AbstractSeveral tests of a model with a cut‐cell representation of orography are presented: a resting atmosphere test, advection across a hill and a warm rising bubble over hills with different gradients. The tests are compared with results from terrain‐following models. Results indicate that errors associated with terrain‐following coordinates are reduced, in some cases greatly reduced, with the cut‐cell approach. In a resting atmosphere, the cut‐cell approach does not generate flow around an isolated hill however steep the terrain. Relative errors in a rising bubble test are an order of magnitude smaller than terrain‐following simulations.