“Our greatest weakness lies in giving up. The most certain way to succeed is always to try just one more time.” Thomas Edison
The marine air-gun source was widely adopted by the seismic industry in the 1960s and soon became the source of choice because of its flexibility and superior safety over explosive sources. The physics of the source is governed by the interaction of the various elements such as number and volume of the guns, and the pressure and depth in terms of hydrostatic pressure. The interaction with the sea-surface which acts as a perfect reflector creates the well-known ghost effect, and this affects effective bandwidth as well as the radiation pattern of the source.
Towards a New Paradigm
Conventionally tuned air-gun arrays consist of numerous single guns and clusters (two or more guns that are fired close together) of different sizes. When fired synchronously, the difference in bubble periods leads to destructive interference and hence primary energy is enhanced and bubble energy is attenuated. While this is desirable from the point of view that we seek a sharp pulse with minimum trailing energy to represent the various echoes from the subsurface, it has the undesirable consequence that it reduces the effective amount of low frequencies. Peak to bubble ratio remains one of the key parameters in marine air-gun design together with absolute strength and source radiation pattern (see Broadband Seismic Technology and Beyond PART I).
With the advent of broadband receiver solutions (Carlson et al.) the industry has started focusing on source solutions that will further increase the bandwidth of seismic data. This is driven by the need for low frequencies for improved deep penetration as well as the recognition of the value of the complement of both low frequencies as well as high frequencies in inversion and estimation of elastic parameters in the earth. New processing techniques like full waveform inversion have further emphasised the need for even lower frequencies all the way down to 1 Hz. In this context, it is paradoxical that we adopt methodologies on the source side which intrinsically diminish the low frequencies through the attenuation of the bubble characteristic of tuned arrays. The industry has, up till now, applied the criterion of peak-to-bubble maximisation in order to create a high resolution wavelet without regard to what it does to the low frequency response. This ignores the fact that the presence of low frequencies greatly reduces the side-lobes of the seismic wavelet and therefore significantly improves the effective resolution power, given that the high frequencies are there.
Optimising Low Frequencies
As indicated in the introduction, the response or strength of air-gun sources is determined by number of guns, volume and pressure, approximately as cube root of the last two. The Rayleigh-Willis approximation describes the bubble period in terms of the factors above and in addition as being inversely related to hydrostatic pressure. This fact is crucial when we attempt to improve the low frequency response of source elements consisting of traditionally tuned arrays, although the second element which determines the air-gun response is the interaction with the free surface, i.e the ghost effect, which means that low frequency response will improve with increasing depth. The net effect can be close to a zero sum game for tuned arrays, albeit somewhat dependent on the nature of the array. In fact if one looks at the response of a single gun of considerable size when the bubble energy is not attenuated, the low frequency energy is best preserved at shallow depths.
Below the frequency of the source-ghost notch, the source spectrum can be shaped to optimise low frequency response as far as signal to noise allows it. Such procedures can be guided by the ghost function. Going across the source-ghost notch is more challenging, just as it is on the receiver side. If we consider source solutions with tuned arrays as a basic building block, there are a couple of methodologies that can be used to de-ghost the source. The ghost reflection can be attenuated by spreading it out in time, or over-under techniques can be used to de-ghost the source, in a similar manner to the techniques used for over-under streamers (see Improving Seismic Quality).
Synchronised Multi-level Source
The source ghost can be attenuated using a beam steering technique originally developed some 60 years ago for dynamite land acquisition (Shock, 1950). The principle is to detonate charges at various depths in a sequence that constructively builds the down-going wave at the expense of the up-going wave. This way the energy of the ghost (surface-reflected down-going wave) is reduced compared to that of the primary pulse. One can adapt the beam steering approach to air-gun arrays in the marine environment. We place guns, clusters of guns or sub-arrays at different depths and fire them sequentially. Contrary to the land dynamite case, the speed of sound in water is well known and varies little at the depth considered, and the trigger-time accuracy is in the order of a fraction of a millisecond. This technique is quite straightforward to implement and requires only minor modifications of the existing gun arrays.
