Toxic Risk Analysis

Toxic hazards associated with rocket launch operations are due primarily to combustion or vaporization of propellants used in the booster or payload. Over the history of sub-orbital and orbital launch vehicle development, many energetic chemicals have been tested for potential use as rocket propellants. Although some exotic propellant chemicals may occasionally be encountered, the vast majority of space launch vehicles in the current international inventory use a relatively small range of propellant types. The widely used propellants can be partitioned into four basic categories; hypergolic liquid propellants, cryogenic liquid propellants, liquid hydrocarbon based fuels, and aluminum based solid propellants. The liquid hypergols and solid propellants can produce airborne chemicals that are toxic at even low concentrations (e.g. below 20 parts per million (ppm)). Cryogenic liquids require special handling due to the low temperature but are otherwise not considered toxic. Likewise hydrocarbon based fuels may pose an environmental cleanup hazard if spilled but are generally not toxic except in very high concentrations (e.g. several thousand ppm). Toxic hazard screening procedures have been developed to help launch ranges and vehicle operators determine whether or not they have a potential toxic hazard and what steps are necessary to contain the hazard or to evaluate the potential risk.

Toxic Hazard Screening Methodology

The United States Range Commanders Council (RCC) (Risk Committee, Range Safety Group, Range Commanders Council, 2010) and the Federal Aviation Administration (FAA) (Federal Register, 2006) have both published essentially the same toxic hazard screening methodology that is intended to provide conservative hazard corridor estimates given various classes of common propellants and the quantity of propellant used on a launch vehicle. The objective of a screening methodology is to provide users a way to determine if a toxic hazard exists and a progressive set of steps to apply if risk management is required. To be useful, a screening methodology should be based on readily available launch vehicle information and general meteorological principles. The RCC and FAA methodologies require the user to know the types and total quantities of propellants used on the launch vehicle, the intended launch site, and the location of populated areas around the launch site. The screening approach can be defined in a six-step procedure.

Step one is to determine whether or not any toxic propellants are used on the launch vehicle. Table 5.1.1 lists commonly used rocket propellants that are considered non-toxic under FAA hazard assessment guidelines.

If the launch vehicle utilizes only non-toxic propellants, then no range safety toxic hazard exists and no further toxic hazard or risk evaluation is required. If toxic propellants are used, then further assessment is required.

Step two of the screening procedure applies when toxic propellants are used on the launch vehicle. Commonly used toxic propellants and recommended toxic threshold concentrations used to define hazard corridors are listed in Table 5.1.2.

This step seeks to determine whether or not potential toxic hazard corridors are contained within a region that is unpopulated. The toxic hazard corridor is defined as the region swept out by the transport of a toxic cloud or plume within which the predicted concentration exceeds a casualty-causing threshold. The length, width and direction of a toxic hazard corridor are highly dependent on the prevailing meteorological conditions at the time of the chemical release. Estimated worst case hazard corridor distances as a function of propellant quantity are provided for common toxic propellants in Table 5.1.3. At this stage in the screening process the distances derived from Table 5.1.3 are used to define circular areas centered on the propellant release site (typically the launch pad). These circular hazard regions assume that on any given day the prevailing wind could be from any point of the compass. If there are no people (mission support or general public) within the circular hazard regions, then the toxic hazard is considered contained and no further range safety analysis of this threat is required. If there are people within the toxic hazard regions, then further analysis is required.

Step three of the screening procedure applies when people are located inside the estimated worst case hazard regions. Under those conditions a variety of options are evaluated to determine if a hazard containment solution may still be viable. One of the first questions asked is if the people inside the hazard region can be evacuated for the launch. Evacuation is feasible if the launch range has control over the affected population locations and disruption of operations at those sites can be tolerated. Evacuation of isolated general public facilities may be tolerable but generally if substantial numbers of people are found within a hazard region, evacuation is not a viable alternative and constraints on the launch itself may need to be considered. Another option that can be applied at this point is to run a more detailed and site specific toxic dispersion analysis using an approved model and local weather data. The lookup table hazard distance values are by design very conservative and a more refined analysis may indicate that hazard distances are shorter than those listed in Table 5.1.3.

Step four of the screening procedure is used to evaluate launch constraints based on weather conditions. If the threatened population centers are confined to a small arc width segment of the circular hazard region, it may be feasible to hold launches when the prevailing winds are blowing toward the populated region but allow launches for all other wind directions. Archived weather balloon data, if available for the site, is useful for estimating the percentage of time that wind directions are favorable. Seasonal or diurnal fluctuations in wind conditions may also be a factor to consider. Toxic releases that are buoyant in nature should consider wind direction from the ground to about 2000 meters altitude to determine plume transport direction. Toxic releases that are non-buoyant stay close to the ground and wind directions from the ground to about 100 meters should be used to determine plume transport direction. If the launch operator can tolerate the estimated launch constraints due to wind direction restrictions, then toxic hazard containment is achieved and no further analysis is required.

Step five of the screening procedure applies when evacuation and wind direction constraints cannot achieve toxic hazard containment. At this point risk assessment and risk mitigation practices need to be evaluated. Most launch ranges operate with some form of acceptable risk policy that requires a rigorous analysis of the launch hazards and the associated risks. In order to perform a toxic risk assessment the analyst must have the following analysis tools and information:

A collective risk estimate expressed as a casualty expectation should be computed by summing the expected risk over all population centers. The worst case individual risk should also be extracted from the ensemble of population center risks. The predicted collective and individual casualty expectations can then be compared with acceptable risk criteria to determine if the mission risk is within acceptable limits and launch is authorized. It is conceivable that a more refined analysis may predict no risk because computed toxic hazard corridors were not as large as the conservative Table 5.1.3 values and no population was predicted to be exposed. Care should be taken to verify that the analysis utilizes an adequate sample of expected weather conditions and that credible worst case toxic dispersion scenarios were defined and evaluated as part of the hazard and risk assessment. If toxic risk is routinely predicted at this stage of the screening procedure, it is practical to expect that a toxic hazard and risk analysis may be needed to evaluate conditions for each launch using day-of-launch weather observations and mission specific population data.

Step six of the screening procedure allows for iteration of the risk analysis process with adjustments to scenario assumptions that may mitigate risk. For instance, sheltering people indoors is often a significant mitigation factor that reduces exposure concentration and therefore reduces risk. If people within a potential toxic hazard corridor can be advised to go indoors, close windows and turn off ventilation, then their risk can often be eliminated. The risk from common toxic propellants to mission support personnel located in the field without the benefit of a shelter may be mitigated by having appropriate breathing apparatus that can be worn in the event of a toxic release.

Selecting an Appropriate Analysis Methodology

Analysis methodologies developed to evaluate toxic hazards associated with rocket propellants can be divided into two categories depending on whether or not propellant combustion is involved in the release. When combustion or explosion processes are involved in the propellant release, the source will be highly buoyant due to the hot combustion gases and modeling the transport and dispersion of the exhaust products in the atmosphere must account for lofting of the buoyant source. Releases not involving combustion are generally spills of liquid monopropellant that result in an evaporating pool of liquid. Explosion of solid rocket motors during the early phase of ascent presents a unique challenge for toxic dispersion modeling because burning solid propellant fragments are ejected at the explosion altitude and they emit a trail of toxic combustion gases along ballistic trajectories to ground impact. Fragments may burn up before ground impact or impact the ground and continue burning producing multiple plumes within the debris impact field. Toxic hazard zones and risk analyses to launch site personnel or the surrounding general public must account for these various types of potential sources. Ultimately any toxic risk model must estimate concentrations of the toxic chemicals at downwind receptors and provide a means to translate the exposure into the probability of an adverse health effect among the exposed public.

Toxic Emissions from Non-Buoyant Sources

The most prevalent non-buoyant emission sources of concern to range safety professionals from a public risk perspective are the hypergols. Hypergolic liquid propellants are widely used in main stages of boosters and in lesser quantities on satellite payloads. Personnel performing hypergol propellant transfers wear full Self-Contained Atmospheric Protective Ensemble (SCAPE) suits to prevent toxicity exposure. Off-site workers and the general public are not provided such protective equipment and therefore operational procedures are typically applied to assess potential hazards to these receptors and facilitate a decision to either evacuate potential hazard areas or to limit propellant handling operations to times when wind directions are favorable.

Hypergols, by definition, react spontaneously when the fuel and oxidizer are mixed and an ignition source is not required to initiate combustion. Small quantities of hypergols may in fact be used to initiate combustion in larger hydrocarbon and liquid oxidizer motors. Hypergols are relatively stable and when stored in tanks and lines of compatible materials can be stored for years before use. The following compounds are the most widely used hypergolic fuels used by the international launch community:

Aerozine-50 is another hypergolic fuel that is a mixture of 50% hydrazine and 50% UDMH by weight.

