Automated seismic survey design and dispersed source array acquisition

Abstract

Reflection seismology is nowadays the preferred technique in the oil and gas industry to estimate the properties of the Earth's subsurface. The method typically includes a series of procedures that fit in three broad categories: • seismic data acquisition; • data processing and imaging; • interpretation and reservoir characterization.

This thesis mainly focuses on the first category and aims at improving both the operational productivity of seismic surveys in terms of costs, and the quality of the data in terms of signal-to-noise ratio and frequency content. Hereafter, we present a novel approach to seismic data collection named Dispersed Source Array (DSA) acquisition. It is proposed to replace traditional broadband sources with a set of devices dedicated to different and complementary frequency bands. Modern multiple driver loudspeaker systems are based on the same key concept and their improved performance is demonstrated.
During field operations, it is often impossible to accurately implement nominal survey geometries in practice. Frequently, acquisition geophysicists are required to cope with unforeseen circumstances such as obstacles in the field and inaccessible or restricted areas. These complications may compromise the quality of the data or lead to delays, and thus extra expenses, during acquisition. In this thesis, we propose two automated approaches to survey design focused on avoiding spatial discontinuities in the recorded
data and on guaranteeing adequate data quality. The two methods are based on the reorganization of regular (centralized) and irregular (decentralized) source acquisition grids, respectively, and provide a practical acquisition plan for seismic crews. In this thesis, based on theoretical considerations and numerical data inversion and imaging examples, the feasibility of Dispersed Source Array acquisitions is demonstrated. Additionally, we show that it is possible to reliably recover subsurface information based on irregularly sampled datasets. We show how, despite the significant mismatch between baseline and monitor survey geometries, decentralized DSA surveys are also suitable for time-lapse studies.