Some Statistical Techniques to Optimize the Search for Stratigraphic Traps on Seismic Data
G. Wesley Rice, Kenneth Harold Waters · Offshore Technology Conference · 1976
ABSTRACT Stratigraphic trap exploration using seismic data rellies on locating significant horizontal changes in a zone of seismic data for relevant parameters. Seismic waveform characteristics and other parameters such as coherence, amplitude, interval P wave and S wave velocity, etc., may be analyze a with multivariate statistical techniques to identify the location of such changes, as well as to provide a mechanism for relating these sequences to known lithologic conditions. The principles described in this paper are illustrated through the use of an onshore field example; however, the same techniques are " applicable offshore. In addition to the delineation and characterization of seismic waveform characteristics, procedures are described to compensate for differences in seismic data acquisition and processing. Band width effects are also considered and may be quantified numerically. INTRODUCTION Until the last several years, the principal use of reflection seismology in hydrocarbon exploration was to define subsurface structure. Except for derived velocities from CDP data, little emphasis was placed on deriving lithologic information from seismic data until the advent of "bright spot" technology in the early 1970' s. This paper relies on the assumption that changes in seismic waveform and associated parameters often can be associated with corresponding variations in lithologic conditions. By identifying and characterizing these changes in the seismic section, a pattern that relates either directly or indirectly to hydrocarbon occurrence can often be identified. Sometimes the variations are readily apparent to the eye, such as amplitude anomalies that are associated with gas reservoirs; however, many of these variations are complex and difficult to analyze, especially where large volumes of data are available in the area of interest. These conditions are well suited to statistical techniques that can identify basic seismic components and provide methods for associating geological data effectively with the seismic control. ORIGINAL DISCRIMINANT METHODS A method of combining geological knowledge with detailed seismic information was reported by Mathieu and Rice. 5 Their procedure makes use of a seismic time zone referenced to some geological marker that is depicted in the seismic cross-section. Each seismic trace provides a suite of variables that are later analyzed using multivariate statistical procedures. The quantitative variables used in their example were directly measured by taking amplitudes of the seismic trace at specified time intervals referenced to a reflection event. Other quantifiable variables relating either directly or indirectly to the acoustic properties of the suspected- anomaly also might be used in place of, or in addition to, the direct digital values. The method assumes that a set of such data can- be measured- for all wells in the area, or can be collected from seismic control in immediate proximity to well locations. The particular method uses linear discriminate analysis to establish classification functions for a priori defined groups of traces.