The following compounds are the most widely used hypergolic oxidizers used by the international launch community:

Several variants of these oxidizers are mixed oxides of nitrogen (MON) and inhibited red fuming nitric acid (IRFNA). MON comes in several grades and is comprised primarily of nitrogen tetroxide with lesser percentages of nitric oxide (NO) and nitrogen dioxide (NO₂). IRFNA is a mixture of predominantly nitric acid with NO₂, N₂O₄, water and a corrosion inhibitor. All of the aforementioned hypergolic liquids are toxic and pose an acute inhalation hazard when released to the atmosphere in gaseous form. Interestingly, complete combustion of hypergolic fuel and oxidizer tends to form benign combustion products as illustrated by the stoichiometric reaction for nitrogen tetroxide and hydrazine:

Thus, from a toxic release perspective, complete combustion of hypergols is a desired outcome relative to partial or no combustion.

Analysis of downwind concentration of a toxic chemical from a non-buoyant release first requires characterization of the initial “source term”. In this context the source term defines the release or evaporation rate, the dimensions of the cloud or plume at the point of origin, the initial concentration of toxic chemical and duration of the release.

Pool evaporation rates are typically computed from theoretical models that attempt to solve the physical relationships governing pool evaporation. There are varying degrees of sophistication among the available evaporation models. Generally, the more sophisticated the model, the more physical data that is needed to define spill and evaporation parameters. The most sophisticated models attempt to perform a full mass and energy balance computation of the pool. The analyst typically provides an estimate of the amount of chemical spilled and a presumed pool depth (typically 1 cm) from which the model determines a pool surface area. For a spill within a known diked area the analyst would provide a spill quantity and the pool area from which the model would determine a pool depth. The analyst may have an option to specify an initial temperature of the spilled liquid propellant (lacking this, most models assume that the initial liquid temperature is equal to the current ambient air temperature). Given the initial pool area, depth and temperature, the model then applies energy balance calculations to arrive at a steady state pool temperature. The energy balance should consider conductive heat transfer between the pool and the ground surface, convective heat transfer from ambient air blowing over the pool, radiant energy exchange with incoming solar radiation and evaporative cooling. For spills on permeable soil surfaces, most models do not consider the complication of the liquid soaking into the ground, but rather conservatively assume that the pool remains above the soil surface.

A full analysis of heat transfer and evaporation requires information on the following chemical, physical and thermodynamic properties, many of which may be characterized as polynomial equations that define temperature dependency:

In addition to the chemical properties, the energy balance computation requires estimates of the thermal conductivity and heat capacity of the ground, which may depend upon the amount of moisture in the soil (e.g. wet or dry). The radiation energy balance requires calculation of the local solar angle, which requires spill site latitude, longitude, date and time of day. Cloud cover and ceiling factors can limit that amount of available solar constant that reaches the earth surface at the spill site and are often required user input factors. The evaporation rate is also dependent upon the wind speed and temperature of the air blowing over the pool.

Obviously full theoretical calculations of pool evaporation rate require a significant amount of user provided information. Much of this information can be predefined in a chemical database that can be linked to the toxic dispersion model. However, there may be cases where an emergency situation arises involving a chemical for which the analyst does not have a complete set of physical parameters. The toxic dispersion model should allow the analyst to specify directly an assumed evaporation rate or execute a simpler evaporation model that requires relatively few analyst input parameters.

An alternative to the theoretical evaporation model approach is to apply empirical evaporation rate equations that are a function of pool area, wind speed and air temperature. These empirical equations are derived from field test measurements using pans of evaporating propellant under varying wind and temperature conditions. Empirical equations exist for most of the hypergolic propellants (Clewell, May 1983; Ille and Springer, August, 1978).

Once the source term is defined, the next phase of toxic hazard corridor analysis requires predicting the transport and dispersion of the initial source cloud or plume. Most atmospheric transport and dispersion models assume that the mass distribution of the toxic chemical within the plume or cloud is most concentrated at the center of the plume or cloud and decreases with distance from the center following an exponential decay profile given by a normal (Gaussian) distribution. This assumption about mass distribution allows the governing differential equation to have a closed form analytical solution that is easily solved in relatively simple computer programs. This so called “Gaussian” modeling assumption has been widely used in the atmospheric dispersion community for many years.

Gaussian dispersion models predict a statistical “average” condition that can be expected over may random samples of a plume or cloud, but at any given time or position within a plume the local instantaneous concentration fluctuates and may for short durations exceed the mean value predicted by the Gaussian model. Gaussian dispersion equations come in two-dimensional plume forms or three-dimensional cloud forms.

For a long duration evaporating pool or gaseous release, the downwind dispersion has the form of an expanding plume of material that is most concentrated at the source origin and grows steadily more dilute moving downwind as the plume grows wider and taller. Growth coefficients for a plume are applied in the crosswind and vertical directions as a function of downwind distance. A short duration release is best characterized as a cloud that moves downwind from the original source and requires growth rate coefficients that are applied in all three dimensions (along wind, crosswind and vertical). The “dispersion coefficient” growth rates used by Gaussian models are typically characterized with empirical data sets derived from field test observations and are partitioned by atmospheric stability classes (Hanna et al., 1982; Center for Chemical Process Safety, 1996).

Atmospheric stability classes are defined in terms of day versus night, cloud cover and wind speed. The rate at which ambient air mixes into the chemical cloud or plume is driven primarily by atmospheric turbulence mixing rather than molecular diffusion along a concentration gradient. Therefore spill site characterization of factors that govern surface turbulence are significant to the dispersion analysis problem. Standard dispersion coefficients are often adjusted by power law scaling factors based on surface roughness for the spill site and height above the ground where the user wind measurement is made.

Many models will allow users to enter measured wind direction standard deviations as a means of defining dispersion coefficients applicable to the site conditions at the time of the release. The measured turbulence intensity is a function of the time duration over which the atmosphere is being measured at the release site. Turbulence eddies in the atmosphere are observed at a wind measurement site as time varying fluctuations in wind speed and direction.

Plume or cloud source size and downwind transport distance of concern must be considered to understand and apply appropriate time scales to the turbulence problem. For example, if an initial source is an instantaneous release that produces a gas cloud that is 50 meters in diameter, eddy scales of a few centimeters will be ineffective in mixing ambient air into the cloud. Eddy scales of hundreds of meters will tend to move the entire cloud rather than cause the cloud to grow (this is referred to as “meander”). Thus, initially, turbulence eddies with a length scale on the order of several meters up to the cloud dimension will be most effective is causing cloud growth. As the cloud moves downwind and grows, it dilutes and at some point the peak concentration at the center of the cloud drops below a threshold where exposure is no longer a concern for adverse health effects. For the sake of example, assume that the cloud diameter has grown to 500 meters when this dilution threshold is reached. At this point turbulence eddies that effectively dilute the cloud have grown to an order of tens to several hundred meters, up to the cloud dimension.

If the lower bound eddy scale of concern with the cloud near the point of origin is 50 meters and the eddy scale of concern with the cloud at the final dilution threshold is 500 meters, these distances can be translated into a time scale given knowledge of the prevailing wind speed. If the wind speed is 4 meters per second, then the time scale for turbulence measurements is in the 10 to 125 second time range. Thus an appropriate turbulence measurement system for the wind tower would be to take a series over 2 minute sampling periods and determine a statistical average of the wind direction and wind speed standard deviations for use with the model. Larger sources with greater transport distances and larger final dimensions might warrant 10 minute or longer averaging times for use with growth dispersion coefficients.

Turbulence scales that are large enough to move the entire cloud or plume should also be considered when considering how to define a toxic hazard corridor. Meander can cause the entire plume or cloud to move crosswind right and left relative to the mean downwind transport direction, thus a receptor downwind might be in the plume for a while and then completely outside of the plume at a later time. Under very light and variable wind conditions the transport direction becomes increasingly uncertain and the source material can stagnate and even move back toward the source of origin. Some dispersion models apply a full 360 degree circle to the toxic corridor for winds that are less that 1 or 2 meters per second. As prevailing winds grow stronger, they traditionally are assigned less directional uncertainty. Vertical turbulence is governed by both frictional shear with the ground surface and by surface heating from solar radiation. On sunny days with light winds solar convection cells drive vertical turbulence. On cloudy days and at night wind shears that induce mechanical mixing drive vertical turbulence. These factors are “built in” to standard stability class dispersion coefficients. Vertical turbulence is a factor in determining the downwind length uncertainty of a toxic hazard corridor. Gaussian models are acceptable for most operational applications.

More sophisticated atmospheric dispersion models are available that fall into categories of specialized applications or research grade tools. These types of models solve the full Navier–Stokes flow equations using various numerical methods and require significantly greater computing hardware. Many of these models generate a three-dimensional wind field from appropriate initial conditions and boundary conditions or accept input data from an external mesoscale wind field model. Most of these models simulate the time evolution of the wind field and can transport and disperse chemical release within the wind field using a variety of techniques from particle-in-cell random walk dispersion to application of transport of Gaussian puffs. Application of such models has received considerable attention within the terrorist threat analysis community that is concerned with release of biological, chemical or radiological agents within large city urban settings where flow fields are greatly complicated by large buildings and “urban canyons” (Annual Conference of Atmospheric and Dispersion Modeling, 2012). Gaussian models do not perform well under these conditions. At launch ranges, payloads that carry significant quantities (kilograms rather than grams) of radioactive isotopes often receive intense launch safety analysis and nuclear safety agencies may seek to work with the range to perform specialized analyses of nuclear material transport and dispersion in the event of catastrophic launch aborts.

