2018/07/05 by Håkon Tjelmeland, Xin Luo, Tjelmeland, Håkon +3
Engineering · Computer Science · #Reservoir Engineering and Simulation Methods #Geochemistry and Geologic Mapping #Hydrocarbon exploration and reservoir analysis
paper · pdf · doi:10.48550/arxiv.1807.01902
We consider a Bayesian model for inversion of observed amplitude variation\nwith offset (AVO) data into lithology/fluid classes, and study in particular\nhow the choice of prior distribution for the lithology/fluid classes influences\nthe inversion results. Two distinct prior distributions are considered, a\nsimple manually specified Markov random field prior with a first order\nneighborhood and a Markov mesh model with a much larger neighborhood estimated\nfrom a training image. They are chosen to model both horisontal connectivity\nand vertical thickness distribution of the lithology/fluid classes, and are\ncompared on an offshore clastic oil reservoir in the North Sea. We combine both\npriors with the same linearised Gaussian likelihood function based on a\nconvolved linearised Zoeppritz relation and estimate properties of the\nresulting two posterior distributions by simulating from these distributions\nwith the Metropolis-Hastings algorithm.\n The influence of the prior on the marginal posterior probabilities for the\nlithology/fluid classes is clearly observable, but modest. The importance of\nthe prior on the connectivity properties in the posterior realisations,\nhowever, is much stronger. The larger neighborhood of the Markov mesh prior\nenables it to identify and model connectivity and curvature much better than\nwhat can be done by the first order neighborhood Markov random field prior. As\na result, we conclude that the posterior realisations based on the Markov mesh\nprior appear with much higher lateral connectivity, which is geologically\nplausible.\n