Wind tunnel experiments made to determine how rapidly dense gas is removed from a topographical depression by a crosswind are reported. The density and flow rate of the gas (input at the bottom of a V-shaped valley in otherwise homogeneous, flat terrain) were together sufficiently low to prevent pooling of the gas on the valley floor. The present work concentrates on cases for which the relevant Richardson number are relatively low. A simple theory is described, based on assumptions about the way in which the (slightly) heavy gas is removed by turbulent entrainment from the separated flow in the valley. (Copyright (c) Elsevier Science Publishers B.V., Amsterdam.)