Quantitative databases storing analog data describing the geometry of sedimentologic features are commonly used to derive input for geostatistical simulations of reservoir sedimentary architecture; however, geometrical information alone is inadequate for the detailed characterization of sedimentary heterogeneity.
A relational database storing fluvial architecture data has been developed and populated with literature- and field-derived data from modern rivers and ancient successions. The database scheme characterizes fluvial architecture at three different scales of observation—recording style of internal organization, geometries, and spatial relationships of genetic units—classifying data sets according to controlling factors (e.g., climate type) and context-descriptive characteristics (e.g., river pattern). The database can therefore be filtered on both architectural features and boundary conditions to yield outputs tailored on the system being modeled to generate input to object- and pixel-based stochastic simulations of reservoir architecture.
When modeling heterogeneity with stochastic simulations, the choice of input parameters quantifying spatial variation is problematic because of the paucity of primary data and the partial characterization of supposed analogs. This database-driven approach permits the definition of various constraints referring to either genetic units (e.g., architectural elements) or material units (i.e., contiguous volumes of sediment characterized by the same value of a given categorical or discretized variable; e.g., same lithofacies type, clay and silt content, and others), which permit the realistic description of fluvial architecture heterogeneity. Applications of this database approach include the computation of relative dimensional parameters and the generation of auto- and cross-variograms and transition-probability matrices, which are necessary to effectively model spatial complexity.