Multitaper spectral analysis of wide-angle refraction data from Costa Rica for t* estimation

The inputs and outputs of the Central American subduction zone have received attention from geoscientists of many disciplines in recent years. In order to have full understanding of subduction, it is important to have constraints on the nature of deep-seated magmatic processes and the composition of newly formed crust. In January to March and July 2005, I was associated with a group of scientists from the United States and Costa Rica that acquired two long seismic refraction lines parallel to and across the arc in Costa Rica to image the physical properties of the entire crust. Line1 runs almost SW-NE, across the arc in central Costa Rica from Pacific coast in the west to Atlantic coast in the east. We deployed 742 seismometers at nominal spacing of 200 m. Reftek “Texans” data loggers were used to record the data with a sampling interval of 4 msec. Twenty explosive shots ranging in size from 200 kg to 1025 kg were used. These sources were detonated at ~ 7.5 km interval along the profile to acquire seismic refraction data on this cross-section of the arc. The ultimate goal of the project is to infer subsurface temperature distribution that can be related to magmatic processes. One approach to estimating temperature distribution is to map seismic attenuation distribution. Towards this goal, I have used the data from this line to map the variation of t* (path integrated attenuation) values across the volcanic arc using a method based on spectral ratios. I have compared several spectrum estimating techniques and demonstrated that a multitaper method is able to obtain robust estimates of spectra. I estimated the source by summing the spectra from near offset stations, a weighted running average of spectra with one km radius was taken to enhance the signal to noise ratio and a Random sampling was used to obtain more realistic estimates of the errors in the spectra. Finally I used a Spectral Ratio’s method to calculate the slopes. Different frequency windows were used to calculate the least squares linear fit. My results show a large variation in the t* values. These fluctuations in the t* values are attributed to the large variations in the near site effects and scattering attenuation associated with them along the profile. Some anomalous t* values may be a result of focusing of seismic energy due to underlying velocity structure. The background noise plays an important role in the t* values as well. At larger source-receiver offsets the signal/noise ratio is lower. Consequently, we overestimate the amplitude of the higher frequency at large offsets. Contrary to the assumption of Spectral Ratio’s method, attenuation seems to be frequency dependent for this data set