3.7 Hazards Associated with CO2 Releases

In the event that the pressure containment is breached, either as a result of damage or operational malfunction, there is the potential for a hazardous situation to arise.

3.7.1 Hazard analysis

Losses from a hazard can often be caused by organisations or individuals failing to use available knowledge to prevent an incident, rather than there being a total lack of knowledge. Hazard analysis makes an important contribution to system safety by making an organisation aware of the hazards, to allow it to apply its knowledge and experience in order to manage safety, or to enable it to seek outside help if the hazard is beyond its expertise or experience.

Various techniques are available for hazard analysis of the risks that may be associated with any particular process or operation. All seek to answer the following questions (in progression):

  1. What undesirable events can happen?
  2. How frequently can they happen?
  3. What are the consequences?
  4. Is the risk from the process or operation acceptable?
  5. What can be done to eliminate the events or reduce the consequences?
  6. Is the risk within acceptable levels?

Hazard identification methods (e.g. 'What if?', 'How can?', 'Hazard and operability study' (HAZOP)) are required to answer the first question. With relevant frequency data, questions two and three are answered by risk analysis methods which may include quantitative risk assessment (QRA). The final questions involve an iterative procedure until an agreed or minimum risk level for the process is obtained.

A work process that can be executed for the hazard analysis of a CCS system is summarised in Figure 3.2. The risk measures, data sources and analysis techniques indicated are discussed in more detail in the following sections.

Figure 3.2 Flow diagram of basic hazard analysis process which is also relevant for a CCS installation

3.7.2 Hazard analysis for offshore CCS

Following the logic described in Figure 3.2, a list of possible events has been drawn up and a coarse assessment made of their frequency and the consequences (low, medium or high: in producing a hazard assessment, statistical data sets, such those described in A.2.3 should be used). These possible events (which are not intended to be exhaustive), are provided in Annex A. From the scenarios listed, those assessed as presenting either a medium or high likelihood or severity were selected as candidates; for example, dispersion modelling, in order to make the models as relevant as possible.

3.7.3 Dispersion modelling

3.7.3.1 Airborne releases

At the initial point of release in air, CO2 will be released in a high velocity jet. This will be a mixture of gaseous CO2 and some fine particles of solid CO2. The characteristics of the initial jet are known as the source terms. They comprise pressure, density, temperature and velocity, which are then used to calculate the mass flow rate and the initial jet momentum within the 'development zone', where the gas expands to atmospheric pressure.

The cloud contains a certain amount of momentum, related to the initial release velocity, which ensures that the cloud starts to move away from the point of release. As it moves, air is entrained into the CO2 cloud, reducing the concentration. Some of the fine solid particles entrained in the cloud of CO2 may rain out onto the ground and form a pile of CO2 snow. Some of the solid could evaporate in the cloud as it takes in heat from the surrounding area. The prevailing wind will also increase the movement of the cloud away from the release point. This may build up when blown towards enclosed and low-lying areas on the platform; it may also spill off the deck of the platform, as a gas, and accumulate in the area below used by ships, and most importantly, the escape pods. As mentioned, fog-like clouds can be formed by the CO2 cooling water from the air, and forming small droplets, which can seriously reduce visibility (this is described more fully in 3.2.4).

As the initial cloud moves away from the release point, further CO2 is released from the pipeline, adding to the cloud. The flow rate from the pipeline decays over time as the inventory is used up. The total inventory released will depend on the pipe from which the release occurs (length, pressure, diameter) and any inventory mitigation in place. During this phase, the cloud is getting larger in overall length and diameter and air entrainment continues around the surface area of the cloud. Rain out of solid CO2 (snow) increases as the release pressure approaches the solid formation pressure, which is approximately 7 barg.

Eventually, the flow rate from the release point decreases and then stops as the total inventory is exhausted. The cloud continues to move downwind, entraining air and eventually dispersing.

A number of different dispersion modelling programmes are available, and guidelines are available to aid the designer to choose one that is suitable52.

When modelling the dispersion of CO2, the models provide different views of the dispersion of the cloud which, when examined together, give a total picture of the cloud. The individual snapshots give the following information:

  1. The mass flow rate over time, demonstrating how the rate of flow of CO2 decays.
  2. The maximum extent of the cloud, which shows for particular concentrations how far the cloud travels in total before it is finally dispersed. As a maximum extent, it gives no indication of the duration of that concentration.
  3. The lethality of the cloud, which combines 'probit figures' with the duration of exposure. This gives an indication of the impact on human health of the cloud at various distances from the release.

