The following chapter is an excerpt from:

Available at: WWW.APPLIEDGEOLOGYBOOK.COM and WWW.AMAZON.COM ISBN: 978-0-89863-399-3 OVER 1000 PAGES ▪ HARDCOVER 49 CHAPTERS ▪ 144 CONTRIBUTORS

APPLIED GEOLOGY IN

CALIFORNIA Robert Anderson Horacio Ferriz Editors

SPECIAL PUBLICATION 26

STAR PUBLISHING COMPANY, INC. PO BOX 5165 BELMONT, CA 94002-5165 Publisher and Managing Editor: Stuart Hoffman Copyeditor: Jennifer Rice

Library of Congress Cataloging-in-Publication Data Names: Anderson, Robert, 1955- editor. | Ferriz, Horacio, editor. | Association of Environmental & Engineering Geologists. Title: Applied Geology in California / Robert Anderson, Horacio Ferriz, editors. Description: Belmont, CA: Star Publishing Company, Inc.; [Zanesville, Ohio]: Published in coordination with the Association of Environmental & Engineering Geologists, [2016] | Series: Special publication; 26 | Includes bibliographical references. ,GHQWL¿HUV/&&1_,6%1 Subjects: LCSH: Geology--California. | Geophysics--California. | Hydrology--California. | Natural disasters--Risk assessment--California. | Earthquake hazard analysis--California. &ODVVL¿FDWLRQ/&&4($_''&GF /&UHFRUGDYDLODEOHDWKWWSVOFFQORFJRY Copyright © 2016 by Star Publishing Company, Inc. All rights reserved. No part of this book may be reproduced, stored in an information storage and retrieval system, or transmitted in any form or by any means, electronic, digital, mechanical, scanning, photocopying, recording, or otherwise, without the prior written permission of the publisher. 7KLVSXEOLFDWLRQLVSURWHFWHGE\WKH)HGHUDO&RS\ULJKW/DZVRIWKH8QLWHG6WDWHVRI$PHULFD7LWOH86& DQGE\,QWHUQDtional Copyright Laws. Reproduction or copying of his publication, in whole or in part, by any means whatsoever, without the H[SUHVVZULWWHQSHUPLVVLRQRI6WDU3XEOLVKLQJ&RPSDQ\,QFLVSURKLELWHGDQGPD\VXEMHFWWKHYLRODWRUV WRFLYLOSHQDOWLHVDQG or criminal penalties. Printed in the United States of America ISBN 978-0-89863-399-3

0987654321

16 17 18 19 20 21 22

CHAPTER 8 San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct Ray Weldon1, Greg de Lamare6, Doug Yule2, William C. Hammond3, Ashley Streig4, Alex Sarmiento5, S. Thomas Freeman5*John Shamma6, and Mohsen Beikae6 , Albert Rodriguez6 1University of Oregon, 2California State University-Northridge, 3University of NevadaReno, 4Oregon State University, 5GeoPentech, Inc. 6Metropolitan Water District of Southern California *Contact: GeoPentech, Inc. [email protected]

ABSTRACT

geodetic and seismological data and interpretations, both published and unpublished, were reviewed and analyzed in completing this modeling exercise. The compilation of this information was used to assess the complex tectonics of the “knot” in the SSAFS through the SGP area and construct a 3D subsurface representation of geometry of the most recently active trace, which is the most in uential parameter driving potential surface fault displacement and the surrounding ground deformation. This 3D subsurface con guration of the fault system was tested and adjusted through several model simulations in comparison with geologic, geomorphic, seismologic and geodetic data to re ne the Coulomb and geodetic block models so that the predicted results agreed with observations. These results were used as input into modeling what is considered to be the most consistent estimate of likely surface fault displacements and ground surface deformation along the CRA from the predicted slip on the SSAFS though the SGP area during a future maximum considered earthquake (up to MW 7.8). Paleoseismic data indicate that surface fault displacements and surface deformation, which would have impacted the CRA in the San Gorgonio Pass area occurred on the SSAFS – South Branch (-SB)

The Metropolitan Water District provides water to approximately 18 million people over 5200 square miles within Los Angeles, Orange, San Diego, Riverside, San Bernardino, and Ventura Counties. Metropolitan has developed an overall approach to system reliability which is encompassed through ve areas: Water Supply Reliability, System Capacity, Infrastructure Reliability, System Flexibility, and Emergency Response. As part of the Infrastructure Reliability function, Metropolitan’s most recent vulnerability assessment looked at the potential impacts to the Colorado River Aqueduct (CRA), from an approximate Maximum Considered Earthquake (up to MW 7.8) on the Southern San Andreas Fault System (SSAFS). One of southern California’s main sources of imported water, the CRA crosses the SSAFS in the area of San Gorgonio Pass (SGP). A key element of Metropolitan’s vulnerability assessment was modeling the potential for surface fault displacements and regional uplift along the CRA in the SGP area using Coulomb 3.3 and a geodetic block motion modeling software. All recent and readily available geological, geomorphical, paleoseismic,

147

148 Applied Geology in California

in ~800 AD and ~1400 AD. The results of the modeling demonstrates that during future large earthquakes (i.e. up to a maximum considered Mw 7.8) surface fault displacements up to ~1 m (~3 ft) vertical and ~3-3/4 m (~12 ft) horizontal could occur across the CRA. The coseismic uplift would gradually diminish over a distance of ~ 48 km (~30 miles) upstream along the CRA. This output from the Coulomb 3.3 and block model are internally consistent, and both models’ output are also consistent with the current geological, geomorphical, paleoseismic, geodetic and seismological evidence. The results of this analysis and modeling effort have provided valuable input into Metropolitan’s assessment of the potential risks to the CRA and their pre-event and emergency response planning.

Introduction Concerns have been raised regarding the seismic vulnerability of the aqueducts conveying water

into the western portion of southern California (Davis and O’Rourke, 2011) in part resulting from the simulation of a future earthquake along the Southern San Andreas Fault System (SSAFS) in “TheShakeOut Scenario” (Jones et al., 2008 and Perry et al., 2008). To address these concerns Metropolitan initiated studies to model future surface fault displacements and ground deformation focusing on where the Colorado River Aqueduct (CRA) crosses the SSAFS in San Gorgonio Pass (SGP). Figure 1 illustrates the relationship between the CRA, the SSAFS, and the crustal strain patterns in the SGP area. Figure 2 highlights the mapped traces of the SSAFS-South Branch (-SB) and North Branch (-NB) through SGP and the key CRA/ SSAFS crossing points. There are signi cant uncertainties in the publicly available data used to estimate potential future fault surface displacements and ground surface deformation. Comparison of surface fault

Figure 1. Southern San Andreas Fault System and the crossing Colorado River Aqueduct in San Gorgonio Pass area and regional faults and crustal strain patterns. Black lines are all Quaternary-active faults from the USGS Fault and Fold Database (USGS, 2010). Colored fault lines and colored hachured patterns between colored fault lines indicating stepover zone in fault system are for illustrative purposes only and do not indicate relative age of Quaternary Fault activity.

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

149

Figure 2. CRA/Quaternary fault crossings in the SGP area. Faults are from USGS Quaternary Fault and Fold Database (USGS, 2010). Shaded relief is from USGS National Elevation Dataset, resolution ~30 meters (USGS, 2013).

rupture lengths and coseismic fault offsets in the commonly used global datasets, such as Wells and Coppersmith (1994) and Wesnousky (2008), with data from California surface-rupturing earthquakes (Biasi et al., 2013 [UCERF3 Appendix F]) suggests that these worldwide regressions may not be as reliable when applied to local areas, particularly in southern California, and that empirical scaling relationships of Shaw (2009, 2013) are more representative. In addition, the ratios of likely amounts of horizontal-to-vertical fault slip are also based on global relationships and not on local fault geometry or loading. The horizontal-to-vertical slip ratios may be better estimated from the interseismic block motion direction that drives the coseismic slip across the 3D geometry of the fault zone. Small differences in the direction of motion between the fault-bounded blocks (Fattaruso, Cooke and Dorsey, 2014) can produce signi cant differences in the amount of horizontal and vertical surface fault displacement and deformation that will occur at or adjacent to a CRA/ SSAFS crossings. These differences in the amount

of horizontal and vertical slip along the trend of the fault will have a direct bearing on the hydraulic capacity of the CRA at or adjacent to a given fault crossing location in the event of slip along the fault. A key uncertainty addressed in this study is the 3D geometric complexity of the SSAFS, particularly near SGP, which controls the interseismic surface deformation near the fault and will control the coseismic horizontal and vertical slip displacements. Coseismic slip on the geometrically complex SSAFS produces oblique displacements at the CRA/SSAFS fault crossings and surface warping that could extend tens of kilometers between the faults and beyond. Because published length-displacement or magnitude-displacement regressions are not based on speci c subsurface 3D geometry beneath SGP the amount of surface displacements or distribution of surface warping cannot be accurately estimated. To more accurately estimate surface deformation along and bordering the fault, a better characterization of the fault geometry at depth and the slip across that geometry are needed.

150 Applied Geology in California

Having an accurate estimate of the amount and distribution of coseismic vertical and horizontal surface deformation is crucial in developing plans to minimize the adverse effects of a future earthquake to the CRA. Expected coseismic displacement will be used in planning measures to evaluate the impact of the ground deformation on the hydraulic capacity of the CRA and to identify options to restore full service as quickly as possible following a rupture of the SSAFS through the SGP. The objective of this study was to model the local aspects of the SGP area and to estimate future ground surface displacements along the CRA at the SSAFS-SB fault crossings. Integrating the local SGP geologic, geomorphic, seismologic and geodetic data in this modeling improves the accuracy of, and provides more con dence in, the estimates of future ground deformation at the CRA-SSAFS crossing. We focused our modeling on the maximum considered earthquake (up to MW 7.8) which would be produced by ~4 m (~13 ft) of fault slip beneath the SGP region. To envelop a range, we also ran models of the surface deformation that would occur along the CRA during more frequent events, such as the 1948 MW 6.2 Desert Hot Springs and the 1986 MW 6.1 Palm Springs earthquakes and a worst-case MW 8.5 event generated by ~8 m (~26 ft) slip on the SSAFS at depth.

Methodology Modeling the future fault surface ruptures and ground deformation that will occur along the CRA where it crosses the SSAFS in SGP required an integration of data from several different geoscience disciplines and two different types of computer software. Three sets of Coulomb 3.3 (USGS, 2011) models and two sets of geodetic block motion models (McCaffrey, 2002; Meade and Hagar, 2005; Hammond and Thatcher, 2007; and Hammond et al., 2011) were developed to estimate the range of fault surface displacements and surrounding ground surface deformation that would envelope the plausible earthquakes on the SSAFS through SGP. The following scenarios of right-lateral strike-slip, imposed on the interface between the blocks located on both sides of and separated by the fault and aligned vertically and horizontally

according to the estimated 3D geometry of the fault beneath SGP were modeled: Scenario 1: ~1 m (~3 ft) of slip above a depth of 12 km while tapering to no slip at the surface and along strike, during a MW ~6.0 to ~6.5 local event with an estimated return period of several decades, similar to a North Palm Springs type event with a rupture patch 8km long and slip patches in 2km down dip increments to a depth of 12 km; Scenario 2: ~4 m (~13 ft) of slip above a depth of 12 km during a MW ~7.0 to ~7.8 event extending from Bombay Beach to Wrightwood or 3-Points (Figure 6) with an estimated return period of 500 to 1,000 yrs; Scenario 3: ~8 m (~26 ft) of slip above a depth of 25 km during a ‘worst-case’ MW ~8.5 event extending from Bombay Beach to Park eld with an estimated return period of >5000 yrs. Modeling these three scenarios involved four steps: Step-1) Analysis and integration of relevant data and interpretations; Step-2) Construction of a 3D graphical representation of the location and subsurface geometry of the SSAFS-SB beneath SGP; Step-3) Trial surface deformation model runs in Coulomb 3.3 and geodetic block motion modeling runs to test the 3D graphical representation; and Step-4) Final Coulomb 3.3 and geodetic block motion models of possible future fault slip and surrounding surface deformation along the CRA in the SGP area and nearby SSAFS. The initial interpretation used relevant published and unpublished data from: a) topography, geology, geomorphology, paleoseismology, hydrogeology, and geophysical studies; b) historical and instrumentally-recorded earthquake records; and c) Global Positioning System (GPS) and Interferometric Synthetic Aperture Radar (InSAR) geodetic surveys. This information was from both state-wide and regional areas as well as from the local area surrounding the CRA at SGP. Together, all these sources of data and information were used in Step- 2 to develop plausible subsurface geometries and level of activity of the SSAFS-SB for use in the computer modeling software. The Coulomb 3.3 computer software (USGS, 2011) was used to model the fault surface displacements and surrounding surface deformation eld

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

along the CRA at SGP resulting from the three plausible subsurface fault ruptures that bracket the earthquake scenarios listed above. Geodetic block motion modeling software (see references cited in Hammond and Thatcher, 2007 and Hammond et al., 2011) was also used to constrain the parameters applied in Coulomb 3.3. Trial model runs were initiated in Step-3 to test the argument that the modeled surface deformation results produced using Coulomb 3.3 are physically plausible when compared with the geodetic data. The Coulomb 3.3 input geometries and properties were compared to the results of the block model by estimating displacements consistent with the geodetic velocities. These two computer programs contain the same underlying physics (i.e., dislocation modeling in an elastic half-space), but calculate deformation differently. Accordingly, similarity between the two results from the applied programs can provide con dence in the results or identify possible problems. The nal surface deformation model runs were completed in Step-4 for the representation of the small, maximum considered, and worst-case ground rupturing earthquake scenarios. Critical reviews of the results of deformation models involved comparison with the results of other research in the SGP area, such as Fialko (2006), Fuis et al., (2012), Yule and Sieh (2003), and more recent published (e.g., Scharer et al., 2014; Yule et al., 2011, 2014; Wolff et al., 2013) and unpublished work (Yule and Scharer, in preparation).

Data Analysis and Integration Tectonic Setting It is understood that the location, con guration, characteristics and rate of crustal strain across the crustal plate boundaries in western North America have been changing at least over the last ~24 million years (Atwater, 1970, Nicholson et al., 1994). This evolution in the crustal plate boundaries has introduced signi cant complexities in the tectonics of western North America and particularly in southern California. These complexities challenge our efforts to characterize the currently active faults at depth within the crust, constrain the distribution

151

the crustal strain between these faults, and model possible future fault ruptures and their generated earthquakes along these faults. Presently, approximately 50 mm/year of relative motion occurs between the Paci c and North American crustal plates in California (DeMets and Dixon, 1999, and many others). Southeast of the Mojave Desert and Cajon Pass this crustal strain is distributed across a broad region with about 20 to 35 mm/year accommodated by the SSAFS and the San Jacinto Fault Zone (SJFZ). The remaining strain is distributed to the east of SGP in the Eastern California Shear Zone (ECSZ) and west of the SJFZ onto the Elsinore Fault and other southern California faults. Generally, the proportion of crustal strain accommodated by each individual fault progressively decreases farther east and west of the SSAFS and SJFZ. South of the “Big Bend” and west of the Tehachapi Mountains, the SSAFS essentially consists of two southwest concave curvatures with Euler rotation poles to the southwest. These arcs are misaligned through the SGP region between San Bernardino Valley to the northwest and Palm Springs on the southeast (Weldon and Humphreys, 1986; Humphreys and Weldon, 1994) producing a transpressive step in the otherwise smoothly curving SSAFS. As illustrated in Figure 1, this ‘step’ produces the most complex pattern of faults, seismicity and ground surface deformation along the entire length of the San Andreas Fault System in California. Cooke (2008), Dair and Cooke (2009) label the complex geology and topography resulting from this ‘step’ the “San Gorgonio Knot”. In this more east-westerly trending portion of the SSAFS the northwest-directed lateral slip between the Paci c and the North American plates produces the complex distribution of faults within what we call here the San Gorgonio Pass (SGP) region. This left step in the SSAFS, as highlighted in Figure 1 and as seen in Figure 2, introduces geometric complexities related to the SSAFS-SB and the SSAFS-NB, resulting in abundant compressional folding, shallower dipping faults and greater vertical components of displacement, and a wide range of levels of activity on different faults.

152 Applied Geology in California

As illustrated in Figure 1, southwest of SGP a zone of transtension occurs through a right-step between the SSAFS and the San Jacinto Fault Zone (SJFZ). A zone of transtension also occurs northeast of SGP where the SSAFS right-steps into the Eastern California Shear Zone (ECSZ), which ruptured during the 1992, MW 7.3 Landers Earthquake and the 1999, MW 7.1 Hector Mine Earthquake. Northwest and southeast of SGP the SSAFS strike becomes progressively more northwesterly, its dip steepens to sub-vertical and its displacement becomes more horizontal (strike-slip). San Gorgonio Pass Geologic Setting At the regional scale, the California Geological Survey’s (CGS) state geological map (Gutierrez et al., 2010) shows that SGP is bordered on the north by Precambrian metamorphic Mesozoic igneous and pre-Cenozoic metasedimentary and metavolcanic rocks that underlie the San Bernardino Mountains where the maximum elevation is at San Gorgonio Peak (3,506 m [11,503 ft]). These rocks also underlie the Little San Bernardino Mountains (maximum elevation 1,772 m [5,814 ft]), which border the eastern portion of the SGP. SGP is bordered on the south by Mesozoic igneous and pre-Cenozoic metasedimentary rocks that underlie the San Jacinto Mountains where the maximum elevation is 3,302 m (10,833 ft). Pliocene and/or Pleistocene sediments and underlying (locally exposed) Mesozoic igneous and pre-Cenozoic metasedimentary rocks form the Badland foothills (maximum elevation 915 m [3,000 ft]) west of the pass. Pliocene and/ or Pleistocene sediments and older alluvial terrace deposits ank the bedrock along the north side of the pass, while late Quaternary alluvial and stream bed deposits oor the pass (elevations of ~150 m to ~825 m [~500 ft to ~2,700 ft]). Fault traces displacing these Late Quaternary deposits offer the strongest evidence of their potential to generate future surface fault displacements and ground surface deformations1. Faults The location where the CRA crosses the mapped surface traces of the Quaternary-active faults within the SGP and the surrounding area are highlighted on Figure 2 by numbered yellow circles. 1

displacing Late Quaternary deposits are mapped along the northern edge of SGP and the southern front of the San Bernardino Mountains. As illustrated on the state geological map (Gutierrez et al., 2010), the northeastern portion of the SGP area is controlled by the buried southwestern bedrock edge of the Little San Bernardino Mountains. The fault color codes in Figure 2 show the time since most recent slip of the faults within the SGP area, as shown in the Quaternary Fault and Fold Database (USGS, 2010, available at http://earthquakes.usgs. gov/regional/qfaults/), which is also generally in agreement with the current CGS database available at http://www.consrv.ca.gov/cgs/information/ publications/Pages/QuaternaryFaults_ver2.aspx. Although not obvious in the way the surface bedrock geology is portrayed by Gutierrez et al., (2010), the bedrock underlying the San Jacinto Mountains, which are part of southern California’s Peninsular Ranges, is not the same as the bedrock underlying the San Bernardino and Little San Bernardino Mountains. For example, the metasedimentary and metavolcanic rocks to the north of SGP are not the same as metasedimentary and metavolcanic rocks found in the San Jacinto Mountains. The rocks mapped as ‘m’ by Gutierrez et al., (2010) to the north of SGP, ~13 km (~8 miles) north of Banning/ Cabazon and ~13 km (~8 miles) southwest of the Mission Creek Fault) are equivalent with the Pelona-Orocopia Schist and are very different from the metamorphic rock of the Peninsular Ranges under the San Jacinto Mountains. The bedrock between the San Bernardino Mountains to the north of SGP and the San Jacinto Mountains to the south of SGP can be separated into three major blocks bounded by the major faults in the SSAFS. Figure 3 illustrates the location of the CRA relative to these three bedrock blocks and their bounding faults and the approximate ages of the bounding faults’ activity. The southsouthwest side of the northernmost block, referred to here as the San Bernardino bedrock block, is separated from the middle block by the old SAF/ Wilson Creek/Mission Creek Fault. The middle block is referred to as the San Gabriel Mountain bedrock block, because the rocks within this block are an offset sliver of San Gabriel Mountains-type

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

153

Figure 3. Older traces of SSAFS through the SGP area interpreted from geology.

basement rocks. The southern boundary between the San Gabriel Mountains bedrock block and the San Jacinto bedrock block to the south is the ‘old Banning Fault’. The ‘old Banning Fault’ is probably a southeastern extension of the 5-10 m.y. old San Gabriel fault, which is located in the San Gabriel Mountains north of Los Angeles, with tens of km of displacement (Powell et al., 1993; Matti et al., 1993). The SSAFS-NB (Mission Creek fault, mainly) is the locus of previous large lateral displacements; the old Banning fault along the SSAFS-SB also carried signi cant lateral displacements. These ancestral strands of the SSAFS were active at different times; the old Banning Fault between ~10-5 m.y. ago, and NB Mission Creek Fault between about 5 and 0.5 m.y. ago (Powell, et al., 1993; Matti and Morton, 1992). Fuis et al., (2012) pointed out an aeromagnetic anomaly following the mapped trend of the bedrock trace of the SSAFS-NB Mission Creek Fault, highlighted in Figure 4. They suggest

that this anomaly is due to the higher abundance of magnetic rocks within the Precambrian igneous and metamorphic rock mass to the north-northeast, in contrast to the nonmagnetic rocks to the southsouthwest in the San Gabriel bedrock block. The more recent, late Quaternary activity, on the SSAFS was shifted by the crustal deformation southward into SGP. The most recent trace of the SSAFS in SGP is very young, only 60,000 to 150,000 yrs old, and has locally reactivated parts of the ‘Old Banning Fault’. This geologic understanding of the SSAFS guided the construction of our 3D geometry for input in the models as discussed in more detail below in the Fault Geometry section. Original CRA Fault Investigations The late Quaternary traces of many strands of the SSAFS through the SGP area were initially mapped by Metropolitan geologists supporting the original design and construction of the CRA in the late 1920s, 1930s and early 1940s

154 Applied Geology in California

Figure 4. SGP area aeromagnetic anomaly map reproduced from Fuis, et al., (2012). Color scale at the bottom represents aeromagnetic anomaly nanoTesla (nT) units. Annotations on map (e.g., black lines, numbers, abbreviations, etc.) pertain to the original publication.

