Stream£Ow Response to the Nisqually Earthquake

Home , 1964 Alaska earthquake

Earth and Planetary Science Letters 209 (2003) 19^28 www.elsevier.com/locate/epsl

Stream£ow response to the Nisqually earthquake

David R. Montgomery a;, Harvey M. Greenberg a, Daniel T. Smith b

a Department of Earth and Space Sciences, University of Washington, Seattle, WA 98195, USA b Water and Land Resources Division, King County, Seattle, WA, USA

Received 22 November 2002; received in revised form 4 February 2003; accepted 4 February 2003

Abstract

An extensive network of stream gages documents regional streamflow response to the M6.8 Nisqually earthquake wherein almost half of the gages analyzed within 115 km of the epicenter exhibited changes in baseflow within 13 h after the earthquake. Rapid streamflow response indicates that the impetus for the increased discharge originated within 100 m of the water table. Distance to the epicenter explained only 13% of the variance in streamflow response and the maximum modeled ground acceleration within 5 km of each gage location was not correlated with increased streamflow. Of those rivers that responded, post-seismic increases in discharge were correlated with pre-earthquake discharge; larger rivers exhibited greater absolute increases in streamflow. Analysis of baseflow recession in the periods 1 month before and 1 month after the earthquake indicates no systematic detectable changes in aquifer properties. Locations with seismically induced increases in streamflow were closer to the epicenter than an empirical limit to the area susceptible to liquefaction based on observations reported for previous earthquakes. In addition, the spatial pattern of streamflow response corresponds to the pattern of near-surface volumetric strain, with decreased streamflow in areas that dilated and substantial increases in streamflow in areas of greatest compression and subsidence. Together these observations suggest that settling and compaction of surficial deposits of the Seattle Basin and liquefaction of partially saturated valley-bottom deposits were responsible for post-seismic increases in streamflow. Compilation of distance^magnitude data for streamflow responses to a wide range of earthquakes show that streamflow changes generally occur in areas susceptible to liquefaction. ß 2003 Published by Elsevier Science B.V.

Keywords: stream£ow; earthquake; hydrologic; liquefaction

1. Introduction water discharge and subsurface groundwater lev- els in wells. Documented hydrological responses Numerous accounts of hydrological response to to earthquakes include changes in the water level earthquakes describe changes in both surface in wells, spring discharge, and stream£ow [1,2]. Various workers attribute such changes to expulsion of £uids from the seismogenic zone [3], pore- pressure di¡usion following co-seismic strain in * Corresponding author. Tel.: +1-206-685-2560; Fax: +1-206-543-3836. the upper crust [1,4,5], compression of con¢ned E-mail address: [email protected] aquifers [6], enhanced permeability of sur¢cial (D.R. Montgomery). materials due to either shaking of near-surface

0012-821X / 03 / $ ^ see front matter ß 2003 Published by Elsevier Science B.V. doi:10.1016/S0012-821X(03)00074-8

EPSL 6576 18-3-03 Cyaan Magenta Geel Zwart 20 D.R. Montgomery et al. / Earth and Planetary Science Letters 209 (2003) 19^28 deposits [7^9] or opening of bedrock fractures base£ow discharge between the time of the earth- [10^13], and liquefaction of sur¢cial deposits quake and midnight that night, 13 h later, so as to [14]. Here we report new observations from an normalize for the wide range of stream sizes and extensive network of gaging stations in the Puget to avoid the in£uence of rainfall the following day Sound region, Washington, to document the re- that would mask evidence for seismically triggered gional pattern of stream£ow response to the Feb- changes in stream£ow. We screened from the rec- ruary 28, 2001, Nisqually earthquake (local time ord those sites where the in£uence of dams pre- of 10:54 a.m. Paci¢c Standard Time). Our obser- cluded evaluation of stream£ow response, as well vations demonstrate an asymmetric pattern of as a few sites in£uenced by seismically triggered stream£ow changes that implicate compaction of landslides and stations noted to be in£uenced by near-surface aquifers and liquefaction of sur¢cial river ice at the time of the earthquake. In addi- deposits as the causes of observed stream£ow in- tion, several stations were omitted because they creases. We also compile observations on the dis- displayed cryptic, short-lived changes many hours tance from the epicenter for stream£ow response after the earthquake. For a subset of the gages to previous earthquakes to further evaluate gen- that demonstrated response to the earthquake eral mechanisms for such response. we also analyzed stream£ow records during periods of sustained base£ow recession in the month before and after the Nisqually earthquake. 2. Nisqually earthquake

