Considering the difference in the information from a borehole and from seismic; the boreholes have very detailed information, but from basically a point, whereas the seismic covers thousands of square kilometers with relatively low-resolution information.
For example, at the Base Triassic in the Eastern Barents Sea, velocity is 4000m/s and the frequency of the data is 25HZ, giving a resolution in depth of around 100 meter. The borehole log-values however, are recorded in decimeters, - unfortunately there is a hundred kilometer between each borehole in this area.
From the perspective of seismic acquisition and processing, the approach would be to boost the frequency content of the seismic. 15 years ago, there was briefly a fad that claimed seismic could be boosted to several hundred hertz, and with the more recent innovation of de-ghosting some set the limit of resolution at the acquisition sampling rate – 2ms giving 250HZ.
But as we know, when you boost frequencies you boost noise as well as signal. And secondly rocks suck up high frequencies. The deeper you get, the more, low frequency of the signal.
This is where inversion comes in. Inversion tries to model the geology that best fits the seismic recordings. In inversion the detectability is higher than the seismic resolution. People who do inversion have a lot of things to bring to the market, but for a seismic processing guy like myself, it is this increased detectability that is interesting. The only disconnect is that inversion seismic is the key ingredient and well is soft data. This sometimes leads to conflict between the log and seismic data wherein seismic reigns supreme.
WCA (Waveform Cluster Analytical) Inversion, claims to increase detectability by a factor of 8 over standard deterministic inversion – which remains to be shown, of course. It is made for reservoir studies in developed fields, and is dependent on a high density of boreholes in clastic depositional environment and specifically works for thin bedded sands and shales.
The WCA builds up a catalogue of wave-forms corresponding to different log sequences and then searches for these inside the constraints of the interpretation of the seismic volume. As I understand it, WCA doesn’t model the response of the acoustic impedance and tries to optimize this, it is more of a pattern-recognition approach where you could search for and model any log value that has some correlation to the seismic signal, for instance SP, GR etc. WCA assures the use of well data to be key for pattern-recognition in the near well-bore area.
For a dilettante like me, who only sees inversion as seismic with colors, this seems like a very fresh and promising approach. And it is still only one solution and is way more cost efficient than your stochastic inversions.
Comments