| Ocean Drilling Program Leg 138 in the eastern equatorial Pacific (EEP) provided a unique opportunity to understand the paleoceanographic significance of seismic reflectors in this climatically sensitive region. Through carefully offset multiple cores and geophysical logging at each site, synthetic seismograms were generated and used to determine the paleoceanographic significance of regionally traceable reflectors by linking variations in the core record to the seismic record. The EEP reflectors are due to changes in density, and in turn, carbonate content, primarily due to variations in surface productivity, as indicated by the presence of mats of the upwelling diatom Thalassiothrix longissima.; A time series of the EEP composite density records was successfully used as a tuning target for density records in the central equatorial Pacific (CEP) to attempt to determine the basin-wide extent of EEP reflectors. It was found that many EEP correspond to reflectors in the CEP. However, some of the CEP reflectors, previously postulated to be the result of periods of enhanced carbonate dissolution, also correspond to diatom mats, requiring the reconsideration of the origin of these reflectors. Further tuning of the density to records from sites in the southern oceans suggest that many of these reflectors correspond to global paleoceanographic events.; As well, by predicting carbonate content from density, and density from the seismic record, the potential exists to map carbonate content, and hence carbonate flux, over both space and time (assuming synchronous, traceable seismic reflectors) in large areas of the equatorial Pacific, at least. Three established methods to predict carbonate content from density were compared using data from the EEP, CEP, and southern oceans. A new method, a multiple regression of density and depth versus carbonate content, proved to be significantly better than the established methods. The applicability of autoregressive deconvolution to invert the seismic record to obtain density was tested over a range of bandwidths and noise levels using model data. This method proved to be very successful, though the results were dependent on both the noise level and source bandwidth. The method was improved by using singular value decomposition to filter small singular values of the autocovariance matrix. |