(Hinds, 1938; Bond, 1939). Metropolitan engineers integrated measures in their engineering designs to minimize the impacts on the ow through the CRA due to future vertical displacements across the key fault traces mapped at that time. The measures included an additional 0.8 m (2.5 ft) of drop beyond that required by siphon losses at the following three fault crossings: 1. The SAFS-NB (i.e., Mission Creek Fault, CRA crossing location #4 in Figure 2); 2. The SAFS-SB (i.e., Banning Fault, CRA crossing location #5 in Figure 2); and 3. The SJFZ (CRA crossing location #8 in Figure 2) Metropolitan engineers also designed the CRA to (1) cross these active faults at the ground surface in inverted siphons and (2) at right angles to the fault traces in order to minimize the length of CRA that would be adversely affected by future horizontal fault slip, and (3) to facilitate access

for repairs. The designers also opted for the more exible siphon design within the SGP area, rather than a more rigid monolithic concrete construction (Bond, 1939). The application of these design measures re ect Metropolitan geologists’ and engineers’ previous knowledge at that time of the ground shaking and deformation that occurred along the San Andreas Fault System during the 1857 Fort Tejon and the 1906 San Francisco earthquakes and their understanding of the active faults in the SGP area gained through their geologic mapping and their analyses of stereo aerial photographs for fault hazards, a technique which was in its infancy at that time. With the implementation of these fault crossing measures during the construction of the CRA, it is our understanding that although some damage will occur in the aqueduct, hydraulically, the CRA can accommodate some vertical coseismic fault displacements at the fault crossings locations #4, #5 and #8 shown in Figure 2, and broad wavelength vertical ground deformation along the aqueduct

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

upstream and downstream of these faults. It is also our understanding that the CRA downstream from the SJFZ (CRA fault crossing location #8 in Figure 2) is able to maintain the minimal hydraulic gradient required to maintain serviceability in the event that vertical fault offsets occur at the fault crossings farther downstream. During the time of the CRA design and construction, paleoseismic data along the SSAFS of the 1812 Wrightwood earthquake and other older events on the SSAFS were not known. Since that time, better techniques for recognizing the more active strands of faults have been discovered and new data have been obtained. Post CRA Design and Construction Fault Investigations Multiple studies of the faults within SGP, such as those performed by Allen (1957) and Dibblee (1991), have been completed since the CRA was rst put into service in the 1940s. Research by Yule and Sieh, (2003), (reproduced hear in Figures 5a through 5d) provides insight into which fault traces in the SGP area are likely to rupture the surface during future earthquakes on the SSAFS. Based on deformed adjacent young alluvial surfaces, and vertically and laterally offset landforms and other geomorphic evidence, their work suggests that the SSAFS-SB San Gorgonio Pass Thrust Fault (SGPTF) and Garnet Hill Fault (GHF) are the most recently active traces of the SSAFS through SGP (CRA fault crossing locations #6 in Figure 2)2. Yule and Sieh (2003) also note that the SSAFS-SB Banning Fault has clear Holocene scarps and laterally offset geomorphic surfaces and vegetation lineaments toward the Indio Hills southeast of CRA fault crossing location #5 (Figure 2) and locally north and west of the CRA. The SSAFS-SB Banning Fault may also break as a subsidiary fault (rather than a primary fault) during future displacements of the GHF. Yule and Sieh (2003) indicate that although the SSAFS-NB Mission Creek Fault (CRA fault Note that CRA fault crossing #6 also involves subsidiary faults in the north-northeast, upper thrown (hanging wall) block of the GHF west of Whitewater Canyon. 2

155

crossing location #4 in Figure 2) clearly exhibits Holocene scarps, laterally offset landforms, deformed adjacent young alluvial surfaces, and vegetation lineaments southeast of the CRA in the Indio Hills area, its activity diminishes toward Hwy 62. The SSAFS-NB Mission Creek Fault bifurcates into the Mill Creek and Mission Creek faults northwest of Hwy 62. Evidence for Holocene activity on the SSAFS-NB Mission Creek Fault in the area of the CRA crossing #4 and farther to the southeast is equivocal due to lack of paleoseismic data between Thousand Palms Oasis on the northeastern edge of the Indio Hills and SGP. Farther to the northwest from CRA crossing #4, evidence for Holocene activity on the SSAFS-NB Mission Creek and Mill Creek faults is also equivocal due to the lack of young surfaces and deposits. Currently unpublished work by Katherine Kendrick and Johnathan Matti of the USGS, recently presented during the Southern California Earthquake Center (SCEC) 2014 Annual Meeting SGP Workshop, indicates that the east-west trending Pinto Mountain Fault offsets the Mission Creek and Mill Creek faults, suggesting that the SGP area may be the terminus point of the SSAFS-NB Mission Creek and Mill Creek faults. However, by the Santa Ana River to the northwest of SGP, the SSAFS-NB is clearly active. Yule and Sieh (2003) show the distribution of key active and inactive faults, exures, and crustal blocks relative to the physiography of the SGP area. Much of the topography is fault controlled. For example, active right-lateral faults of the ECSZ (middle right portion of Figure 5b) are aligned along north/south-oriented valleys. The less active (and perhaps locally inactive) Mill Creek and Mission Creek faults (of the SSAFS-NB) also generally follow east/west-oriented linear valleys. Active sections of the SSAFS to the southeast (the Coachella and Banning faults) bound the Indio Hills, and the active San Bernardino Strand to the northwest also sharply bounds the range front. Active or low angle (dipping) oblique right-lateral reverse faults bound the arcuate range front along the northern edge of the SGP (center of Figure 5b). The absence of active faulting along the southern edge of the SGP (north of San Jacinto Peak), is consistent with

156 Applied Geology in California a.

b.

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

157

c.

d.

(continues)

158 Applied Geology in California d. (continued)

Figure 5. Geologic and geomorphic maps and interpretations of the SGP area reproduced and annotated IURP)LJXUHFRISeeber and Armbruster, 1995] a south-vergent thrust fault, San Gorgonio Pass Fault Zone (SGPFZ), offsets the vertical, strike-slip plane of WKH6DQ%HUQDUGLQRVWUDQGRIWKH6DQ$QGUHDV)DXOW6%6$) F ,QWKHIDXOWEHQGIROGPRGHO>PRGL¿HGIURPAllen, 1957; Dibblee, 1968, 1975, 1982; Matti et al., 1985, 1992a; Matti and Morton, 1993], which we advocate, vertical to moderately dipping strike-slip planes cut the hanging wall block of the SGP thrust fault on the west and approach WKHWKUXVWIDXOWRQWKHHDVW2QO\WKHODWWHUPRGHO¿WVWKHGDWDZHKDYHSUHVHQWHGLQWKLVSDSHU'DVKHGOLQHVLQGLFDWH features in the background. Abbreviations of other strands of the San Andreas Fault System include the Coachella strand (CVSAF) and the, Coachella Valley Banning Fault strand (CVBF). 5d - Structural contour map of the primary fault systems of the San Gorgonio Pass region. These structural contours on the active elements of the San Andreas and San Gorgonio Pass fault zones are constrained by geologic mapping, regional seismicity, and the aftershocks of the 1986 ML 5.9 North Palm Springs and the Mw 6.0 1948 Desert Hot Springs earthquakes (dashed outlines). In this interpretation, the San Bernardino strand of the fault is a tear fault in the hanging wall block of the San Gorgonio Pass fault system. The moderately dipping Garnet Hill Fault and more steeply dipping Coachella Valley Banning fault strands merge at depth and merge with the San Gorgonio Pass fault system from the east.

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

the mature, undisturbed dendritic drainage system evident in the topography. Compressive exures have produced localized blocks that are uplifted or tilted, accommodating strain distributed across the SSAFS en echelon stepover (“knot”) in SGP as Yule and Sieh (2003) illustrated (see Figure 5b). Yule and Sieh (2003) propose a relatively shallow northerly dip on the SGPTF and GH faults as illustrated by their fault structural contour map and by their letter “C” block diagram (Figure 5c). The in uence of the compressional component of strain accumulation across these shallower dipping faults should produce prominent vertical components of displacement on the northern side of these faults. Currently, there is little data available to verify the fault geometry and level of activity of the faults crossing the CRA at points #0 through #3 in Figure 2. The northern portions of the faults at CRA crossing points #1 through #3 ruptured during the 1992 Mw 7.3 Landers Earthquake. The USGS (2010) interpreted their age of last rupture at their CRA crossing points as sometime since, or <15,000 years ago. The USGS (2010) also only interpreted the age of the last rupture on the fault crossing the CRA at point #0 (Figure 2) as sometime since 130,000 year ago. It should be recognized that the USGS database used to construct their fault activity maps (the source of the fault activity legend in Figure 2) is a continuously evolving archive. It is doubtful that the age statuses of the faults at CRA crossing points #0 through #3 have been updated since the Landers Earthquake. Also, USGS map compilers tend to be very conservative (e.g., if there are no clearly deformed dated geologic units less than 130,000 years they classify it as “<130,000 years”, likewise with <15,000 years classi ed faults). This area has only bedrock and undated yet very young alluvium (generally late Holocene) so recent activity on these faults is poorly constrained. Future events are expected to resemble the slip that occurred during the Landers Earthquake, with more horizontal strikeslip surface displacement than vertical throw on these fault strands. Hydrogeological data, which often help de ne the limits of fault traces in bedrock and alluvial sediments, as well as the level of activity in alluvial sediments, was limited to the Coachella

159

groundwater basin in the southeastern portion the SGP study area and the San Bernardino and San Jacinto groundwater basins located west and northwest of SGP (Wisely, 2012; Wisely and Schmidt, 2010). The hydrogeologic information from these groundwater basins was utilized to help interpret the geodetic observations as described below. Paleoseismicity Paleoseismologists have developed a rich history of earthquakes on the SSAFS that extend back >1000 yrs (Figure 6a). The record is well established for the Mojave-San Bernardino and Coachella Valley regions (e.g., Philibosian et al., 2011, Weldon et al., 2004 and Weldon et al., 2013) but has only recently become available from San Gorgonio Pass (Figure 6b) (e.g. Scharer et al., 2014; Wolff et al., 2013; Yule et al., 2011; 2014). The site that is the closest to the CRA is at Cabazon, where the most recent and penultimate earthquakes occurred 590-630 and 1097-1120 yrs BP, respectively (Figure 6c) (K. Scharer, 2014, personal communication). Along the strike of the SSAFS, recurrence intervals of surface rupture vary from northwest to southeast along the fault system from ~100 yrs in the Mojave reach of the fault, to ~200 yrs in the San Bernardino area, to ~500 to ~800 yrs at San Gorgonio Pass based on the ~500 yr. interval between two events and the 600+ yr. open interval since the last rupture event seen in the paleoseismic trench across the SGPTF west of the CRA (Yule et al., 2014 and Yule and Sharer, 2014, in preparation), and then decreases again to ~200 yrs along the interval of the SSAFS in the Coachella Valley to the SE (Figure 6a). Small vertical separations per event at the Cabazon trench site contrast with larger vertical separations of alluvial fan surfaces at Millard Canyon (Figure 6b) (Yule and Scharer, 2014, in preparation). One way to account for this difference is to have had slip occur on a currently concealed fault (‘blind’ fault) that extends south of the mountain front connecting the south SGPTF in Millard Canyon with the GH fault, as has been represented by a dotted red line in Figure 6b (Yule and Scharer, 2014, in preparation). This maintains a 2 km-wide zone of deformation from Millard

160 Applied Geology in California a.

b.

c.

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

161

c.

d.

Figure 6a. Paleoseismic studies in and near SGP and to the northwest and southeast along the SSAFS. Paleoearthquakes and historic horizontal and vertical offsets shown here and single-event offsets from the Appendix R “Offset database” electronic supplement of Madden et al., (in review) and references herein. Slip rates are from McGill (pers. comm. to Weldon). Recurrence intervals (RI) .6b – Local faults and geomorphic surfaces in the SGP area – San Gorgonio Canyon to Whitewater Canyon, reproduced from Yule and Scharer (in preparation). Paleoseismic and slip rate sites (green diamonds). Rupture in last 1200 years: Burro Flats, 7 events; Millard Canyon north, 2 events; Millard Canyon south, >1 event; Cabazon, 2 events. Slip rates: Burro Flats, 5-10 mm/yr.(Orozco, 2004); Millard Canyon north, 2.5-6.2 mm/yr.; Millard Canyon south, 3-6 mm/yr.; Cabazon, >3 mm/yr. 6c - Oblique view of west wall, Trench 5, Cabazon paleoseismic site, reproduced from Yule and Scharer (in preparation). Yellow lines represent unconformity associated with fault rupture at the site. Black numbers 200 through 500a, b and c, and faint black dashed lines label stratigraphic layers. The most recent surface fault rupture event (U4) at this site occurred 630-590 yrs BP, and the penultimate event occurred 1220-1097 yrs BP (Yule and Scharer, in preparation). 6d - Correlation diagram of past 1200 yrs of earthquakes for SSAFS plotted vs distance DORQJWKHIDXOWUHSURGXFHGIURP
162 Applied Geology in California

Canyon to the west to Whitewater Canyon in the east (Figure 6b). The nature of (or even the existence of) any surface deformation along and to the north of the dotted red line (Figure 6b) is unconstrained but potentially could affect the CRA as far west as its crossing at I-10. The paleoseismic data from Cabazon, in combination with data from the SSAFS northwest and southeast of the pass, suggest that few, if any, surface ruptures carry through SGP. Figure 6d is one possible model for fault system behavior since 1200 yrs BP. Green horizontal bars represent quakes restricted to the Mojave and San Bernardino reach of the fault. The pink bars represent quakes restricted to the Coachella Valley. The northern extent of the Coachella Valley segment surface ruptures is not known but may reach the CRA fault crossings at 4, 5 and 6 (Figure 2). The late-14th century orange bar represents the most recent earthquake that could have possibly involved all sections of the SSAFS. If accurate, this late-14th century event would have been the largest major, but least frequent earthquake on the SSAFS that passed through SGP. It is possible that the penultimate event at Cabazon 1100-1200 yrs BP records a second one of these major earthquakes, but these data are preliminary. As seen in Figure 6d many events northwest and southeast of SGP have similar ages (such as that between 1690-1700 AD) so one could speculate that some ruptures passed through the pass without generating surface ruptures at the Cabazon and Millard trench sites or were only expressed as more broadly distributed ground surface deformation in the hanging wall of the SSAFS, Yule et al., (2014) present recently available slip rate data from strands of the San Gorgonio Pass Thrust Fault (SGPTF) located west of the CRA’s crossing of SGP. Based on their work, Millard Canyon preserves at least six generations of alluvial fan formation and a long record of fault displacement (Figure 7a). The oldest landform is known as the Heights Surface (Yule and Sieh, 2003; surface #5 in Figure 7a), which is estimated to be ~60,000 to 150,000 yrs old. Gravels below this surface buttress against the eastern wall of the valley (Figure 7b). The mapped extent of this feature forms a piercing line with 100-120 m (328-656 ft) of vertical separation and 300-450 m (984-1.476 ft) of right-lateral separation across the northern strand

of the SGPTF (Figure 7c). This de nes a postHeights slip vector oriented ~S50°E, sub-parallel to the relative motion across the SSAFS measured by GPS (Figure 10), and a slip rate of 3-6 mm/yr (0.12-0.24 inches) The Heights Surface has also been uplifted >120 m (>394 ft) across the southern strand of the SGPTF (Figure 7c). In addition a Holocene (6,000 – 9,000 yr. old) surface has experienced ~12 m (~39 ft)) of uplift (surface #4 in Figure 7a). Recognizing the uncertainty in these, both the late Pleistocene and Holocene offsets yield ~1-2 mm/yr (~0.04-0.08 inches/yr) of uplift rates and resolve to 3-6 mm/yr (0.12-0.24 inches/yr) of total fault slip rates. The two strands of the SGPTF at Millard Canyon therefore carry a total rate of slip of ~6-12 mm/yr (~0.24-0.47 inches/yr). There is considerable uncertainty in these slip rate estimates considering that the offset measurements, stratigraphic age dates, and number of locations along the faults where these data have been obtained in the SGP area are limited. Considering that the purpose of this study is to estimate the amount of vertical and horizontal fault displacement and surrounding ground deformation at the surface and not to make any temporal forecasts, this local geologic data is most important when estimating the ratio of horizontal to vertical fault slip. Seismicity Recorded seismicity provides data needed to evaluate the location and geometry of the SSAFS at depth below SGP. Fuis et al., (2012) interpreted a changing angle of dip on the SSAFS as the fault progresses from the Tehachapi Ranges to the Salton Sea. Similar to interpretations by Yule and Sieh (2003), Fuis et al., (2012) suggested in their Section #9 that the SSAFS in the SGP area dips ~ 45º to 50º to the north to a depth at least 15 km (9 miles). A more focused examination of the seismicity beneath SGP was completed during this study since the subsurface location and geometry of the SSAFS-SB is a key variable governing the amount and distribution of ground deformation likely to occur during future earthquakes. Catalogs of instrumentally recorded earthquakes in the SGP area with MW 1.25, dating back to 1932

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

a.

b.