The M6.8 Nisqually earthquake was a deep in- 4. Results traslab event within the subducting Juan de Fuca plate with a hypocenter 52 km below the ground On the morning of the earthquake, discharge in surface, 18 km northeast of Olympia, WA [15]. most streams in western Washington was gradu- Shaking lasted about 45 s, with felt intensity of ally decreasing, as is typical for base£ows between up to Modi¢ed Mercali Intensity VII. The event winter storms. Although minor discharge in- involved normal-style displacement on a N^NW creases occurred in some channels on February trending fault, with relative subsidence east of the 27, February 28 was a clear sunny day. At some epicenter. The Nisqually earthquake provides an of the gages that exhibited a response, stream£ow unusual opportunity to investigate seismically in- increased within minutes after the 10:54 a.m. (lo- duced changes in stream£ow because of an exten- cal time) earthquake (Fig. 1A). At other gages the sive regional network of active stream gages and a response was delayed, with a gradual rise that fortuitous lack of rainfall on the day of the earth- started in the minutes after the earthquake con- quake. tinuing to build through the day (Fig. 1B). Some gages peaked and began to decline before midnight, whereas others sustained response to the 3. Methods following day when runo¡ from a typical winter storm obscured recognition of seismically induced Analysis of discharge records from 222 U.S. changes in base£ows. Geological Survey, state, county, and city gaging Stream gages exhibiting no apparent response stations shows both that earthquake-triggered to the earthquake were distributed throughout the stream£ow changes were widespread in western state and included locations near the epicenter. Washington, and that there was a broad range From the time of the earthquake to midnight in the magnitude of stream£ow e¡ects among that night, changes in stream£ow varied from gaging stations that recorded a response to the 1% to s 100% at the stations in western Wash- earthquake. We examined each gage record for ington that exhibited identi¢able response. Given clear changes in stream£ow trends after the earth- the resolution of the hydrologic records and the quake and evaluated the percentage change in magnitude of the variability in pre-earthquake

EPSL 6576 18-3-03 Cyaan Magenta Geel Zwart D.R. Montgomery et al. / Earth and Planetary Science Letters 209 (2003) 19^28 21

Fig. 1. Examples of hydrographs from gaging stations for February 27 and 28, 2001, showing stations that span the range of response to the Nisqually earthquake on February 28 (10:54 a.m. local time): (A) Issaquah Creek, USGS gage 12121600 (20% increase in discharge); (B) Nisqually River near National, USGS gage 12082500 (9% increase in discharge). Vertical bars indicate time of the earthquake. discharges, we consider the signi¢cance of stream- 61 gages farther than 115 km (and up to 414 km) £ow changes of 1^5% as uncertain using this ap- from the epicenter exhibited any discernable re- proach. Within 115 km of the epicenter, 67 out of sponse (Fig. 2). 161 stream gages analyzed responded to the earth- Strong spatial coherence is apparent in the dis- quake, including three gages west of the epicenter tribution of gages that recorded changes in dis- that recorded decreased stream£ow; none of the charge associated with the Nisqually earthquake

Fig. 2. Distance from the epicenter for stream gages that exhibited discernable response (upper) or no apparent response (lower) to the earthquake.

EPSL 6576 18-3-03 Cyaan Magenta Geel Zwart 22 D.R. Montgomery et al. / Earth and Planetary Science Letters 209 (2003) 19^28

Fig. 3. Map showing the percentage change in £ow from the time of the Nisqually earthquake to midnight that night for 161 gaging stations in the Puget Sound region. Data are derived from discharge records from U.S. Geological Survey gaging stations and gages maintained by King, Snohomish, Pierce, Thurston, and Kitsap Counties, the cities of Seattle and Olympia, and the Washington State Department of Ecology. White circles indicate gages with no discernable response (i.e. 6 1%). Black circles represent stations where stream£ow decreased. Colored circles indicate gages for which s 1% increase was identi¢able; color cod- ing corresponds to the percentage change in base£ow. For a few stations for which only stage data were available, the change represents changes in stage rather than discharge. Beachball at the epicenter location represents focal solution for the earthquake. None of the 51 analyzed gages located beyond the area shown and up to 400 km from the epicenter exhibited any discernable response to the earthquake.

