A.3 Event Consequences
In order to understand the consequences of an event, it is necessary to understand the concentrations of CO2 and the duration of the escape. In addition, it is necessary to model the dispersion of the CO2 (giving the concentrations) combined with the lethality of the varying concentrations. Beyond the effects to human health there may be physical effects to include blast and cooling.
The following sections continue with a discussion of release scenarios, the dispersion programmes available to model a compressed dense/supercritical fluid release, plus the specifics required to model CO2. There then follows a review of the data that are required to complete a hazard analysis with specifics on the dose-response relationship for probability of fatality from exposure to CO2.
In order adequately to discuss model outputs and inputs a common software product was used by the authors. Numerous software packages and codes are available. The use and referencing of DNV PHAST101 here does not represent an an official endorsement. Modellers should refer to their software provider to discuss limitations and validity for use with CO2.
A.3.2 Introduction to commercial dispersion models available
A great number of computer models have been prepared to enable prediction of the results of accidental release of dense gases or fluids, drawing on established data from agencies such as NIST102. However, only a small number in common use have been evaluated for accuracy of result against feed test data obtained by experiments for dense gas releases. The American Center for Chemical Process Safety103 has researched on 22 models, some which are publicly available and others that are proprietary products. Of these models only 10 have been tested against selected field test data for the accuracy of predicted results.
PHAST as a widely used proprietary package of predictive risk and consequence calculation programmes was included in the evaluation.At the time, it performed well against the measured data104. For this reason and its general availability, PHAST has been used to carry out the example calculations for the rupture cases described in 188.8.131.52.
The outputs of dispersion CO2 dispersion modelling can be expected to include (not exclusively):
- Mass flow rate vs. time graphs.
- Total mass remaining in the pipe vs. time graphs.
- Pressure at the 'orifice' vs. time graphs.
- Vertical and horizontal view isopleths.
- Limits of lethality from the gas cloud and probability of fatality.
- Release temperature vs. time graphs.
- Post-expansion solid mass fraction.
- Blast-distance relationship.
A.3.3 Modelling source terms for CO2 hazard analysis using commercially available models (excluding PHAST)
Many models allow the source term data to be input directly into the model. This allows the user to place correct physical properties for carbon dioxide within the models and obtain dispersion modelling results.
The selection of the source term parameters used for dispersion modelling should be done with care; Table A.3 sets out bounding assumptions that should be used when carrying out screening calculations.
A.3.4 Combination of dispersion modelling and SLOT/SLOD data to give fatal area predictions
Consequence modelling from the dispersion calculations involves determining the effect distances for any given level of harm which can be caused for each failure case that has been considered for analysis. The effect distances may vary depending on the weather conditions and orientation of release. Unless the worst case effect distance found for all calculated failure cases and release variables is small enough to determine by inspection that the risk is already tolerable (usually because of adequate separation of possible hazard sources from potentially exposed populations), further calculations are required. These calculations should take into account the risk that each failure case, and its effects, can have on any local population. These risks can then be added together from the cases and conditions being considered in order to produce, as appropriate, either the individual or societal risk measures discussed in Annex A.4.2 and Annex A.4.3. Such calculations comprise theas also discussed in Annex A.2.2.
Table A.3 Example source term assumptions
The frequency for each failure case is required, and has to be estimated or obtained from sources such as those in Annex A.2.3. Population data for the potentially exposed area will be required: in the offshore context, this should include shipping movements and/or platform occupancy. Unless the release mechanisms in the individual failure cases are independent of the weather conditions (i.e. they are directional through equipment orientation or extreme terrain effects), probability data on weather conditions and wind directions should be inputted. These latter data may need to be divided between day and night depending on whether the exposed population also varies. Probability data or event trees may need to be constructed to account for possible measures in place, as mitigations of the consequences of the cases considered. This could be the likelihood that release durations are restricted by activation of emergency shutoff valves, or protection available to exposed populations by timely evacuation.
