Space geodetic observation of expansion of the ... - Wiley Online Library

15 downloads 111732 Views 1MB Size Report
7Now at Jet Propulsion Laboratory, Pasadena, California, USA. 8U.S. Geological ..... C. Bowers of the USGS Apple Valley office provided data from the. Baldwin ...
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112, B03409, doi:10.1029/2006JB004448, 2007

Space geodetic observation of expansion of the San Gabriel Valley, California, aquifer system, during heavy rainfall in winter 2004–2005 N. E. King,1 D. Argus,2 J. Langbein,3 D. C. Agnew,4 G. Bawden,5 R. S. Dollar,1 Z. Liu,6,7 D. Galloway,5 E. Reichard,8 A. Yong,1 F. H. Webb,2 Y. Bock,4 K. Stark,9 and D. Barseghian9 Received 14 April 2006; revised 13 October 2006; accepted 27 October 2006; published 24 March 2007.

[1] Starting early in 2005, the positions of GPS stations in the San Gabriel valley region

of southern California showed statistically significant departures from their previous behavior. Station LONG moved up by about 47 mm, and nearby stations moved away from LONG by about 10 mm. These changes began during an extremely rainy season in southern California and coincided with a 16-m increase in water level at a nearby well in Baldwin Park and a regional uplift detected by interferometric synthetic aperture radar. No equivalent signals were seen in GPS station position time series elsewhere in southern California. Our preferred explanation, supported by the timing and by a hydrologic simulation, is deformation due to recharging of aquifers after near-record rainfall in 2004–2005. We cannot rule out an aseismic slip event, but we consider such an event unlikely because it requires slip on multiple faults and predicts other signals that are not observed. Citation: King, N. E., et al. (2007), Space geodetic observation of expansion of the San Gabriel Valley, California, aquifer system, during heavy rainfall in winter 2004 – 2005, J. Geophys. Res., 112, B03409, doi:10.1029/2006JB004448.

1. Introduction [ 2 ] Southern California’s San Gabriel valley is an approximately 200 square km urban area in the Los Angeles metropolitan area (Figure 1). The valley is bounded by the San Gabriel mountains and the Raymond fault to the north, the Repetto and Montebello hills to the southwest, and the San Jose and Puente hills to the southeast [Yeats, 2004]. The Mojave segment of the San Andreas fault lies about 35 km to the north, on the other side of the San Gabriel mountains. Because the direction of plate motion between the Pacific and North American plates is about 25° clockwise from the local strike of the San Andreas fault [DeMets et al., 1990], plate motion in southern California has a significant northsouth compressional component. This compression has raised the east-west Transverse Ranges, including the mountains and the hills that bound the San Gabriel valley. Faults near or within the valley include the reverse Puente Hills blind thrust fault [Pratt et al., 2002; Shaw et al., 2002; 1

U.S. Geological Survey, Pasadena, California, USA. Jet Propulsion Laboratory, Pasadena, California, USA. U.S. Geological Survey, Menlo Park, California, USA. 4 Institute of Geophysics and Planetary Physics, University of California, San Diego, La Jolla, California, USA. 5 U.S. Geological Survey, Sacramento, California, USA. 6 Department of Geophysics, Stanford University, Stanford, California, USA. 7 Now at Jet Propulsion Laboratory, Pasadena, California, USA. 8 U.S. Geological Survey, San Diego, California, USA. 9 Stark Consulting, LLC, Pasadena, California, USA. 2 3

Copyright 2007 by the American Geophysical Union. 0148-0227/07/2006JB004448$09.00

Dolan et al., 2003], Elysian Park blind thrust fault [Oskin et al., 2000], Sierra Madre oblique fault [Crook et al., 1987], and the right-lateral Chino and Whittier and left-lateral San Jose and Raymond faults (see summary by Yeats [2004]). The strike slip faults also have varying degrees of reverse slip. Notable recent earthquakes near the San Gabriel valley include the 1987 M 5.9 Whittier Narrows earthquake on the Puente Hills thrust fault [Hauksson and Jones, 1989], the 1988 M 5.0 Pasadena earthquake on the Raymond fault [Jones et al., 1990], the 1988 M 4.6 and 1990 M 5.2 Upland earthquakes on the San Jose fault [Hauksson and Jones, 1991], and the 1991 M 5.8 Sierra Madre earthquake on a branch of the Sierra Madre fault [Hauksson, 1994]. [3] The valley’s tectonics control topography; topography and geology affect hydrology. The San Gabriel River, Rio Hondo, and other streams have headwaters in the San Gabriel mountains and flow south through the San Gabriel valley [California Department of Water Resources, 1966, 2004]. Alluvial deposits more than 1 km thick bear groundwater. The faults and the southern hills are barriers to groundwater flow, and the only outlet is Whittier Narrows where the San Gabriel River and Rio Hondo flow through the Puente Hills at the valley’s southern edge. [4] Lisowski et al. [1991] used line length changes from trilateration to show that the right-lateral shear strain rate across the Mojave segment of the San Andreas fault is 4  107/yr, and that there is little north-south contraction in the San Gabriel mountains. Argus et al. [1999, 2005] used trilateration and space geodetic crustal deformation results to document about 6 mm/yr of north-south contraction between the San Gabriel mountains and the Palos Verdes

B03409

1 of 11

B03409

KING ET AL.: SAN GABRIEL VALLEY GEODETIC ANOMALY

Figure 1. Views of the study area. Black lines show faults, white squares show water wells, and black circles show continuously operating GPS stations (a subset of the Southern California Integrated GPS Network). (a) Los Angeles metropolitan area. The rectangle outlines the area shown in the lower map. (b) San Gabriel valley. The large white square shows the Baldwin Park Key Well.

peninsula. About 4.5 mm/yr of this north-south contraction occurs in a 25-km-wide east-west trending belt that includes the San Gabriel valley. [5] The San Gabriel valley lies within the Global Positioning System (GPS) array installed by the Southern California Integrated GPS Network (SCIGN), a 1996 – 2005 collaborative project of the U.S. Geological Survey (USGS), the Jet Propulsion Laboratory (JPL), the Scripps Orbital and Permanent Array Center (SOPAC) at the Scripps Institution of Oceanography, and the Southern California Earthquake Center (SCEC). In 2001 SCIGN completed installation of a 250-station array of highprecision, permanent, continuously operating GPS stations in southern California (Figure 1). The purpose of the array is state-of-the-art measurement of crustal deformation due to earthquakes and fault slip [Hudnut and King, 2001; Hudnut et al., 2002a]. The array is particularly dense in the

