Dispersion Modeling

There is an interesting blog post at IEM.com describing some work that organization has done with some dispersion modeling based upon the recent vinyl chloride release as a result of a railroad derailment in Paulsboro, NJ. While such dispersion modeling is an invaluable tool in emergency response planning and execution, this article shows the disparity between predicted exposure levels and actual hazards on the ground.

Plume Concentrations

The blog post reports the plume concentrations reported by the IEM model. While they are careful to use the word ‘may’ in describing the potential area affected by the modeled plume, the numbers are quite extraordinary (see table). The map accompanying the blog post shows a fairly standard airborne dispersion model.

Concentration	Distance	People	Homes
20,000 ppm	0.2 miles	1,056	697
5,000 ppm	0.6 miles	3,236	1,292
500 ppm	1.2 miles	7,137	3,021

The problem with these numbers is that they describe a potentially catastrophic event. The EPA notes [Link Added 12-12-12 10:00 EST] that the LC50 (concentration at which 50% of exposed individuals died) for vinyl chloride is 115 to 225 ppm depending on whether you are looking at mice or rabbits. The EPA web site does not give the time period for the LC50 exposure and the numbers are not specific to humans, but the model data shows that there should have been a significant number of deaths from this incident; there have been none reported.

Simplistic Dispersion Modeling

This is not the first time that dispersion models have been less than adequate predictors of the actual danger associated with a toxic inhalation hazard (TIH) chemical release. I looked at a similar problem with the Graniteville, SC chlorine accident in a blog post in 2010. In that case there was a similar railroad accident (human error instead of old infrastructure, but the release was comparable) involving the release of chlorine gas, a more deadly TIH chemical than vinyl chloride. In that case there were some deaths in the immediate vicinity of the accident, but nowhere near the number predicted by the appropriate models.

Heavier Than Air

I did a subsequent blog post on some of the problems with the dispersion model used for chlorine and much of that discussion would appear to be relevant to this discussion as well. One factor that I skipped over was vapor density. Both chlorine and vinyl chloride vapors are heavier than air (more than twice as heavy in the case of vinyl chloride). Vapors like this have flow characteristics that are very similar to water; they flow very close to the ground and flow towards lower areas. In cases like the New Jersey accident where the accident occurs near a water way, one would expect a significant portion of the flow to be along the surface of that water way, reducing the amount of vapor that would disperse according to the more traditional and simplistic dispersion models.

In urban areas underground utilities would be another place where one would expect the flow to ‘loose’ some of the heavier than air vapors. Utility workers in the areas of spills like this are going to have to be extra careful about their gas monitoring before they enter below ground spaces; simple O₂ and flammability testing will not be adequate. But again, vapors lost to these spaces would be vapors that would not disperse according to the standard model. Downstream workers at sewer treatment plants are also going to be at some risk for vinyl chloride exposure over the coming weeks, particularly chronic effects.

The dispersion model, particularly with these heavier than air vapors, does not adequately address the change in concentration with a change in altitude. Even in the very high concentration areas relatively near the release point, moving to a higher floor in a building or moving to higher ground will significantly reduce the potential exposure. Even so, the dispersion model must take into account the vertical ‘loss’ of heavier than air gasses through diffusion and mixing.

The opposite is, of course, true with lighter than air toxic gasses such as anhydrous ammonia. The dispersion model for these has to take into account the much greater vertical dispersion of the gas as well as the horizontal dispersion. This vertical dispersion will decrease the ground level concentrations from the levels predicted in a standard dispersion model.

Reactivity

Both vinyl chloride and chlorine gas are very reactive chemicals and some amount of the material is ‘destroyed’ by these reactions. Vinyl chloride is interesting in that it is a monomer so that it is expected to react with itself with the oligomers being less hazardous and less volatile. In fact, the EPA web site mentioned earlier notes that the half-life of vinyl chloride is ‘measured in hours’; this is due in large part to its reactivity. Again, anything that removes vinyl chloride from the environment will reduce the amount that will continue to disperse through the area, thus reducing the concentrations in the dispersed cloud.

Models Not Unuseful

While there are significant problems with all of the gas dispersion models currently in use, they are still going to be useful to emergency response planners and initial emergency response personnel. If emergency response plans take into account the worst case scenarios base upon these models, then they are probably going to ensure that the response is more than adequate for the actual incident. And first responders are not going to be far from wrong if their immediate response is based upon such models.

What is clear, however, is that incident commanders are going to have to have a pretty good understanding of the limits of the current set of dispersion models if they are going to adequate deal with the evolving situation. For instance, the high concentrations found in the immediate vicinity of the release point are going to start to dissipate to safer levels as soon as the bulk of the material has been released from the container.