EPSL 6576 18-3-03 Cyaan Magenta Geel Zwart D.R. Montgomery et al. / Earth and Planetary Science Letters 209 (2003) 19^28 23

(Fig. 3). Discharges east of the epicenter increased, with the strongest hydrologic response concentrated in the area northeast of the epicenter. West of the epicenter, increases in discharge s 5% were restricted to proximal locations, and three gages recorded reduced discharge. Even though the stream gage network is not uniformly dispersed throughout the region, there is su⁄cient coverage to conclude that stream£ow response was asymmetrically distributed with respect to the earthquake’s epicenter. Moreover, least- squares linear regression indicates that distance from the epicenter explained only 13% of the variance in normalized stream£ow response among those gages that exhibited identi¢able response to the earthquake (Fig. 4). The post-seismic increase in discharge was better correlated with discharge at the time of the earthquake, which ex- Fig. 5. Post-seismic increase in discharge versus discharge at plained just under half the variance in the the time of the earthquake. Power function regression yields absolute change in discharge, indicating that the y = 0.21x0:67 (R2 = 0.45). degree of response scaled with the size of the river (Fig. 5). storativity due to aquifer compression, settling, Two types of mechanisms could explain in- and/or compaction. Following Manga [14], anal- creased stream£ow in response to earthquakes: ysis of base£ows sustained by groundwater £ow temporary changes in total head such as those between runo¡ producing storm events can be due to dynamic strain during liquefaction, and used to evaluate the degree of seismically induced permanent changes in hydraulic conductivity or changes in aquifer properties. Consider a shallow, one-dimensional, con¢ned aquifer wherein, in the absence of recharge, changes in hydraulic head, h, follow a di¡usion model: Nh=Nt ¼ DN2h=Nx2 ð1Þ where D is hydraulic di¡usivity, t is time, and x is the distance from the aquifer boundary. Adopting the linearized one-dimensional form of the Bous- sinesq equation, the discharge per unit width (q)is given by: q ¼ 3bKNh=Nx ð2Þ where b is aquifer thickness and K is hydraulic conductivity. At long times after surface runo¡ and recharge events, the recession of base£ow discharges for such a system is given by: dlogQ=dt ¼ 3KD ð3Þ where Q is stream discharge, K is a constant that Fig. 4. Percent increase in stream£ow versus distance to epicenter. Least-squares linear regression yields y =3030.28x characterizes the geometry of the aquifer, and for 2 (R = 0.13). a con¢ned aquifer D = K/Ss where Ss is the spe-

