Accidental releases of a toxic, dense gas from a truck or railcar pose a significant health threat to a population in a neighboring community and risk assessments are required. Dispersion models are used for such assessments, i.e., for predicting the cloud behavior and downwind concentration field. One troubling issue from previous accidents was that the dispersion models predicted very high concentrations with significant impacts – severe injury and death – for distances out to a few kilometers from the source. In contrast, “on-the-ground” results showed that the high impact region was much smaller, closer to the source, and limited to only a few deaths. As a result of these differences, the Defense Threat Reduction Agency (DTRA) initiated a series of “Jack Rabbit (JR)” experiments to investigate the behavior of a dense gas cloud from a full-scale chlorine (Cl_{2}) release and the downstream concentration field. The experiments were intended to assess the dispersion model formulations, assumptions, and predictions using the Cl_{2} measurements. The JRI experiment was conducted in 2010 with 1- or 2-ton chlorine releases and JRII in 2015 with 5- to 9-ton releases within a mock urban array, both experiments being conducted at the Dugway Proving Ground. A third series of experiments in 2016 extended the JRII conditions to different emission orientations and included a 20-ton “tanker” release, both over smooth flat terrain.

In support of the DTRA activity, RAL/NSAP initiated a dense gas modeling effort in 2015 aimed at predicting Cl_{2} concentrations in the vicinity of an accidental release and used the JRI data for assessment. Our effort was aimed at worst-case scenarios – stable atmospheric conditions and slow dispersion – that would result in the highest Cl_{2} concentrations.

In 2018, we continued development of a dense gas dispersion model for prediction of the maximum chlorine concentrations for either an instantaneous puff or a short-duration release from a storage tank “blowdown” with application to the JRII 2015 and 2016 experiments. Recent focus has been placed on stable boundary layer conditions and the corresponding JRII tests. The model is an integral approach based on equations for the conservation of puff volume (or mass), radial momentum, buoyancy, and species (i.e., chlorine). The puff is modeled as a squat or shallow-depth circular cylinder and includes: radial spreading driven by the high source momentum flux, gravitationally-induced dispersion by the negative buoyancy due to chlorine, initial “slumping” and height change with time, entrainment of ambient air at the puff top, and transport by the wind. The model is currently aimed only at downward venting releases and is divided into: 1) a short-time regime appropriate for the active or venting release stage, and 2) a long-time regime applicable when the venting is complete and the puff spreads under the action of the total buoyancy released from the tank over the release time.

In the short-time regime, the puff radius is initially dominated by the momentum flux and spreads as t^{1/2} where t is time, whereas at later times in this regime it is dominated by the buoyancy and grows as t^{3/4 }(Fig. 1). The momentum flux leads to a much more rapid puff growth at early times and reduces the near-field modeled chlorine concentrations to 20% -- 50% of the values given by the “buoyancy-only” model. Unfortunately, there are no JRII chlorine measurements in the nearest 30 m downwind to test the near-field predictions. In the long-time regime, the radius grows as t^{1/2 }withthe total buoyancy providing the forcing.

For large distances (x > 50 m), comparisons of the model (blue line) with the maximum observed chlorine concentrations from the JRII experiments in Fig. 1 show that the model exhibits general agreement with the data, mostly within a factor of 2 – 3, and a decrease with distance similar to the observations. In addition, the model is in good agreement (factor of 2) with predictions from a RANS computational fluid dynamics (CFD) model at distances of 200 m and 500 m downwind. The results also show some differences in model behavior at short distances (x < 50 m), where better model-data agreement occurs for Trials 2 and 3 with higher winds (U_{ref} ∼ 4 m/s) than for Trials 1, 4, and 5 with lighter winds (U_{ref} < 3 m/s). The differences with respect to wind speed and wind direction, relative to the mock urban array normal, have been addressed with the CFD model.

- Extend the model to different venting orientations.
- Assess model parameters (entrainment, etc.) for terrain/roughness effects using the 2015 experiment results (mock urban array) and the Jack Rabbit II 2016 experiments, which were conducted over smooth flat terrain.
- Link the integral puff model with a Lagrangian particle model to produce probabilistic concentration predictions.