105 For choked flow the orifice speed (before atmospheric expansion) equals the sonic speed, while for the subsequent atmospheric expansion 'supersonic' flow may occur. As indicated, some models presume a 'pseudo velocity' as input to the dispersion model. It should be realised that it is more important that the near-field jet entrainment is predicted accurately (and therefore the concentrations in the near-field), rather than that the correct value of the post-expansion velocity is chosen.
All of the above can be obtained straightforwardly by QRA analysts. However, data for the level of harm to be considered,which in the case of a carbon dioxide release is the risk of death or irreversible serious injury, are more specialised. Furthermore, they are dependent on the available toxicological research information.
Table 3.1 gives observed exposure reactions of the human body to various concentrations of pure CO2 in air. Section 3.5 describes the benefits of excluding impurities which would allow the effects of CO2 to be masked by those of other gases. However, individuals will have varying responses to concentration levels and the duration of exposure. There can be no discrete predictable point where all individuals will have the identical reaction to any given gas exposure. A more appropriate method is to use probability distribution mathematics on observed or experimental data for exposure of large populations to specific doses. This can be used to obtain a statistical model for assessing a dose-response relationship for a typical population. The probit (probability unit) method is a customary analysis technique used to obtain a generalised time-dependent relationship for any variable that has a probabilistic outcome defined by a normal distribution. (See Lees or similar standard texts for detailed discussion on probit106).
The relationship would be of the form:
Probit = a + b log(dose) where a, b are constants characteristic of the gas (or any other agent).
HSE's SLOT and SLOD levels for CO2 can be used (through the definitions of SLOT and SLOD) to set the threshold and 50 % mortality levels of any exposed population for CO2. The HSE data are given in Table 3.2 text, noting that these have been derived for land use planning, and not for offshore situations.
The UK HSE107 gives the dangerous toxic load (DTL) values for SLOT and SLOD from which the following probit for fatality from exposure to CO2 can be derived. This assumes that SLOT is equivalent (conservatively) to a 1 % probability of mortality in an exposed population.
Y = ln C8.t – 89,8
Y is the probit value, C is concentration of CO2 in air in parts per million by volume, and t is exposure time in minutes.
The probit variable is normally distributed between 2 (zero probability) and 8 (100 % probability of outcome) with a mean value of 5, and a standard deviation of 1. The derived Probit for CO2 gives the following probability results (see Table A.4) that fit the data for SLOT and SLOD.
Table A.4 Derived probability of fatality for CO2
Graphically, the Probit result is as shown in Figure A.1.
Figure A.1 CO2 dose-fatality relationship
A.3.5 The impact of impurities on CO2 consequence modelling
It is unlikely that CCS facilities will be processing or transporting pure CO2. Gas processing of CO2 is an expensive operation, and if the impurities are set at too low a level, removing them will make the CCS process too costly, and producers will choose either to emit or to move their operations elsewhere. Neither of these would deliver the required reduction in global CO2 emissions.
This means that the impact of the impurities on the source term should be assessed. Of particular importance is the change to the predicted mass flow rate. If the impurities are likely to reduce the mass flowrate then, for the purposes of risk assessment, they can be ignored unless the chosen risk criteria cannot be met.
In addition to the impact on the source term, impurities that may be toxic, such assulfide or CO, could be present within a CCS facility or pipeline. It is imperative to check that such impurities do not constitute the 'defining' hazard in terms of consequence modelling from CO2-containing equipment. This aspect has been discussed in 3.5.
101 PHAST is a hazard analysis computer package, used to identify situations which present potential hazards to life, property or the environment.
102 The National Institute of Standards and Technology (NIST) is based in Gaithersburg, (Maryland) and Boulder (Colorado), and provides thermochemical, thermophysical, and ion energetics data compiled under the Standard Reference Data Program.
103 Guidelines for Use of Vapor Cloud Dispersion Models, 2nd Edition 1996 (American Institute of Chemical Engineers, Center for Chemical Process Safety – NewYork, USA).
104 See Figure 8.3 in CCPS reference.
106 Loss Prevention in the Process Industries (ibid) Chapter 9 Section 9.18.3