EPSL 6576 18-3-03 Cyaan Magenta Geel Zwart 24 D.R. Montgomery et al. / Earth and Planetary Science Letters 209 (2003) 19^28 ci¢c storage, whereas for a horizontal uncon¢ned aquifer D = bK/Sy where Sy is the speci¢c yield [16]. This simple 1-D model is justi¢ed because there typically is not enough information available on aquifer properties to justify a more so- phisticated and complicated model for interpret- ing hydrological responses to earthquakes. The slope of a least-squares linear regression of logQ versus time for periods of falling base£ow allows determining 3KD, and therefore estimat- ing aquifer-scale hydrologic properties. If seismically induced changes in stream£ow result from non-recoverable changes in aquifer properties (i.e. K, Ss,orSy), then KD should di¡er for periods of base£ow from before and after the earthquake. Recoverable changes, such as those due to dynamic strain during ground shaking, would not be expected to result in sustained changes in base- Fig. 6. Pre-earthquake versus post-earthquake discharge re- £ow response characteristics. cession constants (KD) for a subset of stream£ow gages that Periods of base£ow recession within a month of exhibited response to the earthquake. Data shown are mean the Nisqually earthquake (February^March, values of KD (Q in cfs and time in days) for those gages for 2001) were identi¢ed from plots of dlogQ/dt ver- which both pre- and post-earthquake values could be deter- sus time for the USGS gaging stations that re- mined. Error bars show range of values for individual events at stations where multiple events were suitable for determin- sponded to the earthquake. The slope de¢ned by ing KD. Power function regression yields y = 1.5x1:1 (R2 = least-squares linear regression of logQ versus t 0.93): 1-to-1 line shown in ¢gure for reference. was used to determine 3KD for each period of falling discharge characterized by steady dlogQ/ dt for a month both before and after the earth- fusion to the water table in near-surface aquifers quake. As found previously for other earthquakes that can be approximated by: [14], discharge recession constants from periods T ¼ 11z2=D ð4Þ before and after the Nisqually earthquake do not exhibit any consistent pattern of change that Rearranging Eq. 4, the maximum depth below would indicate non-recoverable alteration of aqui- the water table that a hydrologic response ob- fer properties (Fig. 6). Although mean recession served at a time T after an earthquake could constants (i.e. KD) decreased for some gages, have originated may be estimated by: others increased and the error bars de¢ned by z ¼ðTD=11Þ0:5 ð5Þ the range of observed values overlap the 1:1 line for all but three gages. Hence, any change due to Incorporating typical hydraulic di¡usivities of seismically induced static strain or changes in per- unconsolidated sands (DW1^10 m2 s31) and the meability was minor or of limited spatial extent, observed stream£ow response within hours of the and did not systematically a¡ect stream£ow in the earthquake into Eq. 5 indicates that excess same way among gaging stations. stream£ow generated by the earthquake originat- The time scale of the observed response pro- ed within 100 m of the water table. vides independent evidence for a near-surface The distance from the epicenter to stream source for the post-seismic increases in stream- gauges that responded to the Nisqually earth- £ow. Roelo¡s [17] derived a relation between quake is consistent with Manga’s [14] interpreta- depth below the water table (z), hydraulic di¡u- tion of dynamic strain due to liquefaction as re- sivity, and the time scale (T) of pore-pressure dif- sponsible for stream£ow response to earthquakes.

EPSL 6576 18-3-03 Cyaan Magenta Geel Zwart D.R. Montgomery et al. / Earth and Planetary Science Letters 209 (2003) 19^28 25

Papadopoulos and Lefkopoulos [18] reported an increases in stream£ow. The composite data set empirical relation that describes the maximum reinforces Manga’s [14] ¢nding that the maximum distance from earthquake epicenters at which distance at which co-seismic stream£ow response liquefaction has been documented: has been reported corresponds to the maximum 3 distance at which liquefaction has been observed. M ¼ 30:44 þ 3U10 8D þ 0:98logD ð6Þ e e The area within which sur¢cial evidence for where M is the earthquake magnitude and the liquefaction was observed after the Nisqually distance to the epicenter (De) is given in cm. Man- earthquake extends to approximately the same ga [14] showed that locations of observed stream- distance as the area within which stream£ow in- £ow response to the 1964 Alaksa, 1952 Kern creases s 5% were observed. Some of the major County, 1959 Hebgen Lake, 1983 Borah Peak, valley bottoms in which no evidence for liquefac- and 1989 Loma Prieta earthquakes occurred close tion was observed correspond to large alluvial enough to the epicenters for those locations to rivers in which no evidence for increased stream- have experienced liquefaction according to Eq. £ow was detected. Ground shaking was highly 6. Stream gages that responded to the Nisqually variable over short distances in the Puget Low- earthquake also were close enough to the epicen- land, and due to di¡erences in the distribution ter for liquefaction to have occurred in their of data we cannot relate ¢eld evidence for lique- watersheds (Fig. 7), whereas sites without detect- faction directly to stream£ow changes. Neither able response extended to beyond the range where the percentage increase nor the absolute increase liquefaction could be expected. We also compiled in stream£ow were correlated with modeled additional data for other earthquakes that caused ground accelerations extrapolated from the rela-

