Three-dimensional models of compressional velocity and azimuthal anisotropy from tomographic inversions using 23,564 ocean bottom seismometer P wave arrivals define systematic lateral variations in seismic structure of the CoAxial segment of the Juan de Fuca Ridge (JdFR). Over much of the segment the across-axis structure is roughly axisymmetric, characterized by a progressive increase in dike velocities moving away from the ridge axis. This trend is most apparent in the basal dikes, where on-axis velocities are about 800 m/s slower than those measured elsewhere within the rift valley. The on-axis sheeted dikes also exhibit ridge-oriented azimuthal anisotropy, with a peak-to-peak amplitude of about 600 m/s. Outboard of the rift valley, beneath ridge flanks with fault scarps, velocities in the upper 1500 m of crust are reduced. The maximum amplitude of this anomaly is about 700 m/s, located near the top of the sheeted dikes. Variations in the three-dimensional velocity model are believed to reflect changes in crustal porosity, from which we infer an axisymmetric porosity model for seismic layer 2 of the CoAxial segment. As the crust ages, the evolution of layer 2 porosity could occur in the following way: (1) the porosity of zero-age, on-axis dikes is set at formation by the contraction of molten material, (2) hydrothermal alteration fills pore spaces as the dikes move away from the center of the axial valley, and (3) normal faulting on the ridge flank scarps opens fractures and increases porosity of the upper dikes as they move off-axis. At the north end of the segment, dike velocities are several hundred meters per second slower, on average, and the across-axis structure is lost. The transition from a coherent, aligned seismic structure to a less distinct pattern with reduced velocities may represent a shift from magmatic to amagmatic extension moving away from the Cobb hotspot on the ridge axis. The porosity structure we have derived for the CoAxial segment suggests an alternative to the usual hydrothermal circulation model of cross-axis convection cells. A circulation model with along-axis convection cells located entirely within the axial valley appears to be more compatible with our data.¿ 1997 American Geophysical Union |