B03409

Los Angeles metropolitan area because this heavily populated region has very high exposure to earthquake risk [Jackson et al., 1995] and previous crustal deformation measurements were sparse. [6] During the selection of 200 new stations in 1998– 2000, SCIGN tried to avoid sites near water wells or other sources of hydrologic signals. However, the difficulties of finding and permitting suitable sites made this almost impossible [King et al., 1998]. Interferometric synthetic aperture radar (InSAR) imagery showed only localized hydrologic deformation in the Los Angeles basin at that time. However, programs beginning in the mid-1990s created economic incentives for water users to purchase surplus surface water during the winter. Users began seasonal groundwater pumping, pumping less water during the winter and more during the summer [Bawden, 2003]. This caused surface deformation, imaged with InSAR and detected in the SCIGN time series that was larger and more widespread than expected [Bawden et al., 2001; Watson et al., 2002; Lanari et al., 2004; Argus et al., 2005]. Groundwater pumping produces geodetically measurable elastic and inelastic surface motion; GPS stations move down and horizontally toward areas of water level decline during pumping, and up and horizontally away from areas of water level rise during recharge. When water levels reach new lows there is subsidence due to inelastic unrecoverable compaction of the sediment. Thus we have known for several years that the GPS station position time series contain quasiperiodic signals (recharge-pumping cycle) and trends (permanent subsidence) that are not related to faulting and earthquakes. [7] In this study we report a large geodetic signal with both vertical and horizontal motion. The signal is neither cyclical nor a trend, and is related to rapid natural aquifer recharge rather than urban water management. The southern California rainy season is in winter, and the 2004 – 2005 season was the second wettest since 1878. A total of 946 mm of rain fell on downtown Los Angeles, just 26 mm short of the record set in 1883– 1884. The 2004– 2005 rainfall is 2.5 times the seasonal average of 381 mm. Water level in wells near the base of the San Gabriel and San Bernardino mountains increased up to 58 m, while water level in wells in the valleys of the greater Los Angeles area increased by up to 21 m (Figure 2). Coincident with these changes, GPS and InSAR studies detected surface deformations. The vertical station position of GPS station LONG, and the areal dilatation inferred from horizontal station positions from stations around LONG, tracked the rapid water level increase from 2005.0 to 2004.5. Water level, the vertical position of LONG, and the areal dilatation remain elevated since 2004.5, and the GPS station positions have not returned to the values extrapolated from pre-2005 time series. To obtain a more coherent picture of the tectonic deformation that the array was installed to study, we will have to learn more about the relationship between aquifer changes and surface deformation.

2. GPS, Water Well, and InSAR Data [8] Each day JPL, SOPAC, and USGS independently invert GPS observables for the positions of the 250 sites

2 of 11

B03409

KING ET AL.: SAN GABRIEL VALLEY GEODETIC ANOMALY

B03409

Figure 2. Water wells and faults in the Los Angeles area. White lines show faults (CF, Chino; EF, Elsinore; NIF, Newport-Inglewood; PVF, Palos Verdes; RF, Raymond; SAF, San Andreas; SGF, San Gabriel; SJF, San Jacinto; SjoF, San Jose; SMdF, Sierra Madre; SMnF, Santa Monica; WF, Whittier), triangles show continuously operating GPS stations, the blue area shows the San Gabriel valley groundwater basin, and blue bars show water level increases during the November 2004 to May 2005 rainy season. The red bar shows water level increase in the Baldwin Park Key Well.

installed by SCIGN. Each daily station file contains 24 hours of raw GPS data sampled at 30 s intervals. [9] Using 5-min means of the original 30-s data, JPL first estimates orbits and satellite clocks using about 50 worldwide sites with the GIPSY data reduction package (http:// gipsy.jpl.nasa.gov/orms/goa/). Next, the positions of about 800 sites are ‘‘precise point positioned’’; the positions are estimated using orbit and clock parameters estimated in the previous step [Zumberge et al., 1997]. The relative positions of the 250 southern California sites are then estimated more accurately by resolving ambiguities; the number of carrier waves in the double-differenced phase delay is determined [Blewitt, 1989]. The southern California station positions are next regionally filtered; the mean misfit of velocities fit to stations positions on one day is subtracted from all position estimates [Wdowinski et al., 1997]. Finally, velocities, intercepts, sinusoids, antenna offsets, and coseismic offsets are estimated [Argus et al., 2005]. Time series are available from http://reason.scign.org/scignDataPortal. [10] SOPAC processes raw continuous GPS data decimated to 120 s with the suite of programs GAMIT version 10.24 [King and Bock, 2002] and GLOBK/GLORG version 5.1 [Herring, 2002] in 24-hour segments. Using distributed analysis, global stations are divided into 5 overlapping subnetworks and regional stations are divided into 30 subnetworks including 10 subnetworks in southern California. Each subnetwork is processed with GAMIT, employing the

double-differencing approach, and estimating station positions, satellite orbital parameters, earth orientation parameters, zenith and slant tropospheric parameters, and including integer-cycle phase ambiguity resolution. The global and regional solutions are then combined with GLOBK/ GLORG to determine a time series of 24-hour positions with respect to the ITRF2000 reference frame. After regional filtering [Wdowinski et al., 1997], the daily position time series are analyzed for interseismic rates, coseismic and postseismic deformation, annual and semiannual fluctuations, and nuisance offset parameters using the method described by Nikolaidis [2002]. Typical daily repeatabilities are 1 mm in the horizontal and 3 to 4 mm in the vertical. (Time series are available at http://sopac. ucsd.edu/cgi-bin/refinedJavaTime series.cgi.) [11] USGS also uses the GAMIT/GLOBK/GLORG software, using the same processing strategies as SOPAC. However, USGS does not perform the subsequent regional filtering or modeling described above. USGS uses orbits from the International GNSS Service (IGS [Beutler et al., 1999]) in a staggered processing scheme that makes its time series the most up-to-date. To be ready for earthquake response, USGS uses the IGS ultrarapid orbit to process data as soon as the 24-hour files close, and later reprocesses each day’s data as the higher quality IGS rapid and precise orbits become available. (Time series are available from http://pasadena.wr.usgs.gov/scign/Analysis/.)

