Single-channel marine seismic profiling and the analysis of reverberatory sequences

Abstract

Analogue recordings of single-channel marine seismic profiles have been used for years to study the thickness and configuration of unconsolidated and semiconsolidated sediments in water-covered areas. If the signals are recorded digitally, it is theoretically possible to estimate some of their physical properties as welL The ability to make such estimates would constitute a means of remotely classifying marine soils and would be a valuable asset in contexts ranging from geotechnical and engineering studies to the monitoring of effluent accumulation. Toward such an end, this dissertation considers the digital recording and analysis of near-to-zero-offset marine reflection data. Under proper conditions, single-channel marine profiles can image sea-floor soil hundreds of metres thick while retaining resolution on the order of a metre or less. Such an image contains not only primary reflections but also multiple reflections. These multiples occur in reverberatory sequences of wavelets that often interfere with, or are mistakenly identified as, primary reflections. For this reason multiple sequences are usually classified as noise and much effort is spent trying to suppress them, often with little success. The thesis propounded here is that, even should it not be practical to "remove" multiple reflections from a data set, it is still possible to use sequences of multiples as a source of information. Examples of multiple sequences observed in industrial waterways are presented. Theory is developed showing how information concerning acoustic impedance contrasts and absorption coefficients is encoded in the rate of sequence decay and the change in shape between successive wavelets. It is recognized that the use of certain field techniques serves to simplify the decoding procedure. Under the assumption that these techniques are followed, a mathematical model is developed to act as a framework for the analysis. A deterministic "dereverberation filter" is identified which would be useful for supressing multiples once estimates of physical properties have been obtained. Its effect is demonstrated on the synthetic examples and the result of such "dereverberation" is considered in detaiL Points of difficulty are discussed and conclusions are drawn.