A method of predicting values of formation parameters (e.g., compressional
velocity, density, pore pressure, and fracture pressure) as a function of
depth includes generating an initial prediction of a profile of the
formation parameters and uncertainties associated therewith using
information available regarding the formation, obtaining information
related to the formation parameters during drilling, and updating the
uncertainties as a function of the first prediction and the information
obtained in a recursive fashion. Known equations are used for finding
initial values, and uncertainties associated therewith are quantified by
using probability density functions (PDFs). A Bayesian approach is
utilized where "prior PDFs" describe uncertainty prior to obtaining
additional information, and "posterior PDFs" account for the additional
information acquired. As additional information is acquired while
drilling, the posterior PDFs are redefined. Uncertainty in the formation
parameters is quantified by sampling posterior PDFs given all the data
with a Markov Chain Monte Carlo algorithm which generates numerous
formation parameter profiles consistent with the data and the computed
Bayesian uncertainties. Histograms of the numerous formation parameter
profiles may be plotted to visualize the uncertainty in the formation
parameters.