(continues)

163

164 Applied Geology in California c.

Figure 7. Geomorphic Analysis reproduced from Yule and Scharer (in preparation). 7a – Oblique, north-northwest aerial view of terraces of Millard Canyon. Image is produced from B4 LiDAR data and color shaded according to slope angle; gentle slopes in blue-green to green and steep slopes in yellow to brown colors. Thin yellow ribbon-like forms that WUHQG16SDUDOOHOWRVWUHDPÀRZDUHULVHUVWKDWVHSDUDWHDOOXYLDOWHUUDFHVRIGLIIHUHQWDJH:KLWHQXPEHUVODEHOWKHVH fan surfaces of increasing age as follows: 0 – modern wash; 1 - <630 yrs old (C-14 charcoal age); 2 - ~1200 yrs old (C-14 charcoal age); 3 – 2000 yrs-3000 yrs old (estimate only, no age data), 4 – (6000 – 9000 yrs old (10Be CRN), and 5 (Heights surface_ - 75000 – 100000 yrs old (estimate only, no age data). Red arrows point to fault scarps. 7b Oblique aerial view of offset buttress unconformity in Heights fanglomerate across northern splay branch of the San Gorgonio Pass Thrust Fault (SGPTF) at Millard Canyon. Image in 7b produced in the same manner as the image LQD5HGOLQHZLWKWHHWKLVVXUIDFHWUDFHRIQRUWKVSOD\RIWKH6*37)PRGL¿HGIURP
San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

and up through December 31, 2013 were collected for the following timeframes: January 10, 1932 through December 31, 1980: ANSS catalog January 1, 1981 through June 30, 2011: Hauksson et al., (2012) catalog (HYS catalog) ‡ July 1, 2011 through December 31, 2013: ANSS catalog The HYS catalog contains waveform-corrected earthquake solutions that produced 3D earthquake locations that are more precisely located than the ANSS catalog. Accordingly, the strike and dip of seismicity lineaments are generally better delineated with the higher precision in the event location in the HYS catalog. Prior to the mid-1970s, the seismometer station coverage in southern California was limited and the instrument quality was not as reliable as modern recordings. These limitations in the older seismicity data are most evident in the focal depths of many ANSS catalog events prior to the mid-1970s, as the catalog contains many earthquake depths at 0 km and 6 km. These events with false or default depths were removed from this analyses. To better understand how reliable the ANSS catalog is after the mid-1970s, we compared the HYS catalog and the ANSS catalog (Figures 8a through 8d and Figure 9). In general we found that

165

while depth and magnitude patterns largely agreed between the two catalogs, the epicentral locations were very different. The correlation between the two catalogs on the magnitudes distribution was expected, as the Hauksson et al., (2012) effort did not re-assign event magnitudes. We interpret the consistency between the two catalogs on the depth to indicate that the earthquake depths in the modern (post mid-1970s) ANSS catalog are adequate, and we interpret the poor agreement between the epicenter locations to indicate the latitude and longitude coordinates in the modern ANSS catalog are less reliable. Accordingly, we recognized in our analyses that the event locations in the ANSS catalog are not as reliable as those in the Hauksson et al., (2012) catalog. In this study, in order to remove the redundancy in the overlapping timeframe between the HYS catalog and the ANSS catalog, we only used the HYS catalog for this portion. The spatial distribution of instrumentallyrecorded seismicity was evaluated in two-dimensional (2D) cross sections. We constructed several cross sections in the SGP area (locations shown on Figure 8a) orthogonal to primary fault orientations, and plotted all earthquakes recorded since 1932 within one mile of each cross section line. Events with depths known to be false (i.e., older events with default depths in lieu of measured depths) were removed for the cross section analyses.

a.

(continues)

166 Applied Geology in California b.

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

c.

167

168 Applied Geology in California d.

Figure 8. SGP Seismicity, 1932 – 2013. 8a - Epicenter Map; red dots from HYS catalog, which includes DOOHYHQWVLQVRXWKHUQ&DOLIRUQLDIURPWKURXJK2QO\0•HYHQWVDUHVKRZQ here. Seismicity (yellow dots) from ANSS catalog for the same magnitude range, but only those outside the HYS catalog time period (i.e., 1932 through 12/21/1980 and 07/01/2011 through 12/31/2013) are plotted here. Epicenters for the 1948 and 1986 earthquakes (shown with white star and labeled in white) are from Nicholson (1996). Topographic data are from the USGS National Elevation Dataset (NED; available at http://ned.usgs.gov). Blue line is approximate location of CRA. Black lines are faults from USGS Quaternary Fault and Fold Database (QFFD), 2010). Location of seismicity cross sections (8b through d) indicated by thick white lines and labeled in small white rectangles. Thin white lines indicate area of seismicity within one mile of cross section line collapsed into cross section. Seismicity outside one mile of the cross section line are not shown in cross sections. 8b - Mission Creek cross section. 8c - Millard Canyon cross section. 8d - Snow Peak cross section. Red dots on cross sections are hypocenters from the HYS catalog and blue dots are from ANSS catalog with hypocenters from the ANSS catalog removed from default depths of 0 km to 6 km). On each cross section, the solid black line represents the surface topography. Short, solid, vertical blue line represents the location of the CRA.

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

169

Figure 9. SGP area earthquake catalog magnitude frequency distribution. Based on this distribution, in general, one can expect to see one M6.5 event in SGP area roughly every 100 years based on 82 years of instrumental recordings.

Interpretations of the seismicity trends in the cross sections are shown in Figures 8b, 8c, and 8d. On some sections, seismicity is clearly clustered on one side of a mapped fault along a strong linear trend that probably follows the dipping fault plane (e.g., the Mission Creek section in Figure 8b). We use these dips to help infer the fault geometries used in the computer models. In other sections, seismicity anks both sides of a fault, and the subsurface fault geometry is not as easily discerned.

The 1948 MW 6.2 Desert Hot Springs and the 1986 MW 6.1 Palm Springs earthquakes3 are the only two signi cant historical earthquakes in the SGP area. Nicholson (1996) analyzed the wave forms from both of these events, which The epicenters of the 1948 MW 6.2 Desert Hot Springs and the 1986 MW 6.1 Palm Springs earthquakes are shown as stars in Figure 8a. The hypocenter of the Palm Spring earthquake is also shown as a star on the cross section in Figure 8b. 3

170 Applied Geology in California

he interpreted as being generated by ruptures on the Banning Fault at depth. Based on his work, the 1986 earthquake was on a 40-50°NE dipping plane that ruptured between depths of 5 and 13 km and did not reach the surface (although minor local surface deformation was reported). The earthquake began with pure right-lateral strike slip and propagated as oblique-reverse slip. As also described by Nicholson (1996), the 1948 earthquake occurred on a steeper-dipping subsurface section of the Banning Fault immediately southeast of the 1986 event in a dominantly strike slip rupture with some reverse motion and likely did not reach the surface. Nicholson (1996) concluded that ruptures on the Banning Fault could be segmented by geometrical discontinuities along the fault or other faults at depth, and we used the next segment to the west to model our Mw 6.0 to 6.5 event. A magnitude-frequency distribution of the instrumentally recorded events in Figure 8a (including the 1948 and 1986 earthquakes) is presented in Figure 9. The rates for both catalogs are shown separately and combined in the gure. As noted in the legend, the curve of the combined catalogs on Figure 9 spans 82 years and is likely complete above ~MW 6 (and perhaps above ~MW 3, based on the slope of the ANSS catalog rate). Based on the combined catalog rate, a MW 6.5 event would be expected roughly every 100 years in the SGP area, and a MW 6.0 earthquake would be expected roughly every 20 years. This MW 6.0 rate is generally consistent with the MW 6.2 1948 Desert Hot Springs and MW 6.1 1986 Palm Springs earthquakes. The higher rate for the Hauksson et al., (2012) catalog could be due to improved seismometer quality, aftershocks associated with the 1986 Palm Springs earthquake, the natural variability of earthquake rates, or a combination of these. Geodesy We focused on two aspects of the active deformation that are relevant to future seismic hazards in the SGP area: 1) horizontal deformation attributable to plate tectonic deformation, and 2) vertical motion that is attributable to thrust or normal slip that is a part of the tectonic deformation but

also to basin-scale response to the extraction and recharge of hydrologic resources. We estimated horizontal deformations from GPS measurements, and the vertical deformation was estimated from a combination of InSAR and GPS. InSAR is a good supplement to the GPS data because it has denser geographic coverage and is sensitive to motion along the space satellite look direction, which is near vertical. GPS provides a strong three-component framework for evaluating crustal deformation in a global reference frame at speci c locations where the ground receivers are installed. GPS networks in southern California are extensive and cover much of the region encircling the SGP area. The Nevada Geodetic Laboratory processes data from all of the public domain networks available (e.g. the EarthScope Plate Boundary Observatory, SCIGN, etc.) as part of the global processing that includes over 12,000 stations. Solutions are aligned to the International Terrestrial Reference Frame (ITRF, Altamimi et al., 2011) and a custom reference frame NA12 (Blewitt et al., 2013), which is most applicable for tectonic studies in North America. Our analysis includes all of the latest data collected at the continuously recording stations and so may be more complete at this geographic scale compared to other datasets. For this project we applied the analysis technique of Hammond et al., (2012) for the natural variability of seasonal oscillations, number of detected discontinuities in the time series, range of vertical rate, and network source on the aftershocks associated with the 1986 Palm Springs earthquake. The GPS network provides the stability and coverage needed to correct for errors that can affect the InSAR LOS (line of sight) rates, e.g. putting them into a consistent reference frame and infer vertical motion. Vertical rates of the solid Earth estimated from the GPS and InSAR data are discussed below. InSAR provides a complementary dense spatial coverage that has higher geographic resolution of rates along the satellite-to-ground look direction, and so can provide constraint where GPS instruments are not present. InSAR rates were aligned

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

with GPS where they overlap, so that the combined velocity eld has accuracy and precision in both the long wavelength signals from GPS and greater completeness in the shorter wavelength signals from InSAR (e.g. Tong et al., 2013; Wei et al., 2010). Analysis of InSAR data also provides a means to identify GPS stations that behave anomalously owing to local conditions and that do not re ect underlying tectonic processes. For this project we processed InSAR data available in the WinSAR archive (hosted by UNAVCO, Inc.), covering the SGP area; focusing on track/ frame sets where the greatest numbers of scenes were available inside speci c frames. The best time series analysis results were obtained when many scenes were available from the same area. For coverage in the SGP area, we used ENVISAT and ERS data from tracks 120, 127, 170, 399, and frames 657, 2907, 2925, and 2943. Many hundreds of scenes were available. We formed as many interferograms as possible for each track/ frame set using the Gamma software (Werner et al., 2000) resulting in between 142 to 2590 interferograms per track/frame set. For each set we used the SBAS algorithm (Berardino et al., 2002) to infer for each pixel the best tting constant rate, plus the amplitude and phase of annual oscillations in position (solving for them reduced bias in the estimates of constant rate). The InSAR line of sight (LOS) rate maps were then transformed into the GPS reference frame by applying a bilinear rate ramp, which minimized the mis t between the InSAR and the GPS rates. We corrected for the effect of horizontal deformation on the radar LOS deformation using the GPS-based horizontal crustal strain rate map of Kreemer et al., (2012) projected into the radar LOS. We unproject the remaining signal, free of horizontal deformation, into vertical by considering the look inclination angle at each pixel. Finally, to compensate for residual errors potentially present in the strain rate map, we perform a nal alignment of the InSAR signal to vertical GPS velocity eld. The result is an estimate of the part of the signal that is attributable to vertical motion, consistent between the InSAR and GPS methods.

171

Horizontal motion from GPS We focus on the signals near and surrounding the SGP area, and for simplicity plot motion with respect to a xed point between the south and north strands of the SSAFS (Figure 10). The results clearly show a snapshot of the interseismic dextral shear strain accumulation across the Paci c/North America crustal plate boundary that will eventually be released as earthquakes. Of immediate interest to this study is whether the strain accumulation rates are symmetric around the SSAFS in the SGP area, and whether the data might reveal localizations in strain rate that can be used to clarify which fault segments are more active. When we view the velocity pro le (Figure 11) derived from the GPS velocities we see a gradient of 3 to 38 mm/yr (0.12 to 1.5 inches/yr) of velocity change, with a budget of about 35 mm/ yr (0.1.4 inches/yr) across the region, consistent with earlier studies. The strain rate in the crust is approximately proportional to the slope of the velocity trend. This simpli ed view of the velocity eld reveals very minor, if any, changes in slope across the fault systems of southern California, and no obvious asymmetry in the trend of the GPS velocities around the SGP section of the SSAFS. Some possible slope changes exist but tend to be larger among the stations with higher uncertainties. To test how signi cant the slope changes are, we highlighted the stations with the highest precision (GPS time series over 5 years long) in Figure 11 with a different color symbol. Focusing on stations with the highest precision shows narrower scatter around the average trend, suggesting that they are the best measure of tectonic deformation, and that the slope changes are small. However, near the Elsinore Fault Zone on the west end of the pro le, the rates are lower, which is consistent with earlier geodetic studies and geologic observations (e.g. see Chuang and Johnson, 2011, for summary). Thus the average shear strain rate is about the same on either side of the SSAFS near SGP. However, when we consider tensor strain rates estimated from GPS velocities, they show that there are changes in the interseismic areal contraction rate along the strike of the SSAFS. In

172 Applied Geology in California

Figure 10. Horizontal GPS velocity around San Gorgonio Pass section of the San Andreas Fault. Magenta box ZDVXVHGWRVHOHFWVLWHVWRVKRZLQSUR¿OHRI*36YHORFLW\VHH)LJXUH 7RKLJKOLJKWWKHGLIIHUHQFHEHWZHHQ *36YHORFLWLHVKRZWKH\YDU\DFURVVWKHSODWHERXQGDU\ WKHYHORFLW\DWWKHUHGVWDULQWKHFHQWHURIWKH¿JXUHZDV VXEWUDFWHGIURPDOORIWKHREVHUYHGYHORFLWLHVWKXVWKHVHUDWHVDUHQRWGHSLFWHGLQD1RUWK$PHULFD¿[HGUHIHUHQFH frame). Gray 4-character IDs are the names of GPS stations used in the analysis.

Figure 12 we show the tensor strain rates estimated from subsets of the horizontal GPS velocities inside circles centered on six locations along the SSAFS. We used the method of Savage et al., (2001) to simultaneously estimate parameters for vertical axis rotation and horizontal tensor strain rate from the GPS velocities. We compute the horizontal dilatation rate as the sum of the principal strain rates, extension reckoned positive. In the vicinity of the SGP and the left step in the SSAFS, the net horizontal dilatation rate becomes negative and contractive. This suggests that the tectonic conditions for compressional faulting exist at SGP, and that we can expect earthquake ruptures that pass through SGP to have a component of compressional slip.

Dilatation rates to the southeast, near the Coachella segment of the SSAFS and the northern Salton Trough, are positive. The along-strike change in sign of geodetic dilatation correlates with the transition to a transpression-dominated section of the SSAFS at SGP (Figure 1), suggesting that geodesy is measuring features of strain accumulation that are expressed in the structure of the SSAFS. Vertical motion from InSAR and GPS The vertical component of motion attributable to strain accumulation generally has rates smaller than the horizontal component, but can place complementary constraints on earthquake cycle models (including estimated slip rates and recurrence

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

173

Figure 11. 3UR¿OHRIKRUL]RQWDO*36YHORFLW\LQ1Û(GLUHFWLRQLHGLUHFWLRQRIPDJHQWDER[LQ)LJXUH  9HORFLW\LVWKHFRPSRQHQWRIPRWLRQLQWKH1Û:GLUHFWLRQ(UURUEDUVLQGLFDWHVLJPDFRQ¿GHQFHLQWHUYDO*UD\ 4-character codes designate names of GPS stations used in the analysis, some have been omitted for clarity. Cyan circles highlight the GPS velocities with the longest observation intervals (over 5 years) and hence the smallest XQFHUWDLQWLHV5HGVWDUQHDUFHQWHURI¿JXUHLVLQVDPHORFDWLRQDVUHGVWDULQ)LJXUH*UHHQGDVKHGOLQHVVKRZ bounds of north and south strands of the SSAFS; magenta dashed lines show SJFZ; grey dashed lines show Elsinore Fault Zone. Yellow line shows the average trend of the observations, which show minor deviations from WKDWDYHUDJHVORSH7KHUHLVDVOLJKWGHFUHDVHLQVORSHRIWKH*36YHORFLWLHVRQWKHZHVWOHIW HQGRIWKH¿JXUH near the Elsinore Fault Zone.

intervals), especially near faults that have an up- or down-dip component of slip such as near the CRA fault crossings in the SGP area. The results show the vertical rate eld has both focused and regionally coherent up/down motions between -3 and 3 mm/yr (-0.12 and 0.12 Inches) (Figure 13). The rates of vertical motion are generally (though not universally) similar based on InSAR and GPS. The blue and red circles in Figure 13 indicate that the GPS velocity tends to occupy areas in the InSAR-derived vertical rate map that are also blue and red, respectively, indicating that GPS and InSAR rates are well-aligned. In Figure 14, the vertical rates along ve pro les oriented approximately normal to the strike of the San Andreas Fault, as shown with the magenta

rectangles on Figure 13. These indicate that vertical rates from InSAR and GPS generally have similar amounts of scatter, and that in almost all cases the GPS rates t inside the scatter of the InSAR rates within the uncertainty. Only one station TMAP (Thermal Airport) inside Pro le 5 appears to be an outlier affected by local subsidence, likely due to utilization of the underlying groundwater basin, and having a gradually wandering vertical GPS time series. The similarity of the InSAR and GPS vertical rates suggests that the internal geometry of the two aligned solutions are in most places measuring the same underlying vertical rates over the decadal time scale of geodetic observation. This suggests that errors that are inherent in InSAR time series analyses

174 Applied Geology in California

Figure 12. Strain rates along the SSAFS. Units are in nanostrain (nstr) = 109; Red rates on map are dilation rates in 109/yr. Thick black dashed lines emphasize northwesterly rotation of right-lateral shear direction.

(e.g. long wavelength components from the atmosphere and spacecraft orbit uncertainties) have been mostly mitigated by the alignment of InSAR rates to GPS. As in earlier studies, we detect signals characteristic of long term changes to ground level in sedimentary basins in southern California (e.g. Bawden et al., 2001; Lu and Danskin, 2001; Brooks et al., 2007; Wisely and Schmidt, 2010; Tong et al., 2013). Both long wavelength and shorter wavelength signals exist. The short wavelength signals are predominantly downward motions related to subsidence from groundwater withdrawals from the sedimentary basins (e.g. the San Bernardino Groundwater Basin studied by Wisely and Schmidt, 2010). These are correlated

with parts of the InSAR/GPS pro les that have larger variance of velocity (e.g. in Pro les 1 and 3, Figure 14) within the SJFZ which may in part re ect annual uctuations in the ground surface due to rising and lowing groundwater levels with in ltrations into and withdrawals from the San Jacinto Groundwater Basin. These subsidence features are commonly bounded by faults, e.g. the right step in the San Jacinto fault system at the San Jacinto Valley section, and the San Bernardino basin between the San Andreas and San Jacinto faults. A longer wavelength signal of about 2 mm/ yr (~0.08 inches/yr) upward gradient from the coast towards the SSAFS is observed. The upward motion in the vicinity of the SSAFS is

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

175

Figure 13. GPS and InSAR Interseismic vertical rates. Vertical rate of Earth’s surface estimated from GPS (circles with 4- character names) and InSAR (background color). Color bar indicates scale of vertical rate in mm/yr, and is the same for GPS and InSAR, blue is downward motion, red is upward motion. Strip of green triangle symbols indicates the path of the CRA. Black lines are faults. Yellow star indicates location of SGP near where the CRA crosses the SSAF. Approximately north-south trending linear discontinuity in vertical rate east of SGP is an artifact of overlaying data from adjacent InSAR. Magenta boxes correspond to vertical rate SUR¿OHVRQ)LJXUH

most convincingly expressed in the vertical GPS velocities because it is a long wavelength signal, which is generally more accurate in GPS data. This motion could be attributable to interseismic deformation of the lithosphere associated with the SSAFS as the strike-slip system bends west in the restraining bend, causing shortening perpendicular to the trend of the fault system (Cook and Dair, 2011; Hammond et al., 2013). In the northern part of the SGP study area vertical motions

may also be attributable, in part, to viscoelastic relaxation following the 1992 Landers and 1999 Hector Mine earthquakes. The location of the uplift is broadly consistent with the location of the southwestern lobe of uplift in a quadrupole pattern around the epicenters northeast of the SGP area, noted to occur months (Pollitz et al., 2001) and years (Freed et al., 2007) after the events. These post-earthquake deformation signals are thought to be transient as they are observed to

176 Applied Geology in California

Figure 14. 3UR¿OHVRIYHUWLFDOUDWHIURP*36EODFNFLUFOHVZLWKVLJPDHUURUEDUV DQG ,Q6$5QXPHURXVJUD\SRLQWVIRUPLQJFORXG 7KHSUR¿OHVFRUUHVSRQGWRWKHPDJHQWD rectangles in Figure 13. Sections through which the San Jacinto and San Andreas faults pass are annotated.

