You can’t have your cake and eat it (too), but you can do two inversions

You will agree…

Seismic inversion methods are powerful tools for reservoir characterization, and when you combine them with rock-physics, the reservoir properties can be directly inferred.

However, Inversion methods got their pros and cons, such as:

 DeliverablesProsCons
Post-stack deterministic inversionP-impedanceQuick and robustLow resolution limited by bandwidth, initial model
Pre-stack deterministic inversionVarious elastic attributes, e.g. VP/VSMore physics, suitable for unconventionalLow resolution limited by bandwidth, initial model
Post-stack stochastic inversionP-impedanceHigh resolution, thin-bed detectionHard to control on the HD results

 

For unconventional, such as shale and carbonate, the geological challenges are more and more complicated, and a single inversion application can no longer solve our technical problems.

Challenges example

  1. Based on the rockphysics analysis, P-impedance itself can hardly separate the reservoir sands and shale, while VP/VS can.
  1. The thicknesses of the thin beds are between 3-5m, maximum 7m, seismic resolution here cannot detect its existence clearly.
  1. Conventional deterministic inversion will fail due to its resolution, and direct application of stochastic inversion is not wise due to lack of control on the sand/shale superstation (low frequency model)

Solution

Two stage inversion strategy.

  1. Use VP/VS inversion to determine the low frequency background of reservoir
  2. Use the HD stochastic inversion to determine the thin beds

Figure. The iRes-Geo two stage inversion workflows

Results and impact

 The Stage 1 VP/VS inversion can separate the sand and shale, providing a “low frequency” earth model at seismic resolution. This is earth model contains realistic distribution of the sand and shale in the reservoir.

Stage 2 Stochastic inversion is run based on this “low frequency” model and the uncertainty is constrained by the deterministic inversion. The whole process converges much faster than unconstrained runs.

With consistently compatibility in both pre-stack and post-stack results, the two stage workflow did a great job here to include as much seismic and well log data as possible, leading to a enhanced resolution and predictability for reservoir prediction.

Figure. Left Sand probability from VP/VS deterministic inversion. Right Sand probability from Stochastic Inversion

iRes-Geo is specialized in seismic reservoir characterization and modelling, please see our comprehensive mature field solutions here, you are only one click away from them.

Author:

Dr. Z.L., Ming. Reservoir Geophysicists Director at iRes-Geo Technology.