Standard Gaussian models also may not perform well if chemical release sites are close to large buildings or geographic obstacles that generate wakes and eddy shedding that propagate through the site. There are a limited number of models available in the public domain dispersion modeling community that attempt to treat building wake effects in a simplified fashion (Schulman et al., 1997); however, most ranges do not apply these models.

Toxic Emissions from Buoyant Sources

Large buoyant clouds of rocket exhaust are produced during normal launch and ascent of rockets. In the event of a catastrophic failure early in flight, all of the propellant on the launch vehicle becomes a potential toxic emission source released within the convective boundary layer of the atmosphere. Heavy lift space launch boosters typically utilize 250,000 kilograms or more of propellant and this amount of propellant can produce toxic hazard corridors that may extend to 20 kilometers from the launch site. Hazardous chemical concentrations can easily extend beyond the blast danger area and vehicle debris field areas that are normally evacuated prior to launches, and consideration should be given to risk presented to mission support personnel and launch spectators that may be located in a toxic hazard corridor downwind from the launch site. From an atmospheric dispersion modeling perspective, evaluation of rocket launch emissions requires a mesoscale type dispersion model that is capable of characterizing the vertical profiles of wind, temperature and turbulence from the ground up to approximately 3000 meters and to distances out to 30 kilometers. Rocket emission sources will typically disperse in less than 1 hour. If prognostic forecast models are used to predict winds in the launch area, the model should have a horizontal grid resolution of no more than 1 kilometer to resolve influences such as land–water interfaces that drive diurnal sea breeze fronts. Many atmospheric dispersion models have been developed for applications in environmental assessment of pollution sources, potential nuclear releases from power plants, chemical and biological weapons and terrorist threats, and various military applications. Some of these models may be suitable for evaluation of rocket launch emissions. These models must be evaluated with regard to several factors before they are applied to rocket launch emission evaluation, including:

A limited number of toxic dispersion models have been tailored specifically to support evaluation of rocket emissions or propellant disposal burns (Nyman, 1999; Herndon et al., 2010). Finding models that can adequately treat the formation of the rocket emission source clouds or plumes and then apply buoyant source calculations is, perhaps, the biggest challenge. Atmospheric dispersion modeling of rocket emissions can be categorized into the following chronological steps:

Most atmospheric dispersion models only compute the estimated hazard, which is expressed in terms of chemical concentration or dosage levels and hazard zones are often defined as regions wherein the predicted concentration exceeds an established allowable threshold. Since the 1990s many countries have moved toward establishing acceptable risk criteria and apply risk management policies as part of the launch decision process.

Relatively few atmospheric dispersion models include the capability to add population data and human vulnerability models that allow predicted chemical exposures to be translated into risk estimates. To arrive at a risk estimate, the safety analyst must also weight the predicted risks for each applicable dispersion scenario by the probability that the scenario could occur.

Since developing the source term is a critical step in performing the dispersion analysis, it is helpful to more fully define the types of sources that can be produced and to identify significant attributes of each source type. In order to support subsequent buoyant cloud rise and dispersion calculations the source term will need to define the following parameters for each source packet (typically a Gaussian puff):

Every successful ignition and lift-off will produce a “normal launch” emission source. Depending on the types of propellants being burned, the normal launch source term may or may not contain highly toxic chemicals. Range safety analysis is concerned with protecting the general public and mission personnel on a launch-by-launch basis. Thus, chemical species such as carbon monoxide or residual unreacted hydrocarbon fuels, which are toxic but require high concentration exposures to be a health risk, are normally not part of launch safety analysis.

Some government regulatory agencies consider normal launch events to constitute a “planned emission” as opposed to accidental releases that are associated with catastrophic launch vehicle failures. Safety analysts are sometimes called upon to conduct normal launch emission studies to support environmental impact studies and to evaluate compliance with general air quality standards as part of a permitting process. This is typically done well before the actual launches and requires use of either worst case or representative meteorological conditions. At the permitting stage, the toxic dispersion modeler may be called upon to evaluate downwind concentrations of all rocket exhaust chemical constituents, including particulate matter under 10 microns in diameter. Proper evaluation of particulate matter transport and deposition requires an atmospheric dispersion model that can predict gravitational settling of particles over various size ranges.

Normal combustion products for commonly used liquid propellants, such as hypergols (MON, hydrazine, UDMH, MMH), RP1 and LOX, or LOX and LH2, are generally of low toxicity and not a concern to range safety analysts. In contrast, the widely used ammonium perchlorate–aluminum based solid propellants produce about 20% by mass toxic hydrogen chloride gas and about 28% by mass aluminum oxide particulates as combustion end products. Range safety analysis will typically address the atmospheric transport and dispersion of solid propellant emissions and associated downwind concentrations of hydrogen chloride, which is an acute inhalation irritant. The need to consider particulate matter hazards may vary with jurisdiction.

At the United States Federal Ranges particulate matter is considered in the permitting and environment impact process and falls under evaluation for compliance with 24 hour national air quality standards. Although rocket emissions may produce a large quantity of output, the emissions are intermittent and of short duration and typically are below levels of concern compared to the 24 standard. Ammonium perchlorate + aluminum + binder (e.g. HTPB, PBAN) solid propellant rocket motors emit substantial particulate loads in the exhaust plume and particles under 10 microns in size are an inhalation health hazard. There may be cases where particulate concentrations need to be considered as part of a launch safety assessment.

Aluminum oxide particulate size ranges are dependent on the propellant formulation and the nozzle flow conditions but a significant percentage of the particulates may be under 10 microns in size (Malcolm et al., Sept. 1999; Sambamurthi, 1995). This type of data must be obtained from the rocket motor manufacturer. When solid propellant motors and liquid engines are burning simultaneously the total mass and heat content of the exhaust cloud is a combination of relatively benign liquid engine products and the more toxic solid propellant products. The analyst must account for the mixture of exhaust types to properly predict cloud rise and initial cloud chemical concentrations.

The exhaust cloud produced by the Space Shuttle, illustrated in Figure 5.1.1, is an example of a vehicle that produces a large toxic cloud that is a combination of solid rocket motor exhaust, liquid main engine exhaust and sound suppression deluge water.

Most launch pad configurations are designed so that a part of the normal launch emission is deflected away from the launch pad through a flame duct or with a flame bucket that deflects the exhaust gases away from the launch vehicle. As the vehicle ascends from the launch pad a certain amount of the expanding launch plume continues to interact with the launch pad and ground surface creating what is referred to as the “ground cloud”. Eventually, the vehicle reaches an altitude where the plume no longer interacts with the ground surface and leaves an exhaust trail that is referred to as the “contrail”.

These elements of the normal launch source are illustrated in Figure 5.1.2, which depicts the emissions from a Titan IVB launch vehicle burning two large strap-on solid rocket motors at lift-off.

The ground cloud generally is larger than the contrail plume width dimension and contains a significant fraction of the normal source emissions that ultimately contribute to downwind ground level concentrations. Launch pads will typically include a water spray system to protect the structures from thermal damage from the plume and may be needed to suppress acoustic vibrations due to the launch plume. The entrainment of water into the ground cloud is important, if it is of significant quantity, because it affects the buoyancy of the cloud and the chemical composition of the exhaust constituents.

Tremendous amounts of water are injected into the Space Shuttle exhaust plume so that the ground cloud reaches saturation conditions and carries an additional load of suspended water droplets. These water droplets scrub hydrogen chloride from the ground cloud and rain out as hydrochloric acid droplets as the cloud rises and droplet settling velocity exceeds the cloud rise velocity. Each Shuttle launch has an associated acid deposition region within a few kilometers of the launch pad that has been routinely evaluated for environmental impact assessment to the surrounding wildlife refuge (Dreschel and Hall, 1990). Mission support personnel are also advised of the predicted acid deposition area and located at alternate sites outside the predicted deposition area.

The contrail portion of the normal launch source term is typically simulated as a series of overlapping Gaussian puffs, or cloud disk elements, with mass content and position derived from the launch vehicle ascent trajectory data. For launch vehicles that experience a catastrophic failure shortly after lift-off, the toxic dispersion analysis should include the normal emission source generated up to the time of failure. There are two reasons to include the normal launch source. First, there may be toxic species in the normal launch source term of the same type as released in the abort (e.g. hydrogen chloride from solid rocket motors ignited at lift-off). Second, even if the normal launch emissions are benign from a toxic stand point, they may add buoyancy to the abort source cloud resulting in a higher stabilization altitude. Additional source term definitions need to be developed to characterize the post-abort chemical releases and combustion conditions.

Launch vehicle failures early in flight typically generate the worst case toxic hazard corridors because, unlike the normal launch emission, all of the propellant on the vehicle is released and becomes involved in the emission. Most catastrophic launch failures that occur early in flight are the result of a self-induced failure such as over-pressurization of a solid rocket motor, burn through of a motor joint, or leakage and ignition of a liquid propellant. A vehicle that lifts off from the pad on an erratic flight trajectory may be quickly destroyed by the range safety officer. Vehicles that lose guidance control may start to pitch over and be destroyed by excessive aerodynamic loads. This latter failure mode typically requires more velocity than the vehicle has achieved in the first 3000 meters of ascent where most toxic hazard analyses are focused. However, vehicles that utilize large solid propellant rocket motors or solid propellant upper stages can fail at higher altitudes and still drop intact stages or large pieces of burning propellant into the launch area that explode or burn on the ground producing a toxic emission source that is displaced from the actual vehicle breakup location.

