Large amounts of chlorine released in a short period of time (less than one hour) from pressurized liquefied storage such as a railcar will form a dense two-phase (gas plus small liquid aerosol drops) cloud. When there are terrain depressions and/or confining boundaries such as buildings and/or vegetation, the dense cloud can be held-up in the depression and the chlorine may be slowly detrained into the ambient air over a period much longer than the initial emission duration. This paper describes a theory for estimating the detrainment rate (i.e., emission into the ambient air) based on the initial cloud mass and Richardson number, and the ambient wind speed. The theory is verified using field observations from the Jack Rabbit (JR) field experiments in 2010 at Dugway Proving Ground, UT, where 1 or 2 tons of chlorine were released over about a 30 sec period into a depression 2 m deep and 50 m in diameter. There were five days of release trials for chlorine, with all releases occurring in the early morning and with wind speeds ranging from 0.6 to 6.2 m/s. For example, the theory matches the observed release duration to the ambient atmosphere due to detrainment from the depression during Trial 2, which was the trial with the lowest wind speed (0.6 m/s). For Trial 6, with the highest wind speed (6.2 m/s) of the five trials, there was very little hold-up of the cloud in the depression and it was transported downwind almost as if the depression wasn't there. A wind speed of about 2 m/s was found to separate the cases with significant cloud hold-up and delayed detrainment from the cases with little cloud hold-up. The estimated emission rates from the dense cloud to the ambient air were input to a standard widely-used dense gas dispersion model (SLAB), for the two trials (2 and 6) with the lowest and highest wind speeds. The primary output of interest was the maximum 20 sec concentration near the ground. The SLAB-predicted concentrations are compared with the observations from several types of chlorine concentration samplers during JR, at downwind distances out to about 500 m. The model simulations agree fairly well with the observations (within about a factor of 3). As a sensitivity study, SLAB was run for several variations in inputs of emission duration (due to detrainment) and stability class. It was found that the variations of maximum concentration with distance are not always what is intuitively expected. For example, for light wind trial 2, and for constant input wind speed (0.6 m/s at z = 2 m) and stability class (F), by the time the cloud reaches downwind distances of 1 to 2 km, the concentrations are not very different for the two cases with release durations of 30 sec and 30 min. Our analysis shows that this is caused by the fact that the cloud from the 30 sec release is behaving as a puff, with significant along-wind dispersion, over most of its trajectory, while the cloud from the 30 min release is behaving as a continuous plume until it reaches distances of more than 1 km. Even on the edges of the depression, the concentrations are calculated to not reflect the factor of 60 difference in release duration. The difference is more like a factor of 20 at that location, primarily due to the larger horizontal spread of the initial dense gas cloud for the shorter release duration. Thus it is important, when carrying out studies of the possible maximum health or environmental impact from a hypothetical chlorine (or other dense gas) release, to carry out several sensitivity studies, varying the inputs within reasonable bounds.