Resumen

DURAN, Evert Leonardo; SOTO, María Beatriz  y  ALDANA, Milagrosa. Metropolis - hasting and gibbs by blocks: use of markov chain monte carlo algorithms to model stratigraphic columns, lama field, Maracaibo lake. Rev. Fac. Ing. UCV [online]. 2014, vol.29, n.4, pp.41-54. ISSN 0798-4065.

In this article we stochastically characterize a reservoir through the modeling of stratigraphic columns using two algorithms: Metropolis-Hasting and Gibbs sampling by blocks; both constitute Markov chain Monte Carlo (MCMC) methods. The stratigraphic unit considered corresponds to the C4 sands of the Misoa Formation, on Lama Field, at the Maracaibo Lake, consisting of inter-bedded sandstones with scarce intercalations of siltstone and shale. Using the information of 11 wells located in the study area, 20 new pseudo-columns were modeled in new locations by means of both algorithms. Combining the information of real columns and the estimated ones, maps of sand content were generated. The Metropolis-Hasting algorithm reproduced, mainly, the presence of sands along the pseudo-sequences generated, which represent more than 70% of the sediment present in the study area. This algorithm suggests the areas where the main variations of sand content could occur, but does not seem to be sensible enough to detect the facies variations of silt and clay. The maps obtained with the Gibbs-by-block algorithm, like that using Metropolis-Hasting, clearly show a NE-SW trend in sand content which correlate with the results of previous works in the area. Furthermore, the Gibbs-by-block algorithm developed here is capable of reproducing the presence of fine layers of the other lithologies observed in the area, if a geologically well thought-out length for the block is used.

Palabras clave : Metropolis-Hastings; Gibbs by blocks; MCMC algorithms; Modeling of stratigraphic columns; Markov; Lama Field.

        · resumen en Español     · texto en Español     · Español ( pdf )