When defining source terms for launch vehicle failures, a distinction is made between liquid propellants and solid propellants. The liquid propellant source term is defined as a large fireball that is produced as the liquid tanks rupture and the propellants mingle and burn at the vehicle failure altitude. Solid propellant motors that are burning and pressurized at the time of failure are expected to shatter the propellant grain in to many pieces of varying sizes and the sudden release of the pressurized combustion gases in the core of the motor (typically on the order of 7 million Pascals) ejects the burning fragments from the launch vehicle failure point. Figure 5.1.3, depicting a Delta II 7925 launch vehicle failure, illustrates the large number of solid propellant fragments ejected from the six pressurized Graphite Epoxy Motors (GEMs) used on this vehicle.

The liquid propellant fireball is perhaps the most challenging source term to define because the dynamics of the vehicle failure and the resulting propellant mixing conditions are poorly understood and highly uncertain. The launch vehicle will carry an amount of fuel and oxidizer in each stage that is tailored to the oxidizer-to-fuel ratio designed for the engine combustion chamber. Propellant flow into the combustion chamber is carefully designed and controlled to achieve combustion stability and, for efficiency reasons, is usually somewhat fuel rich. However as a result of the physical separation of the fuel and oxidizer tanks on the launch vehicle, complete mixing before combustion is initiated is not achieved if the tanks rupture. Liquid hypergols are especially resistant to mixing because combustion is initiated as soon as the fuel and oxidizer contact each other. Sudden expansion of the hot combustion gases in the reaction zone tends to drive the remaining liquid phases of the hypergols apart. This is very significant for launch vehicles containing hypergolic propellants because the toxic hazard arises from any unreacted residual fuel and oxidizer that remain after the fireball burns out.

The source term chemical composition, temperature and heat content for a liquid propellant fireball cannot be properly simulated by entering all of the propellant types and quantities released in the vehicle failure into an equilibrium combustion model of the type widely used to predict combustion products for rocket engines. These combustion models assume all reactants are well mixed and are allowed to react to an equilibrium state that minimizes the Gibbs free energy of the products. Equilibrium combustion models will not support prediction of significant quantities of unreacted vaporized propellant because these products do not support minimization of Gibbs free energy. Figure 5.1.4 depicts the large liquid propellant fireball rising from the catastrophic failure of a Titan 34D-9 launch vehicle at Vandenberg Air Force Base (AFB). The extensive reddish color (shown as dark gray) in the cloud comes from nitrogen dioxide (NO₂) gas released from unreacted nitrogen tetroxide (N₂O₄) oxidizer. The nitrogen tetroxide molecule readily dissociates into two molecules of nitrogen dioxide at ambient temperatures.

At the US Federal Ranges liquid propellant source term chemistry has traditionally been developed by partitioning the fate of released liquid propellants into five possible reaction pathways:

These five reaction pathways must account for 100% of the total liquid propellant released from the vehicle at the time of failure. The challenge lies in selecting reasonable fractions of the propellants to allocate to the various reaction pathways. Analysts should consider the tank and propellant feed designs of the vehicle in question when considering allocation of fractions of propellant to the various pathways. The following assumptions have been used at the US Federal Ranges to characterize Titan and Delta launch vehicles (Prince and Banning, March 4, 1995):

Combustion products from explosive, secondary fireball and afterburning reactions can be computed using an equilibrium combustion model. Thermal decomposition and vaporization reactions are not equilibrium processes and the enthalpy conditions for these processes are defined external to the combustion model from the specific decomposition reactions. The most important thermal decomposition reaction for liquid propellant fireball consideration is the dissociation of nitrogen tetroxide:

A final thermodynamic heat of reaction balance is computed by combining the products from the equilibrium and non-equilibrium pathways. The thermodynamic calculations yield the required total fireball heat content. The thermal decomposition and vaporization reactions “steal” energy from the combustion reactions. The fireball is assumed to be spherical with an initial diameter [m] at cessation of combustion estimated from an empirical formula given in equation 1 as a function of total hypergol propellant mass [kg] (Gayle and Bransford, August 4, 1965).

(1)

The duration of fireball active burning is usually a few seconds or less and the fireball source term is usually simply defined to be equal to the launch vehicle failure time. The empirical formula given in equation 2 can be used to estimate fireball burn duration [s] as a function of total hypergol propellant mass [kg] (Bader et al., 8 1971).

(2)

The source term for fragments of burning solid propellant ejected from a launch vehicle failure should consider the trajectories of the burning fragments and the burn rate of the fragments as they fall to the ground. Some fragments may burn up before impacting the ground and the remaining fragments will produce plumes of emissions as they burn out within a scattered debris field. Figure 5.1.5 shows the plumes of toxic gas emitted by burning propellant fragments that survived to impact following explosion of the Delta II 7925 vehicle illustrated in Figure 5.1.3.

Development of the source term for solid propellant fragments requires trajectory calculations to determine the flight path of each fragment and the impact location. Most atmospheric dispersion models that are tailored for industrial smoke stack type pollution assessments do not have the capability to simulate fragment trajectories and associated emissions. Dispersion models that can accept an input file of Gaussian puffs with different locations, sizes and formation times could potentially be used to model the source term for falling burning propellant fragments. The safety analyst would need to generate the input data external to the dispersion model using range safety debris impact analysis tools. There are several factors that analysts should keep in mind when computing the source term for burning solid propellants:

Once the source terms are defined, the toxic dispersion model must be able to predict how high the source puffs or plumes will rise. Many dispersion models include capabilities to treat buoyant sources; however users should be aware of the details of how these models are formulated. Many dispersion models are designed for use with air pollution smoke stack type emissions where the emitted gases are only 10 to 20 degrees Celsius warmer than the ambient air. Rocket exhaust plumes and fireballs have adiabatic flame temperatures on the order of 2500 K. If temperature gradients within these two types of sources are very different and the internal velocity gradients that evolve can lead to different turbulence and mixing conditions. If the dispersion model in question uses empirical relationships based on smoke stack type sources, the cloud rise equation may not be suitable to predict the rise of very large and highly buoyant sources.

Most physically based cloud-rise algorithms use conservation of momentum and buoyancy calculations and an accurate solution requires an iterative process that accounts for the entrainment of ambient air into the cloud or plume as it rises. This requires the user to provide the dispersion model with a temperature profile of the atmosphere in the region where the source formed and near the time that the source formed. This is typically acquired from a rawinsonde weather balloon sounding taken at the launch site just prior to the launch.

Atmospheric stability is a critical factor affecting predicted cloud rise. Air temperature naturally tends to decrease with altitude because expansion of the air to a lower pressure causes a temperature decrease. The rate of temperature decrease solely due to the pressure change with altitude is referred to as the standard lapse rate and an actual atmospheric sounding that closely follows the standard lapse rate is a neutral atmosphere (as opposed to stable or unstable). A warm cloud rising in a neutral atmosphere will tend to decrease in temperature at the same rate as the surrounding air and hence will maintain a positive temperature difference and will tend to keep rising. In actuality, ambient air is continuously entrained into the rising cloud causing the average temperature to decline and approach that of the surrounding air, but the cloud will maintain a net positive energy surplus relative to the surrounding air (conserving buoyancy). Stable layers in the atmosphere, where the temperature change with increasing altitude is less than the standard lapse rate, have a retarding effect on cloud rise and act as a barrier to mixing of ambient air across the stable layer. Stable air layers are associated with top of convective boundary layers during the day and can be very strong in regions where persistent high pressure systems lead to large scale subsidence of air. Shallow but strong stable layers also tend to form at night when skies are clear and the ground cools more rapidly than the air by radiating energy to the night sky. These nocturnal stable layers can persist into the early morning hours and may pose an adverse atmospheric condition for toxic dispersion scenarios where burning solid propellant fragments produce weakly buoyant plumes that get trapped in the stable layer near the ground.

A critical parameter in all cloud rise models is the air entrainment rate. More sophisticated models may attempt to internally compute an entrainment rate from cloud turbulence estimates, but many models assign an empirical constant for the entrainment rate. This may be a user input or it may be embedded in the model and not readily accessible to the user.

For a spherical cloud, a constant entrainment rate, gamma, leads to a linear growth in cloud radius as a function of cloud rise distance (Δz) as expressed in equation 3.

(3)

Photographic imagery of Titan vehicle exhaust clouds taken from calibrated cameras at multiple viewing angles was post-processed to estimate cloud volume rate of change as a function of altitude during the cloud rise phase and the empirical data exhibited a linear growth rate in keeping with equation 3 (Abernathy, March 1997; Abernathy, June 1997; Abernathy, February 1996; Abernathy, February 1997). The slope of the linear growth rate phase is the gamma entrainment rate and the intercept with the y-axis serves to define a reasonable approximation of the initial cloud dimensions. Figure 5.1.6 illustrates the cloud growth data derived from a Titan launch cloud imagery case.