3 of 11

B03409

KING ET AL.: SAN GABRIEL VALLEY GEODETIC ANOMALY

B03409

[12] The signal presented in this paper was first observed in the USGS daily station position time series. Since it was a change in the station position rates of change, it took some time to notice. No such signal had been observed in southern California GPS positions before, so we verified the sense and magnitude of the changes in the SOPAC and JPL time series (Figure 3). For simplicity, and because USGS, SOPAC, or JPL station positions would yield essentially the same signal, most of the results we show are from SOPAC time series. King et al. [2001, 2002] showed that JPL and SOPAC GPS station positions (in the same reference frame) agree to within better than 1 mm in the north and east components and 1 to 3 mm in the vertical; baselines, which do not depend on reference frame, agree to within 1 to 2.5 parts in 109. The REASoN project [Webb et al., 2004; http://reason.scign.org/scignDataPortal] compares and integrates station positions from JPL, SOPAC, and USGS, and finds that they are consistent for the purposes of this study. [13] During the 2004 – 2005 rainy season, water levels increased in Los Angeles area wells (Figures 1 and 2). In the San Gabriel valley, the Baldwin Park Key Well provides the most detailed water level data. Since June 1932 water level has been measured, in feet, to the nearest 3 mm (0.01 foot). The sampling rate was every few days until 1970, and approximately monthly since then. [14] Interferometric Synthetic Aperture Radar (InSAR) imagery was used to map spatial and temporal variations in ground motion across metropolitan Los Angeles and the San Gabriel valley. The May 1998 to May 2000 data were collected by the European Space Agency satellites ERS-2 (G. Peltzer electronic communication, 2003). The January 2005 to July 2005 ENVISAT data are ascending satellite tracks. In July 2005, after the signal was observed in the GPS time series, the satellites were specifically tasked to image the San Gabriel valley. The SAR imagery were processed using GAMMA InSAR software with DELFT orbits [Werner et al., 2000].

3. Time, Statistical Significance, and Size of the Signal [15] The unprecedented signal appeared as rate changes, starting around the beginning of 2005, in the San Gabriel

Figure 3. Detrended SOPAC time series (red) for San Gabriel valley stations, with the water level record from the Baldwin Park Key Well (black). Vertical red lines mark rapid water level rise from 2005.0 to 2005.4. The left axis shows mm for GPS time series; trends and offsets have been removed, and a constant amount has been subtracted from each time series. Because of power problems, there are no recent data at LPHS. The right axis shows water level elevation, in m above mean sea level, in the Baldwin Park Key Well; these values are absolute. (a) North GPS component. JPL (green) and USGS (blue) time series are shown for station VYAS. (b) East GPS component. JPL (green) and USGS (blue) time series are shown for station VYAS. (c) Vertical GPS component. JPL (green) and USGS (blue) time series are shown for station LONG. 4 of 11

B03409

KING ET AL.: SAN GABRIEL VALLEY GEODETIC ANOMALY

Figure 4. Rate change as a function of hinge point (date of rate change), expressed in standard deviations, for the north, east, and vertical components of stations CVHS, LPHS, VYAS, WCHS, and WNRA.

valley GPS station position time series (Figure 3). The onset was abrupt, and the new rates persisted until about 2005.4. We seek to estimate the time, statistical significance, and size of the signal. [16] The GPS station position time series contain linear trends due to tectonic deformation and reference frame representation, hydrologic signals including trends and quasiperiodic signals [Bawden et al., 2001; Watson et al., 2002; Lanari et al., 2004; Argus et al., 2005], offsets [Williams, 2003], and fluctuations due to noise. Langbein and Johnson [1997], Langbein [2004] and Williams et al. [2004] showed that noise models are complex and that the calculated uncertainties of signals inferred from these time series are too small if inappropriately simple models are used. [17] For the north, east, and vertical components of stations CHVS, LPHS, VYAS, WCHS, and WNRA, we used the maximum likelihood method described by Langbein [2004] to identify the appropriate noise model, (described in the frequency domain as a power spectral density function P( f ), where f is frequency), estimate its parameters, find the occurrence time that maximizes the statistical significance of the signal, and calculate its size. Results varied among stations and components, but in general the white noise (P( f ) = P0) is 1 to 1.5 mm for the north and east components and 3 to 4 mm for the vertical. The nonwhite part of the error spectrum is typically power law noise (P( f ) = P0/f N, where N need not be an integer) with exponents between 0.8 and 1.1 and amplitudes of 2 to 4 mm for the north and east and 8 to 10 mm for the vertical. In a few cases, the preferred model is flicker noise (power law noise with N = 1) plus quasiperiodic noise. [18] Using these noise models, we estimated the time and significance of the signal by calculating the north, east, and _ D_e, D_v) and their standard vertical rate changes (Dn, deviations (sDn_ ; sD_e ; sD_v ) at a ‘‘hinge point’’ date and moving the hinge point through the time series. We defined 1.5-year subsections of the time series with the beginning at t0, the hinge point at th = t0 + 1 year, and the end at te = t0 + 1.5 years. We moved the hinge point through each north, east, and vertical time series in increments of 30 days from