Fig. 7. Distance from epicenter versus earthquake magnitude for locations that exhibited seismically induced stream£ow response. Line represents the empirical relation reported by Papadopoulos and Lefkopoulos [18] to describe the distance from the epicenter beyond which liquefaction has not been observed (i.e. Eq. 6). Diamonds represent data compiled by Manga [14], squares represent data from the Nisqually earthquake, and circles represent additional data compiled on increases in spring £ow or stream£ow at Alum Rock Park, CA, reported by King et al. [9]; response of Waddell Creek to the Loma Prieta earthquake reported by Briggs [10]; the Kern County earthquake derived from ¢gures 1 and 2 in Briggs and Troxell [7]; the Loma Prieta earthquake derived from ¢gure 1 of Rojstaczer et al. [12]; spring and river £ow changes from the 1995 Kobe earthquake derived from ¢gure 1 of Sato et al. [21]; surface water response reported for the 1992 Landers [22] and 1994 Northridge earthquakes [23]; the distance to which stream£ow e¡ects were reported by Waller [8,24] for the 1964 Alaska earthquake; and the 1906 San Francisco earthquake based on anecdotal accounts of changes in stream and spring £ow reported in Lawson [25].

EPSL 6576 18-3-03 Cyaan Magenta Geel Zwart 26 D.R. Montgomery et al. / Earth and Planetary Science Letters 209 (2003) 19^28

Fig. 8. Preliminary vertical displacement ¢eld modeled for the Nisqually earthquake showing contours of uplift and subsidence [26] and the location of stream gages that decreased (black), increased more than 5% (white) or had between 1 and 5% response (open circles). Contours from http://www.panga.cwu.edu/olympia_eq/nisqually.jpg. tively sparse pattern of instrumentally determined This broad spatial correspondence points to a ground motion [19,20] (R2 of 0.05 and 0.04, re- connection between the style and magnitude of spectively). However, the in£uence of channel size stream£ow response and the pattern of near-sur- on the post-seismic increase in stream£ow sug- face crustal strain. gests the in£uence of the size of uncon¢ned aquifers along riverine valley bottoms. In addition, it is likely that the asymmetry in stream£ow re- 5. Discussion sponse re£ects to some degree the distribution of potentially lique¢able near-surface deposits in the The simplest explanation for the observed Puget Lowland. In particular, sandy deposits of stream£ow response to the Nisqually earthquake the Seattle Basin to the northeast of the epicenter is that it re£ects both expulsion of water from should be more susceptible to liquefaction than shallow aquifers due to compaction of unconsoli- coarser, predominantly gravel deposits found west dated near-surface deposits and potentially lique- of the epicenter. ¢able sur¢cial deposits in response to ground However, areas with stream£ow response s 5% shaking. Locations near the epicenter that did also correspond to areas to the northeast of the not respond to the earthquake presumably re£ect epicenter predicted to have subsided s 1mmon local geologic conditions less susceptible to set- the down-dropped hanging wall of the triggering tling and liquefaction. However, the correspon- fault (Fig. 8). All of the observed drops in stream- dence of areas exhibiting the greatest stream£ow £ow occurred on the upthrown side of the fault. response with areas that subsided indicates a sub-

