The San Francisco Public Utilities Commission is seismically retrofitting the water delivery system at San Andreas Lake, San Mateo County, California, where the reservoir intake system crosses the San Andreas Fault (SAF). The near-surface fault location and geometry are important considerations in the retrofit effort. Because the SAF trends through highly distorted Franciscan mélange and beneath much of the reservoir, the exact trace of the 1906 surface rupture is difficult to determine from surface mapping at San Andreas Lake. Based on surface mapping, it also is unclear if there are additional fault splays that extend northeast or southwest of the main surface rupture. To better understand the fault structure at San Andreas Lake, the U.S. Geological Survey acquired a series of seismic imaging profiles across the SAF at San Andreas Lake in 2008, 2009, and 2011, when the lake level was near historical lows and the surface traces of the SAF were exposed for the first time in decades. We used multiple seismic methods to locate the main 1906 rupture zone and fault splays within about 100 meters northeast of the main rupture zone. Our seismic observations are internally consistent, and our seismic indicators of faulting generally correlate with fault locations inferred from surface mapping. We also tested the accuracy of our seismic methods by comparing our seismically located faults with surface ruptures mapped by Schussler (1906) immediately after the April 18, 1906 San Francisco earthquake of approximate magnitude 7.9; our seismically determined fault locations were highly accurate. Near the reservoir intake facility at San Andreas Lake, our seismic data indicate the main 1906 surface rupture zone consists of at least three near-surface fault traces. Movement on multiple fault traces can have appreciable engineering significance because, unlike movement on a single strike-slip fault trace, differential movement on multiple fault traces may exert compressive and extensional stresses on built structures within the fault zone. Such differential movement and resulting distortion of built structures appear to have occurred between fault traces at the gatewell near the southern end of San Andreas Lake during the 1906 San Francisco earthquake (Schussler, 1906). In addition to the three fault traces within the main 1906 surface rupture zone, our data indicate at least one additional fault trace (or zone) about 80 meters northeast of the main 1906 surface rupture zone. Because ground shaking also can damage structures, we used fault-zone guided waves to investigate ground shaking within the fault zones relative to ground shaking outside the fault zones. Peak ground velocity (PGV) measurements from our guided-wave study indicate that ground shaking is greater at each of the surface fault traces, varying with the frequency of the seismic data and the wave type (P versus S). S-wave PGV increases by as much as 5–6 times at the fault traces relative to areas outside the fault zone, and P-wave PGV increases by as much as 3–10 times. Assuming shaking increases linearly with increasing earthquake magnitude, these data suggest strong shaking may pose a significant hazard to built structures that extend across the fault traces. Similarly complex fault structures likely underlie other strike-slip faults (such as the Hayward, Calaveras, and Silver Creek Faults) that intersect structures of the water delivery system, and these fault structures similarly should be investigated.