This talk will present results from preliminary simulations of flow over complex, self-similar boundaries with a dynamically determined drag coefficient. In traditional simulations of flow over complex terrain with unresolved surface features, the effect of the small-scale geometry is often parameterized using a drag coefficient determined in an ad-hoc fashion. Conversely, by defining an appropriate test-filtering operation, we show that the drag coefficient can be obtained dynamically from the smallest resolved scales of the surface, in a similar framework to that used in the dynamic model of Germano et al. (1991). The approach assumes scale-similarity of the boundary and also assumes a simple Reynolds number dependence for the drag coefficient. Results are obtained by performing numerical simulations of flow over Koch curve-shaped geometries of varying order, over a range of Reynolds number. In the high Reynolds number cases, a simple mixing-length turbulence model is used tentatively, along with an assumed inertial scaling (i.e. Reynolds number independence) of the drag coefficient. A priori tests done with fully resolved geometry show reasonable agreement between exact drag coefficients and dynamically determined values, at certain scales. A posteriori tests done with partially resolved geometry and a dynamically determined drag coefficient will also be discussed. The present preliminary simulations are limited to 2-D flow over 2-D geometry, and are performed on a structured grid. Imposed body forces (Mohd-Yusof, 1997) are used to represent the effect of the geometry. Several interpolation schemes are tried with the body forces to accurately represent the desired boundary conditions.