EPSL 6576 18-3-03 Cyaan Magenta Geel Zwart D.R. Montgomery et al. / Earth and Planetary Science Letters 209 (2003) 19^28 27 stantial role for settling and compaction of sur¢- tude, post-seismic changes in stream£ow are con- cial deposits. Given that the base£ow recession sistent with empirical limits on the distance to analysis shows no systematic evidence for changes which liquefaction occurs from epicenters. For in basin-scale hydraulic conductivity, we conclude the Nisqually earthquake, the in£uence of pre- that the increased stream£ow following the earth- earthquake discharge on the response together quake must have originated either from dynamic with the spatial correspondence between increased strain accompanying liquefaction, or settling and stream£ow and near-surface volumetric strain compaction of near-surface aquifers that did not support the interpretation that stream£ow re- strongly in£uence base£ows. The strong in£uence sponse was caused by settling or compaction of of the pre-earthquake base£ow discharge on the unconsolidated shallow aquifers and dynamic re- magnitude of observed response argues in favor sponse to shaking of saturated valley-bottom de- of the interpretation that stream£ow changes were posits. caused by shaking of variably saturated near-surface deposits along river valleys and by compaction of shallow aquifers developed in Pleistocene Acknowledgements outwash sands of the Seattle Basin. Other potential explanations for the increased We thank David Mitchell (Snohomish County), stream£ow in response to the Nisqually earth- Mark Biever (Thurston County), Casey Clishe quake are not supported by our observations. In (Washington Department of Ecology), John Col- particular, the rapid response indicates a shallow lins (Pierce County), Jim LeCuyer (Kitsap PUD), source within 100 m of the water table for the Mark Mastin and Luis A. Fuste¤ (United States increased £ow. This precludes both expulsion of Geological Survey), Beth Schmoyer (City of Seat- over-pressured £uids in the seismogenic zone and tle), and Tim Tayne (City of Olympia) for provid- pore-pressure di¡usion following co-seismic strain ing data from their stream gage networks. We in the upper crust as dominant mechanisms for also thank Michael Manga, Norio Matsumoto, stream£ow changes associated with the earth- Tom Lisle, Lori Dengler, and an anonymous re- quake. The near-surface origin of the hydrologic viewer for critiques of draft manuscripts that response is consistent with the deep focus of the helped shape the analysis presented here. Focal Nisqually earthquake because low-frequency, non- solution for the Nisqually earthquake and GIS directional shaking at the ground surface would coverages of modeled peak ground accelerations tend to favor liquefaction as a mechanism for provided by UW Seismological Station. Ralph stream£ow changes. We note, however, that fre- Haugerud of the U.S. Geological Survey guided quent runo¡-producing events in the region lim- us to the compilation of locations of ¢eld evidence ited our ability to detect any signature of delayed for liquefaction observed after the Nisqually response from deeper sources. earthquake at http://www.geophys.washington. edu/SEIS/EQ_Special/WEBDIR_01022818543p/ maps/gd.jpg. Vertical displacement pattern shown 6. Conclusions in Figure 8 is from modeling by K.E. Austin and M.M. Miller (http://www.panga.cwu.edu/ The magnitude of hydrologic response to earth- olympia_eq/nisqually.jpg).[RV] quakes is inherently site-speci¢c due to local geological conditions such as the presence of unconsolidated or lique¢able deposits. But our ¢ndings indicate that local near-surface strain and re- References sponse to strong ground shaking can explain [1] R. Muir-Wood, G.C.P. King, Hydrological signatures of stream£ow response to the Nisqually earthquake. earthquake strain, J. Geophys. Res. 98 (1993) 22035^ Moreover, our analysis provides further support 22068. for the observation that, given earthquake magni- [2] E.A. Roelo¡s, Persistent water level changes in a well near

EPSL 6576 18-3-03 Cyaan Magenta Geel Zwart 28 D.R. Montgomery et al. / Earth and Planetary Science Letters 209 (2003) 19^28

