Seismic Attenuation in the Subduction Zone of Costa Rica

Newman and others (GRL, 2002 and this meeting) using permanent and temporary land stations as well as Ocean-Bottom Seismometers, have shown a significant change in depth to the top of the seismogenic zone at the boundary between EPR and CNS crust. These plate partitions differ in heat flow, so their thermal structures should be different, although there is little difference in water depth. Attenuation of seismic waves, especially S, is increased by heating so attenuation becomes a useful proxy for temperature. Attenuation studies usually use the spectral ratio method, in which spectra observed at two stations are divided, so that common elements in the product-of-transfer-functions equation (source spectrum, instrument corrections) are removed. There is, however, a way of treating many observations from a single source in a single solution, rather than pairwise. By taking the logarithm of the transfer function equation, the product becomes a sum, and can be treated as a row in a matrix equation containing data at multiple stations at multiple frequencies (Dorman, JGR, 1968, 1969). The bulk of the data processing is being done within the framework of the Antelope relational database system, which offers several programs to calculate spectra and provides comprehensive housekeeping tools. The system, however, is intricate and we have not yet been successful in making all the programs operate with the waveform and response files we generated for the OBSs. The result of this is that we have to do more processing station-by-station than we had hoped. The solution for the matrix equation described in the third paragraph is stable, and eigenvector expansion techniques yield insight to the process. The solution vector is tripartite, consisting of the source spectrum, the range dependent amplitude terms, and 1/Q. The value of 1/Q is associated with a large eigenvalue, so it is well-determined. Preliminary analysis shows values of about 250 for Q_p. This result is subject to revision as we examine more data.