B03409

th = 2002.0 to th = 2005.0, and normalized each rate change by its standard deviation. This analysis tests how common it is to detect a statistically significant signal in 6 months of a station position time series, based on the trend established by the previous year. As the hinge point moves toward _ Dn_ , D_e/sD_e , D_v/ 2005.0, the normalized rate changes (Dn/s sD_v ) increase sharply (Figure 4). For stations CVHS, LPHS, VYAS, WCHS, and WNRA the maximum or minimum values are 5.1, 8.1, 6.8, 7.7, and 6.2, respectively, _ Dn_ , 7.5, 6.4, 3.4, 4.7, and 2.3 for D_e/sD_e , and for Dn/s 5.3, 1.4, 2.6, 3.3, and 4.1 for D_v/sD_v . Only 9.4%, 5.3%, and 2.4% of the north, east, and vertical normalized rate changes are larger than ±3, and only 6.1%, 2.0%, and 0.4% are _ Dn_ , D_e/sD_e , D_v/sD_v larger than ±5. All the values of Dn/s in excess of ±4 occur as the hinge point approaches 2005. Therefore the 2005 changes in slope are larger, and more statistically significant, than any previously seen in these time series. To within about 30 days, the best fitting time is 2005.0. [19] To quantify the signal at the best fitting hinge point th = 2005.0, we removed gross outliers, the linear trend, seasonal oscillations, and offsets. For each component of each station, we performed a least squares fit to the model X i ðt Þ ¼ Ai1 þ Ai2 t þ Bi1 cos 2pt þ Bi2 sin 2pt n X   þ Cki H t  tki

ð1Þ

k¼1

where t  2005.0, i indicates the station, Xi is the north, east, or vertical component, t is time in years from the beginning of each time series to 2005.0, Ai1 and Ai2 are the intercept and slope of a linear trend, Bi1 and Bi2 describe the annual oscillation, H(t) is the Heaviside function (H(t) = 0 for t < 0, H(t) = 1 for t  0), and the Cik are the amplitudes of offsets at times tki for k = 1, 2,. . ., n. The unknowns are Ai1, Ai2, Bi1, Bi2, and the Cki, while the offset times tik are specified in advance. We extrapolated this pre-2005.0 model to the end of the time series and calculated the residuals over the entire time series. The pre-2005.0 residuals are flat, because the model is based on the pre-2005.0 time series. If there is no signal, the post-2005.0 residuals are also flat. If there is a signal, the post-2005.0 residuals have a nonzero trend and the magnitude of the signal is the 2005.4 value predicted by that trend. [20] We applied this analysis to time series from all stations in southern California, except those affected by postseismic deformation due to the 1999 Hector Mine [Hudnut et al., 2002b; Owen et al., 2002] and 2004 Parkfield earthquakes and those with extremely sparse or noisy station position time series. Only the San Gabriel valley stations display a significant signal. These stations exhibit a regional signal consisting of a 47-mm vertical uplift at station LONG and, at nearby stations, horizontal displacements of about 10 mm away from LONG (Figures 5a and 5b). [21] InSAR imagery from January to July 2005 corroborates the signal observed in the GPS time series (Figure 5c). Since the largest measured GPS signal is in the vertical at LONG, and InSAR is most sensitive to vertical motion, we assumed that the predominant surface deformation measured by InSAR is vertical. Therefore we projected the unwrapped InSAR range changes onto the InSAR unit

5 of 11

KING ET AL.: SAN GABRIEL VALLEY GEODETIC ANOMALY

B03409

B03409

Figure 5. The 2005.0– 2005.4 GPS signals in the San Gabriel Valley. The signal is the cumulative discrepancy between the post-2005.0 trend and the extrapolation of the trend and annual oscillation established by data before 2005.0. For comparison, signals calculated using the same method are shown for other stations in southern California. Note the different scales of Figures 5a, 5b, and 5c. (a) Vertical signal, (b) horizontal signal, (c) horizontal signal, with January 2005 to July 2005 InSAR interferogram. The interferogram has been converted to vertical deformation.

vector (north = 0.069, east = 0.315, vertical = 0.947) for the center of the San Gabriel valley. The image shows about 30 mm of uplift over an area about 20 km across by 14 km wide, with additional uplift of 15 mm west of LONG. Station LONG, near the center of the uplift, moved upward, while the GPS stations in a ring around LONG (Figure 5) moved away from the center of the uplift.

4. Cause of the Signal [22] Since the beginning of geodetic observations in the San Gabriel valley in the mid-1990s, there has been a striking correlation between groundwater elevation and surface deformation measured by both GPS and InSAR. Water level in the Baldwin Park Key Well increased from 1932 to the mid-1940s, decreased rapidly as development put pressure on the valley’s water resources, and has oscillated since the late 1950s (Figure 6a). Just before the very wet 2004 – 2005 rainy season, water level in the Baldwin Park Key Well was at its lowest value since measurements began in 1932. GPS observations in the

San Gabriel valley began in 1995 at station LONG and in mid-1998 at most other stations. Since installation, these stations have measured both the interseismic tectonic trend and a hydrological deformation caused by the long, steady decrease in water level. In particular, InSAR imagery between May 1998 and May 2000 shows that the region near LONG subsided by about 15 mm, [Argus et al., 2005, Figure 6] in good agreement with the LONG vertical position time series in Figure 6. [23] During the 2004 – 2005 rains, water levels in the Baldwin Park Key Well and other San Gabriel valley water wells rose by about 16 m (Figures 2 and 6). As Figure 5 shows, the accompanying change in the GPS station positions is both vertical (at LONG) and horizontal (at surrounding stations). We express horizontal deformation in the San Gabriel valley as areal dilatation (11 + 22)/2, where 11 and 22 are north and east extension, respectively) computed from the north and east components of stations CIT1, CVHS, LONG, RHCL, SGHS, VYAS, and WHCS. Both dilatation and the vertical station position component of LONG correlate well with water level since mid-1998,

6 of 11

B03409

KING ET AL.: SAN GABRIEL VALLEY GEODETIC ANOMALY

B03409

Figure 6. (a) Water level elevation, in m above mean sea level, from the Baldwin Park Key Well since 1932. The horizontal black bar shows the time period during which GPS observations were made. (b) Water level elevation in m above mean sea level, from the Baldwin Park Key Well (black), vertical component of LONG (red), and areal dilatation ((11 + 22)/2 (blue), where the 11 and 22 are east-west and north-south extension, respectively) for the San Gabriel valley, from 1995.5. The inner and outer left axes show mm for the LONG vertical time series and areal dilatation in parts per million, respectively; a constant amount has been subtracted from these time series to enable stacking. The right axis shows water level elevation in m above mean sea level; these values are absolute.