Park¢eld, California, due to local and distant earth- [15] Sta¡ of the Paci¢c Northwest Seismograph Network quakes, J. Geophys. Res. 103 (1998) 869^889. (PNSW Sta¡), Preliminary report on the Mw = 6.8 Nis- [3] R.H. Sibson, Fluid £ow accompanying faulting: Field qually, Washington earthquake of 28 February 2001, Seis- evidence and models, in: D.W. Simpson, P.G. Richards mol. Res. Lett. 72 (2001) 352^361. (Eds.), Earthquake Prediction, American Geophysical [16] W. Brutsaert, J.P. Lopez, Basin-scale geohydrologic Union Maurice Ewing Series 4, 1981, pp. 593^603. drought £ow features in riparian aquifers in the southern [4] A. Nur, Matsushiro, Japan, Earthquake swarm: Con¢r- Great Plains, Water Resour. Res. 34 (1998) 233^240. mation of the dilatancy-£uid di¡usion model, Geology 2 [17] E. Roelo¡s, Poroelastic techniques in the study of earth- (1974) 217^221. quake-related hydrologic phenomena, Adv. Geophys. 37 [5] H. Wakita, Water wells as possible indicators of tectonic (1996) 135^195. strain, Science 189 (1975) 553^555. [18] G.A. Papadopoulos, G. Lefkopoulos, Magnitude-distance [6] S.H. Wood, C. Wurts, T. Lane, N. Ballenger, M. Shaleen, relations for liquefaction in soil from earthquakes, Bull. D. Totorica, The Borah Peak, Idaho earthquake of Oc- Seismol. Soc. Am. 83 (1993) 925^938. tober 28, 1983 - Hydrologic e¡ects, Earthq. Spectra 2 [19] R.A. Haugerud, G. Thomas, S.P. Palmer, P. Lombard, (1985) 127^148. Regional map view of instrumentally-determined ground [7] R.C. Briggs, H.C. Troxell, E¡ects of the Arvin-Tehachapi motions, Nisqually earthquake of 28 February 2001, Seis- earthquake on spring and stream £ow, in: G.B. Oake- mol. Res. Lett. 72 (2001) 393. shott (Ed.), Earthquakes in Kern County California, dur- [20] http://spike.geophys.washington.edu/shake/0102281854/ ing 1952, California Division of Mines and Geology Bull. products.html. 171, 1955, pp. 81^97. [21] T. Sato, R. Sakai, K. Furuya, T. Kodama, Coseismic [8] R. Waller, E¡ects of the March 1964 Alaska earthquake spring £ow changes associated with the 1995 Kobe earth- on the hydrology of south-central Alaska. U.S. Geologi- quake, Geophys. Res. Lett. 27 (2000) 1219^1222. cal Survey Professional Paper 544-A, 1966. [22] E. Reolo¡s, W.R. Danskin, C.D. Farrar, D.L. Galloway, [9] C.-Y. King, D. Basler, T.S. Presser, C.W. Evans, L.D. S.N. Hamlin, E.G. Quilty, H.M. Quinn, D.H. Schaefer, White, A. Minissale, In search of earthquake-related hy- M.L. Sorey, D.E. Woodcock, Hydrologic e¡ects of the drologic and chemical changes along Hayward Fault, June 28, 1992 Landers, California, earthquake, U.S. Geo- Appl. Geochem. 9 (1994) 83^91. logical Survey Open File Report 95-42, 1995. [10] R.O. Briggs, E¡ects of Loma Prieta Earthquake on sur- [23] E.G. Quilty, C.D. Farrar, D.L. Galloay, S.N. Hamlin, face waters in Waddell Valley, Water Resour. Bull. 27 R.J. Laczniak, E.A. Roelo¡s, M.L. Sorey, D.E. Wood- (1991) 991^999. cock, Hydrologic e¡ects associated with the January 17, [11] S. Rojstaczer, S. Wolf, Permeability changes associated 1994 Northridge, California, earthquake, U.S. Geological with large earthquakes: An example from Loma Prieta, Survey Open File Report 95-813, 1995. California, Geology 20 (1992) 211^214. [24] R. Waller, E¡ects of the March 1964 Alaska earthquake [12] S. Rojstaczer, S. Wolf, R. Michel, Permeability enhance- on the hydrology of the Anchorage area, U.S. Geological ment in the shallow crust as a cause of earthquake-in- Survey Professional Paper 544-B, 1966. duced hydrological changes, Nature 373 (1995) 237^239. [25] A.C. Lawson (Chairman), The California Earthquake of [13] T. Tokunaga, Modeling of earthquake-induced hydrolog- April 18, 1906, Report of the State Earthquake Investiga- ical changes and possible permeability enhancement due tion Commission, Vol. 1, Carnegie Institution of Wash- to the 17 January 1995 Kobe Earthquake, Jpn. J. Hydrol. ington, Washington, DC, 1908. 223 (1999) 221^229. [26] M.M. Miller et al., Geodetic signature of the February 28, [14] M. Manga, Origin of postseismic stream£ow changes in- 2001 Nisqually Earthquake: PANGA’s earthquake re- ferred from base£ow recession and magnitude-distance sponse, Seismol. Res. Lett. 72 (2001) 391. relations, Geophys. Res. Lett. 28 (2001) 2133^2136.

EPSL 6576 18-3-03 Cyaan Magenta Geel Zwart