The major scientific issues concerning source emissions models and dispersion models for Toxic Industrial Chemicals (TICs) are discussed, with emphasis on chemicals stored and/or transported as pressurized liquefied gases (e.g., chlorine, anhydrous ammonia, and sulfur dioxide). Many tons of a gas/aerosol mixture can be released and have been released into the atmosphere in one or two minutes in railcar accidents. Some recommendations are given for specific source emissions equations, with scientific rationale provided. Field experiments on TIC source emissions are reviewed and some results of evaluations of models for droplet formation in flashing jets are presented. The characteristics of various dispersion models for TIC releases in urban areas are reviewed. A hypothetical chlorine railcar release in Chicago is simulated with a CFD model (FLACS) and with several widely-used simpler models and the results compared. The CFD modelsimulated effects of the urban buildings on the dense gas transport and dispersion are discussed, such as constraints by buildings near the source, reductions in transport speed, diversion of the dense gas down drainage slopes, increases in turbulence intensities, and hold-up in building wakes after the main cloud has passed. 1. OBJECTIVES AND BACKGROUND The authors have been carrying out research on Toxic Industrial Chemical (TIC) source emissions models and dispersion models for several years. The source emissions model improvements have been under study for two years, where the main objective is to suggest improvements for the toxic industrial chemical (TIC) source emissions models in HPAC (DTRA, 2008). Currently these source emission models are included in the Industrial Facilities (IFAC) and Industrial Transportation (ITRANS) modules (DTRA, 2004) and in the SCIPUFF transport and dispersion module (Sykes et al, 2007). The study team is making use of TIC emissions models suggested by Corresponding author address: Steven R. Hanna, 7 Crescent Ave., Kennebunkport, ME 04046-7235, [email protected] the chemical industry (CCPS, 1996) and detailed field experiments such as those involving twophase jets (CCPS, 1999, Witlox et al., 2007). The highest priority scenarios concern releases of many tons of pressurized liquefied gases such as chlorine, anhydrous ammonia, and sulfur dioxide. In most cases, these releases become dense due to their high molecular weight, their cold temperature, and or/their imbedded liquid droplets. For most scenarios, the worst case will be when all of the released material quickly ends up in the gas phase as it is transported downwind. The existing TIC source emissions equations have been reviewed (Hanna et al., 2008b), and some recommended equations are being evaluated with data from TIC field experiments, emphasizing recent experiments with pressurized liquefied gases. In addition, the source inputs required by SCIPUFF are being reviewed to ensure a smooth transition to SCIPUFF (Hanna et al., 2008c). An expert workshop on the topic was held in March, 2008, and resulted in a set of conclusions and recommendations described by Hanna and Britter (2008).
The authors have also been developing and testing transport and dispersion models for dense TIC releases, with recent focus on complex urban and industrial areas (Hanna et al., 2008d). Most TIC accidents and high-priority scenarios involve the presence of industrial buildings, tanks, and other obstacles, or urban areas with complex building shapes. The buildings obstruct the flow and cause changes in the rates of dispersion, and any ditches or slopes may enhance drainage flows. Hanna et al. (2009) applied a CFD model (FLACS) to a real chlorine accident at a chemical processing plant in Festus, Missouri, and to a hypothetical chlorine release at a railroad crossing in Chicago. The intent was to investigate the effects of the buildings and of terrain on the dense gas cloud. Also, several widely used simpler dense gas TIC models were applied to the Chicago scenario. 2. SOURCE EMISSIONS MODELS The highest priority TIC release scenario involves storage of pressurized liquefied gases, such as chlorine, anhydrous ammonia, and sulfur dioxide. These TICs are important because they are inhalation hazards, they are frequently stored and transported around the U.S. in large quantities (say 50 to 100 tons per container vessel), and they have low boiling points (i.e., high saturated vapor pressures). Because of their low boiling points and the fact that they are stored as pressurized liquids, when the vessel ruptures, the TIC “flashes” into a gas-aerosol mixture with high velocity. For holes of diameter 10 cm or larger in the vessel, nearly all of the 50 to 100 ton contents may be emitted to the atmosphere in a matter of minutes. Some major TIC source emission model concerns are: thermodynamic state of the stored material, rupture size(s) and type, response of materials within the storage vessel including rapid phase change, heat transfer processes and foaming (level swell), phase of release, flashing depressurizing jet, and aerosol drop size distribution and subsequent proportions of rainout and suspended small aerosol drops. Figure 1 is a schematic diagram showing the regions of importance in a flashing jet and the symbols used, going from the storage vessel to the point where the jet pressure decreases to nearly ambient values. Figure 2 shows the temperature-entropy curve for chlorine at various pressures, illustrating how an isentropic release could involve all-liquid, all-gas, or two phase processes, depending on the storage conditions.
