WEB-AVO Wisdom #1.0: What makes WEB-AVO unique when compared to conventional AVO

All right, recently we have promised to publish a number of technical articles in order to explain the unique features of WEB-AVO in more detail, so here we go.

Let’s start by the very basics: What does the acronym WEB-AVO stand for and what justifies the right of this technology to exist in a field in which conventional AVO technologies are well established since several decades?

WEB-AVO stands for “Wave-Equation Based Amplitude Versus Offset” and it is the wave-equation that plays an integral part for any technology/service we develop and offer at Delft Inversion. In the following I’ll talk (or more precisely write) you through the above Figure to elaborate on the role of the wave-equation in the context of seismic AVO inversion.

First off, it should be realised that the colours in Fig.1 are indicative for different data types. Turquoise refers to seismic wave-fields that “live” in the time-domain, red describes different processes and the yellow box stands for the desired reservoir model, hence elastic subsurface properties in the depth domain. This already brings us to an important point of WEB-AVO. While input to the technique is seismic data in the time-domain, output is elastic subsurface properties in the depth-domain. It is the wave-equation that connects both worlds and intrinsically converts time-domain data into depth-domain properties. Most of you may be more familiar with the procedure from wave-equation based forward modeling techniques, e.g. finite-difference (FD) modeling. Here, input to FD modelling is subsurface models in depth and output is synthetic seismic data in the time-domain. Again, it’s the wave-equation doing the magic just that for WEB-AVO the concept is applied in an inverse sense.   

The two turquoise ellipses on top of Fig.1 describe the input to WEB-AVO (or any other AVO technique, since the data requirements are close to being identical). On the left we have an incident seismic wave-field that propagates in a smooth background model. The incident field requires seismic wavelets coming from well-ties and various techniques exist to create a smooth background from well data (structurally guided interpolation, depth trends, etc.). On the right we have migrated seismic data, which most probably has undergone a sequence of seismic processing steps, ideally with a focus on preserving the true AVO of the data.

These datasets are then input to some AVO inversion (red box on the right) which usually is based on a linearised data model, hence propagation of waves in the non-scattering smooth background model and consequently primary reflections only. Various commercial/non-commerical implementations of seismic AVO inversion exist and they all may differ in the parameterisation of the desired reservoir model (AI, SI, Vp/Vs, and others) and/or deployed regularisations (sparse spike, blockiness, etc.) but, and let me be provocative here, all of them are based on the assumption of primary reflections only. As a consequence, the obtained reservoir model will be an approximation of reality; since seismic field data is inevitably much more complex than could be explained by this simplification. I am fully aware that a significant budget is spent on data processing with the aim to remove multiples from the seismic data but there is a difference between making your gathers look flat and making them consistent with the assumption of linearity. Forgive me for not getting into this topic in great detail right now but we may make this a specific topic of a future contribution to this series of technical articles.

In the remainder of the current article I prefer to focus on how WEB-AVO aims to overcome this assumption by introduction of the wave-equation into the field of seismic reservoir characterisation. Having an initial reservoir model we can use these properties to update the seismic wave field within the subsurface (red box at bottom). We do this by adding another order of scattering to the incident wave field that originally propagated in a smooth background medium only. At this stage I hope the following remarks to be valuable:

• While in the current context, “orders of scattering” refers to a mathematical term, effectively physical multiple scattering, including mode conversions and transmission, will be included in the underlying data model.
• The so-called seismic field update will be performed for each grid cell within the reservoir model. The resulting field can be interpreted in its own right and gives interesting insights into the potential presence of multiple scattering and mode conversions including the respective geological interfaces that create these wave complexities.
• WEB-AVO will equally well function in an environment with low contrasts between geologically units and where linearity may be a valid approximation. The wave-equation intrinsically tells if and to what degree seismic complexities are relevant in the current play by making wave-propagation consistent with the subsurface properties. 
• By solving the wave-equation WEB-AVO becomes a true amplitude inversion technique (within the limits of the implementation) with the ambition to offer the most realistic physical AVO engine to the seismic reservoir characterisation communities within the oil&gas, geothermal and CO2 storage industries. 

After the first order of scattering has been included in the total field, the linear AVO inversion is repeated and an updated reservoir model is obtained based on the latest known total field. This alternating procedure of including higher-order terms in the seismic wave-field followed by AVO inversion of the seismic dataset is repeated until neither the total field nor the reservoir properties do change any more and consequently convergence is reached. For synthetic data the procedure may be repeated up to 15 or even 20 times but on field data experience tells us that five iterations are in most cases sufficient.

While WEB-AVO is computationally several times more expensive than conventional, linearised AVO techniques, there are several business implications that create value and justify the additional effort:

• More accurate and reliable reservoir models can be obtained by extracting more information from existing seismic data. Or phrased differently, by turning what is considered as noise these days into useful information. In practice this feature has led to improved interpretability of areas where interference between some shallower multiple generators and the reservoir section has originally created difficulties.
• WEB-AVO has the potential to retrieve spatially broadband (quantitative) subsurface properties from temporally band-limited seismic. This bandwidth extension is a unique feature of solving the wave-equation and cannot be achieved by linear AVO techniques where the bandwidth of the reservoir model is directly determined by the bandwidth of the input seismic. In practice this feature has been used to successfully discriminate residual gas from commercial gas which could previously not be achieved using linear AVO techniques.
• Successful estimation of quantitative amounts of reservoir fluids from surface seismic data. This can potentially have a severe impact on field development planning and production strategies.

While this article is supposed to be the kick-off for an entire technical series about WEB-AVO I hope that you enjoyed reading and that, despite the shortness, I could trigger some interest in what we at Delft Inversion consider to be the next-generation AVO technology.

Bye for now,

Peter

If you are interested in hearing more about wave-equation based AVO inversion and how its unique technological features can potentially create value within your asset portfolio, please contact me or my colleague Panos Doulgeris or visit www.delft-inversion.com for more information. Furthermore, in case you find this article informative, I would appreciate if you could share the post within your network and if you would become a LinkedIn follower of Delft Inversion to be notified about future developments.