A conventional air-gun array is made of several sub-arrays each containing a number of guns, or clusters of guns. All guns are at the same depth (typically between 5 and 10m) and fire simultaneously. This provides constructive down-going energy but also constructive up-going energy, and the ghost therefore has the same energy as the direct wave. The multi-level source concept puts guns, clusters or sub-arrays at different depths and fires them sequentially so that only the down-going waves build up constructively, as shown in the figure to the right. The up-going wave does not build constructively and the ghost effects are consequently reduced. The pure ghost effect will favour low and high frequencies whereas the middle frequencies are attenuated.
An important issue to consider with the multi-level source is the radiation pattern. The illustration clearly shows that a direction other than down-going also benefits from the beam steering. Although conventional source arrays have a radiation pattern and are far from being isotropic, this issue is more pronounced with the multi-level source. Array modelling is required to ensure the spectral benefits are not offset by unintended consequences.
Beyond the change in ghost behaviour, the air-gun signature is also affected by this new design. First, the source ghost is an effective attenuator of bubble pulses. Consequently, we expect the multi-level source to deliver a lower peak-to-bubble ratio (PTB) than a conventional source. Secondly, since the guns are towed deeper, they operate under higher hydrostatic pressure, which reduces the size of their bubble. Smaller bubbles mean less low frequency content, which can partly offset the gains demonstrated in the ghost signatures illustration. Thirdly, the reduced diversity in bubble sizes requires less parameter modifications to play with in array design and bubble pulse attenuation. Therefore, modelling is the key to ensuring an effective multi-level source array design.
A Fully De-Ghosted Source
The fact that a synchronised multi-level source just attenuates the ghost and creates an anisotropic radiation pattern encourages us to look for alternatives. What springs to mind is the over/under principle applied earlier for de-ghosting of streamers. One could achieve this by shooting a survey twice, using a shallow source and then a deeper source, but this would be quite costly. Alternatively, a deep and a shallow source could be fired in turn at the same surface location but this would then reduce the fold by 50% which is also highly undesirable. A fully blended solution with a randomised firing scheme offers an attractive alternative (Parkes et al.), as de-blending techniques have been in rapid development over the last decade (see e.g. Van Borselen et al.).
The GeoSource is a time- and depth-distributed source. The source array is divided into sub-sources, each of which consists of one or more conventionally tuned sub-arrays towed at different depths. A typical configuration is two sub-arrays at 9m depth and a single one at 5m, which creates a high degree of complementarity in the frequency domain. The arrangement of the sub-arrays is such that air in the water from the arrays fired first can only affect the subsequently fired arrays at very high angle, thus not impacting the emitted wave-field directed into the earth. At least one sub-source is fired with a time delay that is varied in a random fashion from shot to shot within a typical 1 sec. window. The randomised fire time delays allow us to separate the wave-fields emitted by the sub-sources, and the depth distribution of the sub-sources allows us to recombine these wave-fields in such a way that the source ghost is removed (Posthumus).
The methodology allows the source ghost to be removed in a robust way at an early stage in the processing, i.e. pre-stack. When combined with GeoStreamer, the resulting image is broadband and ghost free. After the source response has been compensated for, the wavelet in the resulting data has a flat amplitude spectrum without notches, and the phase spectrum is zero.
The intrinsic source response, which is now ghost-free, remains in the data. The form of this response is well-known and results from the oscillatory nature of air-gun bubbles. Due to inaccuracies in numerical modeling towards the very low frequencies, accurate near field measurements are needed to calibrate the modelled results. This applies to any source configuration trying to preserve the low frequency integrity.
Due to the randomised firing scheme it may also be possible to reduce the shot-generated noise towards the lower frequencies. GeoSource is an acquisition-based methodology and does not make any restricting assumptions about sea state. The challenge lies in the source separation, especially in shallow water and in the presence of strong diffractions.
GeoSource exhibits some similarities to, but also differences from a synchronised multi- level source. Due to complete de-ghosting, the radiation pattern is much more uniform and it represents a more robust alternative for source de-signature.
Challenging the Old Paradigm
Both solutions described above are based on tuned sub-arrays. The main contributor to low frequency response, the bubble, is intrinsically attenuated. Alternative array designs and tuning strategies may offer new opportunities to boost the low frequency part. One strategic element in such novel designs will be an accurate representation of the far-field signature through high quality near-field measurements. Technology is moving rapidly in this direction and will offer new opportunities. Alternative sources such as marine vibrators may complement our search for more low frequencies. The rapid development of techniques like full waveform inversions compels us to keep up the good work.(Source: Jan Langhammer)