechanics inEngineering Practice, 3rd edn. New York: Wiley.

Thurber C, Roecker S, Ellsworth W, Chen Y, Lutter W, andSessions R (1997) Two-dimensional seismic image of theSan Andreas fault in the Northern Gablian range, CentralCalifornia: Evidence for fluids in the fault zone. GeophysicalResearch Letters 24: 1591–1594.

Tokunaga T (1999) Modeling of earthquake-induced hydrologi-cal changes and possible permeability enhancement due to17 January 1995 Kobe earthquake, Japan. Journal ofHydrology 223: 221–229.

Townend J (1997) Subducting a sponge; minimum estimates ofthe fluid budget of the Hikurangi Margin accretionary prism.Geological Society of New Zealand Newsletter 112: 14–16.

Trique M, Richon P, Perrier F, Avouac JP, and Sabroux JC(1999) Radon emanation and electric potential variationsassociated with transient deformation near reservoir lakes.Nature 399: 137–141.

Tsunogai U and Wakita H (1995) Precursory chemical changesin ground water: Kobe earthquake, Japan. Science269: 61–62.

Unsworth M, Bedrosian P, Eisel M, Egbert G, andSiripunvaraporn W (2000) Along strike variations in the elec-trical structure of the San Andreas Fault at Parkfield,California. Geophysical Research Letters 27: 3021–3024.

Wakita H (1975) Water wells as possible indicators of tectonicstrain. Science 189: 553–555.


Wakita H (1996) Geochemical challenge to earthquake predic-tion. Proceeding of the National Academic Science93: 3781–3786.

Wakita H, Nakamura Y, and Sano Y (1988) Short-term andintermediate-term geochemical precursors. Pure andApplied Geophysics 126: 78–89.

Waller RM (1966) Effects of the March 1964 Alaska earthquakeon the hydrology of south-central Alaska. USGS ProfessionalPaper 544A, pp 28.

Wang C-H, Wang CY, Kuo C-H, and Chen W-F (2005c) Someisotopic and hydrological changes associated with the 1999Chi-Chi earthquake, Taiwan. Island Arc 14: 37–54.

Wank C-Y (1984) On the constitution of the San Andreas fault.Geophysical Research Letters 89: 5858–5866.

Wang C-Y, Manga M, and Wong A (2005a) Floods onMars released from groundwater by impact. Icarus175: 551–555.

Wang C-Y, Cheng L-H, Chin C-V, and Yu S-B (2001) Coseismichydrologic response of an alluvial fan to the 1999 Chi-Chiearthquake, Taiwan. Geology 29: 831–834.

Wang C-Y, Dreger D, Manga D, and Wong A (2004c)Streamflow increase due to rupturing of hydrothermalreservoirs: Evidence from the 2003 San Simeon, California,earthquake. Geophysical Research Letters 31: L10502 (doi:10.1029/2004GL020124).

Wang C-Y, Dreger DS, Wang C-H, Mayeri D, and Berryman JG(2003) Field relations among coseismic ground motion,water level change and liquefaction for the 1999 Chi-Chi(Mw¼7.5) earthquake, Taiwan.Geophysical Research Letters17, doi:10.1029/2003GL017601.

Wang C-Y, Rui F, Yao Z, and Shi X (1986) Gravity anomaly anddensity structure of the San Andreas fault zone. Pure andApplied Geophysics 124: 127–140.

Wang C-Y, Wong A, Dreger D, and Manga M (2006)Liquefaction limit during earthquakes andunderground explosions – implications on ground-motionattenuation. Bulletin of the seismological Society of America96: 355–363.

Wang C-Y, Wang C-H, and Kuo C-Y (2004a) Temporal changein groundwater level following the 1999 (Mw ¼ 7.5) Chi-Chiearthquake, Taiwan. Geofluids 4: 210–220.

Wang C-Y, Wang C-H, andMangaM (2004b) Coseismic releaseof water from mountains: Evidence from the 1999 (Mw ¼ 7.5)Chi-Chi, Taiwan, earthquake. Geology 32: 769–772.

Wang R, Woith H, Milkereit C, and Zschau J (2004d) Modellingof hydrogeochemical anomalies induced by distant earth-quakes. Geophysical Journal International 157: 717–726.

Wang HF (2000) Theory of Linear Poroelasticity, 287 pp.Princeton, NJ: Princeton University Press.

Whitcomb JH, Garmany JD, and Anderson DL (1973)Earthquake prediction: Variation of seismic velocities beforethe San Fernando earthquake. Science 180: 632–635.

White DE (1967) Some principles of geyser activity, mainly fromSteamboat Springs, Nevada. American Journal of Science265: 644–684.

Wibberley C and Shimamoto T (2003) Internal structure andpermeability of major strike-slip fault zones: The MedianTectonic Line in W. Mie Prefecture, S.W. Japan. Journal ofStructural Geology 25: 59–78.

Wibberley CAJ and Shimamoto T (2005) Earthquake slipweakening and asperities explained by thermal pressuriza-tion. Nature 436: 689–692.

Wolf LW, Rowe CA, and Horner RB (1997) Periodic seismicitynear Mt. Ogden on the Alaska–British Columbia border: Acase for hydrologically-triggered earthquakes? Bulletin ofthe seismological Society of America 87: 1473–1483.

Wong A (2005) Relations between the Spectral Composition ofGround Motion and Hydrological Effects, and a RobustnessTest of a Finite-Fault Model in the 1999 Chi-Chi (Taiwan)Earthquake. Masters Thesis, University of California,Berkeley.

Wyss M and Booth DC (1997) The IASPEI procedure for theevaluation of earthquake precursors. Geophysical JournalInternational 131: 423–424.

Yan HR (2001) Water problems in the Yuenlin Mountains:Changes in groundwater after the Chi-Chi earthquake,Taiwan. Chinese Daily (in Chinese).

Youd TL (2003) Liquefaction mechanisms and induced groundfailure. In: Lee WHK, Kanamori H, Jennings PC, andKisslinger C (eds.) International Handbook of Earthquake andEngineering Seismology, Part B, pp. 1159–1173. San Diego:Academic Press.

Zoback MD and Zinke J (2002) Production-induced normalfaulting in the Valhall and Ekofisk oil fields. Pure and AppliedGeophysics 159: 403–420.

Zoback MD, Zoback ML, Mount VS, et al. (1987) New evidenceon the state of stress of the San Andreas fault system.Science 238: 1105–1111.


Documents

4.10 Earthquake Hydrology - Berkeley Seismological Laboratory