A new onshore-offshore seismic dataset from the Cascadia subduction zone was used to characterize mantle lithosphere structure from the ridge to the volcanic arc, and plate interface structure offshore within the seismogenic zone. The Cascadia Initiative (CI) covered the Juan de Fuca plate offshore the northwest coast of the United States with an ocean bottom seismometer (OBS) array for four years; this was complemented by a simultaneous onshore seismic array. Teleseismic data recorded by this array allows the unprecedented imaging of an entire tectonic plate from its creation at the ridge through subduction initiation and back beyond the volcanic arc along the entire strike of the Cascadia subduction zone. Higher frequency active source seismic data also provides constraints on the crustal structure along the plate interface offshore. Two seismic datasets were used to image the plate interface structure along a line extending 100 km offshore central Washington. These are wide-angle reflections from ship-to-shore seismic data from the Ridge-To-Trench seismic cruise and receiver functions calculated from a densely spaced CI OBS focus array in a similar region. Active source seismic observations are consistent with reflections from the plate interface offshore indicating the presence of a P-wave velocity discontinuity. Until recently, there has been limited success in using the receiver function technique on OBS data. I avoid these traditional challenges by using OBS constructed with shielding deployed in shallow water on the continental shelf. These data have quieter horizontals and avoid water- and sediment-multiple contamination at the examined frequencies. The receiver functions are consistently modeled with a velocity structure that has a low velocity zone (LVZ) with elevated P to S-wave velocity ratios at the plate interface. A similar LVZ structure has been observed onshore and interpreted as a combination of elevated pore-fluid pressures or metasediments. This new offshore result indicates that the structure may persist updip indicating the plate interface may be weak. To focus more broadly on the entire subduction system, I calculate phase velocities from teleseismic Rayleigh waves from 20-100 s period across the entire onshore-offshore array. The shear-wave velocity model calculated from these data can provide constrains on the thermal structure of the lithosphere both prior to and during subduction of the Juan de Fuca plate. Using OBS data in this period band requires removal of tilt and compliance noise, two types of water-induced noise that affect long period data. To facilitate these corrections on large seismic arrays such as the CI, an automated quality control routine was developed for selecting noise windows for the calculation of the required transfer functions. These corrections typically involve either averaging out transient signals, which requires the assumption of stationarity of the noise over the long periods of time, or laborious hand selection of noise segments. This new method calculates transfer functions based on daily time series that exclude transient signals, but allows for the investigation of long-term variation over the course of an instrument’s deployment. I interpret these new shoreline-crossing phase velocity maps in terms of the tectonics associated with the Cascadia subduction system. Major findings include that oceanic plate cooling models do not explain the velocities observed beneath the Juan de Fuca plate, that slow velocities in the forearc appear to be more prevalent in areas modeled to have experienced high slip in past Cascadia megathrust earthquakes, and along strike variations in phase velocity reflect variations in arc structure and backarc tectonics.