Bayesian sequential indicator simulation of lithology from seismic data

摘要

Method for generating simulated models of discrete lithoclass values on a regular array of pixels by combining known lithologic data at a number of distributed control pixels corresponding to well locations, with seismic attributes gathered from stations associated with the pixels, comprises: (a) measuring petraphysical properties of several earth layers; (b) calculating the conditional probability distribution of the seismic attributes for each lithoclass; (c) randomly selecting an as-yet unsimulated pixel; (d) estimating a prior probability distribution function of the lithoclasses at the selected pixel by combining lithologic data at nearby control pixels; (e) calculating a likelihood function of the lithoclasses in the selected pixel from the lithoclass-conditional distribution of seismic attributes; (f) calculating a posterior distribution function for the lithoclasses in the selected pixel by combining the prior distribution function with the likelihood function; (g) drawing a simulated lithoclass valve in the selected pixel by sampling at random from the posterior probability distribution function; (h) entering the simulated lithoclass value into the array as an additional control pixel; (i) repeating steps (b) through (g) until lithoclass values are simulated for all pixels of the array; and (j) generating a lithological model of the earth's subsurface.