Seismic lacks low frequencies, so for an absolute seismic inversion a low-frequency model (LFM) is required. Starting from an empty LFM, interpreters would like to post sand velocity values where there is sand, shale velocity values where there is shale, etc.

But unless there are very favorable circumstances, interpreters don’t know in any great detail where the facies differentiations are located in the subsurface. Therefore, an LFM is typically compromised, and during the inversion the seismic cannot “fix” a compromised LFM because it lacks low frequencies.

Facies-based inversion

An advanced approach to a better LFM is to input an LFM for each of the facies expected and to let the inversion decide what the ultimate LFM is. In other words, the LFM used is an output, not an input.

This new inversion involves two steps:

• Invert the seismic for impedances given the facies; and

• Invert the impedances for facies.

The two steps help one another. Clearly, better impedances lead to a better facies estimate. The reverse is also true because in each execution of the impedance inversions, the LFM is recreated based on the facies estimate.

The implementation is a Bayesian one. Therefore, multiple equiprobable facies realizations can be generated for subsequent uncertainty analysis.

Application from exploration to production

The parameterization of facies-based inversion changes from exploration to production. In exploration the inversion window is usually quite large and the number of zonations in that window are quite small, so the number of facies inverted will be small as a result. In production the reverse is true: a small inversion window targeted on the reservoir, many zonations and a larger number of specific facies (Figure 1a).

FIGURE 1a. Left, the log and seismic traces are shown at the well location. The middle trace shows an exploration inversion, in which the interpreter may decide on only two faces (sand and shale), the trends/LFMs of which are shown. Note there are three zonations. Right, in a development or production setting, interpreters may decide on only four facies: clean sand, dirty sand, soft shale or hard shale. (Source: Ikon Science)

In one exploration case study Ikon inverted the Willem survey offshore Northwest Australia. The dataset consisted of 2,400 sq km (927 sq miles) of seismic (four partial angle stacks) with only two wells with elastic logs (and a further three wells without elastic logs and one well, Pyxis-1, which very little was known other than it was a gas discovery).

Figure 1b (top) shows the facies-based inversion result (impedances are not shown are the discrete facies) on an arbitrary line through the Pyxis-1 discovery. For the sake of comparison, the company also ran a model-based simultaneous inversion, which requires an LFM as an input. This, however, only gives impedances, and so interpreters subsequently derived facies using Bayesian inversion (Figure 1b/bottom).

FIGURE 1b. The facies-based inversion (top) and model-based simultaneous inversion followed by Bayesian classification to facies (bottom) of the Willem 3-D survey are shown. Grey is shale, yellow is water-bearing sand, red is gas-bearing sand, blue=limestone and purple is marl. (Source: Ikon Science)

The facies-based inversion result looks more credible. This can be substantiated by 1) inspecting the two ellipses (the faciesbased inversion identifies a gas-water contact) and 2) looking at the two arrows. The facies-based inversion images a continuous gas-bearing sand and predicts an 18.2-m (60-ft) gas column at Pyxis-1, and later the interpreters learned that they were only 1 m (3 ft) short. Simultaneous inversion followed by Bayesian classification does find the gas leg, but it is broken up, has water-bearing sand (yellow) above and below the gas-bearing sand (breaking hydrological rules), and the gas column is too small.

Production setting

In the mature Forties Field the objective of seismic inversion is to assist in locating untapped hydrocarbons (bypassed pay, undrilled fault blocks, etc.). The first 3-D survey was shot in 1988 and forms a baseline for 4-D studies. Five monitor surveys have been acquired, the last of which was in 2013.

Facies-based inversion as discussed so far is a 3-D inversion, and therefore this technique was applied to the 1988 and 2013 surveys individually. Figure 2 shows a map view of the facies distribution in 1988 (top) and 2013 (bottom).

The massive sweep signature to the southwest is evident, but detailed analysis shows finer details also (e.g., “halos” around water injectors).

FIGURE 2. This facies image shows the top 20 m (66 ft) of the Forties reservoir in 1988 (top) and 2013 (bottom). (Source: Ikon Science)

Look ahead

Facies-based seismic inversion is powerful because facies typically have distinct elastic properties, but they usually have distinct resistivity, too. Therefore, a facies-based seismic and controlled-source electromagnetic (CSEM) inversion makes sense.

The industry has seen a rapid adoption of facies-based inversion over the full life cycle of an asset since it opens the door to new subsurface workflows. Facies-based seismic and CSEM inversion as well as facies-based 4-D inversion and facies-based anisotropic inversion have been successful, and more are in the pipeline.

References available. Contact Rhonda Duey at rduey@hartenergy.com.

Acknowledgment: Ikon Science thanks Tullow Oil Plc and CSIRO for their cooperation.