tracking the 7-year decrease, the abrupt increase in 2005.0 – 2005.4, and the sustained new levels since 2005.4. [24] The ratio of surface vertical displacement to groundwater elevation change provides an estimate of the elastic storage coefficient. From 2005.0 to 2005.4, the water level increased by 16 m while LONG rose 47 mm. The 1998.5 – 2005 drawdown in the Baldwin Park Key Well is 21 m; a linear fit to the vertical changes at LONG yields average long-term subsidence of 34 mm during the same time period (Figure 6). The estimated elastic storage coefficients are 0.0029 and 0.0016 for the 2005.0 to 2005.4 water level increase and the 1998 – 2005 drawdown, respectively. Because the drawdown brought water levels down to historic low values, it is likely that some inelastic compaction occurred [Leake and Prudic, 1991]. [25] The relationship between vertical displacement and water level change is consistent with results from other basins in California and Nevada. In the San Bernardino valley, just east of the San Gabriel valley, Lu and Danskin [2001] used InSAR to detect uplift of about 30 and 70 mm in two areas in 1993. The ratio of surface uplift to increase in groundwater elevation was 0.0030 to 0.0050 in one area

(Lytle Creek) and 0.0006 and 0.0008 in the other (Santa Ana River area). Our estimated ratios are similar to those observed for Lytle Creek. Hanson et al. [2004] used extensometer data and water levels in the Santa Clara valley to estimate elastic storage coefficients of 0.0012 and 0.0062. Hoffmann et al. [2001] used InSAR and water level changes to estimate elastic storage coefficients for six locations in the Las Vegas valley. Values were highest (0.0015 to 0.0034) in the center of the valley. These values are similar to those estimated for the San Gabriel valley. The Baldwin Park Key Well and the LONG GPS station are a few km apart in the center of the basin, where deposits are thickest and are likely to have higher portions of finer grained materials. In Las Vegas, the estimated aquifer thickness in the center of the valley was 160– 180 m. In the center of the San Gabriel valley, the thicknesses of the recent and older alluvium are estimated to be as high as 1200 m [California Department of Water Resources, 1966]. [26] Seasonal and annual fluctuations in aquifer-system pore fluid pressures (measured as groundwater levels in open wells) accompanying the development of groundwater resources typically cause measurable ground displacements

7 of 11

B03409

KING ET AL.: SAN GABRIEL VALLEY GEODETIC ANOMALY

in unconsolidated basin fill groundwater basins [Galloway et al., 1999]. Elastic deformation of aquifer systems accompanying changes in pore fluid pressure is described by a fully coupled poroelastic response [Biot, 1941], and the principle of effective stress [Terzaghi, 1925]. Hydrologists have typically simulated subsidence and uplift associated with changes in regional, confined groundwater flow based on the storage coefficient S defined by Jacob [1940], where vertical displacement of an aquifer uzz can be represented as uzz ¼ Sk Dh

ð2Þ

where Sk represents the volume of water released from or taken into storage in the aquifer related to vertical

B03409

compression/expansion of the aquifer and Dh is the change in head or water level change in the aquifer. [27] We used equation (2) to simulate the geodetically observed uplift following heavy rainfall as a hydrologic response to water level increase in multiple wells. The spatial extent and magnitude of the uplift can be explained as an elastic aquifer system response associated with the recovering groundwater levels from a pumped aquifer (Figure 7a). The water level recovery and uplift range from about 20 m and 0.06 m, respectively, near the wells and diminish exponentially away from wells to 0 near the periphery. This is in good agreement with the geodetic and groundwater level measurements. For the specified aquifer system hydraulic parameters, the vertical displacement field is highly sensitive to the location of the wells. Heterogeneity and anisotropy of aquifer system properties would result in more irregular and diffuse distribution of the vertical displacement, as seen in the InSAR imagery. The radial horizontal deformation observed in Figure 5 can be approximated as an elastic response to the inflation and doming of the ground surface associated with aquifer recovery (Figures 7b and 7c). Modeled horizontal displacements are largest near the basin margins, where they reach one third of the uplift over the aquifer, similar to the ratio observed in the GPS time series. [28] Although correlations between GPS, InSAR, and water well levels strongly suggest that this is a hydrologic signal, we evaluated the possibility of an aseismic slip transient on local faults. Such transients, or ‘‘silent earthquakes,’’ have been observed in subduction zones [Dragert et al., 2001; Freymueller et al., 2001; Hirose et al., 1999; Lowry et al., 2001; Ozawa et al., 2001]. We applied the Extended Network Inversion Filter developed by McGuire and Segall [2003] to invert position time series for slow slip transients. It is possible to find a complex model that

Figure 7. Modeled aquifer system recovery for vertical hydrologic uplift and approximated horizontal and vertical deformation approximated with an elastic half-space model. (a) Idealized aquifer system response to seasonal (6 month) recovery following a period of seasonal groundwater extraction from nine arbitrarily spaced pumping centers in a hypothetical 50  50 km groundwater basin. The recovery of hydraulic head accompanying the elastic expansion of the aquifer system is computed using the nonequilibrium equation of Theis [1935]. The vertical displacement (uplift) of land surface is computed using the one-dimensional (no horizontal strains) storage coefficient (S) typically used by hydrologists in analyses of regional groundwater flow. Grid spacing 100 m; transmissivity 500 m2/d; storage coefficient 0.0035 (accounting for the compressibility of water, assuming porosity 0.20, and aquifer system thickness 500 m); extraction at each pumping center 10,900 m3/d, time 182.5 days. (b) A thin 20 km by 20 km aquifer embedded at a depth of 0.5 km in an elastic half-space with Young’s modulus 104 MPa and Poisson’s ratio 0.25, roughly approximating the San Gabriel basin. Arrows show horizontal displacements in map view. (c) Displacements (arrows) in the 20 km by 20 km aquifer, for cross section A-A’. 8 of 11