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

decay in amplitude over time, leaving a small but detectible strain signal in the far eld ve years after the events (Hammond et al., 2010). Along the alignment of the CRA, the vertical rates near SGP are upward in the global reference frame (red areas in Figure 15), but do not vary rapidly except near the boundaries of the subsiding basins. Along the length of the aqueduct, the most rapid subsidence is in the San Jacinto Valley, near -2 mm/yr (-0.08 inches/ yr), and has the strongest gradient along the aqueduct (Figure 15). Farther to the east of the SSAFS the rates are

177

generally near -1 to 2 mm/yr (-0.04 to 0.08 inches/yr) and vary less drastically. There is a slower subsidence, -1 to 0 mm/yr (-0.04 to 0 inches/yr) south of the SGPTF that could possibly be a vertical interseismic strain signal associated with these faults, but is located inside another fault-bounded groundwater basin of Beaumont and Banning, and so may be recording hydrologic compaction of valley sediments (Figure 13). There are no strong gradients in vertical rate near the SSAFS east of SGP. However, the peak of the vertical rates lies just north of the SSAFS-NB

Figure 15. GPS and InSAR interseismic vertical rates shown on Figure 14 zooming into area near SGP (yellow star). The prominent blue area of subsidence is south of station P584 in the San Jacinto Valley. Lighter blue to white area east of GPS station BMRY and northeast of GPS station P584 is the basin containing the cities of Banning and Beaumont. Approximately north-south trending linear discontinuity in vertical rates near east edge of ¿JXUHLVDQDUWLIDFWRIRYHUOD\LQJGDWDIURPDGMDFHQW,Q6$5VFHQHV

178 Applied Geology in California

(Figure 16), and could be a signal associated with a thrust component of strain accumulation on a NE dipping strand of the fault. This signal is of interest because it could help localize the accommodation of active contraction across the SAF restraining bend through SGP. However, the combination of noise in the rates and proximity of this feature to the seam in the vertical rate eld (Figures 13 and 15) make this local uplift signal dif cult to uniquely ascribe to strain accumulation. Other Uncertainties The scatter of the InSAR and mis t between GPS and InSAR are less than 1.0 mm/yr. giving an estimate of the mean aleatory uncertainty of the vertical rates in the combined InSAR/GPS technique. Some seams in the vertical rate eld are noticeable where adjacent track/frame pairs overlap, e.g. in the north-central section of Figure 13, and the north-east corner of Figure 15. These seams arise from various uncertainties in alignment of InSAR relative to the GPS data that create slight discontinuities in the rate eld that are near the level of the aleatory uncertainties but are easy to identify by eye in some cases. In many cases, the seams are not visible because the alignment is accurate. However, where the motion

is not constant over time (e.g., in sedimentary basins or in areas affected by decaying post seismic relaxation), seams will occur when GPS and InSAR data are collected during different time spans. Thus, interpretation of the vertical rate eld should proceed with caution and knowledge of the additional epistemic contributions to uncertainty in the rate elds. For example, motion in structurally-bounded sedimentary basins is often strongly seasonal, and some observation time intervals may show subsidence while others show uplift, e.g. in the San Bernardo, Los Angeles, and Santa Clara basins (Bawden et al., 2001; Lu et al., 2001; Schmidt and Bürgmann, 2003; Wisely and Schmidt, 2010; Amos et al., 2014). Subsidence rates can change from one decade to another, especially in the alluvial basins when climate and/or municipal water management changes can affect pumping, recharge or compression rates. We can demonstrate that such rate changes are likely present in our data and affect the quality of the alignment between InSAR and GPS. Figure 17a and b show scatterplots of the vertical rate from GPS versus the vertical rate from the aligned InSAR rate eld. If there were perfect agreement between

Figure 16. 3UR¿OHRI,Q6$5HVWLPDWHGYHUWLFDOUDWHDORQJWKH&5$SURMHFWHGRQWRORQJLWXGHJUD\GRWV %ODFN line connects black dots representing vertical rates at locations of the aqueduct. Red line connects points giving average vertical rate inside a circle of radius 1 km centered on each location. Subsidence approaching 2 mm/yr near Longitude -117 is in the San Jacinto Valley. Sections through which the San Jacinto and San Andreas faults pass are shown with magenta and green dashed lines respectively. Note broad uplifted centered on northern edge (near Longitude -116.5) of San Andreas is consistent with a north-easterly-dipping fault.

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

179

Figure 17. Scatterplot of InSAR-derived vertical rates. 17a - At locations where there are estimates of GPS vertical rate (mm/yr) for all data in this analysis. 17b - Considering only ENVISAT data gathered between 2003 and 2010, a time overlapping the epochs when most of the GPS data were collected.

the two datasets all the points would plot on the diagonal where the two rates are equal. We see that the correlation between the rates is especially good when we compare rates from ENVISAT InSAR data collected between 2003 and 2010 to rates from the GPS data that were collected most often after 2003. When the entire InSAR dataset is compared the correlation degrades especially at the GPS stations that go downward the most quickly. This suggests that downward rates in alluvial basins tended to increase their subsidence rate between the two intervals of observation, so the best match between rates is obtained after 2003, when the InSAR and GPS data have better temporal overlap. Currently the available InSAR and GPS datasets cover overlapping but not identical time intervals, so our ability to achieve perfect alignment is inherently limited when using all the data by the assumption of constant rates over time. In the future, improved modeling strategies, the evolution of precise geodetic networks in the western U.S., and increased availability of data from future InSAR satellite missions promise to improve efforts to obtain integrated InSAR+GPS vertical rate maps.

Fault Geometry Capturing the possible vertical components of fault slip and ground deformation along the CRA requires focusing on the con guration of the active Late Quaternary traces of the SSAFS-SB at depth below and through the SGP area. In general, the fault can be considered a ‘propeller-like’ or ‘Möbius-like’ geometry, curving vertically and horizontally from its near-vertical linear traces northwest and southeast of SGP (e.g., Fuis, et al., 2012). However, currently there is some debate in the geoscience community regarding the speci cs of the fault’s subsurface geometry at depth through the SGP area and representing this curving geometry with the simple planar geometries used in Coulomb 3.3 and in geodetic block motion modeling aggravates the issue. Furthermore, the expected vertical coseismic motions predicted from these models are sensitive to these geometric assumptions. Little argument exists over where Yule and Sieh (2003) mapped the active fault traces at the surface and their projections of relatively shallow north-northeast dipping upper portions of the fault (Figures 5a through 5d). The main difference in con guring the

180 Applied Geology in California

fault’s subsurface geometry is in the deeper portions of the fault. Figure 18 presents two simple 2-D schematics of the basic alternative fault geometries. Model 1 - A northeasterly-dipping SSAFS that continues dipping ~45° to the north-northeast in SGP and ~75° to the east in Coachella Valley through the entire crust (basically the Yule and Sieh (2003) and Fuis et al., (2012) models); and Model 2 - A SSAFS that dips about 45° north to northeast in the upper crust that links to/merges with a vertical trace of the SSAFS-NB at depth in the lower crust inside the SGP area below the sub-vertical bedrock trace of the North Branch zone at the surface, and is steeplydipping outside the SGP area. Both models were investigated during this CRA analysis. With the same upper crust con guration in both Model 1 and Model 2, in Figure 19 we highlighted

how Yule and Sieh (2003) mapped surface traces which were simpli ed for input into the Coulomb 3.3 modeling. Considering the difference between Model 1 and Model 2 in the deeper crust, data and interpretations presented by Yule and Sieh (2003) (summarized in Figure 5c) and by Fuis et al., (2012) (Figures 4 and 20a) indicated that they preferred Model 1, in which the SSAFS dips to the north-northeast indefinitely through the seismogenic layer. However, we suggest that the weight of the data supports Model 2 with its vertical to sub-vertical geometry of the SSAF through the lower seismogenic crust as shown on Figure 18. Below are summaries of our observations that support Model 2. Observations from Geology, Geophysics, and Seismicity In the surface bedrock geology (Gutierrez et al., 2010), pre-Cenozoic metasedimentary and metavolcanic rocks (which are typical southwest of the SSAFS) are separated from the Precambrian igneous and metamorphic rocks (which are typical

Figure 18. Alternate models of the SSAFS-SB geometry beneath the SGP area.

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

181

Figure 19. Surface Traces of SSAFS-SB that were modeled in this study are highlighted in yellow on these geology maps of the SGP area that were reproduced and annotated from Yule and Sieh (2003). The location of CRA is highlighted with a line of blue dots. Disregard boxes on map; these pertain to the original publication.

northwest of the SSAFS). This boundary is highlighted in Figure 3. This location is consistent with the Fuis et al., (2012) interpretation of a deep crustal boundary along an aeromagnetic anomaly (Figure 4), but would be vertical if it connects to the surface trace. We suggest that the sharp step in the maximum depth of seismicity presented by Fuis et al., (2012) (Figure 20a) can support Model 2’s vertical to subvertical geometry of the SSAF through the lower seismogenic crust, as illustrated on Figure 20b where we superimposed Model 2’s con guration onto the original interpretation presented in Fuis et al, (2012). In addition, the vertical to sub-vertical seismicity lineaments and steps in the maximum depth of the seismicity we observed during our review of the seismicity patterns, as summarized in the cross sections shown on Figures 8b, 8c, and 8d, support a vertical “deep trace” in the geometry of the SSAFS that generally falls east of the mapped surface traces of the SSAFS in the SGP area. Close investigation of the GPS velocity gradients near the fault by (Fialko, 2006) corroborate the suggestion that the deep, creeping trace of the SSAFS is north-northeast of its mapped surface traces in the SGP area. The results of our geodetic analyses,

described above in the Geodesy sub-section, are consistent with the notion of localized creep north-northeast of the mapped surface traces of the SSAFS in the SGP area. In Figure 21 we clearly demonstrate the very good agreement between the interpreted locations and geometries of the lower crustal portion of the SSAF in the SGP area based on the bedrock boundary, the aeromagnetic anomaly, the deep sub- vertical seismicity lineaments and step-overs in the maximum depth of the seismicity and the geodetic data as discussed above supporting Model 2. We also considered the tectonic evolution of southern California to elucidate the dip (angle) and strike (azimuth) of the SSAFS plane at depth beneath the SGP area. As shown on the panels in Figure 22a thru c, we envision the SSAFS beginning as a straight or smoothly curving, steeply dipping strike-slip fault (schematic 1. Figure 22a), as it presently is northwest and southeast of SGP. Block motion, in uenced by the Eastern California Shear Zone, offset this simple trace, generating oblique compression near the fault that was initially accommodated by warping and folding (schematic 2. Figure 22a). Eventually, the folds gave way to a shallower dipping thrust or reverse faults and

182 Applied Geology in California a.

b.

Figure 20. $OWHUQDWLYH66$)6PRGHOVEHQHDWK6*3DUHDVXSSHULPSRVHGRQWRSUR¿OHUHSURGXFHGIURP)XLV HWDO DQGPRGL¿HGZLWKPDJHQWDDQGEOXHOLQHV'LVUHJDUGSULRUDQQRWDWLRQVRQWKHSUR¿OHLH\HOORZOLQHV green lines, etc.), these pertain to the original publication. 20a - Model 1, alternative model by Fuis et al., (2012); northeast-dipping active trace of SSAFS in SGP area continues with moderate dips to depth (i.e., no at depth subvertical SSAFS trace). 20b - Model 2, our preferred SSAFS model; northeast-dipping active faults in SGP area merge into sub-vertical SSAFS creeping deep trace.

Figure 21. Comparison of deep trace interpretations of the SSAFS beneath SGP area.

possibly the incorporation of the old San Gabriel Fault Zone (schematic 3. Figure 22a). As displacement increased on the SSAFS, the shallower dipping thrust fault became the more ef cient path for the main strike-slip trace to accommodate the motion through the area, and the shallower dipping

thrust fault became an oblique-slip fault while the original more steeply dipping strike-slip portion of the main fault was abandoned (schematic 4. Figure 22a). These models are similar to those presented by Cooke et al., (2013) to support the abandonment of vertical fault for dipping oblique slip fault

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

a.

183

b.

c.

Figure 22. SSAFS evolution and modeling. 22a - (1) The straight strike-slip fault begins deforming into a smoothly curving strike-slip fault, with (2) compression initially accommodated by warping and folding. (3) Eventually thrust faults form, and as strike-slip displacement increases (4) the main strike-slip trace begins using the thrust fault, as is presently occurring in San Gorgonio Pass. 22b - With the main fault outside SGP remaining a vertical or very steeply-dipping strike-slip fault the lower angle oblique-slip reverse faults in the Pass formed using a thrust component of displacement to accommodate strain; and because there is no compression driving the thrust component of displacement from the Y-Y’ direction, the thrust must merge with the vertical main fault at depth, as shown in Model 2 on Figure 18. For the Coulomb 3.3 model, we developed a 3D geometric model of the fault plane with increasing depth using Model 2) and the available data and information. We also take the azimuth of the block motion at each point on the constructed fault plane and slip the fault consistent with the three-dimensional fault geometry Figure 22c at that point to get the horizontal and vertical components of displacement applied to the Coulomb 3.3 program, as listed in Table 1.

in the SGP. In this tectonic progression, the SSAFS outside SGP to the northwest and southeast should have remained and continued to be a vertical or very steeply-dipping strike-slip fault. The weight of the geologic, seismic and geodetic data and information presented above leads to the following three considerations: 1. The shallower dipping oblique-slip faults in SGP formed and began using thrust to accommodate the strain as the main fault, 2. There is no compression driving the thrust from the east as seen in Figure 22b, and 3. The shallower dipping oblique-slip faults in SGP must merge with the vertical or very steeply-dipping strike-slip fault to the northwest and southeast of SGP, also shown in Figure 22b.

These three considerations demonstrate that the shallower dipping oblique-slip faults beneath SGP area must also merge with a sub-vertical, strikeslip main fault at depth as included in Model 2. We preferred Model #2 based on the above presentation of the data and discussion. However, before we ran the models we did not know how signi cant the difference between these two models would be on the deformation eld, so we spent considerable effort running the two models through trial runs and comparisons with the geologic, geomorphic, paleoseismic, seismic and geodetic data. After we completed the model trial runs we found that because the difference was so deep in the crust there was little impact on the surface deformation and the CRA crosses the SSAFS where there is the least difference between the two models. Details on

184 Applied Geology in California

what the fault does northwest and southeast of SGP, are not directly pertinent to this project, but we note it is possible that the SSAFS south of Indio could be sub-vertical in the upper crust, although seismicity and geodesy leads us to our preferred model. It is also possible that the seismicity is occurring on the older traces of the fault and the young fault is subparallel but seismically quiescent. In this case, the deep trace could be essentially beneath the surface trace, and the geodesy could be explained by a contrast in elasticity across the fault (Fialko, 2006). Regardless, whether or not the Coachella segment of the SSAFS is sub-vertical does not affect the CRA where it crosses the fault zone in the SGP area and would result in only a minor decrease in deformation where the CRA is parallel to the Coachella segment of the SSAFS (Fattaruso et al., 2014). Accordingly, a simpli ed con guration of the curving subsurface geometry of the SSAFS through SGP and beneath the CRA was developed using Model 2. Figure 22c also illustrates how the changing strike (H=horizontal), dip (V=vertical), and rake (net) components of the slip across the changing geometry of the simpli ed fault plane were calculated (results listed in Table 1) and applied to the Coulomb 3.3 software. Figure 23 and Figure 24 show the con guration of the simpli ed geometry of the fault model that was used as the basis for the Coulomb 3.3 analysis. Based on available geologic, geomorphic, seismologic and geodetic data, variations in the strike and dip of the fault’s 3-D geometry were made along its trend, primarily in its shallow portions, to be compatible with the modeling programs. The shallower portions of the fault con guration presented in Figure 24, as discussed above, were simpli ed from the surface traces mapped by Yule and Sieh (2003) (Figures 5a through 5c and Figure 19) near CRA fault crossings #5 and #6 (Figure 2).

Fault Displacement and Strain Modeling Software Coulomb 3.3 We calculated interseismic rates and coseismic static displacements in the horizontal and vertical directions at the Earth’s surface caused

by slip along the SSAFS in the SGP area using Coulomb 3.3 (USGS, 2011). Strain and stress changes on other planes other that the surface are potential outputs of the Coulomb 3.3 but were not included here because we were focused on fault slip and horizontal and vertical deformations of the ground surface. Coulomb 3.3 calculations are made in an elastic halfspace with generally uniform isotropic elastic properties, following Okada (1992). At the scale of the depth of fault ruptures during events with magnitudes considered in this study topography and gravitational forces related to topographic and rock type differences have very limited impact on the results so for simplicity the ground surface is modeled as at. The 3D geometry of the SSAFS-SB fault interface that is slipped in the model was represented in the Coulomb 3.3 model as planar fault sections. Where faults intersect or abruptly change dip at depth we added small intermediate oriented fault sections or panels and each “panel” was given a separate strike, dip, and rake where appropriate. The coordinates of all fault sections are included in Table 1. Other input parameters include the fault rupture depth, and elastic half-space parameters de ned by Young’s Modulus, Poisson’s ratio and the coef cient of friction. The three scenario events discussed in the Introduction and Methodology, were applied across the 3D modeled fault geometry interface of the SSAFS-SB shown on Figures 23 and 24 centered beneath SGP. Based on the geology, seismicity, and geodesy in and around the SGP area, the maximum rupture depth in the Coulomb 3.3 model for Scenario 1 and 2 was 12 km (7.5 miles), and the rupture depth was 25 km (15.5 miles) for Scenario 3. We believe that 12 km (7.5 miles) is the likely rupture depth in SGP because that is where the fault changes from dipping to vertical, and 12 km. (7.5 miles) rupture depths are customary in California hazard models, such as WGCEP (2008) UCERF 2 and WGCEP (2013) UCERF 3. Because seismicity at SGP extends down to almost 25 km (as shown in Figures 8b, 8c and 8d) and because there is some evidence that very large ruptures

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

Table 1: Interpolated Coulomb 3.3 fault node input parameters.