An average entrainment rate of 0.36 with a standard deviation of 0.02 was derived from the images of seven launch clouds as shown in Figure 5.1.7. This entrainment rate is consistent with approximate rates applied by other atmospheric dispersion model researchers who have evaluated buoyant cloud rise applications. An entrainment rate of 0.36 is recommended as a best estimate for both normal launch and launch abort clouds; however, very limited field test data is available for actual launch vehicle abort exhaust clouds. Consequently, some ranges have used 0.5 as a more conservative entrainment rate for launch abort clouds. The higher entrainment rate results in lower cloud stabilization height predictions and higher ground level toxic chemical concentration predictions.

One of the challenges with modeling atmospheric dispersion of rocket emissions is that the sources are highly buoyant. Even those sources that initially form at or near ground level often rise to stabilization altitudes of hundreds to several thousand meters above the ground. The downwind concentration versus distance profile for these elevated sources typically has a high concentration near the source where the buoyant cloud at formation time is still in contact with the ground, followed by a zone of very low or zero concentration in the region where the lofted cloud passes overhead. Eventually the exhaust cloud material mixes back down to the ground. This “far field” concentration is lower in magnitude but ramps up to a peak and then declines as the exhaust cloud continues to expand and dilute. This concept is illustrated in Figure 5.1.8.

In contrast, sources that are non-buoyant, or have very low stabilization heights, produce concentration versus distance profiles that are highest at the source origin and steadily decrease with increasing downwind distance. The dispersion model used to predict downwind concentrations from rocket exhaust plumes and clouds should have the following capabilities:

Direct measurement of turbulence is usually restricted to the lower 30 meters or so of the atmosphere where instrumented wind towers are erected. Most dispersion models resort to theoretical derivations of horizontal and vertical turbulence as a function of altitude. Models that apply an average turbulence and wind transport condition over the entire mixing boundary layer may oversimplify the dispersion problem; thus, analysts should verify the validity of such models for range safety applications. Some models define the source cloud or plume as an ensemble of Gaussian puffs and apply local wind speed, direction, and turbulence intensity at the altitude and geographic position of each puff. These puff models also tend to employ puff-splitting and puff-merging algorithms. Puff splitting is initiated when the puff growth exceeds a predefined size or when shears distort the puff beyond a predefined limit. Puff merging is done when puffs overlap each other sufficiently to be considered to effectively represent a single puff.

The dispersion model should allow the user to track cloud or plume dispersion until the predicted toxic chemical concentration drops below a user defined level of concern. Many models partition the large initial source cloud or plume into many overlapping ellipsoidal or spherical puffs or other geometric elements such as “disks” defined by slices of meteorological wind measurement levels through a source cloud. Each puff or disk can be treated as being independent of other puffs or disks and concentration calculations at a point are made by summing the concentration contribution from all puffs or disks that are close to the selected point at any given time. Most models will keep a record of the maximum concentration predicted over a grid of receptor points as the cloud or plume passes through the calculation domain. This type of calculation produces a grid of concentration values that can be contoured to show regions where predicted concentration exceeded a defined threshold at some point in time during the cloud passage. When coupled with an estimate of the model prediction uncertainty, concentration contour plots become a very useful tool for safety personnel to define a toxic hazard corridor within which a prescribed action should be taken, such as evacuation, sheltering indoors, or equipping personnel with protective breathing apparatus.

Maximum concentration contour plots do not indicate the predicted position of the cloud or plume as a function of time and may not provide information on the expected exposure duration. This type of information can be extracted from dispersion model results if the model outputs concentration versus time information, such as “snapshots” of the cloud position at times after the release or concentration versus time history at receptor locations. Knowing the predicted position of the toxic cloud as function of time may be important if emergency responders need to move into the toxic corridor as quickly as possible.

The duration of exposure is also an important consideration for comparing model concentration exposure predictions with different published allowable exposure standards. Chemicals that cause an acute physiological response typically have exposure standards based on short duration exposures with specified maximum allowable “ceiling” concentration values. Hydrogen chloride and nitrogen dioxide, perhaps the two chemicals of greatest concern in rocket launch toxic analyses, are acute irritants that cause almost immediate coughing, choking and burning sensation responses upon exposure. Hydrogen chloride reacts with moisture to form hydrochloric acid, and nitrogen dioxide reacts with moisture to form nitric acid. Cases of severe exposure can result in hospitalization with lung damage and pulmonary edema. Chemicals that accumulate in the body and cause a delayed adverse health response generally have allowable exposure standards that are defined in terms of total dosage and the dispersion model exposure calculations need to be able to integrate concentration exposure over time to calculate exposure dosage. Other exposures standards define permissible thresholds in terms of a time weighted average (TWA) concentration for a specified duration. If the range safety analyst needs to estimate TWA concentration exposures, the applied dispersion model should be required to estimate the maximum TWA concentration by identifying the time interval with maximum dosage accumulation and dividing that dose by the averaging time.

For example, if a chemical of concern has a 15 minute TWA concentration exposure standard and the range dispersion analysis predicts overall exposure duration at a receptor of 40 minutes, the dispersion model should keep a running 15 minute accumulated dosage versus time calculation across the 40 minute exposure period. The maximum dosage during any 15 minute period is then divided by 15 minutes to arrive at the desired TWA concentration to be used in comparison with the toxicological standard.

Many toxicological standards have been developed by different regulatory agencies and jurisdictions. Range safety personnel needing to make decisions on allowable exposure standards are advised to review the Acute Exposure Guideline Levels (AEGLs) published by the United States Environmental Protection Agency as one very useful source of toxicology information (United States Environmental Protection Agency Acute Exposure Guideline Levels, March 2012). The toxicological community has published a number of chemical safety exposure standards to help governments and private industry regulate and control chemical exposures of workers and the general public. Standards developed for the workplace presume long duration exposure of workers and are not suitable standards to be applied for occasional or accidental releases that apply to rocket launch flight safety. The more appropriate standards are those developed for emergency conditions involving accidental releases. The AEGLs have been developed for short duration accidental release exposures. Other US standards of this type are the Emergency Response Planning Guidelines (ERPGs) published by the American Industrial Hygiene Association (American Industrial Hygiene Association, March 2012) and the Temporary Emergency Exposure Limits (TEELs) published by the Department of Energy (United States Department of Energy web site for TEEL information, March 2012). Each of these standards defines recommended chemical exposure thresholds associated with onset of health response symptoms that can be approximately categorized as mild, moderate and severe. Each organization publishes specific definitions of their thresholds. Similar standards are in use by individual countries and the European Union. The AEGLs are of interest here because they are current, provide guidance for exposure durations of 10, 30 and 60 minutes and have established guidelines for HCl, NO₂, hydrazine, MMH and UDMH. To take advantage of the full extent of information in the AEGLs, a range safety dispersion model needs to keep track of predicted exposure duration as well as peak concentration at receptor sites.

The primary use of flight safety toxic dispersion models is to predict potential exposures of mission support personnel and the general public near the launch facility. These receptors of concern are generally at ground level and therefore usually protected from direct exposure to high concentrations that can occur in the middle of the elevated source cloud. There are some exceptions to this general rule that range safety analysts may need to consider that could levy additional requirements on their toxic dispersion computer program. The first special consideration is for aircraft or helicopters that may be operating in the launch area and could fly through a toxic cloud. For example, at least one range has used open cockpit helicopters with special camera mounts to video launches and conduct site surveillance. With the helicopter fuselage doors open, the pilot and crew could be directly exposed to high concentrations of toxic chemicals if they flew through a launch plume or abort cloud. A second consideration is if the launch site has nearby steep mountainous topography. Most dispersion models assume that the wind flow is approximately terrain-following and that a toxic cloud stabilized above the ground remains at that height above the ground as the cloud transports downwind over topography with shallow slopes. This assumption is not necessarily valid in steep terrain and an elevated highly concentrated cloud can impinge on receptors located on ridge tops with potentially much higher concentrations than receptors on the ground in flat terrain. Safety analysts should carefully review the capabilities of the dispersion model to treat toxic source transport and dispersion and receptor concentrations if hills or steep topography characterize the launch area.

Evaluating Toxic Risk

Often the flight safety analyst is interested in characterizing the risk from a toxic hazard in addition to estimating hazard corridors and concentrations at selected receptor locations. The launch decision may be based on assessing the risk from all vehicle hazards, including potential toxic exposures. In order to translate predicted downwind chemical concentrations into a risk estimate, the analyst needs to incorporate the following additional elements into the toxic dispersion analysis:

The US Federal Ranges have adopted Exposure Response Functions (ERFs) as a means to estimate probability of mild, moderate and severe health effects for both sensitive individuals as well as healthy adults given a predicted exposure concentration to specific chemicals (HCl, NO₂ or HNO₃) (National Research Council, 1998). The exposure response function presumes that health response in a population follows a log-normal distribution and that the probability of an individual in a population subgroup experiencing such a response can be estimated using a log-probit probability distribution. The difficulty with this method is that to define the probability distribution, toxicologists must provide an estimate of a lower bound concentration (1 percentile) below which essentially no one in the population experiences the adverse effect and an upper bound (99 percentile) above which essentially everyone in the population experiences the adverse effect (or worse). This type of information was elicited from toxicologists for HCl, NO₂ and HNO₃ and ERF curves, such as those shown in Figure 5.1.9, were developed for these chemicals.