B03409

KING ET AL.: SAN GABRIEL VALLEY GEODETIC ANOMALY

fits the GPS observations; slip is less than 100 mm everywhere, and must be distributed between the Puente Hills thrust and the Raymond, Sierra Madre, and Whittier faults. The fit to the GPS station positions is unsatisfactory because such a slip transient predicts other signals which are not observed. [29] We cannot rule out aseismic slip triggered by groundwater changes, or a combination of hydrologic and tectonic events. However, given the difficulties in fitting the station positions with the Extended Network Inversion Filter, the requirement that aseismic slip occur on several faults, previous hydrologic signals reported by Bawden et al. [2001] and Argus et al. [2005], the corroboration of the uplift on the January 2005 to July 2005 InSAR interferogram, the coincident timing of the GPS and well level signals, and the good agreement between the observed signal and the hydrologic simulation (Figure 7), our preferred explanation is that the signal is purely hydrologic. Although this is the first such signal observed since GPS observations began in the San Gabriel valley, we will probably observe additional signals in the future; the Baldwin Park Key Well record (Figure 6a) shows that abrupt water level increases like the one observed in 2005 have occurred several times in the past and therefore are likely to occur in the future. Geodetic time series will therefore contain both hydrologic and tectonic signals. Unlike previous hydrologic signals, the signal presented here cannot be modeled with simple trends or sinusoids. To understand the tectonics and earthquake hazard of the San Gabriel valley, we must learn more about its hydrology and the relationship between groundwater and surface deformation. 4.1. Other Aquifers [30] All of southern California received near-record amounts of rain during the 2004 –2005 rainy season. Water levels rose in wells all over the greater Los Angeles area. Indeed, the largest water level increases were observed in three wells at the base of the San Gabriel and San Bernardino mountains (Figure 2), probably due to runoff from the mountains during the unusually wet winter. However, the GPS signal is observed only in the San Gabriel valley (Figure 5). The January to July 2005 InSAR interferograms show that a very broad uplift occurs only the San Gabriel valley. Other areas, such as the San Bernardino valley, have much more localized areas of uplift. The difference may be due to the structural setting of the San Gabriel valley. The San Gabriel valley groundwater basin appears to be less compartmentalized internally. Furthermore, it is generally bounded on all sides by bedrock or faults with only one narrow down-gradient outlet, the Whittier Narrows at the valley’s southern edge.

5. Conclusions [31] 1. An unprecedented, statistically significant signal was observed in GPS station position time series from the San Gabriel valley. The signal is a 47 mm uplift at station LONG, with stations around LONG moving horizontally away from it by about 10 mm. The signal begins at 2005.0, the same time as a 16-m increase in water level at a nearby well. A January 2005 to July 2005 InSAR image also shows

B03409

uplift in the same part of the San Gabriel valley. The maximum uplift is about 45 mm centered just west of station LONG. Neither GPS nor InSAR detected any significant deformation elsewhere in southern California. [32] 2. Since mid-1998, the vertical position of station LONG has tracked both the steady 1998 – 2005 drawdown in the Baldwin Park Key Well and the abrupt 16 m water level recharge in 2005. [33] 3. The GPS signal and the water level in the Baldwin Park Key Well began leveling off during the 2005 summer dry months. Water level, vertical station positions at GPS station LONG, and areal dilatation inferred from the horizontal positions of GPS stations around LONG have not returned to pre-2005 levels. [34] 4. Our preferred explanation is deformation due to recharging of the aquifer after near-record rainfall in 2004– 2005. This is supported by the coincident timing of the GPS and well level signals and the good agreement between the observed signal and a hydrologic simulation. We cannot rule out an aseismic slip event occurring in addition to the hydrologic event, but we consider such an event unlikely because it requires slip on multiple faults and predicts other signals that are not observed. [35] Acknowledgments. Lucile M. Jones of USGS, Michael Shulters of USGS, and Thomas Jordan of the Southern California Earthquake Center organized a special study group to study this signal. Thomas Herring of MIT provided valuable feedback about the GPS processing and ideas about the cause of the signal. We thank Evelyn Roeloffs, Gary Fuis, Kristine Larson, and an anonymous reviewer for helpful advice and reviews. James C. Bowers of the USGS Apple Valley office provided data from the Baldwin Park Key Well. California American Water, the city of Whittier, Orange County Water District, and the Water Replenishment District of Southern California provided water level data. Tyler Johnson of USGS compiled water level data, and Peter Martin, Stan Leake, Randy Hanson, and Wes Danskin (USGS) provided input on hydrogeology. The continuously operating GPS stations in southern California were installed by the Southern California Integrated GPS Network (SCIGN), a project of the Southern California Earthquake Center (SCEC) funded by the U.S. Geological Survey, NASA, the National Science Foundation, and the W. M. Keck Foundation. Satellite radar imagery was obtained through the European Space Agency ERS1, ERS2, and ENVISAT satellites on ESA CAT-1 project 2766 and the WinSAR archive. Research at USGS is funded by the National Earthquake Hazard Reduction Program and the USGS California Water Science Center. Research by D. C. Agnew was in part supported by NSF EAR 04-5447. Research by D. F. Argus and F. H. Webb was performed at the Jet Propulsion Laboratory (JPL), California Institute of Technology, under contract with NASA.

References Argus, D. F., M. B. Heflin, A. Donnellan, F. H. Webb, D. Dong, K. J. Hurst, D. C. Jefferson, G. A. Lyzenga, M. M. Watkins, and J. F. Zumberge (1999), Shortening and thickening of metropolitan Los Angeles measured and inferred using geodesy, Geology, 27, 703 – 706. Argus, D., M. B. Heflin, G. Peltzer, F. Crampe, and F. H. Webb (2005), Interseismic strain accumulation and anthropogenic motion in metropolitan Los Angeles, J. Geophys. Res., 110, B04401, doi:10.1029/ 2003JB002934. Bawden, G. W. (2003), Separating ground-water and hydrocarbon pumping effects from tectonic contraction measurements across Metropolitan Los Angeles, CA, U.S. Geol. Surv. Open File Rep., 03-0308. Bawden, G. W., W. Thatcher, R. S. Stein, K. W. Hudnut, and G. Peltzer (2001), Tectonic contraction across Los Angeles after removal of groundwater pumping effects, Nature, 412, 812 – 815. Beutler, G., M. Rothacher, S. Schaer, T. A. Springer, J. Kouba, and R. E. Neilan (1999), The International GPS Service (IGS): An interdisciplinary service in support of earth sciences, Adv. Space Res., 23, 631 – 635. Biot, M. A. (1941), General theory of three-dimensional consolidation, J. Appl. Phys., 24, 155 – 164. Blewitt, G. (1989), Carrier phase ambiguity resolution for the Global Positioning System applied to geodetic baselines up to 2000 km, J. Geophys. Res., 94, 10,187 – 10,203.