(continues)

185

186 Applied Geology in California Table 1: Interpolated Coulomb 3.3 fault node input parameters. (continued)

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

187

Figure 23. 66$)66%&RXORPEPRGHOQRGHVDQGNP§PLOHV GHSWKFRQWRXUV'HSWKFRQWRXUVUHODWLYHWR sea level. Contours are color-coded by composite surfaces; see inset map (upper right) for fault names. Sampling JULGXVHGLQ&RXORPEPRGHOVKRZQKHUH LVNP§PLOHV E\NP§PLOHV &RXORPEPRGHOSRO\JRQ coordinates shown as gray dots with identifying three-digit numbers. Modeled fault dips also listed.

extend deeper than smaller ones (e.g., King and Wesnousky, 2007), we used 25 km (15.5 miles) for our “worst case scenario”. We used a piecewise planar fault geometry without tapering slip to zero for two reasons: 1) although recent research work on shallowly dipping faults (like subduction zones – see Burgette et al., 2009) indicated that tapering has a signi cant impact on surface deformation, early trial runs during this study, showed that tapering with depth had virtually no impact for steeply dipping faults like the SSAFS; and 2) Coulomb 3.3 involves a Greens function approach and discontinuities are not a modeling problem (see Okada, 1992). The surface deformation is simply the sum of deformation contributed by all the midpoints or independent rectangular portions of the model fault and there is no

requirement that adjacent points or rectangles have the same slip; each contributes independently. Constant slip was assigned to each segment or planar surface of the modeled 3D geometry of the SSAFS beneath SGP, both vertically and horizontally calculated, as illustrated above in Figure 22c. At each point on the modeled SSAFS through SGP listed in Table 1 and shown in Figure 23, we give an azimuth for the direction of motion of the blocks across the modeled fault plane, based on a smoothly curving SSAF orientation and conrmed by model inversions of the geodesy. For example, it would be N47W at the southeast end of the model and N60W at the northwest end. At each point on the fault plane model, a vector with the appropriate orientation (between N47W and N60W) and 4 m. ( 13 ft.) (or 8 m. [ 26 ft.) for

188 Applied Geology in California

Figure 24. Modeled geometry of SSAFS-SB beneath the SGP area.

the worst case and 1 m (3 ft.) for the M6 event) of slip was projected onto the 3D geometry of the fault model. From this vector, the horizontal and vertical components of slip were calculated. For example, if the fault is parallel to the block motion and vertical, (as it basically is at the southeast and northwest ends of the model) one would get 4 meters (13 feet) of purely horizontal slip. As the fault dips and becomes more westerly in trend than the slip vector representing the block motion across it, we get a greater and greater component of vertical slip (and correspondingly less horizontal). For the North Palm Springs-like event we applied 1 m ( 3 ft.) of slip at the base of the seismogenic zone at about 12 km (7.5 miles). In contrast to the two larger fault rupture scenarios, and to capture the lack of clear evidence of fault surface rupture in the SGP area during the past two MW 6 events, we tapered slip from 1 m ( 3 ft.) at 10 – 12 km (6.2 - 7.5 mile) depth to 0 slip at ~2 km (~1.2 mile) depth where the Banning

and Garnet Hill faults bifurcate in the modeled 3D geometry of the SSAFS-SB as shown on Figures 8b and 8c. The 4 m (~13 ft.) and 8 m (~26 ft.) of slip for each respective earthquake was applied along the rupture length within the modeled region in map or plan view (noting that because the fault orientation and dip change from segment to segment the exact value for the horizontal and vertical components change as described above). There is constant slip along the fault’s strike (corrected for the changing fault geometry) so there is no apex in the slip distribution along the strike of the fault. Ideally, we could have modeled a square root of a sine rupture shape, but because we are only interested in slip on the fault very near (within tens of kilometers on a rupture that is hundreds of kilometers long) the CRA crossing, we did not do so. If we were interested in the total deformation eld from Bombay Beach on the southeast shore of the Salton Sea to Wrightwood northwest of San Bernardino and Cajon Pass, we

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

would need to model some sort of tapered rupture pro le for the current project, we simply assume that the tapers near the end of the fault are far from the CRA crossing. With evidence that the empirical relations of Well and Coppersmith (1994) may not apply as well with the strike-slip tectonics of California, and the lack of evidence in California of the amount of coseismic slip on events in the MW 8+ range, we chose to use 8 m (~ 26ft.) of slip, as discussed below under the section and subsection of this report titled Discussion of Model Results and Coulomb 3.3 models. Even with the simplicity in the Coulomb 3.3 modeling software and the limitation in representing the fault’s actual subsurface 3D geometry and its physical properties, the deterministic values of deformation generated by Coulomb 3.3 are considered reasonable to be a representation of the surface deformation from earthquakes generated by 4 m. (~ 13 ft.), or 8 m. (~ 26 ft.) of maximum slip at depth on the SSAF-SF below SGP. The model allows us to evaluate distributed deformation at the surface from slip on the fault at seismogenic depths. The sur cial distribution of both vertical and horizontal components of deformation for prescribed ‘scenario event’ slip on the fault are the key output to evaluate structural performance and inform design along the CRA crossing for the scenario events used in this study. Coulomb 3.3 Interseismic Strain Model An interseismic strain model was developed using Coulomb 3.3, to compare with the interseismic ground deformation recorded with the GPS and InSAR. We made this model to simulate how deep interseismic deformation might manifest itself at the ground surface in a transpressive stepover zone like San Gorgonio Pass. Coulomb is not designed to model interseismic slip or strain; it is a rupture model. However, by placing slip at depth, where creep occurs interseismically, one can get a crude idea of what the interseismic surface deformation might look like based on the use of Coulomb 3.3 software. For this Coulomb 3.3 interseismic strain model it was assumed that the interseismic deformation would occur below 12 km (~7.5 mile) for the 4-m (~ 13 ft.)/12-km (~ 7.5 mile) deep event

189

and below 25 km (~ 15.5 mile) for the 8-meter (~ 26 ft.)/25-kilometer (~ 15.5 mile) deep event. In the 4-m (~ 13 ft/12-km (~7.5 mile) deep interseismic Coulomb 3.3 model we placed constant slip with depth on the fault plane from 12 km. (~ 7.5 miles) to 50 km (~ 31 miles). In the 8-meter (~ 26 ft.)/25-kilometer (~ 15.5 mile) deep interseismic Coulomb 3.3 model we placed constant slip with depth on the fault plane from 25 km. (~ 15.5 miles) to 50 km (~ 31 miles). For these Coulomb 3.3 strain models the interseismic slip was applied for each individual fault segment of the SSAFS (such as the Coachella, Indio Hills, Garnet Hills, etc.). The amount of horizontal (and vertical) slip for each segment was calculated (as discussed above toward the end of the previous section, and illustrated in Figures 22c and 23, and listed in Table 1) based on its 3D geometry and the direction of slip across the fault segment from the block movement across the fault zone. This calculation was done by taking the 4 m. (~ 13 ft.) horizontal displacement vector (or the 8 m. (~ 26 ft.) horizontal displacement vector for the “worst case” scenario) oriented in a progressively curving direction to mimic the rotating block motion (315° for the Coachella section, located southeast of SGP, and 300° for the southern end of the Mojave section, northwest of the modeled area). The Coulomb 3.3 program calculated the surface deformation due to the above scenarios by an Okada/Green’s function approach to solving for the deformation due to a dislocation embedded in an elastic media. Since we are representing an interseismic model, the upper crust, i.e., down to 12 km (~ 7.5 miles) in the 4-m (~13 ft) model and down to 25 km (~ 15.5 mile) in the 8-m. (~ 26 ft) model) does not slip. Theoretically slip at depth simply loads or strains the upper crust, which then suddenly and coseismically slips along the fault plane relaxing the elastic strain in the upper crust around it during the earthquake. Commonly the horizontal extent of the Coulomb 3.3 model boundaries are established to a distance equivalent to several fault depths in order to minimize the programs limitations at its lateral boundaries. The larger horizontal extent also minimizes the effects of topography so that a at ground

190 Applied Geology in California

surface can be assumed when modeling the surface deformation patterns. However, for simulating the ruptures along the SSAFS-SB within the SGP area, this model’s boundaries were extended much farther in each direction along the fault (i.e., about 100 km (~ 62 miles) each way for a total of 200 km) (~ 124 miles) and about 100 km (~ 62 miles) each way across the fault to capture the very broad curvature of the SSAFS through this region. Since the Coulomb 3.3 program simulates surface deformation based on an assigned amount of coseismic crustal slip along a 3D model of the fault beneath the area of interest, the extent of the model boundaries along the fault do not need to correlate with the total length of rupture used to estimate the magnitude or amount of slip during the earthquake. The Coulomb 3.3 program calculates the deformation at each grid point in the fault model (Table 1 and Figure 23) resulting from the slip on each individual portion of the modeled fault, and then adds the contributions up to give the net deformation at that point in the model. Slip on portions of the fault that are farther and farther away from the area of interest contribute less and less to a point’s net deformation, so the slip on portions of the fault distant from the area of interest have trivially small impact on the net result within the model area. The rule of thumb is that your model should extend at least a fault depth or so beyond the region of interest. Also, if one looks at the horizontal deformation eld in the SGP area (compare Figure 10 with Figure 23 and 24), one will see how the vectors start to curve and become nonparallel at the edges and corners of the Coulomb 3.3 modeled area; so as long as these “edge effects” are beyond the area of interest one can be comfortable that the model is large enough. The base boundary conditions were set differently between the interseismic and coseismic Coulomb 3.3 models. Interseismic deformation models imparted slip on the fault below 12 km. (~7.5 miles) (or 25 km. (~ 25.5 miles) for the worst case model) for comparison with geodetic results. Because Coulomb 3.3 is a half-space modeling program, in the interseismic deformation models the slip can be thought to extend inde nitely below 12 km (~7.5 miles), or 25 km (~ 25.5 miles)

in the crust. In the coseismic models a standard fault dislocation model was utilized where the slip was imposed above a locking depth (~12 km (~ 7.5 miles) in the 4 m (~ 13 ft) displacement models, and ~25 km (~ 25.5 miles) in the 8 m (~ 25 ft) displacement models). Other Coulomb 3.3 Parameters The strength properties along the SAFS-SB fault planes are represented in the model with a coef cient of friction (per Byerlee’s law [Byerlee, 1978], 0.4 is typical, where 0.0 to 1.0 are the limits of the crust). The elastic properties that de ne the halfspace, Young’s modulus and Poisson’s ratio (PR), were set to 8 × 105 bars (8 × 1010 Pa) and 0.25 (PR), respectively. A PR of 0.2 to 0.3 is common for most rocks. There is considerable debate about these values; for example, Fialko (2006) argues that these properties vary across the fault. Because the Pelona Schist is known to occur in the study area at depth, one could make some very simple variations in these properties, at depth, across the SAFS-SB and other nearby faults in the modeled area. However, given the limited time and the overwhelming impact of the location and 3D fault geometry on the model results variations in the elastic properties were not pursued. For the larger earthquake rupture scenarios that extend far beyond the modeled area surrounding SGP, we imposed the appropriate amount of slip for the given earthquake magnitude/length on a simple con guration of those portions of the SAFS-SB northwest and southeast of the SGP fault complex. For smaller ruptures that will not extend beyond SGP, we prescribed slip on appropriate sized portions of the fault system, to explore the consequences of different rupture locations. Due to the regional scale of the modeling (i.e., the decreased detail that comes with its 200 km and 100 km boundaries) a number of factors (i.e., thermal, groundwater, microcracking and other factors in uencing the physical properties of the underlying sediments and bedrock) were beyond the scope of this particular study. The Coulomb 3.3 program is set up to effectively utilize different values of the coef cient of friction and normal stress, so the role of uid pressure can be introduced, but it is

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

currently not known how uid pressures vary with depth within the sediments and bedrock at the multi-kilometer depth scale utilized in this particular model con guration. Coulomb 3.3 is a simple kinematic/static stress change model calculator that does not account for dynamic stresses, preexisting stress heterogeneities (i.e. due to past earthquakes), pore- uid diffusion, and viscoelastic rebound. As stated by the USGS (2011) in their introduction, “Further, elastic stiffness differences between basins and crustal layering modify the stresses in comparison to the elastic halfspace implemented in Coulomb. Nevertheless, we believe that a simple tool that permits exploration of a key component of earthquake interaction has great value for understanding and discovery.” The point is that the world is not an elastic halfspace, so a simple Coulomb 3.3 model is a necessary simpli cation of the world useful for exploring the major parameters, like fault geometry and large displacements that occur on fault interfaces. One could make a more complex model that takes into account variations in stresses and elastic properties but 1) we don’t know these variations well enough to con dently add them, and 2) these variations are second order to things like geometry and slip style on faults, so uncertainty in the rst order things would swamp out any value added by trying to ne tune the second order issues. If the distribution of stress in the crust (and the details of subsurface fault geometry and distribution of strength variations) were really known, a more sophisticated modeling approach, as one does for an engineered structure where one knows enough to justify this approach could be applied. Unfortunately, the materials (i.e., geology) and distribution of stress in the Earth, especially beneath SGP, are simply not known well enough to warrant more detailed dynamic simulations like boundary element or nite element models. If attempted, it may generate unrealistic precision. Coulomb 3.3 generates kinematic/static results simply to investigate slip on faults and displacement (horizontal and vertical displacement vectors) at the ground surface on a regional scale. Thus, Coulomb 3.3 is an appropriate platform for estimating surface displacements in the SGP area.

191

Geodetic Displacement Model To place constraints on the rate and style of earthquake slip on the SSAFS and other faults near SGP, we used the interseismic GPS velocity data to constrain models of crustal block motion. During the interseismic period faults that separate crustal blocks are (in most cases) locked at the surface and slipping at depth. We assumed the fault geometry as described above and used the regional GPS data (not just data near the faults) to constrain the relative direction and rates of the motions of the blocks that are separated by the segments of our assumed fault geometry. Through this kind of modeling the data provide strong constraints on the budgets of deformation that occurs over time, and reconciles the measured interseismic deformation with the average long term moment release rate that will occur over many future seismic cycles. The geodetic block motion models rely on the accuracy of assumed fault geometry, and thus uncertainties in that assumed geometry are a part of the uncertainty in estimates of future coseismic slip rate and style of motion along the fault and the coseismic surface deformation in the crustal blocks bordering the fault. Block rotations are estimated simultaneously with slip rates using a methodology that expresses interseismic velocity as rigid block motions minus the average rate of deformation from coseismic slip over time. The geodetic block motion models are inversions of the GPS data and thus present the most likely motion of blocks and slip rates given the data provided. The methodology has been described in many studies (see e.g. McCaffrey et al., 2002; Meade and Hagar, 2005; Hammond and Thatcher, 2007; Hammond et al., 2011). In our model we separate blocks with rectangular dislocation patches that slip in an elastic half-space (Okada, 1985), similar to the Coulomb 3.3 modeling. We use an elastic shear modulus of 3.0 × 105 bars (3.0 × 1010 Pa) and a Poisson’s ratio of 0.25. In the Earth’s crust, these parameters vary with pressure, temperature and mineralogy, and may also possibly vary horizontally across the fault. While the Earth’s crust is clearly more complicated than an elastic half space, the simpli cation is justi ed in light of 1) the scope of this analysis effort, 2)

192 Applied Geology in California

the uncertainties that we have about how the physical properties of the Earth’s crust vary at seismogenic depths in vicinity of the SSAFS at SGP, 3) how these physical properties translate to the elastic moduli that are effective during a large seismic event, and 4) uncertainty in the geometry of the fault that slips inside this medium. Creating a more physically detailed model should be possible in future studies when consensus on these features emerges. Various fault geometries were tested that are consistent with the scenarios considered in the Coulomb 3.3 models, such as the differing amounts of horizontal to vertical slip implied by different geometric con gurations of Model 1 and Model 2 (Figure 18). For this study we have designed two end-member structural-kinematic models of the SSAFS through the SGP area to explore the sensitivity of the models to changes in the fault’s subsurface geometry. Each geodetic block motion model has four blocks, separated by the SSAFS, the San Jacinto Fault Zone, and the Pinto Mountain Fault. As illustrated on Figure 18, the two models are end-members in the sense that Model 1 has the maximum expected thrust component of slip while Model 2 has almost no interseismic thrust component of slip below 8 to 12 km (5 to 7.5 miles). For Model 1, the dip and locking depth of the SSAFS are speci ed by similar structural parameters used in the Coulomb 3.3 modeling shown in Figure 23 (the fault is a segmented set of planar surfaces that intersect the surface along nodes). The fault dips to the northeast along its entire depth including the deep creeping section that extends into the lower crust. This surface trace follows the SSAFS Coachella Valley Banning Fault to the intersection with the SSAFS San Bernardino section, where it bends to the northwest, west of the CRA. Model 2 is similar except that the SSAFS dips 85 (near vertical) below a depth of 8 km to 12 km as shown in Figures 8c and 8d and illustrate in Figure 18. Since the fault turns near-vertically downward near the bottom of the locked dipping section, the surface projection is northeast of the San Gorgonio/Banning Faults, following the Mission Creek and Mill Creek faults before rejoining the

San Andreas Fault and heading northwest. Because these models (Model 1 and Model 2) only account for interseismic deformation they do not estimate where future coseismic deformation may occur, e.g. in the locked sections up-dip of the vertical or dipping deep zones, but the interseismic strain rate remains fastest above the deep slipping section of the fault. We use the interseismic motion obtained from the block models to estimate the coseismic slip that is expected for future scenario earthquakes. The models provide the average slip rate that must be accommodated by earthquakes over time, not coseismic slip for any particular earthquake. However, if we assume a speci c time interval, e.g. the average recurrence interval of maximum magnitude earthquakes, we can estimate the slip on that event to be that which is required to release all strain on the fault that has been stored in the relative motion of the blocks over that time. This is the average earthquake of an assumed magnitude and location on the fault. Smaller earthquakes could occur that release less of the total moment budget, or larger earthquakes could occur e.g., if a longer than average time elapses between events. For the coseismic displacements estimated using the geodetic block motion modeling and discussed in the next section, the input motion is based on the GPSestimated motion of the crustal blocks adjacent to the SSAFS, and not a deterministic value of slip applied to the fault.

Discussion of Model Results The Coulomb 3.3 model runs included the 4-m(~13 ft) slip and 12-km (~7.5 miles) depth earthquake scenario consistent with a MW 7.0 to 7.8 maximum credible event through the pass as discussed in the Methodology section, the “worst case” 8-m (~26 ft) slip and 25-km (~15.5 miles) depth scenario consistent with a MW 8.5 event, and the smaller more frequent MW 6.0 to MW 6.5, 1-m (~3 ft) slip and 12-km (~7.5 miles) event similar to the 1948 MW 6.2, Desert Hot Springs and 1986 MW 6.1 Palm Springs earthquakes; all applied on the SSAFS-SB in a progressively curving direction from the southeast to northwest. The progressively curving direction is supported by the rotation of principal strains

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

along the fault evident in the geodetic data (Figure 10 and 12) and rst documented by Weldon and Humphreys (1986). The input geometries and conditions were also checked in the geodetic block motion modeling software using GPS data to assess if the applied fault subsurface geometry is consistent with the ongoing interseismic accumulation of strain. Coulomb 3.3 Model To convert the earthquake scenarios presented in this report into rupture displacements that can be modeled in Coulomb 3.3 (and other elastic halfspace models) to quantify surface deformation, one must rst understand the empirical relationships between magnitude, rupture length, and average displacement. The largest earthquake in California’s historic record is the M7.7-7.9 1906 San Francisco Earthquake, which ruptured 477 km (~296 miles) of the northern San Andreas Fault System. The next largest was the MW7.6-7.8 1857 Fort Tejon Earthquake that ruptured 334 km (~208 miles)of the SSAFS. Widely cited scaling relationships (such as Wells and Coppersmith, 1994) that are generally based on linear regressions of small datasets dominated by ruptures shorter than 200 km (~124 miles) suggest that the 1857 event should have an average slip of more than 6 m (~20 ft) and 1906 more than 8 m (~26 ft) (Figure 25) whereas the actual average values are just over 4 m (~13 ft). Modern scaling relationships (e.g. Shaw, 2009 and 2013) and the California data set (Figures 25 and 26) suggest that ruptures growing up to lengths of ~200 km (~124 miles) rapidly increase in their average displacement; however the rupture lengths >200 km (~124 miles) do not signi cantly increase the average displacement above 4 m (~13 ft). Thus, for scenario earthquakes that exceed MW ~7.3 or 215 km (~134 miles) we would expect an average rupture displacement of ~4 m (~13 ft). We take as our “worst case” scenario a MW ~8.5 earthquake with 540 km (~336 miles) of rupture with average slip of twice this, 8 m (~26 ft). This earthquake is considered very unlikely because: 1) it would require rupture of the entire SSAFS from the Creeping Section in Central California near Park eld to the southern end of the fault near the

193

Salton Sea at Bombay Beach; and 2) it would require scaling, such as proposed by Wells and Coppersmith (1994), that few (if any) modern investigators believe. An alternative way to look at a hypothetical 8 m (~26 ft) displacement could be to consider variability of displacement along the strike of the fault. Given a likely 4 m (~13 ft) average displacement, perhaps it is possible to have a larger displacement locally along some small portion of the fault (such as the CRA crossing point). To address this possibility, consider Figure 27, which shows how displacement varies along strike for an average of strike-slip ruptures longer than 200 km (124 miles) (UCERF3, Field et al., [2014). The dashed lines show the ±1 sigma range in the along-strike displacement relative to the average; the 2 sigma range (not shown) is approximately twice this. Thus, at low probability, it is possible to have a sustained stretch of an individual rupture (up to ~10% of the rupture’s length) where the actual displacement is twice the average. So, even with a likely 4 m (~13 ft) average displacement for a through-going southern California rupture, a 5 km (~ 3 mile) stretch of the rupture that could fortuitously include the CRA crossing at the 2 sigma level of uncertainty could approach 8 m (~26 ft) of slip. In an attempt to better match the interseismic pattern of uplift and assess which model for the deep geometry of the San Andreas is most likely to be correct, we made a model in Coulomb to simulate deep interseismic creep that is largely lateral, parallel to the SSAFS, but also contains a component of compression across the fault which is consistent with the block motions across the fault zone. Because compression (or extension) across a plane is not possible in a simple modeling program like Coulomb 3.3, we introduced it by placing oppositely dipping faults (45° perpendicular to the local strike of the SSAFS, see Figure 28a and b) at depth symmetrically about the vertical deep trace of the SSAFS (the dashed green line in Figure 29) and input slip values consistent with our 4-meter (13-ft) and 12-kilometer (7.5 ft) model (see Figure 28) along strike at a depth of 25 to 50 km (15.5 to 31 miles). This

194 Applied Geology in California

Figure 25. Displacement/Rupture Length scaling for California earthquakes. Blue diamonds are for well-studied historic California earthquakes, red boxes are for early historic ruptures studied with LiDAR (gray boxes connected with blue lines are alternative interpretations of the earthquakes’ lengths). California ruptures longer than 100 NP§PLOHV IDOOEHORZWKHZLGHO\FLWHG:HOOVDQG&RSSHUVPLWK UHODWLRQVKLS%HFDXVHRIWKLVDQGRWKHU discrepancies, including outdated datasets, such as Wells and Coppersmith [1994]) UCERF3 used the scaling relationships of Shaw (2009, 2013) which are consistent with the orange line.

