Stratigraphic modeling based on physical and geologic principles has been i
mproved by more sophisticated process models and increased computer power.
However, such efforts may reach a limit in their predictive power because o
f the stochastic, multiscaled nature of the physical processes involved. Bu
ilding on techniques from the geostatistical literature, a conditional simu
lation method, dubbed "SimStrat," has been developed to improve predictions
of stratigraphic architecture from limited ata. No physical processes are
invoked. Rather, the prediction is based solely on geometric and statistica
l principals. The method takes as input sonar bathymetry, seismically defin
ed stratigraphic horizons, and core-defined horizons. Each stratigraphic ho
rizon is characterized using spectral modelling and coherence modeling for
adjacent horizons. Predictions of subsurface horizons are improved where se
afloor bathymetry conforms with the underlying strata. Conditional simulati
ons can then be generated that conform to available data constraints and st
atistical characterization. Tests with synthetic data in one and two dimens
ions for differing spectral models confirm the reliability of the method.