Statement
n an partly cloudy night with light winds, a tank truck filled with liquid chlorine to its full
O 16 metric ton capacity, jack-knifes, then overturns in tall grass adjacent to the highway. The tank ruptures and empties in 3 0 minutes in a manner so that the rapidly expanding fluid forms an
. . . and E
Hazard: . . . . ~I the appropriate monitoring instruments were available at the ~ ~ j ~
Averaging time: 900 s If a release such as exemplified here were to be made, and . . .
I b
served. At 4 d s wind, the cloud fiom the 30-minute release would travel 7,200 m during that 30 minutes. Even if the wind direction and speed stayed constant for that long (within the averaging time concepts), then for longer times the cloud undoubtedly would start to change course along its path. Also, no terrain is completely plane, so the cloud, or parts of it, would be diverted by terrain features, trapped into low lying areas, etc. Then, if the period of interest included the transition from night to day, sun heating effects will change the surface temperature and winds, etc., so as to change the meteorological conditions from those postulated.
Analysis
fl right places and times, a number of phenomena would be ob-
Source Characterization
The fraction aerosol in the released mixture could be calculated for this pure component by Equation 3-52, which (along with other physical properties for the two phases) could initialize a boiling pool area source directly modeled by HEGADAS and SLAB. The triples input could be used with DEGADIS. However, experience from field tests and modeling indicates that aerosols are completely vaporized in less than 100 meters, so the distance error caused by assuming the source is initially all vapor should be negligible.
Therefore, for the worst case purpose of this scenario modeling, it should be adequate to model the source as a cloud at its boiling point being discharged over a relatively small ground area. If the speciủed area is small enough, the dense gas dispersion models will cause the cloud to spread out until the product of maximum take-up flux coupled with initial cloud surface area equals the discharge rate, thus providing a correct steady state source area size. It follows that the dense gas, area source dispersion models, HEGADAS, DEGADIS, and SLAB can be used directly.
Atmospheric Parameters
The given nighttime conditions indicate a PG stability class E (See Table 4-1); a 3 m/s wind was used because it is the minimum for the class under the < 3/8 cloudiness condition. Similarly, sta- bility class C was indicated for an average wind speed of 4 d s . These two sets of meteorological conditions should be representative enough to give the worst case estimates required.
Copyright American Petroleum Institute
--`,,-`-`,,`,,`,`,,`---
A P I PUBLx4h28 ù b 0732290 0560103 6 5 2
S5-2 Chapter 6
The given rural terrain conditions indicate the roughness length parameter should be 0.03 m (Appendix I). An averaging time of 900 s was used.
Simulation
Figures SS-1 and S5-3 show the resulting centerline concentrations vs downwind distance, while Fzpres S5-2 and.S.5-4 show 1 ppm chlorine crosswind concentration isopleths. Because SLAB uhvuys averages the concentra- tions with respect to release duration, only one simulation was run for each of the two sets of meteorological condi- tions; the results are labeled “SLAB 1800 s” on the fig- ures. To obtain the finite duration results for HEGADAS, it was necessary to run the steady-state HEGADAS-S pro- gram, then use the HSPOST post-processor program to time-average the results with respect to downwind travel
MODELING PARAMETER RECAP Released Gas
Molecular weight 70.9
Specific heat vapor,
c,, KJI[kg.Kl 1.58
Release rates, kg/s 8.9
Source area input, m2 20
Normal boiling point =
gas temperature, K 239
Release duration, s 1800
Roughness length, z,, rn 0.03 AtmosDheric Variables
Stability C Stability E
temperature, K 303
wind speed, mis 4
temperature, K 293
wind speed, rn/s 3
time for the 1800 s release duration (curves labeled “HEGADAS 1800 s”). The steady state results for DEGADIs are not corrected for downwind travel time; as noted elsewhere, that requires an external program not currently available.
As expected, the models predict very long distances to the points where chlorine concentration falls below 1 ppm. Note in Figure S5-1 that the distance to 1 ppm chlorine for HEGADAS without correcting for release duration is about 100 km, whereas the corrected curve shows the distance to be about 46 km!
As expected, the chlorine concentration drops below 1 ppm at long distances, compared with distances where source modeling effects are important. For the C stability cases, the 1 ppm points are about 38 km downwind for both the corrected and uncorrected curves. Therefore, it can be concluded that more detailed source modeling is not justified for this case.
Discussion
Using these models to predict concentrations out to such long distances must be considered semi- quantitative, at best. The farther the distance from the source, the more uncertain the estimates will become. Ifsimulations were to be made for F stability, the distances would have been much greater than for the E stability case shown here.
--`,,-`-`,,`,,`,`,,`---
Scenario 5: Liquified Chlorine Tank Truck Accident s5-3
o > w
o T
Copyright American Petroleum Institute
--`,,-`-`,,`,,`,`,,`---
A P I PUBLx4bZB 96 = 0732290 0560305 425
Release time W e : Finite duration Turbulent jet: . . . . . . . . Horizontal
"Cloud height: . . Initially elevated Roughness vpe: . , . . , . /ndust/fa/
Stabil@: . . . . . . . . , . D
Averaging time: . . . Various Hazard: . . . . . . . . Toxic
Upstream automatic control systems are being designed so that ifa release such as this occurs, flow can be stopped in a very short time (Several minutes or less). To assist the con- trol system designers and management in deciding the design maximum automatic shut-off time, fiom a safety and best