A new approach to describing the associated topography at different
scales in computational fluid dynamic applications to gravel bed rivers
was developed. Surveyed topographic data were interpolated, using
geostatistical methods, into different spatial discretizations, and
grain-size data were used with fractal methods to reconstruct the
microtopography at scales finer than the measurement (subgrid) scale.
The combination of both scales of topography was then used to construct
the spatial discretization of a three-dimensional finite volume
Computational Fluid Dynamics (CFD) scheme where the topography was
included using a mass flux scaling approach. The method was applied and
tested on a 15 m stretch of Solfatara Creek, Wyoming, United States,
using spatially distributed elevation and grain-size data. Model runs
were undertaken for each topography using a steady state solution. This
paper evaluates the impact of the model spatial discretization and
additional reconstructed-variability upon the spatial structure of
predicted three-dimensional flow. The paper shows how microtopography
modifies the spatial structure of predicted flow at scales finer than
measurement scale in terms of variability whereas the characteristic
scale of predicted flow is determined by the CFD scale. Changes in
microtopography modify the predicted mean velocity value by 3.6% for a
mesh resolution of 5 cm whereas a change in the computational scale
modifies model results by 60%. The paper also points out how the
spatial variability of predicted velocities is determined by the
topographic complexity at different scales of the input topographic
model.