The distribution of channel deposits in fluvial reservoirs is commonly modeled with object-based techniques, constrained on quantities describing the geometries of channel bodies. To ensure plausible simulations, it is common to define inputs to these models by referring to geologic analogs. Given their ability to reproduce complex geometries and to draw upon the analog experience, object-based models are considered inherently realistic. Yet this perceived realism has not hitherto been tested by assessing the outputs of these techniques against sedimentary architectures in the stratigraphic record.

This work presents a synthesis of data on the geometry of channel bodies, derived from a sedimentologic database, with the following aims: (1) to provide tools for constraining stochastic models of fluvial reservoirs in data-poor situations, and (2) to test the intrinsic realism of object-based modeling algorithms by comparing characteristics of the modeled architectures against analogs.

An empirical characterization of the geometry of fluvial channel bodies is undertaken that describes distributions in (and relationships among) channel-body thickness, cross-stream width, and planform wavelength and amplitude. Object-based models are then built running simulations conditioned on six alternative, analog-informed parameter sets, using four algorithms according to nine different approaches. Closeness of match between analogs and models is then determined on a statistical basis.

Results indicate which modeling approaches return architectures that more closely resemble the organization of fluvial depositional systems known from nature and in what respect. None of the tested algorithms fully reproduce characteristics seen in natural systems, demonstrating the need for subsurface modeling methods to better incorporate geologic knowledge.