can be thought of as the amount (and correct style) of interseismic strain required to produce our preferred 4-meter (~13 ft) and 12-kilometer (~7.5 mile) coseismic displacement model. The pattern of interseismic strain accumulation generated by this Coulomb 3.3 approximation (as shown in Figure 29) is quite similar to the observed pattern discussed in the subsection on GPS/InSAR data (Figure 13 and 15). The problem with the Coulomb 3.3 model of the interseismic strain accumulating across a deep SSAFS, is that it produces a thrust fault/subduction zone type of interseismic uplift pattern, with high uplift rates on the northeast side of the fault and subsidence on the southwest, and not the observed pattern shown in Figures 13 and 15. In addition, if our preferred 4-meter (~13 ft) and 12-kilometer (~7.5

miles) displacement model is the dominant rupture mode on the SSAFS we can scale the interseismic uplift by the time it would take to generate 4 m (~13 ft) of elastic strain, approximately 200 years, and determine the uplift rate by dividing our hypothetical model’s uplift by 200 (0.5 m (~1.6 ft)/200 yrs) to estimate a peak interseismic uplift rate of 2.5 mm/yr (0.1 inches), consistent with the GPS/InSAR data. However, accounting for the effects of gravity would reduce the uplift rate estimated from the model. This correction can be as high as 90% for broad regions in isostatic balance (Bowie, 1922), but the actual correction will be less owing to exural rigidity that supports a portion of relief generated by uplift. Furthermore the viscous relaxation time scale of the underlying mantle and time since recent

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

195

Figure 26. Documented California single event offsets. The most recent compilation of single event offsets from KLVWRULFDQGUHFHQWSUHKLVWRULFUXSWXUHVGHPRQVWUDWHVWKDWZHOOFKDUDFWHUL]HGUXSWXUHVJUHDWHUWKDQP§IHHW  DUHVLPSO\QRWVHHQ2IHYHQWVLQWKHGDWDEDVHWKHUHDUHSHUFHQW JUHDWHUWKDQP§IHHW DQG both of them may have been caused by multiple earthquakes. It should be noted that it is widely (but mistakenly) EHOLHYHGWKDWWKHUXSWXUHRQWKH6DQ$QGUHDV)DXOWZDVDVVRFLDWHGZLWKaP§IHHW RIIVHWVLQWKH Carrizo Plain from the early work of Sieh (1978); it has since been shown that those offsets are caused by repeat UXSWXUHVZLWKP§IHHW RIVOLS=LHONHHWDO 

Figure 27. 5XSWXUHSUR¿OHVKDSHIRUVWULNHVOLSHDUWKTXDNHV6ROLGGDUNOLQHLVWKHDYHUDJHRIWKHEHVWVWXGLHGVWULNH VOLSHDUWKTXDNHVZLWKOHQJWKJUHDWHUWKDQNLORPHWHUV§PLOHV SUHYLRXVVWXGLHVRIODUJHUGDWDVHWVRIOHVVZHOO characterized earthquakes show the same results). Dashed line is a standard deviation of the offsets and the red line is DVTXDUHURRWRIVLQH¿WWRWKHGDWDRIWHQXVHGIRUPRGHOLQJSXUSRVHVHJ8&(5))LHOGHWDO>@ 7KHGDWDDUH QRUPDOL]HGE\OHQJWKDQGDYHUDJHGLVSODFHPHQWVRIRUH[DPSOHDNP§PLOHV ORQJUXSWXUHZLWKDYHUDJH GLVSODFHPHQWRIPHWHUV§IHHW ZRXOGKDYH PHWHUV§IHHW RIRIIVHWDWLWVFHQWHUSRLQW WDSHULQJWRPHWHUV §IHHW DWNLORPHWHUV§PLOHV IURPLWVHQGVDQGPHWHUVDWHDFKHQG

196 Applied Geology in California a.

b.

Figure 28. Coulomb 3.3 model of deep interseismic creep. 28a because deep trace is steep, there must be interseismic convergence at depth to drive oblique coseismic slip observed during earthquakes. Convergence shown here as broad zone of distributed deformation away = ⊗ and towards =‫ݪ‬WKHSDJH%HFDXVH LVGLI¿FXOWWRPRGHOZHFRQVLGHUHGWKHYHFWRUVROXWLRQLQ  and divided the total convergent vector into strike slip (SS) and compression. Then, in the models we represented the interseismic compression along the deep vertical trace as two symmetrical, dipping reverse faults as shown in (3) on 28b and the coseismic deformation with oblique slip on the thrust and lateral slip on the deep vertical trace as shown in (4) on 28b.

earthquakes will play a role in how long the crust needs in order to nd balance that accounts for gravity and exure, and the rate of contemporary uplift measured geodetically. As a result we do not expect exact agreement between measured vertical rate and that estimated from mean event

displacement and recurrence interval, but nd the similarity encouraging. The preferred model for through-going slip on the SSAF through the study area was 4 m (~13 ft) of right-lateral slip, applied in a direction that continuously varies from parallel to the SSAFS

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

197

Figure 29. Coulomb 3.3 interseismic model: vertical surface deformation. In an attempt to better match the interseismic pattern of uplift (as shown on Figure 15) and assess which model for the deep geometry of the San Andreas is most likely to be correct, we made a model in Coulomb 3.3 to simulate deep interseismic creep that is largely lateral, parallel to the SSAF, but also contains a component of compression across the fault that is consistent with the block motions across the fault zone. Because compression (or extension) across a plane is not possible in a simple modeling program like Coulomb, we introduce it by placing oppositely dipping faults (45 degrees perpendicular to the local strike of the SSAF) at depth symmetrically about the vertical deep trace of the 66$)GDVKHGJUHHQOLQH DQGVOLSWKHPE\DPRXQWVFRQVLVWHQWZLWKRXUPHWHUV§IHHW PRGHODORQJVWULNHDW DGHSWKRIWRNP§PLOHV 7KLVFDQEHWKRXJKWRIDVWKHDPRXQWDQGFRUUHFWVW\OH RILQWHUVHLVPLF VWUDLQUHTXLUHGWRSURGXFHRXUSUHIHUUHGPHWHUV§IHHW FRVHLVPLFGLVSODFHPHQWPRGHO

in the Coachella Valley to the south to parallel to the Mojave SSAF to the north. The horizontal and vertical components of the 4 m (~13 ft) of right-lateral slip were projected onto the panels re ecting the 3D modeled geometry of the fault (Figures 23 and 24) according to the orientation of the panels. Surface deformation is calculated by Coulomb 3.3 by imposing the slip on each element of the model. Seismogenic depth is 12 km (~7.5 miles); the SGPTF and the GHF dip 30° from 0 – 2 km (0 - ~1.2 mile) depth (top – bottom depths de ning fault plane width) where they join the SB-Banning Fault, which dips 45° from 2 km (~1.2 mile) – 12 km (~7.5 miles) depth. The

SB-Banning Fault width varies from 2 km (~1.2 mile) – 8 km (~5 miles) from east to west, longitude -115.6° to -116.4°; 2 km (~1.2 mile)-10 km (~6 miles) depth from -116.4° to -116.6°; and 2 km (~1.2 mile) – 12 km (~7.5 miles) from -116.6° to -116.8°, and the SB-Banning Fault joins the deep trace of the SSAFS at these bottom depths. Because the SGPTF and GHF connect to the deeper SB Banning Fault and in a rupture model have the same amount and style of slip at their junction at ~ -2 km (~1.2 mile) they were assigned the same lateral and vertical slip as the SB-Banning Fault rather than projecting the regional slip vector onto those planes like all the others.

198 Applied Geology in California

The results of the 4-meter (~13 feet) and 12-km (~7.5 mile) displacement Coulomb 3.3 model run (see Figure 30a) show that upward vertical deformation is concentrated north of the most westerly striking portion of the SSAFS. Three areas of relatively high uplift occur along the CRA route; right at the surface trace of the SSAFS, about 10 km (~6 miles) west (along the CRA route) above where the SB-Banning Fault and the GHF join at depth, and northeast of the southeast termination of the SB-Banning Fault and signi cant bend of the SSAFS to the south near the Indio Hills. The horizontal components of deformation of the 4-meter (~13 ft) and 12-km (~7.5 mile) model results are presented on Figure 30b. We note a large horizontal shear between the SB-Banning and the deep trace of the SSAFS (below the SSAFS-NB). The displacement vectors highlight that this region is essentially a shallowly dipping ap that roots into the deeper fault, and highlights that this region will be subjected to intense shear and likely broadly distributed lateral faulting, as documented in the mapping of Yule and Sieh (Figures 5a through 5c). Curvature in the displacement vector eld at ends of SSAFS plots are due to ends of the model. The vertical and horizontal surface deformation pro les along the CRA from the 4-meter (~13 ft) and 12-km (~7.5 mile) model are shown in Figure 30c. These pro les were plotted by taking the nodal point results from the Coulomb 3.3 model’s km grid (shown in Figure 23) and associating these km grid nodal point results with the closest lat/ long coordinate of the CRA alignment provided by Metropolitan. The results show a high frequency of scatter in the points on the pro les as the smoothly varying model is projected onto irregularly spaced points along the CRA. In the upper pro le in Figure 30c we can see the 3 regions of vertical uplift shown in Figure 30a. As seen in the upper pro le, the vertical surface fault displacement of the SSAFS, here represented by the SB-GHF (CRA/SSAFS crossing point 6), is just under 1 m (~3 ft) compared to almost 4 m (~13 ft) of horizontal surface fault displacement seen in the lower pro le on Figure 30c. In addition to the displacement at the GHF (CRA/SSAFS crossing point 6) there is a very steep gradient in

the vertical component in the approximately 5 km (~3 miles) between the SB-GHF and the SB-Banning Fault. There is also a 25 km (~15.5 mile) length of the alignment west of the fault that subsides (up to 20 cm or ~0.7 ft along the SB-GHF). Figure 30c (lower panel) shows the approximately 4 m (~13 ft) of right-lateral shear across the SB-GHF fault and a ~20 km (~12 mile) wide zone of high horizontal shear gradient on each side of the fault, which could result in horizontal displacements of up to a meter if the shear is concentrated into a discrete fault zone. As noted by Yule and Sieh (2003), the surface trace of the SSAFS is not sharp and should not be viewed as a discrete fault surface. Detailed eld investigations will be needed to determine the exact location and style of shearing within this fairly wide zone. As seen in Figure 31a, the Coulomb 3.3 model results from the “worst case” 8 m (~26 ft) and 25 km (~15.5 mile) scenario have twice the horizontal and vertical displacement values of the 4 m ( 13 ft) and 12 km (~7.5 mile) scenario. The rupture depth was extended two ways; rst, by extending the dipping fault in the upper crust to 25 km (~15.5 mile) (Figure 18, Model 1), and second (Figure 18, Model 2) by extending the rupture onto a deep vertical trace from 12 km (~7.5 mile) to 25 km (~15.5 mile) constructed as discussed above. The two models were essentially identical in shape and basically double the surface deformation of that of the 4 m (~13 ft) model. Here we show the results of Model 2 because it produced slightly greater deformation, and thus is the “worst case”, probably due to the fact that the deformation is more concentrated below the surface trace. The only signi cant difference beyond doubling the amplitude of deformation in the 4 m (~13 ft) model is an increase in the very broad wavelength deformation. In Coulomb 3.3 the crust is modeled as a linear elastic material. As such, modeled deformation values will scale linearly with displacement. Thus, one could just decide on any displacement desired for the SSAFS and scale the resultant deformation from the preferred 4 m (~13 ft) model. In addition, the similarity of the 4 m (~13 ft) and 8 m (~26 ft) models demonstrates that the depth of fault rupture only minimally affects the model, because surface

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

199

a.

b.

(continues)

200 Applied Geology in California c.

Figure 30. Maximum considered event Coulomb 3.3 model results. 30a - Coulomb 3.3 vertical surface GHIRUPDWLRQPRGHOP§IHHW DQGNP§PLOHV XSWR0W ~7.8) scenario. Thin black lines are Quaternary-active faults from USGS Quaternary Fault and Fold Database (QFFD: USGS. 2010). Datum/ projection is NAO 1983, UTM11 North. Longitude/latitude tick marks are approximate. Standard (normalized) color bar scale utilized for comparison purposes. 30b - Coulomb 3.3 horizontal surface deformation model: P§IHHW DQGNP§PLOHV XSWR0Wa VFHQDULR/LJKWJULGOLQHVDUHNP§PLOHV VSDFLQJ Datum/projection is NAO 1983, UTM11 North. Longitude/latitude tick marks are approximate. 30c - Coulomb PRGHOP§IHHW DQGNP§PLOHV 8SWR0Wa VFHQDULRSUR¿OHVDORQJ&5$URXWH

deformation is strongly driven by the near surface slip on the faults. Since the 4 m (~13 ft) and 8 m (~26 ft) models have identical geometry down to approximately 12 km (~7.5 miles), extending the depth of rupture, whether dipping or vertical, only produces a very long wavelength (distributed) affect at the surface. Like the vertical deformation, the horizontal deformation (Figure 31b) is approximately double that of the 4 m (~13 ft) model. However, this doubling has a more signi cant spatial impact in the horizontal than the vertical because it extends high shear farther from the fault zone. This can be seen more clearly in the pro les (Figure 31c). As noted in Figures 31a and b, the pro les are very similar in shape to the 4 m (~13 ft) model and the amplitude of uplift (or subsidence) is essentially doubled, with an offset at the fault of almost

2 meters (~6.5 ft). However, because the width of signi cant deformation for the vertical component is approximately the same the gradients are also doubled. For the horizontal component, however, not only are the offsets and gradients doubled, but the width of signi cant shearing is much greater for the 8 m (~26 ft) model. In the 4 m ( 13 ft) model displacement greater than 1 m (~3 ft) occurs within 60 km (~37 miles) of the fault to the northeast but in the 8 m (~26 ft) model it is essentially doubled to 120 km (~75 miles). This may generate secondary shear faulting to the CRA route where it is subparallel to the SSAFS for almost 100 km (~62 miles) (Figure 31b). The most likely event in the region will be a MW 6-MW 6.5 earthquake on the deep part of the SB-Banning Fault, like the 1948 Desert Hot Springs and 1986 North Palm Springs earthquakes.

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

201

a.

b.

(continues)

202 Applied Geology in California c.

Figure 31. Worst-case event Coulomb 3.3 results. 31a - Coulomb 3.3 vertical surface deformation model: P§IHHW DQGNP§PLOHV XSWR0W 8.5) scenario. Thin black lines are Quaternary-active faults from USGS Quaternary Fault and Fold Database (QFFD: USGS, 2010). Datum/projection is NAO 1983, UTM11 North. Longitude/latitude tick marks are approximate. Standard (normalized) color bar scale utilized IRUFRPSDULVRQSXUSRVHVE&RXORPEKRUL]RQWDOVXUIDFHGHIRUPDWLRQPRGHOP§IHHW DQGNP §PLOHV XSWR0W 6FHQDULR/LJKWJULGOLQHVDUHNP§PLOHV VSDFLQJ'DWXPSURMHFWLRQLV1$2 8701RUWK/RQJLWXGH/DWLWXGHWLFNPDUNVDUHDSSUR[LPDWHF&RXORPEPRGHOP§IHHW   NP§PLOHV 8SWR0W VFHQDULRSUR¿OHVDORQJ&5$URXWH

It will probably not be associated with surface displacement large enough to signi cantly affect ow in the CRA and will have approximately a meter of slip at depth tapering to zero slip at the surface. We placed this event immediately northwest of where the aqueduct crosses the SSAFS, because it is the most likely segment to rupture given the progression of the 1948 and 1986 events. Slip on all faults in the model is set to zero, except the Banning Fault from east to west, -116.55° to -116.73° longitude. This fault section dips 45° degrees, and extends from 2 km (~1.2 miles) to 12 km (~7.5 mile) depth. We subdivide the fault into six 2 km (~1.2 mile) sections from 0-2 (~1.2 mile), 2 km (~1.2 mile)-4 km (~2.5 mile), 4 km (~2.5 mile)-6 km (~3.7 mile) etc. to 12 km (~7.5 mile) depth. Lateral and vertical slip on the fault is zero at the surface, and increases

incrementally up to 1 m (~3 ft) horizontal, and 0.58 m (~2 ft) vertical on the 10 km (~6 mile) to 12 km (~7.5 mile) depth section. This horizontal to vertical ratio is based on the projection of block movement onto this portion of the SSAFS. The surface deformation associated with this type of event is extremely small, in part due to its fairly low magnitude of slip and the fact that most of the slip is lateral, but most importantly due to the depth of most of the slip. The resulting event produces a 10 km bull’s eye of uplift with a peak of about 10 cm (see Figure 32a). Like the vertical component of this model the surface horizontal deformation (Figure 32b) is in the cm range and largely displaced to the north of the CRA route due to the locus of most signi cant slip at the base of the Banning Fault. The classic

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

a.

b.

Figure 32. Most frequent event Coulomb 3.3 results. 32a - Coulomb 3.3 vertical surface deformation PRGHOP§IHHW DQGNP§PLOHV 0W 6.0 to MW 6.5) scenario. Thin black lines are Quaternary-active faults from USGS Quaternary Fault and Fold Database (OFFD: USGS, 2010), Datum/projection is NAO 1983, UTM11 North. Longitude/Latitude tick marks are approximate. Standard (normalized) color bar scale utilized for comparison purposes. Deformation scenario is similar to the 1948 Mw 6.2 Desert Hot Springs earthquake and the 1986 Mw 6.1 North Palm Springs earthquake. E&RXORPEKRUL]RQWDOVXUIDFHGHIRUPDWLRQPRGHOP§IHHW DQGNP§PLOHV 0W 6.0 to MW 6.5) Scenario./LJKWJULGOLQHVDUHNP§PLOHV VSDFLQJ'DWXPSURMHFWLRQLV1$2870 North. Longitude/latitude tick marks are approximate. Deformation scenario is similar to the 1948 Mw 6.2 Desert Hot Springs earthquake and the 1986 Mw 6.1 North Palm Springs earthquake.

203

204 Applied Geology in California

“double couple” pattern is due to the limited extent of the rupture. Because the deformation was so small we did not make pro les of this event. The horizontal displacements that result from the Coulomb 3.3 model (Figures 30b and 31b) can be used to estimate the rake and amount of coseismic slip across the fault by comparing vectors on either side of the fault. In addition, the region of large horizontal motion predicted along the northern edge of the SSAFS-SB would be expected to produce many secondary faults. Yule and Sieh (2003) have mapped such structures in this area (e.g., the Cox Ranch Fault and other sub-parallel faults), these are superimposed on the Coulomb 3.3 result on Figure 33a and b. We also compared the Coulomb 3.3 results to uplifted geomorphic surfaces between faults and what little is currently known by the geoscience community about 3D slip per event on faults in SGP for ground truth. Yule and Sieh (2003) provide the most precise mapping of fault offset and uplifted geomorphic surfaces in SGP as illustrated in Figure 7b. In Figure 34, we compared the geomorphic surface mapping in Yule and Sieh (2003) with the Coulomb 3.3 - 4 m (~13 ft) model’s vertical displacement. This was simply to qualitatively compare the modeled coseismic slip and surface deformation to markers of longer term uplift adjacent to the SSAFS. As seen on Figure 34, the results of the Coulomb 3.3 modeling are very similar with the existing geomorphic data, suggesting that the subsurface 3D geometry of the SSAFS-SB and the 4 m (~13 ft) of slip utilized in the Coulomb 3.3 model is consistent with naturally occurring uplift. Geodetic Displacement Model The geodetic block models show that, as expected from earthquake geology, most of the slip expected on the SSAFS in future earthquakes will be predominantly dextral strike slip. Figure 35 shows that the dextral slip rates on all faults are higher (between 13 and 22 mm/yr, (~ 0.5 and 0.87 inches) depending on the fault geometry at any given latitude) than the thrust rates. The expected thrust/normal rates, which depend on the local strike of the fault, are generally less than 5 mm/yr (~ 0.2 inches/yr) but are highest (a maximum of 9 mm/yr;, 0.35 inches/yr) through

SGP in Model 1. There are even some sections where the expectation for fault slip is oblique normal where the strike of the fault changes to make a right step, e.g. in Model 1 where the SSAFS-SBBanning Fault transitions into the San Bernardino section of the SSAFS. The geodetic block motion modeled coseismic slip is obtained by scaling the maximum coseismic slip rate obtained in the block model by the amount of time necessary to obtain a maximum of 4 meters (~13 ft) of slip. This strategy has the advantage of honoring geodetic constraints on active motion of adjacent crustal blocks and estimates the average slip needed to release strain accumulation, predicting an earthquake that may not have occurred in the historic past. For the maximum credible 4 m (~13 ft) event on the SSAFS-SB, the result of the modeling shows <1 2 m (< ~1 ft) of vertical ground deformation (Figures 36a,c and d) and ~2 1 2 to ~2 3 4 m (~8 ft to ~9 ft) of horizontal slip (Figures 36b,c and d) along the CRA. The 8 m (~26 ft) maximum slip block model shows ~6 m (~20 ft) of horizontal slip and <2 m (~6 ft) of vertical slip along the CRA (Figures 37a, b, c and d). The slip based on the block Model 1 shows considerable convergence in both the horizontal and vertical motions across the Coachella strand of the SSAFS that does not seem likely given the lack of such convergence seen between the Indio and Mecca hills (note the topographic expression of the hills are associated with bends in the fault, not regional compression across the fault). This convergence was not produced in the Coulomb 3.3 model, and the difference is due to different driving mechanisms in the models. In the Coulomb 3.3 model, we projected 4 m (~13 ft) of 8 m (~26 ft) of slip onto the fault with the direction smoothly curving from the southeast to northwest. This is an appropriate method because we know the SSAFS is basically pure strike slip along its arcs south of 3 Points, with the convergence localized across SGP due to the left-step in the fault zone (see Figure 1). An additional, small difference between the geodetic block motion model and the Coulomb 3.3 model results is related to the input geometries. In

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

a.

b.