Sheltering people indoors is an effective risk mitigation technique for short duration exposures (e.g. under 1 hour) because the air exchange rate for most buildings limits the amount of contaminated air that can infiltrate into the building. Figure 5.1.10 illustrates a conservative example in which the outdoor concentration is assumed to immediately rise to a peak concentration of 100 ppm at a receptor site and persist at that level for the 60 minutes. The indoor concentration assumes the interior air is well mixed and rises exponentially at a rate that depends on the air exchange rate expressed in building volumes per hour. To increase the effectiveness of a toxic shelter, the ventilation system should be shut down with windows and doors closed during the exposure period. After the exterior toxic exposure has passed, occupants should evacuate the building and the building should be actively ventilated.

Range safety computer programs capable of predicting toxic risk estimates are often statistical in nature and run many random samples with perturbations applied to model uncertainties in such parameters as vehicle failure time, wind and temperature profiles, cloud rise air entrainment rate, turbulence intensities and source term mixing assumptions. Although the primary use of flight safety toxic dispersion models is to predict hazard corridors and risks in the launch area in the final hours of a countdown, these models have applications in the early phases of mission planning. If the range maintains an archive of historical weather balloon soundings, the toxic dispersion model can be executed for large numbers of weather samples to determine the likelihood of exceeding permissible toxic thresholds and the prevailing direction and size of toxic corridors. This type of information may help range personnel make decision among alternate site for a new launch pad or launch facility. Annual or diurnal weather patterns may indicate the most favorable time of day or season to conduct certain launch operations.

References

1. Abernathy Dr R. Ground Cloud Dispersion Measurements During the Titan IV Mission #K23 (14 May 1995) at Cape Canaveral Air Station, Volume I – Test Overview and Data Summary. El Segundo, CA: Report No. TR-97(1410)-1, The Aerospace Corporation; February 1996.

2. Abernathy Dr R. Ground Cloud Dispersion Measurements During the Titan IV Mission #K15 (5 December 1995) at Cape Canaveral Air Station, Volume I – Test Overview and Data Summary. El Segundo, CA: Report No. TR-97(1410)-3, The Aerospace Corporation; February 1997.

3. Abernathy Dr R. Ground Cloud Dispersion Measurements During the Titan IV Mission #K16 (24 April 1996) at Cape Canaveral Air Station, Volume I – Test Overview and Data Summary. El Segundo, CA: Report No. TR-97(1410)-4, The Aerospace Corporation; March 1997.

4. Abernathy Dr R. Ground Cloud Dispersion Measurements During the Titan IV Mission #K22 (12 May 1996) at Vandenberg Air Force Base, Volume I – Test Overview and Data Summary. El Segundo, CA: Report No. TR-97(1410)-5, The Aerospace Corporation; June 1997.

5. American Industrial Hygiene Association web site for ERPG information. < http://www.aiha.org/INSIDEAIHA/GUIDELINEDEVELOPMENT/ERPG/Pages/default.aspx; >, Tested March 2012.

6. Annual Conference of Atmospheric and Dispersion Modeling. George Mason University, Fairfax, VA < http://camp.cos.gmu.edu/15th-announcement.html; >, Tested March 2012.

7. Bader BE, Donaldson AB, Hardee HC. Liquid-Propellant Rocket Abort Fire Model. Albuquerque: Sandia National Laboratories; December, 1971; NM. Journal of Spacecraft, Vol. 8.

8. Center for Chemical Process Safety. Guidelines or use of vapor cloud dispersion models. 2nd ed. New York, NY: American Institute of Chemical engineers; 1996.

9. Clewell III Harvey J. A Simple Formula for Estimating Source Strengths From Spills of Toxic Liquids. Florida: Engineering and Services Laboratory, air Force Engineering and Services Center, Report No. ESL-TR-83–03, Tyndall Air Force Base; May 1983.

10. Dreschel Thomas, Carlton Hall. Quantification of Hydrochloric Acid and Particulate Deposition Resulting from Space Shuttle Launches as John F Kennedy Space Center, Florida, USA. Environmental Management. 1990;14(4):501–507.

11. Register Federal. Appendix I of Part 417 – Methodologies for Toxic Release Hazard Analysis and Operational Procedures. 2006; Vol. 1, No. 165.

12. Gayle JB, Bransford JW. Size and Duration of Fireballs From Propellant Explosions. Huntsville, AL: NASA Technical Memorandum TM X-53314, George C. Marshall Space Flight Center; August 4, 1965.

13. Hanna SR, Briggs GA, Hosker Jr RF. Handbook on Atmospheric Diffusion. Oak Ridge, TN: DOE/TIC-11223, Technical Information Center, U.S. Department of Energy; 1982.

14. Herndon, et al. LATRA3D Technical Description Manual. Torrance, CA: Report No. 10-142/2.1-02, prepared for Department of Air Force, 30^th Space Wing, Vandenberg AFB, by ACTA Inc.; 2010.

15. Ille Gerhard, Charles Springer. The Evaporation and Dispersion of Hydrazine Propellants from Ground Spills. Civil and Environmental Engineering Development Office, Air Force Systems Command, Tyndall AFB, Report No. CEEDO-TR-78–30 1978.

16. Jay Sambamurthi. Plume Particle Collection and Sizing from Static Firing of Solid Rocket Motors. Huntsville, AB: NASA-TM-111873, NASA Marshall Space Flight Center; 1995.

17. Malcolm Ko, Shia Run-Lie, Weisenstein Debra, et al. Global Stratospheric Impact of Solid Rocket Motor Launchers. TRW Space and Electronics Group Sept. 1999.

18. National Research Council, committee on Toxicology. Assessment of Exposure-Response Functions for Rocket-Emission Toxicants. Washington D.C: National Academy Press; 1998.

19. Nyman Randolph L. REEDM Version 7.09 Technical Description Manual. Torrance, CA: Vandenberg AFB: Report No. 99-400/11.3-02, prepared for Department of Air Force, 30^th Space Wing, by ACTA Inc.; 1999.

20. Prince SP, Banning DW. Launch Vehicle Abort Source Strength Model – Final Report-Source Characterization. Denver, CO: Report No. MCR-94-506, Martin Marietta Astronautics; March 4, 1995.

21. Risk Committee, Range Safety Group, Range Commanders Council. Common Risk Criteria Standards for National Test Ranges: Supplement. New Mexico: RCC 321–10, U.S. Army White Sands Missile Range; 2010.

22. Schulman LL, Strimaitis DG, Scire JS. Addendum to ISC3 User’s Guide – The PRIME Plume Rise and Building Downwash Model. Concord, MA: for the Electric Power Research Institute, by Earth Tech Inc.; 1997.

23. United States Department of Energy web site for TEEL information. < www.atlintl.com/DOE/teels/teel.html; >, Tested March 2012.

24. United States Environmental Protection Agency Acute Exposure Guideline Levels web homepage. < www.epa.gov/oppt/aegl/; >, Tested March 2012.

Description of the Phenomenology

A variety of launch accident scenarios can produce significant explosions. In particular, substantial explosions can result if: (1) a vehicle failure early in flight leads to an intact impact with the ground or tower; (2) a vehicle breakup produces ground impact of liquid propellant tanks or solid rocket motor segments; (3) a vehicle tips over at the pad (e.g. due to one or more rocket motors not firing properly or improper release of the mechanisms used to hold the rocket in place prior to flight); or (4) activation of a Flight Termination System (FTS) triggers a propellant explosion.

Under certain weather conditions, which are described below, the shock waves emanating from an explosion may be refracted by the atmosphere back towards the ground as shown in Figure 5.2.1. Depending on variations in: the sonic velocity as a function of altitude, the shock wave can be refracted horizontally and return to the ground so that it impacts localized areas much further away than would be affected by blast waves that normally attenuate rapidly with distance from the source. For large explosions, some effects may occur hundreds of kilometers from the blast. Thus, focusing of the shock waves can produce significant overpressures on distant communities beyond the boundary of the launch facility and result in window breakage and the potential for serious injuries.

Explosive effects can be roughly categorized as either “near field” or “far field”. The near-field effects are attributed to the direct expansion of the shock front. The two most important parameters for characterizing the magnitude, and hence the impact of an explosion, are the peak overpressure and the impulse, which attenuate approximately with the cube of the distance. The overpressures may fall in the 1 to 10 pounds per square inch (psi) range or higher and can cause partial or complete collapse of building walls, break windows or physically throw people off their feet or cause head, chest, abdomen injuries and ear drum rupture. These near-field effects are relatively unaffected by meteorological conditions and are treated as part of the explosive debris hazard and risk assessment. The far-field effect of a large explosion occur at much lower overpressures, below one psi, when a shock front propagates as an acoustic wave through the atmosphere.

