Processing of Interferometric Time-Series acquired with Ground-Based Radar for Mapping Deformation of the Calaveras Fault at Coyote Dam

A creeping section of the Calaveras fault passes directly through Coyote Dam, an earthen dam located 13 km northeast of Gilroy, CA. Aseismic deformation causes measurable deformation of the dam between 10 and 15 mm/year. Assessment of fault related deformation and determining the fault location relative to the dam structure are essential for evaluation of dam safety and planning a seismic retrofit. Terrestrial Radar Interferometry (TRI) has the capability to generate spatially continuous deformation maps due to fault creep and other processes with accuracy on the order of 0.1mm along the line of sight (LOS). TRI also permits multiple illumination directions required for 3D deformation mapping, the ability to acquire large numbers of observations (>500/day) to mitigate the effect of atmosphere on the interferometric phase, and near zero-baseline geometry reducing the need for high accuracy digital elevation models. The first phase of our radar investigation of Coyote Dam took place during February through July 2015. On the upstream face of the dam, the fault surface LOS deformation was very clearly delineated and the deformation was consistent with existing fault model. On the downstream side, the motion field was shown to be more complex with LOS motion from a single observation position that was difficult to reconcile with the existing model. This motion includes possible rotation of a large silicate/carbonate block within the fault zone. A follow-on campaign, begun in May 2016, aims to elucidate the 3D motion field on the down-stream side of the dam using 5 different observation positions including 4 new positions on the crest of the dam. This campaign aims to develop and demonstrate 1) improved methods for mitigation of atmospheric phase error using both radar reflector measurements and meteorological in-situ data, 2) a 3D least-squares solution for deformation using interferometric data from multiple observation points. A network of corner reflectors has been deployed to ensure exact co-registration and localization of the data as well as assist with estimation of atmospheric phase gradients. During the current 6-month campaign observations are being acquired, on average every 3 weeks. We present initial results from processing of data acquired since the start of the study.