Wind tunnel experiments were made to determine how rapidly dense gas trapped in a topographic depression could be removed by an entraining crosswind. The two-dimensional outflow volume flux, vo, was assumed equal to the inflow rate during 92 steady-state experiments with CO2 continuously supplied into the bottom of two-dimensional, V-shaped valleys. As predicted by theory, at large Reynolds numbers it was found that vo is approximately equal to Us(3)/gi, where Us is the speed just above the dense gas pool and gi' is gravity times the relative density difference. The width of the pool, w, does not affect vo when the primary Froude number < 1, except at low Reynolds numbers; in this case the data suggest vo is approximately equal to (Us wK)1/2 as an asymptote, where k is the molecular diffusivity. A universal relationship is suggested for vo bridging these two asymptotes. Transient experiments were conducted by filling a valley with dense gas, turning it off, then quickly removing a sliding cover; vo was measured as a function of time with an array of samplers downwind. These experiments essentially confirmed predictions based on the steady-state results, even when SF6 was substituted for CO2.