Numerical simulators of upslope flow forming on the lee side of a heated mountain ridge showed this flow to be a transient phenomenon, in agreement with observations. The simulations, performed with a two-dimensional, dry version of the cloud model of Tripoli and Cotton, included a nocturnal inversion layer (cold pool) and ridgetop winds. Cross sections of potential temperature and vertical profiles of potential temperature and horizontal winds reproduced observations well. Runs with and without the inversion layer showed that this layer was important in the formation of upslope winds near the ground when ridgetop-level winds were present: the runs without an inversion layer never developed upslope flow on the lee slopes. Runs in which the upper-level winds were varied showed that the duration of the upslope flow on the lee slope bore an inverse relationship to upper-level wind speed: runs with stronger winds had shorter-lived upslope flow. A set of observations from South Park, Colorado supported this laser conclusion. The conclusion implies that, on days with strong ridgetop-level winds, the leeside convergence zone mechanism proposed in the earlier observational study will not be as effective in initiating and sustaining deep convective clouds in the mountains.
An evaluation of the terms in the horizontal equation of motion showed that the initial push starting the upslope winds along the lee slope in all of the runs came from the pressure-gradient force, as expected, and a period of steady-state upslope then followed this push. The subsequent transition to downslope flow described in the observational studies, however, occurred through different processes, depending on upper-level wind speed. With stronger winds aloft, downslope winds first mixed downwards and then advected horizontally from higher elevations. With weaker winds aloft the transition occurred mostly because of the downslope propagation of a surface pressure minimum (low). Following the transition in all of the runs, near-steady downslope existed until the end of the simulation.