The simpler all-gas or all-liquid cases have well-known analytical solutions. Unfortunately, real scenarios always have complications such as jagged holes at unknown locations on the tank, length of the pipe extending a short distance from the tank, release rate that varies in time, and poorly-known weather conditions. Relevant field and laboratory experiments have been reviewed and a subset chosen for source emission model evaluation, including: Modelers’ Data Archive (MDA), RELEASE (flashing jets and droplet sizes), FLADIS and URAHFREP, FLIE, DNV JIP flashing jet studies (Witlox et al, 2007), and those from Richardson (2006). Project Madison (1 ton chlorine cylinders) field data should be available within one year. Figures 3, 4, and 5 provide examples of the comparisons of droplet size model predictions with observations. The two models are the CCPS (1999) RELEASE model and the Witlox et al. (2007) JIP model. The two data sets are the CCPS (1999) RELEASE data and the Witlox et al. (2007) JIP data package. It should be mentioned, though, that the JIP data contain direct observations of droplet diameter, while the CCPS RELEASE data use the CCPS RELEASE model to back-calculate the droplet diameters from observations of the rate of deposition of liquid on the ground surface. The vertical axis in the three figures is the ratio of predicted to observed droplet diameter, Dpred/Dobs. In Figure 3, Dpred/Dobs is plotted versus superheat, ΔT, which is the difference between the ambient temperature and the TIC boiling point.
Smaller droplets are expected at larger superheat because the jet will be released with more force. It is seen that the two models tend to overpredict the droplet diameter at small superheat and underpredict it at large superheat. The model error is less than a factor of two for about 60 % of the points. Figures 4 and 5 plot Dpred/Dobs versus orifice velocity for the JIP droplet model and the CCPS RELEASE droplet model, respectively. Figure 4 shows that the JIP droplet model tends to overpredict the droplet diameter at small orifice velocities and underpredict (by about a factor of two) at larger velocities. Figure 5 shows that the CCPS RELEASE model has little bias for the CCPS data, but that is expected since the model was tuned to those data. The CCPS RELEASE model tends to overpredict the observed JIP droplet sizes by 40 % to a factor of 3.4. The research is also addressing the hand-off from the source emission model to the HPAC/SCIPUFF dispersion model. Hanna et al. (2008c) present a review of general methods for the transition (mostly based on physics-based dimensionless criteria), as well as a review of the methods in HPAC (mostly based on arbitrary criteria concerning jet velocity and excess temperature). It is concluded that, while the current methods in HPAC are adequate, they are subjective in many places, and could be improved by having the transition criteria based on the fundamental physics relations. New criteria are being recommended, but must account for the fact that SCIPUFF currently handles only the accelerations in a vertical jet and not the accelerations in jets at other angles. SCIPUFF is being modified, however, to treat accelerations of jets at any angle. To attempt to identify topics in the area of TIC source emissions where there is a consensus, an expert workshop was held in March 2008 and a summary and set of recommendations written (Hanna and Britter, 2008). The experts used the Festus, MO, and the Project Madison videos of chlorine releases to provide a focus for discussions. In addition, the group agreed that the foaming inside the vessels was important and its parameterization considered a key element of the TIC source emission model. 3. DISPERSION MODELS FOR URBAN AND INDUSTRIAL SCENARIOS Most accident or terrorist scenarios involving TIC releases include complications due to complex terrain and/or obstacles such as urban or industrial buildings and obstacles. Most of the widely-used dense gas dispersion models generally do not account for the presence of obstacles, other than perhaps through an enhanced surface roughness, and account only for simple linear slopes, at best. HPAC/SCIPUFF is in this category (Sykes et al., 2007). However, the HPAC system does have the capability to use mesoscale meteor