Meteorological conditions strongly influence the far-field overpressure time history, and thus the damage caused by an explosion (Reed, 1992; Boyd and Wilfong, 1988). Far-field overpressure levels can be modeled based on the blast wave rays that emanate from the blast source (Rayleigh, 1896). If the surrounding atmosphere is calm (no wind), steady (no temporal thermodynamic property variations in the time of interest), and uniform (no spatial thermodynamic property variations in the space of interest), then the sonic velocity is uniform and steady. As shown in Figure 5.2.2 when there is the idealized atmosphere just described, the blast rays move radially away from the blast site because the sonic velocity is uniform and steady. In this case, the blast wave intensity (overpressure) decreases approximately as the square of the distance from the source. In general, the blast wave intensity decreases as the distance between adjacent rays increases. Variations in the atmospheric properties which affect the sonic velocity determine the direction of the blast rays under actual atmospheric conditions. The sonic velocity is primarily a function of temperature and wind velocity, although humidity also has a small influence.

Figure 5.2.2 illustrates the influence of the sonic velocity profile on the path of blast rays under various atmospheric conditions. A standard atmospheric condition corresponds to a gradient condition when the sonic velocity decreases with altitude. A gradient atmosphere bends the blast rays upward. Conversely, an inversion condition, where the sonic velocity increases with altitude, bends the blast rays toward the ground. A focusing phenomenon is caused by atmospheric conditions that produce a sonic velocity profile that decreases and then increases with altitude (ANSIS2.20-1983). Focusing occurs when the blast rays initially bend upward and then bend back toward the ground, producing areas of elevated overpressure at a remote location. Notice in Figure 5.2.2 that the blast rays are closely spaced in the focusing region.

Figure 5.2.2 also shows that focusing regions are separated from the source by a “zone of silence” where the overpressure is relatively low.

An American National Standard (ANSI S2.20-1983) and many papers have been published on the subject of blast propagation over the past 150 years (Stokes, 1857; Reyolds, 1900; Milne and Philos, 1921; Taylor, 1939; Perkins et al., 1960). The influence of meteorological conditions on blast wave propagation has also been investigated thoroughly (Berning, 1948; Cox et al., 1954; Cox et al., 1949). Focusing is known to produce greatly amplified overpressures at locations that are relatively remote from the blast site. Overpressures up to 9.6 times greater than expected, relative to a uniform standard atmospheric condition, have been reported (Reed, 1992).

A well-documented case of focusing occurred at the Aberdeen Proving Ground on 14 January 1948 (Berning, 1948). In this instance, a 12,000 lb charge was detonated. Observations made at 4500 yards and 12,000 yards from the blast source were consistent with expected decreases of the intensity of the blast wave; at 4500 yards only a faint sound occurred, at 12,000 yards no sound was heard. However, at 25,000 yards, a sharp crack followed by a prolonged rumble was observed. A weather sounding indicated a sonic velocity profile corresponding to focusing conditions. The US government devoted significant resources to develop the capability to predict safe firing conditions for the Aberdeen Proving Ground.

An accidental detonation of 111,500 lb of chemical high explosive at the Medina Facility (near San Antonio, Texas) on 13 November 1963 provides one of the best examples of focusing conditions leading to window breakage (Reed et al., 1968; Pape et al., 1966). A special weather balloon observation was made 30 minutes after the explosion at the San Antonio International Airport. The weather conditions indicate that focusing conditions existed in the direction of the nearby communities. Insurance claims and physical surveys of the area showed that window breakage occurred in scattered pockets. A total of 2786 windows were broken, yet no injuries were reported. Analysis of the evidence indicates that all of the window breakage resulted from rather low peak overpressures (0.03 to 0.14 psi).

More evidence of the potential for blast wave focusing to contribute to window breakage was generated by a series of accidental explosions of ammonium perchlorate (AP) at the PEPCON (Pacific Engineering Company) plant in Henderson, Nevada, on 8 May 1988. The largest single blast at the PEPCON plant was estimated to be equivalent to a surface burst of 103,000 lb of trinitrotoluene (TNT). Unfortunately, very scant weather data are available for the region around Henderson, Nevada, on 8 May 1988. Window breakage occurred up to about 7 miles from the explosion source.

The hazard from this distant focused overpressure is not to people in the open but to people inside buildings who are standing or sitting near windows that can be shattered, with glass shards causing laceration and penetration injuries. Safe design of launch operations requires an analysis to determine if near-pad vehicle accidents can result in explosions large enough to pose a distant focusing overpressure (DFO) risk to receptors beyond the facility boundary. Small launch vehicles (such as sounding rockets) will usually not pose a significant DFO threat and therefore simplified screening analyses can first be used to determine if a more detailed DFO approach is required. A detailed DFO analysis can include acquiring real time launch weather data and using specialized models to: (a) perform atmospheric ray tracing to determine overpressure focusing areas; (b) calculate the probability of window breakage; and (c) estimate the risk to building occupants from glass shard throw. This section describes two simplified screening methods and a much more detailed DFO modeling approach currently used by several US ranges.

Simplified Screening Models

This subsection presents two screening procedures for determining the potential for DFO. The two methods are: (1) a deterministic window breakage screening analysis, and (2) a simplified risk-based screening analysis. If the screening analyses indicate no potential hazards or risk or if mitigations can be implemented to reduce the hazards/risk then the detailed DFO analysis approach is not required. The DFO screening analyses are described below.

Detailed DFO Risk Analysis

If the simplified screening analyses described in the previous subsection indicate that the DFO hazards/risks are not acceptable and mitigations cannot be implemented to reduce the risks, then the implementation of a high fidelity DFO risk analysis should be considered. A detailed DFO risk analysis has been used to support the launch of large expendable launch vehicles (ELVs) from the US Eastern and Western Ranges using specialized software developed to help automate such analyses. This subsection summarizes the steps involved in performing a detailed DFO risk analysis.

Figure 5.2.7 summarizes the physical processes that need to be simulated for a detailed DFO risk analysis:

To account for uncertainties in the yield levels, weather conditions, and overpressure focusing, hundreds to thousands of random Monte Carlo simulations should be performed to sample the yield histogram, vary the sonic velocity profile, estimate the amplified overpressure, and predict the average expected number of casualties and maximum individual risk. The entire process is summarized in the top level flowchart shown in Figure 5.2.8. The following sections provide information on each step of the process.

References

1. ANSIS2.20-1983. Estimating Air Blast Characteristics for Single Point Explosions in Air, with a Guide to Evaluation of Atmospheric Propagation and Effects. New York: Acoustical Society of America; 1983.

2. Baker W, et al. Workbook for Predicting Pressure Wave and Fragment Effects of Exploding Propellant Tanks and Gas Storage Vessels. San Antonio, TX: NASA Contractor Report No. 134906, Southwest Research Institute; 1977.

3. Bankston LT. Sound-Focusing on a Non-Reflecting Flat Earth in a Stratified Atmosphere. U. S. Air Force, Pacific Missile Range Technical Memorandum No. PMR-TM-62-8 1962.

4. Berning WW. Investigation of the Propagation of Blast Waves Over Relatively Large Distances and the Damaging Possibilities of Such Propagation. Maryland: Report No. 675, Ballistic Research Laboratories, Aberdeen Proving Ground; 1948.

5. Boyd BF, Wilfong TL. Impact of Weather on Acoustic Propagation Forecasts for Launch Support. AIAA 26th Aerospace Sciences Meeting (AIAA Paper No 88-0199) 1988;11–14.

6. Cox EF, et al. Upper-Atmosphere Temperatures from Helgoland Big Bang. J of Meteor. 1949;6:300–311.

7. Cox EF, et al. Meteorology Directs Where Blast Will Strike Bull of Amer. Meteorological Soc. 1954;V35.

8. HAZL (Window Fragment Hazard Level Analysis) Version 1. U.S. Army Corps of Engineer Research and Development Center June 2000.

9. Milne EA, Philos. Mag & Journal of Science. 1921;42:96–114.

10. Pape BJ, et al. Evaluation of Window Pane Damage Intensity in San Antonio Resulting from Medina Facility Explosion on November 13, 1963. Report for Sandia contract No. August 1966;74–0812.

11. Perkins Jr B, et al. Forecasting the Focus of Air Blasts due to Meteorological Conditions in the Lower Atmosphere. Maryland Report No. 1118, Ballistic Research Laboratories, Aberdeen Proving Ground 1960.

12. Rayleigh JWS. The Theory of Sound. 1896;VII.

13. Reed JW, et al. Evaluation of Window Pane Damage Intensity in San Antonio Resulting from Medina Facility Explosion on November 13, 1963. Annals of the New York Academy of Science. 1968;152:565–584.

14. Reed JW. Explosion Airblast Predictions on a Personal Computer and Application to the Henderson, Nevada, Incident, Minutes of the 23^rd Explosive Safety Seminar. 1988; Atlanta, GA, August 1988.

15. Reed JW. Injuries from the Pepcon Explosion [1988] and Other Incidents, Minutes of the 26^th Explosive Safety Seminar. 1994; Miami, FL, August 1994.

16. Reed JW. Analysis of the Accidental Explosion at PEPCON, Henderson, Nevada, on May 4, 1988. Propellants, Explosives, Pyrotechnics. 1992;17 pp. 88–95.

