Seismic reflection imaging of near surface structures using the Common Reflection Surface (CRS) Stack Methodud

摘要

This PhD thesis aims to evaluate whether the udCommon-Refludection-Surface (CRS) stack method can be considered as a more cost efficient processing alternative to the conventionally used Common Midpoint (CMP) stack method for processing of shallow and ultrashallow reflection data.udThe CRS stack is a seismic imaging method established for oil and gas exploration that is similar in concept to the conventional CMP stack method. Unlike the CMP stack,udthe CRS stack process is not confined to single CMP gathers (offset direction), but also includes neighbouring CMPs (midpoint direction) into the so-called CRS supergathers.udThe use of CRS supergathers enables stable data-driven", i.e. without human interactions, velocity analysis and residual static corrections, avoiding the poorly-automatedudand time-consuming processing steps that are instead required when implementing conventional CMP processing. This makes the seismic imaging process more compatibleudwith budgets available for near-surface geophysical investigations. Improving seismicudimaging of near-surface reflection data, while at the same time reducing processing costs and human interaction during processing was the principal objective which guidedudmy work. To investigate the advantages and limitations of exporting the CRS stack from the hydrocarbonudexploration field to the near-surface scale, I have firstly analysed and adapted the characteristics of the CRS to the requirements of near-surface reflection data. Then, I have compared the results (seismic sections and velocity fields) obtained by processingudwith the CMP and the CRS stack methods for two real field datasets (P- and SH-wave)and two synthetic datasets that exhibited very large vertical velocity changes. Finally,udI have proposed some original solutions that overcome several of the issues encountered when using CRS stack with near-surface data. The P-wave dataset was collected as part of a hydrogeological investigation with the aimudof delineating the hydrogeological framework of a paleolake environment to a depth of few hundred metres. Using the CMP method, several nearly horizontal reectors with onsets from 60 to about 250 ms were imaged. The CRS stack produced a stacked section with greater coherency and lateral continuity than the CMP section, but also spuriousudalignments of seismic energy which hinder interpretation. Weighing the CRS stacked section with the corresponding CRS coherence and number of CRS stacked traces leadsudto a considerable reduction of the spurious alignments, resulting in a seismic section more suited to delineate the aquifer and its confining units. The SH-wave ultrashallow dataset was collected to support a geotechnical study to auddepth of 10 m. The obtained CMP stacked section imaged a dipping bedrock interface below four horizontal reudectors in unconsolidated, very low velocity sediments. Theudvertical and lateral resolution was very high, so that despite the very shallow depth the resulting CMP stacked section showed the well-defined pinchout of two layers at less than 10 m depth. The CRS stack improved the continuity of the shallowest reector but showed an excessive smearing effect with some reector portions, including the pinchout,udunresolved and not as well defined as in the (very detailed) CMP counterpart. Restricting the CRS stack process to single CMP gathers, preserving the CRS-supergather for the search of stacking parameters, produced a time section very similar to the CMP counterpart. In both cases, I swiftly obtained the CRS stacked sections in a fully automatic way,udso with a cost/benefit ratio considerably more advantageous than that of the CMP sections, which required time-consuming prestack velocity analysis as well as residualudstatic corrections. Moreover, using the kinematic wave field attributes determined for each stacking operation I reconstructed velocity fields matching the ones estimated with the CMP processing, even if this required a greater amount of work than that required to produce the CRS stacked sections. Finally, using two synthetic datasets, I addressed the issue of the crossing reection events that appear in data acquired in soils characterized by strong vertical velocity gradients. Although a matter debate for decades, this is an issue still unresolved by useudof the conventional CMP method. Using the first synthetic dataset I showed that unlike in conventional CMP processing which cannot accurately process crossing reflection events without generating distortions and artefacts, the data-driven CRS stack imaging process considerably restricts their generation, limiting the reduction of signal-to-noise ratio and of temporal resolution in stacked traces. With the second synthetic dataset I simulated a data acquisition reproducing a case history with a high-velocity contrast in the first 5 m depth. The CRS results that I obtained from the modelled data demonstratedudthat the CRS stack method may be a reliable alternative for processing crossing reection events, definitely easier and faster than the construction of complicated velocityudfunctions and/or the separated processing of the crossing events. By comparing these results with those obtained using the CMP method I obtained other interestingudresults, which, however, to validate would necessitate the use of real datasets. The findings of this present study demonstrates that the CRS stack could represent audsignificant step forward for the reduction of the costs involved in shallow and ultrashallowudseismic reflection data processing, one which does not compromise the quality of results. Both these conditions being essential to the increased acceptance of the seismic reection method as a routine investigation method for use in shallow and ultrashallow seismics.