Figure 33. Modeled horizontal offsets versus mapped shear zones in region of predicted horizontal motion during earthquake. 33a – Maximum considered event. 33b – Worst-case event. Vector arrows show distribution of magnitude and direction of horizontal displacement. *UHHQOLQHUHSUHVHQWVPRGHO¶VVLPSOL¿HGVXUIDFHWUDFHUHGOLQHVDUHIDXOWVVKHDUVIROGVIURP Yule and Sieh (2003).

205

206 Applied Geology in California

Figure 34. &RPSDULVRQRI&RXORPEPRGHOP§IHHW DQGNP§PLOHV XSWR0W ~7.8) scenario vertical deformation results with geomorphology and surface features. Image shows the vertical displacement JHQHUDWHGE\WKHPHWHU§IHHW PRGHOVHH)LJXUHD ,PDJHLVDPRGL¿HG)LJXUHERI
the Coulomb model, we added bridging surfaces where changes in strike and dip were discontinuous. This effort smoothed out the discontinuities and removed effects produced by kinks and sharp bends in the SSAFS. Notwithstanding such effects, the differences in the displacement elds produced by the two different models are minimal, and the horizontal and lateral motions agree well. We also compared the mapped uplifted geomorphic surfaces from Yule and Sieh (2003) to the vertical deformation results from the geodetic block motion model of the most likely 4 m (~13 ft) slip event. This comparison is shown on Figure 38 and the region of relative uplift from the Coulomb 3.3 results (Figure 34) is very similar to the slope of the existing geomorphic surfaces suggesting that the subsurface 3D geometry of the SSAFS-SB used in our modeling is an

accurate representation of the active fault that is likely to slip in the next major earthquake and poses a hazard to the CRA.

Conclusions and Limitations Metropolitan geologists and engineers realized and accounted for the risks of crossing the active SSAFS in the SGP area during the design and construction of the CRA back in the 1930s and early 1940s. Over the decades since then there has been considerable debate over alternative interpretations of the characteristics and properties of the SSAFS in this area and likely surface deformation associated with future seismic events. These alternative interpretations are due to the unique complexity of the geology and characteristics of the SSAFS through the SGP area and the limitations in each of the geoscience, seismology and geodetic disciplines used

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

207

Figure 35. Two Geodetic Block Motion models of interseismic deformation. Model 1 has fault dipping northeast WKURXJKRXWWKHFUXVW0RGHOKDVQHDUYHUWLFDOGLSSLQJIDXOWEHORZNP§PLOHV GHSWKHJVHH)LJXUH 

to collect and interpret relevant data. However, in the last couple of decades signi cant advances have been made in our understanding of the faults in the area and their future seismic risks. In this study, existing geologic mapping, seismicity, paleoseismology, geodesy, and SSAFS rupture scenarios were used to construct simpli ed 3D geometries and likely displacement amounts and directions for the SSAFS. These geometries and displacements were used to construct elastic models to investigate how the ground surface would respond to slip on the SSAFS. The results of these computer model runs were then reviewed against the available geoscience data from SGP. The ruptures used in the models are comparable with California statewide seismic risk assessment (WGCEP, 2013; e.g., The Uniform California Earthquake Rupture Forecast, Version 3 [UCERF 3] dataset).

Based on the results of this study, the following conclusions have been reached: 1. Review of the local SGP geoscience data suggest that the most likely large earthquake capable of surface ruptures and ground surface deformation across and along the CRA in SGP would be a 4 m (~13 ft) slip event, such as a Wrightwood to Bombay Beach MW 7.0 to MW 7.5 event or a 3-Point to Bombay Beach MW 7.7 to MW 7.8 event; the latter event would have displacements consistent with the dynamic ground motion simulations used by Jones et al., (2008) and Perry et al., (2008) in the November 2008 ShakeOut Scenario and modern displacement scaling relationships (UCERF 3). Earthquakes of this size passing through SGP could have an average recurrence interval of ~500 to 800 years based on

208 Applied Geology in California a.

b.

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

209

c.

d.

Figure 36. 0D[LPXPFRQVLGHUHGHYHQWPRGHOUHVXOWVEDVHGRQVFHQDULRZLWKP§IHHW RIPD[LPXPVOLSGRZQ WRNP§PLOHV GHSWKXSWR0W ~7.8). The slip vector is derived from fault geometry and relative motion of blocks adjacent to SSAFS constrained by geodetic data through block model. 36a coseismic vertical displacement. Thin black lines are Quaternary-active faults from USGS Quaternary Fault and Fold Database (QFFO; USGS,  E&RVHLVPLFKRUL]RQWDOGLVSODFHPHQW3UR¿OHDQG&5$V\PEROVDUHVDPHDVLQDF3UR¿OHVRI vertical and horizontal coseismic displacement along CRA route. 36d - Vertical and horizontal displacement from PRGHOZLWKYDOXHVH[WUDFWHGDORQJUHGEODFNDQGEOXHSUR¿OHVFURVVLQJ6*3DUHDDVLQDDQGE

210 Applied Geology in California a.

b.

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

211

c.

d.

Figure 37. :RUVWFDVHHYHQWEDVHGRQVFHQDULRZLWKP§IHHW RIPD[LPXPVOLSWKURXJKRXWVHLVPRJHQLFFUXVW (up to MW 8.5). The slip vector is derived from fault geometry and relative motion of blocks adjacent to SSAFS constrained by geodetic data through block model. 36a - Coseismic vertical displacement. Thin black lines are Quaternary-active faults from USGS Quaternary Fault and Fold Database (QFFO; USGS, 2010). 37b - Coseismic KRUL]RQWDOGLVSODFHPHQW3UR¿OHDQG&5$V\PEROVDUHVDPHDVLQDF3UR¿OHVRIYHUWLFDODQGKRUL]RQWDO coseismic displacement along CRA route. 37d Vertical and horizontal displacement from model, with values H[WUDFWHGDORQJUHGEODFNDQGEOXHSUR¿OHVFURVVLQJ6*3DUHDDVLQDDQGE

212 Applied Geology in California

Figure 38. Correlation between modeled coseismic uplift and regional geomorphology. Shown is vertical FRVHLVPLFGLVSODFHPHQWIURPPRGHOGULYHQE\*36PHDVXUHGEORFNPRWLRQVHH)LJXUHD ZLWKP§IHHW RI PD[LPXPVOLSGRZQWRNP§PLOHV XSWR0W 7.8) with geomorphology and surface features superimposed. ,PDJHLVPRGL¿HGYHUVLRQRI)LJXUHERI
paleoseismic investigations just completed in SGP to the west of the CRA (Yule and Sharer 2014, in preparation). According to this paleoseismic data the last earthquake of this size occurred in SGP around 1400 A.D. The results of the Coulomb 3.3, 4 m (~13 ft) slip model suggest that at a depth of 100s of meters to a kilometer below the surface, ~1 m (~3 1/3 ft) of vertical and ~ 33/4 m (~12 ft) of horizontal fault displacement could occur where the CRA crosses the SSAFS-SB GHF during this future earthquake. The modeled maximum vertical coseismic offset of ~2 m (~6.5 ft), in the portion of the SGP area to the west of the CRA, is generally near the uplifted late Pleistocene geomorphic surfaces mapped in the area (Figure 6b)

(Yule and Scharer, in preparation). Depending on how this fault displacement propagates to the surface, it could result in surface rupture at the inverted siphon at CRA SSAFS-SB Banning Fault crossing point #5, or at CRA SSAFS-SB GHF crossing points #6. It is possible that most of this displacement would propagate to the surface along the GHF especially if the GHF extends to the west and connects with the SGPTF as discussed in the Geoscience subsection and illustrated in Figure 6b. This could affect the buried canal portion of the CRA between the Cottonwood Canyon stream channel and the I-10 Freeway. The displacement might also be distributed in a broad zone including subsidiary faults

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

bordering both the Banning Fault and the GHF and the region in between. This could also affect the buried canal and tunnel between the Cottonwood Canyon stream channel and Whitewater Canyon as well as the inverted siphon at CRA SSAFS-SB Banning Fault crossing point # 5. Our simulations of coseismic surface displacement based on 4 m (~3 ft) slip indicate that the amount of vertical ground displacement along the CRA would progressively decrease to negligible values of subsidence ~30 to 40 km (~19 to 25 miles) downstream of the peak at the CRA SSAFS-SB GHF crossing points (Figure 30c). In the upstream direction displacement becomes negligible ~ 60 km (~37 miles) from the crossing point. Two additional peaks in the vertical uplift along the CRA upstream of the SSAFS-SB GHF crossings are seen in Figure 30c, one ~ 2/3 m (~2 ft) high just upstream of the CRA SSAFS-SB Banning Fault crossing #5 and the other ~ m (~1 2/3 ft) high, ~30 km (~19 miles) farther upstream near CRA fault crossing point # 1. The horizontal ground deformation in the 4 m (~13 ft)modeled taper off to ~1/3 m (~1 ft) by the end of the model at CRA Station 5894+29.29 (Mile Marker 122.54) as shown in Figure 30c. 2. Based on current geoscience data along the SSAFS and consistent with the work by WGCEP (2013), a “worst-case” 8 m (~26 ft) slip earthquake, such as an imagined MW 8.5, Park eld to Bombay Beach event, is very unlikely to occur. However, it is worth exploring such a model to place an upper constraint on possible ground surface deformation. If this “worst-case” hypothetical earthquake were to occur, the modeled results indicate such an event would generate ~2 m (~6 ft) of vertical uplift and ~7 m (~25 ft) of horizontal displacement across the SSAFS-SB GHF (Figures 31a-c, and 37a-c). This deformation would extend to a depth of at least 100s of meters to a kilometer below the surface. Similar to the 4 m ( 13 ft) model results,

213

depending on how this slip propagates to the surface, it would result in surface rupture at CRA SSAFS-SB Banning Fault crossing point # 5 and CRA SSAFS-SB GHF crossing points #6; or this amount of slip could be distributed between the two faults and subsidiary faults around them. The model results also indicate that along the CRA the amount of vertical ground subsidence during this 8 m (~26 ft) modeled earthquake would progressively decrease away from the CRA SSAFS-SB GHF crossing points # 6 for ~30 to 40 km (~19 to 25 miles) on the downstream side. Uplift would progressively decrease for ~60 (~37 miles) upstream of these fault crossing points. Two additional peaks in the uplift along the CRA resulted from modeling this larger size event; one ~1¼ m ( ~4 ft) high, ~5 km (~3 miles) upstream of the GHF, at CRA SSAFS-SB Banning Fault crossing point # 5 and the other ~1 m (~3 1/3 ft), ~40 km (~25 miles) upstream of the GHF. The modeled horizontal coseismic deformation will diminish to zero where the CRA crosses the upper part of Chuckwalla Valley north of the Eagle Mountain Pump Station and possibly to Lake Mathews on the downstream side. This shear could produce secondary faulting well beyond the SSAFS primary trace, perhaps by reactivating inactive or rarely active bedrock faults over a broader region. 3. The most likely earthquake to affect the SGP region would be a MW 6.0 to MW 6.5 earthquake, similar to the 1948 MW 6.2 Desert Hot Spring 1986 MW 6.1 Palm Springs, which will likely occur on average every ~50 to 100 years. Experience from 1948 and 1986 earthquakes and the results of the Coulomb 3.3 modeling suggests that an earthquake in the MW 6.0 to MW 6.5 range will generate ~1 cm (~ ”) to 6 cm (~3”) of vertical displacement along the CRA alignment with a peak at ~1 to 2 km (~ to 1 mile) west of the siphon at fault crossing point #4. Also according to the model results this vertical ground deformation along the CRA would diminish downstream

214 Applied Geology in California

to a negligible value at the syphon at CRA fault crossing point #5. It would diminish to a negligible value upstream ~5 km (~3 miles) from CRA fault crossing point # 4 (Figure 2 and 32a). The corresponding modeled horizontal ground deformation also would be in the cm range and nearly negligible at fault crossing points 4 and 5 (Figure 32b). The results of the modeling completed during these studies are consistent with other geoscience data collected in the area of SGP and also with more regional and state-wide data and modeling exercises. While the geodetic block motion model (Figures 36a, b, c and d and 37a, b c and d) uses GPS data to drive the deformation, and the Coulomb 3.3 modeling uses prescribed slip across the fault zone, the results are very similar. The GPS block modeling does not de ne the geometry of the SSAFS as precisely as the Coulomb 3.3 model and it utilizes a different computer software, so the calculated displacements and deformation values from the two programs are not precisely the same, but their similarity provides con rmation that both codes are performing as expected and generating results consistent with simple elastic models. The ratio of horizontal to vertical displacement, ~4 to 1, resulting from the two modeling methods is consistent with the ratio of horizontal to vertical slip evident in the offset 100,000 to 150,000 year old geomorphic terrace surface mapped in Millard Canyon approximately 6 km (~3.7 miles)west of the CRA crossing of I-10 (Yule and Sharer, 2014, in preparation). Furthermore the amount of vertical displacement (0.5 to 1 m) 1 to 3 ft) visible in the paleoseismic trenches recently excavated in Millard Canyon and north of Cabazon CRA (Yule and Sharer, 2014, in preparation) is similar to the result of our modeling of the 4 m (~13 ft) event. Comparing between the results of the Coulomb 3.3 and GPS block models, to the uplifted geomorphic surfaces mapped by Yule and Sieh (2003) found that they were very similar, as seen in Figures 34 and 38. These comparisons suggest that the subsurface 3D geometry of the SSAFS-SB and the slip distribution utilized in the models is consistent with what it is naturally occurring below SGP.

The results of this study are not precise enough for speci c CRA stationing or mile-marker location retro- t designs and construction, but are robust enough to provide information on the uplift gradients for engineers to model ow in the CRA past the fault zones. They also give reasonable estimates for the total deformation across the fault zones but these model results cannot tell us if the deformation will occur on a single fault plane, or will be distributed across a zone of broken ground spanning 10s to 100s of meters (~10s to 100s of yards) or even broader, producing warping and cracking of the ground surface if the rupture remains blind (i.e., terminating just below the surface). The deformation associated with a speci c fault or zone of faults along the CRA must be evaluated from detailed studies to see how past ruptures manifested themselves. The consistency between the two different models and the mapped geomorphology illustrates the importance of combining all different forms of geoscience data in formulating the input to these types of modeling exercises. Although, considerable improvements have become available in the geoscience data along the SSAFS through SGP, there remains a gap in the geoscience data along the SSAFS between Cabazon and the Indio Hills. Filling in this geoscience data gap and improvements in the seismic and geodetic databases around SGP and along the CRA will greatly improve the accuracy and reliability of estimates of fault surface displacements and ground surface deformation along the CRA during future earthquakes.

REFERENCES Advanced National Seismic System (ANSS), 2013, ANSS Catalog Search, available at [http://www. ncedc.org/anss/catalog-search.html], extracted January 2014. Allen, C.R., 1957, San Andreas fault zone in San Gorgonio Pass, southern California: Geological Society of America, Bulletin v.68, no.3, p.315–350. Altamimi, Z., Collilieux, X., Métivier, L., 2011. ITRF2008: an improved solution of the international terrestrial reference frame. Journal of Geodesy 85, 457–473. doi:DOI 10.1007/s00190-011-0444-4 Amos, C.B., P. Audet, W.C. Hammond, R. Bürgmann, I.A. Johanson, G. Blewitt, 2014, Uplift and seismicity driven by groundwater depletion in central

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

California: Nature, v. 509, p. 483–486, available at [http://dx.doi.org/10.1038/nature13275]. Atwater, T.M., 1970, Implications of plate tectonics for the Cenozoic tectonic evolution of western North America: Geological Society of America, Bulletin, v.81, p.329–332. Bawden, G.W., Thatcher, W., Stein, R.S., Hudnut, K.W., and Peltzer, G., 2001, Tectonic contraction across Los Angeles after removal of groundwater pumping effects: Nature, v. 412, p. 812–815. Blewitt, G., Kreemer, C., Hammond, W.C., and Goldfarb, J., 2013, Terrestrial reference frame NA12 for crustal deformation studies in North America: Journal of Geodynamics, v. 72, p. 11–24, ISSN 0264-3707, doi:10.1016/j.jog.2013.08.004. Berardino, P., Fornaro, G., Lanari, R., and Sansosti, E., 2002, A new algorithm for surface deformation monitoring based on small baseline differential SAR Interferograms: IEEE Transactions on Geoscience and Remote Sensing, v. 40, p. 2375–2383. Biasi, G.P., Weldon, II, R.J., and Dawson, T.E., 2013, Appendix F – Distribution of Slip in Ruptures in The Uniform California Earthquake Rupture Forecast, Version 3 (UCERF 3), USGS OFR 20131165; CGS Special Report 228; SCEC Publication #1792, 41 pp. Bond, 1939, Siphons – Problems Involved and Methods Used in the Construction of the 144 Inverted Siphons on the Main Aqueduct, in The Great Aqueduct The Story of the Planning and Building of the Colorado River Aqueduct by The Metropolitan Water District of Southern California, 32 pp. Bowie, W., 1922, The Earth’s Crust and Isostasy, Geographical Review, v. 12, no. 4, p. 613–627. Brooks, B.A., Merri eld, M.A., Foster, J., Werner, C.L., Gomez F., Bevis, M., and Gill, S., 2007, Space geodetic determination of spatial variability in relative sea level change, Los Angeles basin: Geophysical Research Letters, v. 34, no. L01611, doi:10.1029/2006GL028171. Burgette, R.J., Weldon, R.J., and Schmidt, D.A., 2009, Interseismic uplift rates for western Oregon and along strike variation in locking on the Cascadia subduction zone: Journal of Geophysical Research, v. 114, no. B01408, doi:10.1029/2008JB00579. Byerlee, J.D., 1978, Friction of Rocks: Pure and Applied Geophysics, v. 116, p. 615–626. Cooke, M., 2008, Analog and Numerical Simulation of Fault Complexity of the San Gorgonio Knot, Final Project Report for 08HQGR0041, University of Massachusetts, Amherst, MA, 18 pp.