17. Reyolds O. Scientific Papers. 1900;1 pp. 89–106.

18. Stokes GG. British Assoc Report 1857; Trans. of Sect., p.22.

19. Taylor GL. The Propagation and Decay of Blast Waves. R.C. 1939;39.

20. Tomei EJ. Propellant Explosive Hazards Study. El Segundo CA: TOR-0089 (4025-04)-1, Vol. 2, Aerospace Corp.; 1989.

21. Tomei EJ. Explosive Equivalence of Liquid Propellants, JANNAF conference paper presented in Houston. Vandenberg AFB, CA: TX on April 21–23, Aerospace Corp.; 1998.

22. TTU GRTL. Misty Picture Data: Glass Window Experiment. Lubbock, Texas: Glass Research and Testing Laboratory, Texas Tech University; 1987.

23. Wang J, McDonald JR. Fracture and Stress Pattern Correlations in Window Glass Plates. Lubbock, Texas: Glass Research and Testing Laboratory, Texas Tech University; 1986.

24. Ward, J., Implementation of Preliminary PIRAT Explosion Model, Memo for Record, dated 22 Oct 1998, Research Triangle Institute, Cocoa Beach, FL.

25. Wilde PD, Anderson M. Development of a Yield Histogram for Space Shuttle Blast Risk Analyses. San Diego, CA: JANNAF Safety and Environmental Protection Subcommittee Meeting; 1999.

26. Wilde PD, Collins JD. Computation of Explosion Induced Window Breakage and Associated Casualties. Orlando, FL: Proceeding of the 28^th Explosive Safety Seminar; 1998.

27. Wilde PD, Philipson L. Titan IV Yield Histogram Analysis. Torrance, CA: Report No. 97-350/4.1-02, ACTA Inc.; 1997.

28. Wilde PD, Philipson L, Anderson M. Launch Vehicle Yield Histogram Analysis for BLASTX/C. Torrance, CA: Report No. 98-370/9.1-02, ACTA Inc.; 1998.

29. Willoughby AB, Wilton C, Mansfield J. Liquid Propellant Explosives Hazards. Burlingame, CA: Volumes 1–3, URS Research Co.; 1968.

30. WINGARD – Version 4.1. Window Glazing Analysis Response and Design, developed for General Services Administration, Public Building Service, Office of the Chief Architect, developed by: Applied Research Associates, Security Engineering and Applied Sciences. 2003; Vicksburg, MS, 39180.

31. Young LA, Becvar K, Meyer S. Multi-Hit Glass Penetration Software Development – Volume 1 Technical Report. Task Report Prepared By Applied Research Associates, Inc. for the Combating Terrorism Technology Support Office Technical Support Working Group, Contract Number N41756-01-C-7448 2002.

Unguided Rockets

There are two common classes of unguided rocket, small rockets launched by amateur rocketeers and sounding or research rockets designed to accommodate measurement instrumentation or perform scientific experiments during their sub-orbital flights. This narrative focuses on sounding rockets.

Sounding rockets are typically unguided, solid-fueled, two stage vehicles (other configurations also exist) with a nose tip, which may or may not separate from the second stage, depending on mission objectives. They are spin-stabilized by deployed fins but may incorporate thrusters for despinning during coast and before descent. The first stage separates after burnout via drag, and there may be a substantial coast phase before second stage ignition.

Unguided vehicles are typically launched from a rail oriented to point them in the right direction to accomplish their mission. Adjusting the rail orientation for the winds at the time of launch is typically accomplished either by “wind-weighting” or by using a trajectory optimization computer program. Unguided vehicles are most commonly flown without any range safety Flight Termination System. Consequently, risk management consists of determining hazardous areas and applying an exclusion policy.

Many ranges consider it sufficient to determine the regions in which the spent stages or other planned jettisons will impact and to protect those regions. The size of these regions is typically developed by a combination of experience and analyzing the uncertainties in thrust and wind-weighting adjustments (Hennight et al.,1964; de Jong, 1963).

Less frequently, ranges attempt to model the risks that may occur as a result of failed unguided rockets. Absent a guidance system, the collection of trajectories defining the performance envelope of the vehicle is assumed to include all breakup state vectors. The performance envelope is typically defined to be large enough to accommodate the range of trajectories the rocket may have to fly to accommodate the wind variations that occur during the month of the planned launch. Obviously, after application of wind-weighting factors on the day of launch the envelope is much smaller. While off-nominal trajectories for a sounding rocket may be generated by a variety of failure modes, such as fin failures or propagation of a failure of the motor grain to a case burn through, the magnitude of the off-nominal excursions is typically limited by the spin stabilization of the rocket.

These two factors, the limited magnitude of the off-nominal excursions and the large trajectory envelopes, typically allow the hazard posed by a failed sounding rocket to be characterized by only considering breakup state vectors from trajectories within the performance envelope. These breakup state vectors are typically combined with selected breakup modeling hypotheses to estimate an upper bound to the risk resulting from the flight of the sounding rocket. Thus, for example each breakup state vector may consider the vehicle re-entering intact, breaking into stages, and breaking up as if one of the motors were overpressurized.

Airplanes

Launches from an airplane differ from land-based launches in several important ways:

Air-launched rockets benefit from the flexibility of being able to be launched from locations remote from population centers and at an altitude above significant portions of the atmosphere. The rocket is dropped from the aircraft and, after achieving a safe separation distance, it is ignited and it pitches up to begin its flight. Protection of the aircraft and its crew requires defining procedures for addressing a “hang fire” where the rocket is not ejected as intended. Additional operational concerns include assessing how long after the rocket is dropped it can initially hazard population centers. This is to assure that there is adequate time for the vehicle to be acquired by range safety tracking equipment before forcing the vehicle to be destroyed because positive control of the vehicle is not possible.

From a risk analysis perspective, the major difference between an air-launch and a land-based launch of a vehicle is the additional uncertainty introduced into the launch point position and velocity. This information is incorporated in the dispersed trajectories characterizing the vehicle guidance and performance uncertainties.

Sea Launches

Sea launches, like air launches, provide the capability to select a launch point at some distance from land-based population centers. In addition, it may be feasible to accommodate larger space boosters for a sea launch; they do not, however, offer the launch point altitude advantages of air launches.

Sea launches also pose similar concerns to air launches: The operational concept for the launch must protect the ship and its crew from vehicle failures resulting in the rocket falling back and impacting on or near the ship. There may be special concerns of range safety tracking and control of the vehicle. There is a need to assure vehicle range safety control and a need to protect the ship from early launch failures by using strategies such as inhibiting destruct action until the rocket has advanced far enough along its trajectory so that there is little chance of hazarding the ship.

As with the launches from an airplane, the most important factors that distinguish sea launches from land based launches for the risk analyst is the need to account for the additional uncertainty in launch position and initial velocity induced by the ship’s motion. This is normally treated by incorporating this additional uncertainty into the collection of dispersed trajectories used to define the launch vehicle’s guidance and performance envelope.

Scientific Balloons

Scientific balloon flights differ in several significant ways from launch of a space booster or a sounding rocket. The mission time lines for scientific balloon flights are significantly longer than those for rockets. Because balloon lateral motion is governed by the wind, the time to assess anomalies and react to them is much larger than for rockets. Typically, balloon risk analyses consider three phases: the ascent, float, and descent phases of flight (Dargelos-Descoubez, et al, 2007; Fuentes, 2008).

Risks are driven by a combination of the balloon flight path and the area hazarded by the impacting balloon and its payload. The hazarded area is based on the physical dimension and the horizontal distance the object will travel due to wind drift during descent from the top of a person’s head to the ground. The region at risk is dependent on the direction of the prevailing winds and the wind variability.

Risks in the launch area are commonly managed by a combination of selection of the location for the balloon release and constraining balloon release conditions to assure that the immediate ascent path is clear of populations. In general, the entire path is selected to minimize overflight of more densely populated areas. Stricter controls are applied to the ascent and descent phases.

While balloons are not equipped with destruct systems, hazard containment is often achieved by cutting down the balloon if the winds change so that the balloon might drift over densely populated areas. Protection is required for both the balloon and the payload. The lethal area of the balloon depends on altitude at the time of termination. At higher altitudes the balloon is expected to have collapsed on itself presenting a compact mass. At lower altitudes it will be more extended. Sheltering is considered for participants. Allowable launch winds are determined based on the size of the balloon. A hazard corridor is developed based on population density to assure compliance with the Ec limits. In addition, overflight of larger towns is avoided. As the balloon is tracked, the effects of unanticipated winds that might cause it to drift beyond the corridor are limited by the ability to cut it down.

References

1. Dargelos-Descoubez, Fuentes Nathalie. On-ground Risk Estimation for Scientific Balloon Flights in France. Proceedings of the 2nd IAASS Conference, Space Safety in a Global World 2007; ESA SP-645, July 2007.

2. de Jong JL. Review of Some Methods for Dispersion Calculation and Impact Prediction for Sounding Rockets, High Altitude Engineering Laboratory. Ann Arbor, Michigan: Department of Aeronautical Engineering, University of Michigan; 1963.

3. Fuentes Nathalie. On-ground Risk Estimation for Scientific Balloon Flights over the world. Proceedings of the 3rd IAASS Conference, Building a Safer Space Together 2008; ESA SP-662, July 2009.

4. Hennight Keith. Field Wind Weighting and Impact Prediction for Unguided Rockets. NASA Technical Note TN D-2142 1964.