9 of 11

B03409

KING ET AL.: SAN GABRIEL VALLEY GEODETIC ANOMALY

California Department of Water Resources (1966), Planned utilization of ground water basins: San Gabriel valley, Appendix A: Geohydrology, Bull. 104-2, Sacramento. California Department of Water Resources (2004), San Gabriel valley groundwater basin, Bull. 118, Sacramento. Crook, R., Jr., C. R. Allen, B. Kamb, C. M. Payne, and R. J. Proctor (1987), Quaternary geology and seismic hazard of the Sierra Madre and associated faults, western San Gabriel Mountains, in Recent Reverse Faulting in the Transverse Ranges, California, U.S. Geol. Surv. Prof. Pap., 1339, 27 – 64. DeMets, C., R. G. Gordon, D. F. Argus, and S. Stein (1990), Current plate motions, Geophys. J. Int., 101, 425 – 478. Dolan, J. F., S. A. Christofferson, and J. H. Shaw (2003), Recognition of paleoearthquakes on the Puente Hills blind thrust fault, California, Science, 300, 115 – 118. Dragert, H., K. Wang, and T. S. James (2001), A silent slip event on the deeper Cascadia subduction interface, Science, 292, 1526 – 1528. Freymueller, J., C. Zweck, H. Fletcher, S. Hreinsdottir, S. C. Cohen, and M. Wyss (2001), The great Alaska ‘‘earthquake’’ of 1998 – 2001, Eos Trans. AGU, 81(47), Abstract G22D-11. Galloway, D., S. A. Ingebritsen, and D. R. Jones (1999), Land subsidence in the United States, U.S. Geol. Surv. Circ., 1182. Hanson, R. T., Z. Li, and C. C. Faunt (2004), Documentation of the Santa Clara valley regional ground-water/surface water flow model, Santa Clara valley, California, U.S. Geol. Surv. Sci. Invest. Rep., 2004-5231. Hauksson, E. (1994), The 1991 Sierra Madre earthquake sequence in southern California: Seismological and tectonic analysis, Bull. Seismol. Soc. Am., 84, 1058 – 1074. Hauksson, E., and L. Jones (1989), The 1987 Whittier Narrows earthquake sequence in Los Angeles, southern California: Seismological and tectonic analysis, J. Geophys. Res., 94, 9569 – 9589. Hauksson, E., and L. Jones (1991), The 1988 and 1990 Upland earthquake: Left-lateral faulting adjacent to the central Transverse Ranges, J. Geophys. Res., 96, 8143 – 8165. Herring, T. (2002), Global Kalman Filter VLBI and GPS Analysis Program, Mass. Inst. of Technol., Cambridge. Hirose, H., K. Hirahara, F. Kimata, N. Fujii, and S. Miyazaki (1999), A slow thrust slip event following the two 1996 Hyuganada earthquakes beneath the Bungo Channel, southwest Japan, Geophys. Res. Lett., 26, 3237 – 3240. Hoffmann, J., H. A. Zebker, D. L. Galloway, and F. Amelung (2001), Seasonal subsidence and rebound in Las Vegas Valley, Nevada, observed by synthetic aperture radar interferometry, Water Resour. Res., 37, 1551 – 1566. Hudnut, K., and N. King (2001), SCIGN – New Southern California GPS Network advances the study of earthquakes, USGS Fact Sheet 069-01, U.S. Govt. Print. Off., Washington, D. C. Hudnut, K. W., Y. Bock, J. E. Galetzka, F. H. Webb, and W. H. Young (2002a), The Southern California Integrated GPS Network (SCIGN), in Seismotectonics in Convergent Plate Boundary, edited by Y. Fujinawa and A. Yoshida, pp. 167 – 189, TERRAPUB, Tokyo. Hudnut, K. W., N. E. King, J. E. Galetzka, K. F. Stark, J. A. Behr, A. Aspiotes, S. van Wyk, R. Moffitt, S. Dockter, and F. Wyatt (2002b), Continuous GPS observations of postseismic deformation following the 16 October 1999 Hector Mine, California, earthquake (M 7.1), Bull. Seismol. Soc. Am., 92, 1402 – 1422. Jackson, D. D., K. Aki, C. A. Cornell, J. H. Dieterich, T. L. Henyey, M. Mahdyiar, D. Schwartz, and S. N. Ward (1995), Seismic hazards in southern California: Probable earthquakes, 1994 to 2004, Bull. Seismol. Soc. Am., 85, 379 – 439. Jacob (1940), On the flow of water in an elastic artesian aquifer, Eos Trans. AGU, 2, 574 – 586. Jones, L. M., K. Sieh, E. Hauksson, and L. K. Hutton (1990), The 3 December 1988 Pasadena, California earthquake: Evidence for strike-slip motion on the Raymond fault, Bull. Seismol. Soc. Am., 80, 474 – 482. King, N. E., J. Behr, and J. Galetzka (1998), Southern California Integrated GPS Network Expansion: Site specifications meet reality in Los Angeles, Eos Trans. AGU, 79(45), Fall Meet. Suppl., F190. King, N. E., K. Hurst, J. Langbein, and M. van Domselaar (2001), Comparison and combination of solution from the Southern California Integrated GPS Network, Eos Trans. AGU, 82(47), Fall Meet. Suppl., Abstract G51B-0247. King, N. E., M. Heflin, T. Herring, K. Hurst, S. Kedar, J. Langbein, and L. Prawirodirdjo (2002), Toward an ITRF200 combined solution for the Southern California Integrated GPS network, Eos Trans. AGU, 83(47), Fall Meet. Suppl., Abstract G21A-0951. King, R. W., and Y. Bock (2002), Documentation for the GAMIT GPS Analysis Software, Dep. of Earth, Atmos., and Planet. Sci., Mass. Inst. of Technol., Cambridge.

