Every interpreter knows that the bandlimited nature of seismic data limits its resolving power. We have all seen the picture of a rock face with a seismic wavelet superimposed, giving a salutary reminder of just how difficult the task of connecting seismic with geology is going to be. But what is resolution? Much of the time we think of ‘vertical resolution’, the notion that there is a minimum thickness of a geological unit for which the top and base can be uniquely identified. Of course, Widess (1973) is required reading for every young interpreter and the quarter-wavelength criterion for tuning thickness should be seared into their brains. Sheriff (1977) showed that there is also a spatial dimension to the problem, that the seismic reflection from a boundary is the result of constructive interference over an area of the wavefront (the Fresnel zone) defined by the quarter wavelength. This was a serious issue when we generated prospects using unmigrated 2D data, requiring map migration by hand!
So it is all down to wavelength; no wonder then that the favourite mantra of the processing gurus at Britoil in the early 1980s was ‘resolution is bandwidth’.‘Must remember that’ I thought, but what is bandwidth? ‘Octaves,’ they muttered and the seismic inversion experts nodded sagely, knowing that more octaves on the low end means better inversions. But, like all fundamental issues, there is more than one way of looking at it. Octaves, yes, but also (and critically) wavelet shape. Koefed (1981) taught us that. It has also become clear from an understanding of AVO that resolution can also depend on stacking angle. Fascinating…
3D acquisition and migration was a massive game changer, apparently consigning the Fresnel zone to history. The dramatic improvements in imaging can be considered an improvement in resolving power, but the main benefit of 3D is the increased spatial context of one reflection to another. All the advantages of 3D analytical techniques have accrued from the ability to present the data in map form. Increasingly, it has become clear that the very definition of ‘resolution’ needs to be broadened to include the detection of features on maps which can be assigned a geological significance.
A favourite illustration is an early time lapse model published by Archer et al.(1993).A simple synthetic 3D model was constructed to model the effect of movement in the gas–oil contact (GOC) in the Oseberg reservoir. To make the results realistic, noise was added to the model such that the signal:noise ratio is 1.5 (pretty severe but never mind).The ‘signal only’ difference section shows that the expectation is a simple tuning response, but on the noisy section it is difficult to appreciate quite where the signal is. However, a time slice through the difference cube (position shown by the dashed line in the section) shows that even in the presence of noise the map lineaments associated with the contact rise can be interpreted (almost exactly).This is a great illustration of the increased dynamic range of maps over sections, and it is a good lesson to learn that interpretation is an iterative process between vertical sections and maps. Don’t treat 3D as glorified 2D and spend months looking at vertical sections before making maps of amplitude and other attributes.
Archer, S, G King, R Seymour, and R Uden (1993). Seismic reservoir monitoring — the potential. First Break 11, 391–397, DOI 10.3997/1365-2397.1993020.
Koefed, O (1981). Aspects of vertical seismic resolution. Geophysical Prospecting 29, 1–30, DOI 10.1111/j.1365-2478.1981.tb01008.x.
Sheriff, R (1977). Limitations on resolution of seismic reflections and geological details derivable from them. In: Payton, C ed., Seismic stratigraphy — Applications to hydrocarbon explorations. AAPG Memoir 26, 3–14.
Widess, M (1973). How thin is a thin bed? Geophysics 38, 1176–1180, DOI 10.1190/1.1440403.