One Dimensional Stochastic Inversion (ODiSI) is changing the game of inversion techniques by combining different approaches to estimate reservoir properties with limited information.
“Man muss immer umkehren," loosely translated into “invert, always invert,” has been the strategy for solving problems by great thinkers and mathematicians for centuries. Within reservoir characterization, inversion has also taken a similar position, as inversion techniques, to replace seismic signature by a blocky – more geological – response that corresponds to acoustic impedance in the layer boundaries, has been the preferred method of choice for geologists and geophysicists for decades.
Coming into popularity sometime in the late 1970s, seismic inversion takes all information provided about an oilfield through previously drilled wells – rocks encountered, geological make-up of rocks, inherent reservoir properties, and fluid properties – and piece together a hypothetical reservoir model to gain insight into how the reservoir may look like in unpenetrated locations.
However, any attempt to characterize the subsurface always brings along a high degree of uncertainty. Specifically, because seismic inversion is not a unique process, geoscientists regularly face significant difficulties when determining the level of confidence they can place in the predictions based on seismic inversion.
Transitioning from Traditional to Sophisticated Seismic Inversion Methods
Traditional seismic inversion methods transform seismic data into acoustic impedance, a measure of the ease with which a sound wave propagates through a particular medium. As acoustic impedance is the result of the product of two values, one for density and one for velocity, this transformation brings along specific implications, particularly a chance of biasing the assumption. As you never know how much there are of each value, there can be infinite possibilities.
More sophisticated seismic inversion methods, on the other hand, transform seismic traces into geological properties and add additional information to reduce uncertainties. Prior information stemming from interpretation at the well site is used as input, converting the seismic data not into acoustic impedance but into something more geological – a lithology volume – as well as additional information that controls the validity of the information.
Compared to traditional methods, this approach delivers a more detailed picture of possible lithologies, like sands or shales, which allows you to estimate future positions for drilling wells better. In other words, it reduces uncertainty.
Introducing One-Dimensional Stochastic Inversion (ODiSI)
This approach to seismic inversion has been dubbed One-Dimensional Stochastic Inversion or ODiSI for short. It combines the best of both deterministic and stochastic techniques to provide a robust estimate of your reservoir properties. ODiSI takes all the information from your known locations and builds it into a framework to produce a detailed model of your well, including how the seismic response should look at that location – not dissimilar from the conventional techniques.
However, ODiSI doesn’t just do this once. For each trace location, ODiSI creates a very large number of Pseudo-Wells (PW). The PW’s are generated from stochastic 1D stratigraphic profiles that drive the creation of a full log suite, much like a real well. Each PW is calibrated to real wells using a sophisticated model for the vertical statistics, a rock physics model, and then further constrained by prior expectations of the reservoir.
It does these computations a few tens of thousand times to provide a statistically robust understanding of the wells or vertical columns. ODiSI then takes a few hundred of the best outputs for that location, averages them, and validates this to the seismic information to ensure you get the best match for the seismic at that specific trace location. And because the process is averaging the best match pseudo-wells at each trace location, this also provides the associated uncertainties around each of the estimated reservoir properties.
ODiSI changes the game of seismic inversion by taking the regular spatial interval of a seismic survey, say every 25 meters, and repeats this process again and again for each trace location. At each trace position, ODiSI builds up a picture of the entire reservoir, one trace at a time, and independent of the previous trace.
Strictly speaking, ODiSI has now proven to deliver an extremely robust approach to estimating reservoir properties with the limited information available. So instead of letting your next reservoir characterization project run into the sand, why not use ODiSI to actually find that sand?