Date(s) - 05/12/2020
11:30 am - 1:00 pm
Rescheduled for September 1, 2020
Geological facies extracted from seismic data guided by well information and the geologist’s insight
Kim Gunn Maver
Seismic AVO data is the preferred technology for mapping the subsurface between wells. Acquisition has become very cost effective and, with recent advances like broad-band seismic and full azimuth ocean bottom seismic, have made it possible through seismic inversion to predict subsurface rock and fluid properties as well as pressure and fluid changes from time-lapse seismic data.
When doing AVO seismic inversion for elastic properties the variation in rock properties are only partly resolved and as a result many facies and fluid configurations are similar in the seismic domain. The seismic domain space is also horizontally and vertically “unaware” of for example the geological ordering (young above old bedding), porosity distributions, depth trends and fluid ordering (gas/oil/water). Geological knowledge is also difficult to quantify and in a consistent manner integrate in the seismic inversion process. Finally, reservoir related decision-making and risk analysis requires an increasingly degree of assessment of the uncertainties associated with any interpretation and statements based on the seismic data, which is not an integral part when inverting seismic data for elastic properties.
Direct Probabilistic Inversion makes it possible to design a geological framework of prior information within which the seismic AVO data is transformed to provide reliable results by handling many of the described limitations and uncertainties in the seismic AVO data when performing an inversion for elastic properties.
It is a one-step inversion process, which honors multi domain inputs and assumptions respecting the confidence in these inputs by using a fully Bayesian probabilistic formulation. A key attribute of the Direct Probabilistic Inversion process is the definition of a geological framework of prior information, which is flexible and defined as probability density functions, hence the more likely solution has a higher probability and the less like solution has a lower probability. This enables an optimal propagation of uncertainty and handling nonuniqueness by
The probabilistic inversion problem is solved by localization of the problem by defining natural neighborhoods in the rock physics-, elastic- and seismic- domain, which influence a certain point in space the most. In the initial state a subsurface model is defined based on the data from the different domains. This is updated with the information from angle stack data, which measures, in terms of probability, the misfit between forward modelled seismic data and the processed angle stacks. The likelihood of the model being correct contains, in addition to a seismic noise model, the combination of a statistical rock physics model from facies to elastic property domain, and a seismic convolutional AVO forward model from elastic properties to the seismic angle stack domain.
The result of the Direct Probabilistic Inversion is a probability volume for each of the defined facies and all thefacies can also be combined into a final volume of the most likely facies distribution with corresponding probabilities.
Case studies will be used to present Direct Probabilistic Inversion process.
Kim Gunn Maver has extensive global oil and gas service company experience within management, strategy and sales. He has an Ph.D. in geology from Copenhagen University and MBA from Copenhagen Business School. He was managing director of the seismic inversion company Ødegaard, when the company was acquired by Schlumberger in 2006 and he stayed on in various management positions. In 2009 he started as VP of sales and Marketing at the OBC company RXT and later moved to Spectrum where he served in various Vice President positions. Since 2018 Kim has work as an executive consultant and are currently working with Qeye Labs developing the US market for QI consultancy services.