215

Cooke, M. L., and L. C. Dair, 2011, Simulating the recent evolution of the southern big bend of the San Andreas fault, Southern California, Journal of Geophysical Research, Solid Earth, v. 116, B04405, doi:10.1029/2010JB007835. Cooke et al., 2013 Chuang, R.Y. and Johnson, K.M., 2011, Reconciling geologic and geodetic model faults slip rate discrepancies in Southern California: Consideration of non-steady mantle ow and lower crustal fault creep: Geology, v. 39, no. 7, p. 627–630, doi:10.1130/G32120.1. Dair, L., and Cooke, M. L., 2009, San Andreas fault geometry through the San Gorgonio Pass, California: Geology, v. 37, no. 2, p. 19–122. Davis, C. and O’Rourke, T., 2011, ShakeOut Scenario: Water System Impacts from a Mw 7.8 San Andreas Earthquake: Earthquake Spectra, v. 27, no. 2, p. 459–476. DeMets, C., and Dixon, T.H., 1999, New kinematic models for Paci c-North America motion from 3 Ma to present, I: evidence for steady motion and biases in the NUVEL-1A model: Geophysical Research Letters, v. 26, no. 13, p. 1921–1924. Dibblee, T., 1991, Geologic Map of the (South ) Quadrangles, Dibblee Geologic Foundation Map #DF-30, Map Scale=1:24,000. Fattaruso, L., Cooke, M. and Dorsey, R., 2014, Sensitivity of uplift patterns to dip of the San Andreas fault in the Coachella Valley, California: Geosphere, v. 10, no.6, p.1235–1246. Fialko, Y., 2006, Interseismic strain accumulation and the earthquake potential on the southern San Andreas fault system: Nature, v. 441, p. 968–971. Field, E.H., Arrowsmith, R.J., Biasi, G.P., Bird, P., Dawson, T.E., Felzer, K.R., Jackson, D.D., Johnson, K.M., Jordan, T.H., Madden, C., Michael, A.J., Milner, K.R., Page, M.T., Parsons, T., Powers, P.M., Shaw, B.E., Thatcher, W.R., Weldon, R.J., II, and Zeng, Y., 2014, Uniform California Earthquake Rupture Forecast, Version 3 (UCERF3) — The TimeIndependent Model: Bulletin of the Seismological Society of America, v. 104, no. 3, p. 1122–1180. Freed, A.M., Bürgmann, R., and Herring, T., 2007, Far-reaching transient motions after Mojave earthquakes require broad mantle ow beneath a strong crust: Geophysical Research Letters, v. 34, L19302, doi:10.1029/2007GL030959. Fuis, G.S., Scheirer, D.S., Langenheim, V.E., and Kohler, M.D., 2012, A New Perspective on the Geometry of the San Andreas Fault in Southern

216 Applied Geology in California California and its Relationship to Lithospheric Structure: Bulletin of the Seismological Society of America, v. 102, no.1, p. 236–251. GeoPentech, Inc., 2013, Draft Technical Memorandum: Seismic Event Vulnerability Study, Induced Ground Deformations, Colorado River Aqueduct, Coachella-San Gorgonio Pass-Hemet Area, prepared for Metropolitan Water District of Southern California, 51 pp. Gutierrez, C., Bryant, W.B., Saucedo, G., and Wills, C., 2010, Geology Map of California, California Geological Survey, scale 1:750,000. Hammond, W.C., Johnson, K., Weldon, R.J., Blewitt, G., and Burgette, R., 2013, A new look at vertical motion around the San Andreas Fault in the Southern California from integrated GPS and InSAR measurements [abstract]: Fall AGU session G41C-02, San Francisco, CA, 9-13 December 2013. Hammond, W.C., Blewitt, G., Li, Z., Kreemer, C., and Plag, H.-P., 2012, Contemporary uplift of the Sierra Nevada, western United States, from GPS and InSAR measurements: Geology, v. 40, p. 667-670, doi:10.1130/G32968.1. Hammond, W.C., Blewitt, G., and Kreemer, C., 2011, Block modeling of crustal deformation of the northern Walker Lane and Basin and Range from GPS velocities, Journal of Geophysical Research, v. 116, no. B04402, doi:10.1029/2010JB007817. Hammond, W.C., C., Blewitt, G., and Plag H.-P., 2010, Effect of viscoelastic postseismic relaxation on estimates of interseismic crustal strain accumulation at Yucca Mountain, Nevada: Geophysical Research Letters, v. 37, no. L06307, doi:10.1029/2010GL042795. Hammond, W.C. and Thatcher, W., 2007, Crustal Deformation across the Sierra Nevada, Northern Walker Lane, Basin and Range Transition, western United States Measured with GPS, 2000– 2004: Journal of Geophysical Research, v. 112, no. B05411, doi:10.1029/2006JB004625. Hauksson, E. Yang, W., and Shearer, P.M., Waveform Relocated Earthquake Catalog for Southern California (1981 to 2011), 2012, Bulletin of the Seismological Society of America, v. 102, no. 5, p.2239–2244, doi: 10.1785/0120120010. Hinds, J., 1938, Major Problems of Aqueduct Location: Engineering News Record, 650 pp. Humphreys, E. D., and Weldon, R.J., 1994, Deformation across the western United States: a local estimate of Paci c-North American transform deformation,

Journal of Geophysical Research, Vol. 99, NO. B10, p. 19,975–20,0120. Jones, L.M., Bernknopf, R. Cox, D. Goltz, J. Hudnut, K. Mileti, D. Perry, S. Ponti, D. Porter, K., Reichle, M., Seligson, H., Shoaf, K., Treiman, J., and Wein, A., 2008, The ShakeOut Scenario, US Geological Survey Open File Report 2008-1150, 308 pp., available at [http://pubs.usgs.gov/of/2008/1150/]. King, G.C.P., and Wesnousky, S.G., 2007, Scaling of fault parameters for continental strike-slip earthquakes: Bulletin of the Seismological Society of America, v. 97, p. 1833–1840. Kreemer, C., Hammond, W.C., Blewitt, G., Holland, A., and Bennett, R.A., 2012, A geodetic strain rate model for the southwestern United States: Nevada Bureau of Mines and Geology, M178. Lu, Z., and Danskin, W., 2001, InSAR analysis of natural recharge to de ne structure of a ground-water basin, San Bernardino, California: Geophysical Research Letters, v. 28, no. 13, p. 2661–2664, doi:10.1029/2000GL012753. Madden, C., Haddad, D.E., Salisbury, J.B., Zielke, O., Arrowsmith, J.R., Weldon, R., and Colunga, J., 2013, Appendix R — Compilation of Slip-inthe-Last-Event Data and Analysis of Last Event, Repeated Slip, and Average Displacement for Recent and Prehistoric Ruptures in The Uniform California Earthquake Rupture Forecast, Version 3 (UCERF3), USGS OFR 2013-1165; CGS Special Report 228; SCEC Publication #1792, 65 pp. Matti, J.C., and Morton, D.M., 1993, Paleographic evolution of the San Andreas fault in southern California: A reconstruction based on a new crossfault correlation, in Powell, R.E., Weldon, R.J., and Matti, J.C., eds., The San Andreas fault system: displacement, palinspastic reconstruction, and geologic evolution: Geological Society of America Memoir 178, p. 107–159 McCaffrey, R., 2002, Crustal block rotations and plate coupling, in Plate Boundary Zones, AGU Geodynamics Series, ed. Stein, S.A. and Freymueller, J., p. 101–122, American Geophysical Union. McGill, S., 2013, personal communication to R. Weldon. Meade, B.J., and Hager, B.H., 2005, Block models of crustal motion in southern California constrained by GPS measurements: Journal of Geophysical Research, v. 110, no. B03403, 19 pp., doi:10.1029/2004JB003209. Nicholson, C., 1996, Seismic Behavior of the Southern San Andreas Fault Zone in the Northern Coachella

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

Valley, California: Comparison of the 1948 and 1986 Earthquake Sequences: Bulletin of the Seismological Society of America, v. 86, no. 5, p. 1331–1349. Nicholson, C., Sorlien, C.C., Atwater, T., Crowell, J.C., and Luyendyk, B.P., 1994, Microplate capture, rotation of the western Transverse Ranges, and initiation of the San Andreas transform as a low-angle fault system: Geology, v. 22, p. 491–495. Okada, Y., 1992, Internal deformation due to shear and tensile faults in a half-space: Bulletin of the Seismological Society of America, v. 82, no. 2, p. 1018–1040. Okada, Y., 1985, Surface deformation due to shear and tensile faults in a half-space: Bulletin of the Seismological Society of America, v. 75, p. 1135–1154. Orozco, A.A., 2004, Offset of a mid-Holocene alluvial fan near Banning, California: Constraints on the slip rate of the San Bernardino strand of the San Andreas Fault, Master’s Thesis, California State University, Northridge, 56 pp. Perry, S., Cox, D., Jones, L., Bernknopf, R., Goltz, J., Hudnut, K., Mileti, D., Ponti, D., Porter, K., Reichle, M., Seligson, H., Shoaf, K., Treiman, J., and Wein, A., 2008, The ShakeOut Earthquake Scenario – A Story That Southern Californians Are Writing, US Geological Survey Circular 1324 and California Geological Survey Special Report 207, available at [http://pubs.usgs.gov/circ/1324/]. Philibosian, B., Fumal, T.E., and Weldon, R.J., II, 2011, San Andreas fault earthquake chronology and Lake Cahuilla history at Coachella, California: Bulletin of the Seismological Society of America, v. 101, doi:10.1785/0120100050. Pollitz, F.F., Wicks, C.W., and Thatcher, W., 2001, Mantle Flow Beneath a Continental Strike Slip Fault: Postseismic Deformation After the 1999 Hector Mine Earthquake: Science, v. 293, p. 1814–1818. Powell, R.E., 1993, Balanced palinspastic reconstruction of pre-Late Cenozoic paleogeology, southern California: geologic and kinematic constraints on evolution of the San Andreas fault system, in Powell, R.E., Weldon, R.J., and Matti, J.C., eds., The San Andreas fault system: displacement, palinspastic reconstruction, and geologic evolution: Geological Society of America Memoir 178, p. 1–106. Savage, J.C., Gan, W., and Svarc, J.L., 2001, Strain accumulation and rotation in the eastern California shear zone: Journal of Geophysical Research, v. 106, p. 21,995–922,007.

217

Scharer, K., 2014, personal communication to D. Yule. Scharer, K., Weldon, R., Streig, A., Yule, D., Wolff, L., 2014, Comparison of fault behavior across the Big Bend of the San Andreas Fault, California, submitted to Seismological Society of America Meeting, Anchorage, AK. Schmidt, D. and Bürgmann, R., 2003, Time-dependent land uplift and subsidence in the Santa Clara valley, California, from a large interferometric synthetic aperture radar data set: Journal of Geophysical Research-Solid Earth, v. 108, no. B9, doi:10.1029/2002JB002267. Shaw, B.E., 2013, Earthquake Surface Slip-Length Data is Fit by Constant Stress Drop and is Useful for Seismic Hazard Analysis: Bulletin of the Seismological Society of America, v. 103, no. 2A, p. 876–893. Shaw, B.E., 2009, Constant stress drop from small to great earthquakes in magnitude-area scaling: Bulletin of the Seismological Society of America, v. 99, no. 2A, p. 871-875, doi:10.1785/0120080006. Sieh, K., 1978, Slip along the San Andreas fault associated with the great 1857 earthquake: Bulletin of the Seismological Society of America, v. 68, p.1421–1428. Tong, X., Sandwell, D.T., and Smith-Konter, B., 2013, High-resolution interseismic velocity data along the San Andreas Fault from GPS and InSAR: Journal of Geophysical Research Solid Earth, v. 118, doi:10.1029/2012JB009442. United States Geological Survey (USGS), 2013, The National Map Viewer, National Elevation Datasets (1-second arc-grid), accessed December 2013. USGS, 2011, Coulomb 3.3 Graphic-Rich Deformation and Stress-Change Software for Earthquake, Tectonic, and Volcano Research and Teaching-User Guide, Open-File Report 2011-1060, 63 pp. USGS, 2010, Quaternary fault and fold database for the United States, updated November 2010, [http// earthquakes.usgs.gov/regional/qfaults/]. Wei, M., Sandwell, D., and Smith-Konter, B., 2010, Optimal combination of InSar and GPS for measuring interseismic crustal deformation: Advances in Space Research, v. 46, no. 2, p. 236–249, doi:10.1016/j.asr.2010.03.013. Weldon, R.J., II, Dawson, T.E., Biasi, G., Madden, C., and Streig, A.R., 2013, Appendix G – Paleoseismic Sites Recurrence Database in The Uniform California Earthquake Rupture Forecast, Version 3 (UCERF 3), USGS OFR 2013-1165; CGS Special Report 228; SCEC Publication #1792, 73 pp.

218 Applied Geology in California Weldon, R.J, II, Scharer, K., Fumal, T., and Biasi, G., 2004, Wrightwood and the earthquake cycle—what the long recurrence record tells us about how faults work: Geological Society of America Today, v. 14, p. 4–10, doi:10.1130/1052-5173(2004)014. Weldon R.J., II, and Humphreys, 1986, A Kinematic Model of Southern California: Tectonics, v. 5, no. 1, p. 33–48. Wells, D.L., and Coppersmith, K. J., 1994, New empirical relationships among the magnitude, rupture length, rupture width, rupture area, and surface displacement: Bulletin of the Seismological Society of America, v. 84, p. 974–1002. Werner, C., Wegmüller, U. Strozzi, T., and Wiesmann, A., 2000, Gamma SAR and interferometric processing software: ERS-ENVISAT Symposium, European Space Agency, Gothenberg, Sweden. Wesnousky, S.G., 2008, Displacement and geometrical characteristics of earthquake surface ruptures: Issues and implications for seismic-hazard analysis and the process of earthquake rupture: Bulletin of the Seismological Society of America, v. 98, p. 1609–1632. Wisely, B.A., 2012, Geophysical and Hydrogeologic investigations of two primary alluvial aquifers embedded in the southern San Andreas fault system: San Bernardino Basin and upper Coachella Valley, unpublished Ph.D. Dissertation: University of Oregon, 192 pp. Wisely, B.A. and Schmidt, D., 2010, Deciphering vertical deformation and poroelastic parameters in a tectonically active fault-bound aquifer using InSAR and well level data, San Bernardino basin, California: Geophysical Journal International, v. 181, p. 1185-1200, doi: 10.1111/j.1365-246X.2010.04568.x. Wolff, L., Yule, D., Scharer, K., Witkosky, R., Desjarlais, I., and Huerta, B., 2013, Evidence for ve paleoearthquakes on the San Gorgonio Pass Fault Zone in the last 6000 years, SCEC Annual Meeting, Palm Springs, CA. Working Group on California Earthquake Probabilities (WGCEP), 2013, The Uniform California Earthquake Rupture Forecast, Version 3 (UCERF 3), USGS OFR 2013-1165; CGS Special Report 228; SCEC Publication #1792, 97 pp. WGCEP, 2008, The Uniform California Earthquake Rupture Forecast, Version 2 (UCERF 2), USGS OFR 2007-1437; CGS Special Report 203; SCEC Contribution #1138, version 1.0, 104 pp. Yule, D., and Scharer, K., 2014, in preparation.

Yule, D., Scharer, K., Sieh, Wolff, L., McBurnett, P., Ramzan, S., Witkosky, R., and Desjarlais, I., 2014, Paleoseismology and Slip Rate of the San Andreas Fault system at San Gorgonio Pass, submitted to joint GSA Cordilleran/Rocky Mountain Section Meeting, Bozeman, MT. Yule, J.D., McBurnett, P., and Ramzan, S., 2011, Long Return Periods for Earthquakes in San Gorgonio Pass and Implications for Large Ruptures of the San Andreas Fault in Southern California, AGU, 92, Fall Meeting Supplemental Abstract S21A-2141. Yule, D., and Sieh, K., 2003, Complexities of the San Andreas fault near San Gorgonio Pass: Implications for large earthquakes: Journal of Geophysical Research, v. 108, no. B11, 2548, 23 pp. Zielke, O., Arrowsmith, J. R., Grant Ludwig, L., and Akciz, S. O., 2010, Slip in the 1857 and Earlier Large Earthquakes Along the Carrizo Plain, San Andreas Fault: Science, v. 327, p. 1119–1122. Ray J. Weldon II is a professor of structural geology, neotectonics, and Quaternary geology in the Department of Geological Sciences at the University of Oregon. He received his Ph.D. in Geology at the California Institute of Technology in 1986 and his B.A. (cum laude) in Geology at Pomona College in1977. Since joining the faculty at the University of Oregon in 1987, Ray’s teachings and research have focused on active faults in the eld and he works with seismologists, geodesists, and hazard analysts to build forecasts of seismic hazard. Ray serves on the Executive Committee of the Uniform California Earthquake Rupture Forecast (UCERF), the Steering Committee for the US National Seismic Hazard Map (that sets national building codes, among many other things), and advises a number of public and private entities on seismic hazard to critical infrastructure. Greg de Lamare is a Principal Engineer at the Metropolitan Water District of Southern California and is responsible for oversight of Engineering Services’ Infrastructure Reliability programs. In this role he has led multiple seismic vulnerability assessments of Metropolitan facilities, including the Colorado River Aqueduct. He received a Bachelor of Science in Engineering from California State University, Northridge, and is a registered Professional Engineer with the state of California. Doug Yule is a professor of geological sciences at California State University, Northridge. He teaches courses in introductory and advanced eld mapping, structural geology and tectonics, and earthquake geology. His research is eld based and involves active tectonics, structural geology, and paleoseismology and focuses on documenting paleoearthquakes on

San Andreas Fault - South Branch Surface Deformation Modeling and Risk to the Colorado River Aqueduct

the San Andreas Fault system and the Himalayan Main Frontal thrust system. He earned his BA in Geology from Pomona College, and MSc in Geology from the University of Wyoming, and PhD in Geology from Caltech. William C. Hammond is an associate professor of geodesy and geophysics working in the Nevada Bureau of Mines and Geology at the University of Nevada, Reno. He uses geodetic techniques to measure active crustal deformation, mountain building and associated seismic hazards. He operates the Nevada Geodetic Laboratory’s MAGNET GPS network, a tectonics observatory and research infrastructure that extends into ve of the western United States. He received a B.A. in applied mathematics from U.C. Berkeley, and a Ph.D. in geophysics from the University of Oregon. Ashley Streig is an Assistant Professor of geological sciences at Portland State University, Oregon. Her research and courses focus primarily on paleoseismology, earthquake recurrence, fault interaction and seismic hazard, she has worked in the United States, Japan and Taiwan. She received a National Science Foundation Earth Sciences postdoctoral fellowship, and worked with Dr. Chris Gold nger at the College of Earth, Ocean, and Atmospheric Sciences at Oregon State University in Corvallis, Oregon. In 2014, Ashley received her Ph.D. in geosciences from the University of Oregon under the leadership of Professor Ray Weldon. She earned an MSc in Geology from Central Washington University, and a BA in Geology from Occidental College. In addition she worked as a project geologist with William Lettis & Associates before pursuing her PhD. Alexandra Sarmiento is an engineering geologist at GeoPentech in southern California. She holds a B.S. in Geological Engineering and an M.S. in Geology from the University of Nevada, Reno. Her experience and interests are in seismic source characterization and ground motion analysis. S. Thomas Freeman is a California registered professional geologist and certi ed engineering geologist and hydrogeologist. His 44 yrs of experience spans the geologic and seismic hazards through-out California, North America, and internationally; focused on Metropolitan’s and other water district’s aqueducts, feeders, reservoirs and treatment plants in southern

219

California, including the entire length of the CRA. He holds a B.S. in Geology from the U.C. Santa Barbara and a M.S. in Geological Engineering from U.C. Berkeley. John E. Shamma is the Facility Planning Team Manager for the Metropolitan Water District of Southern California overseeing Metropolitan’s infrastructure reliability and vulnerability investigations. He served as Project Manager on a variety of projects including the Central Pool Augmentation Project and the Inland Feeder Water Conveyance System’s Arrowhead Tunnels. Prior to his work on the Inland Feeder Project, he served as design manager for a multitude of projects within Metropolitan’s Engineering Group. A registered professional engineer in California, he earned Bachelor of Science in Engineering and a Master of Science in Civil Engineering from the California State University, Los Angeles. He also earned a Master of Science in Environmental Engineering from the University of Southern California. He serves as an adjunct professor of civil engineering at California State University - Los Angeles. Dr. Beikae is a senior engineer and a member of the Safety of Dams Team at Metropolitan Water District of Southern California. He is experienced in stability and deformation analyses of dams/structures under static and seismic loading conditions as well as developing computer programs for geotechnical/structural applications. He holds a Ph.D. degree in earthquake engineering with minors in structural dynamics and mathematics from University of California, Berkeley (UCB). He also received a M.S. degree in geotechnical engineering from UCB and M.S. and B.S. degrees in structural/civil engineering from Tehran University, Iran. Albert Rodriguez is an Engineer in the Facility Planning Department for the Metropolitan Water District of Southern California. Albert’s responsibilities include the assessment of Metropolitan’s facilities and systems against various hazards including seismic hazards. Albert received his Bachelor of Science in Civil Engineering from the California State University in Los Angeles where he graduated with honors. He is a registered Professional Engineer in the state of California.