B03409

Lanari, R., P. Lundgren, M. Manzo, and F. Casu (2004), Satellite radar interferometry time series analysis of surface deformation for Los Angeles, California, Geophys. Res. Lett., 31, L23613, doi:10.1029/ 2004GL021294. Langbein, J. (2004), Noise in two-color electronic distance meter measurements revisited, J. Geophys. Res., 109, B04406, doi:10.1029/ 2003JB002819. Langbein, J., and H. Johnson (1997), Correlated error in geodetic time series: Implications for time-dependent deformation, J. Geophys. Res., 102, 591 – 604. Leake, S. A., and D. E. Prudic (1991), Documentation of computer program to simulate aquifer-system compaction using the modular finite-different ground-water flow model, U.S. Geol. Surv. Open File Rep., 88-482. Lisowski, M., J. C. Savage, and W. H. Prescott (1991), The velocity field along the San Andreas fault in central and southern California, J. Geophys. Res., 96, 8369 – 8389. Lowry, A. R., K. M. Larson, V. Kostoglodov, and R. Bilham (2001), Transient fault slip in Guerrero, southern Mexico, Geophys. Res. Lett., 28, 3753 – 3756. Lu, Z., and W. R. Danskin (2001), InSAR analysis of natural recharge to define structure of a ground-water basin, San Bernardino, California, Geophys. Res. Lett., 28, 2661 – 2664. McGuire, J. J., and P. Segall (2003), Imaging of aseismic fault slip transients recorded by dense geodetic networks, Geophys. J. Int., 155, 778 – 788. Nikolaidis, R. (2002), Observation of geodetic and seismic deformation with the Global Positioning System, Ph.D. thesis, Univ. of Calif., San Diego. Oskin, M., K. Sieh, T. Rockwell, G. Miller, P. Guptill, M. Curtis, S. McArdle, and P. Elliot (2000), Active parasitic folds on the Elysian Park anticline: Implications for seismic hazard in central Los Angeles, California, Geo. Soc. Am. Bull., 112, 693 – 707. Owen, S., G. Anderson, D. C. Agnew, H. Johnson, K. Hurst, R. Reilinger, Z. K. Shen, J. Svarc, and T. Baker (2002), Early postseismic deformation from the 16 October 1999 M 7.1 Hector Mine, California, earthquake as measured by survey-mode GPS, Bull. Seismol. Soc. Am., 92, 1423 – 1432. Ozawa, S., M. Murakami, and T. Tada (2001), Time-dependent inversion study of the slow thrust event in the Nankai trough subduction zone, southwest Japan, J. Geophys. Res., 106, 787 – 803. Pratt, T. L., J. H. Shaw, J. F. Dolan, S. A. Christofferson, R. A. Williams, J. K. Odum, and A. Plesch (2002), Shallow seismic imaging of folds above the Puente Hills blind-thrust fault, Los Angeles, California, Geophys. Res. Lett., 29(9), 1304, doi:10.1029/2001GL014313. Shaw, J. H., A. Plesch, J. F. Dolan, T. L. Pratt, and P. Fiore (2002), Puente Hills blind thrust system, Los Angeles, California, Bull. Seismol. Soc. Am., 92, 2946 – 2960. Terzaghi, K. (1925), Erdbaumechanik auf bodenphysikalischer Grundlage, Deuticke, Vienna. Theis, C. V. (1935), The relation between the lowering of the piezometric surface and rate of duration of discharge of a well using ground-water storage, Eos Trans. AGU, 2, 519 – 524. Watson, K. M., Y. Bock, and D. T. Sandwell (2002), Satellite interferometric observations of displacements associated with seasonal groundwater in the Los Angeles basin, J. Geophys. Res., 107(B4), 2074, doi:10.1029/2001JB000470. Wdowinski, S., Y. Bock, J. Zhang, P. Fang, and J. Genrich (1997), Southern California Permanent GPS Geodetic Array: Spatial filtering of daily positions for estimating coseismic and postseismic displacements induced by the 1992 Landers earthquake, J. Geophys. Res., 102, 18,057 – 18,070. Webb, F., et al. (2004), GPS data products for solid earth science, paper presented at 2004 SCEC Annual Meeting, South. Calif. Earthquake Center, Palm Springs, Calif. Werner, C., U. Wegmu¨ller, T. Strozzi, and A. Wiesmann (2000), GAMMA SAR and interferometric processing software, paper presented at ERSENVISAT Symposium, Gothenburg, Sweden, Oct. Williams, S. D. P. (2003), Offsets in Global Positioning System time series, J. Geophys. Res., 108(B6), 2310, doi:10.1029/2002JB002156. Williams, S. D. P., Y. Bock, P. Fang, P. Jamason, R. M. Nikolaidis, L. Prawirodirdjo, M. Miller, and D. J. Johnson (2004), Error analysis of continuous GPS position time series, J. Geophys. Res., 109, B03412, doi:10.1029/2003JB002741. Yeats, R. S. (2004), Tectonics of the San Gabriel Basin and surroundings, southern California, Geol. Soc. Am. Bull., 116, 1158 – 1182. Zumberge, J. F., M. B. Heflin, D. C. Jefferson, M. M. Watkins, and F. H. Webb (1997), Precise point positioning for the efficient and robust analysis of GPS data from large networks, J. Geophys. Res., 102, 5005 – 5017. 

D. C. Agnew and Y. Bock, Institute of Geophysics and Planetary Physics, Dept. 0225, University of California, San Diego, La Jolla, CA 92093-0225, USA.

10 of 11

B03409

KING ET AL.: SAN GABRIEL VALLEY GEODETIC ANOMALY

D. Argus, Z. Liu, and F. H. Webb, Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Dr., Pasadena, CA 91109, USA. D. Barseghian and K. Stark, Stark Consulting, LLC, 525 S. Wilson Ave., Pasadena, CA 91106, USA. G. Bawden and D. Galloway, U.S. Geological Survey, 3020 State University Drive East, Modoc Hall Suite 4004, Sacramento, CA 95819, USA.

B03409

R. S. Dollar, N. E. King, and A. Yong, U.S. Geological Survey, 525 S. Wilson Ave., Pasadena, CA 91106, USA. ([email protected]) J. Langbein, U.S. Geological Survey, 345 Middlefield Road, MS977, Menlo Park, CA 94025, USA. E. Reichard, U.S. Geological Survey, California Water Science Center, 4165 Spruance Road, Suite 200, San Diego, CA 92101, USA.

11 of 11