It is only by understanding the combined information produced by dispersion models and matching it with a probit function that the likely impact of the scenario can be modelled and understood.

3.7.3.2 Subsea releases

In this situation, at the initial point of release, CO2 will be released into the seawater at depth in a high velocity jet. This will be mostly in the liquid form, but as heat is absorbed from the surrounding water it will form a mixture of gaseous CO2 bubbles and possibly some fine particles of solid CO2.

Dependent on the orientation, the bubbles will have a little horizontal momentum, related to the initial release velocity, but this may be damped rapidly by the presence of the surrounding water. The density of both the liquid droplets and the bubbles is lower than that of seawater and they will start to move upwards from the point of release, and soon become gaseous bubbles. As they move upward, some of the CO2 will dissolve into the seawater (see 2.2.6); the rest will probably emerge at the surface as a relatively cold 'gas pool'.

This will warm as it takes in heat from the sea beneath and the air in the surrounding area, at which point air flow from wind will also increase the movement of the cloud away from the release point.

3.7.4 Hazard modelling examples

This section describes the input and results for a small hazard modelling exercise intended to illustrate several techniques used to predict the consequences of dense phase CO2 pipeline rupture. It also provides a basic guide to understanding the output from dispersion models.

3.7.4.1 Scenarios modelled

As previously described in 3.7.2, the scenarios outlined in Annex A.2.3 were used as a basis for examples to demonstrate the modelling of CO2 releases. Five scenarios were modelled, as shown in Table 3.12.

Table 3.12 Summary of the scenarios modelled

Note 1 ESDV is emergency shutdown valve.

Note 2 + with the ESDV shutting and in working order.

Note 3 – with the ESDV means that it fails and does not shut as it is supposed to.

3.7.4.2 Scenarios 1 and 2, platform pipe leak

Figures 3.3 and 3.4 depict the scenarios.

The assumed conditions are:

- A vertical riser pipe 12" (300 mm) in bore has developed a horizontal leak from a 100 mm (4") hole, a height of 1 m above the platform deck.

- The pipe is located 20 m from the edge of the platform.

- The leak is pointing along the platform deck which measures 80 m x 80 m and is 25 m above sea level.

- The pipe contains pure dense phase CO2 at a temperature of 4 ˚C and a pressure of 150 barg, and has an internal roughness consistent with drawn steel, Ra = 40 microns (Rubert N6).

- Any ongoing flow in the pipeline, prior to rupture, can be ignored as it is negligible compared to the initial flow rates following the rupture.

- The inventory within the riser pipe is sufficient for steady state conditions to be reached.

- Ambient air temperature is 0 °C.

- Atmospheric stability is Pasquill Class D.

- The relative humidity is 70 %.

- Two wind speeds, 1,5 m/s and 5 m/s, are modelled, with the wind direction blowing across the platform deck.

- The surface roughness value for the deck was chosen to take account of the level of floor equipment on a typical production rig installation.

Figure 3.3 Platform pipeline leak scenario, wind speed 1,5 m/s

Figure 3.4 Platform pipeline leak scenario, wind speed 5 m/s

3.7.4.3 Scenarios 3 and 4: sea surface pipeline leak, ESDV operates (limited inventory)

Figure 3.5 shows an artist's impression of the scenarios. The conditions assumed are:

- A vertical riser pipe 12" (300 mm) in bore suffered a guillotine fracture at, or close to, the surface of the water, such that an unimpeded vertical jet is produced.

- The pipe is located 20 m from the edge of the platform.

- The subsea pipe contains pure dense phase CO2 at a temperature of 4 ˚C and a pressure of 150 barg, and has an internal roughness consistent with drawn steel, Ra = 40 microns (Rubert N6).

- Any ongoing flow in the pipeline, prior to rupture, can be ignored as it is negligible compared to the initial flow rates following the rupture.

- The inventory within the riser pipe is sufficient for steady state conditions to be reached.

- Ambient air temperature is 0 °C.

- Atmospheric stability is Pasquill Class D.

- The relative humidity is 70 %.

- There is an ESDV, 600 m away from the guillotine fracture, and that this closes two minutes after the guillotine fracture takes place (in other words, a limited inventory discharge).53

- Dispersion at two wind speeds, 1,5 m/s and 5 m/s, are modelled.

- The surface roughness value chosen was that appropriate to the sea at the two wind conditions.

Figure 3.5 Sea surface pipeline leak scenario: constrained inventory

3.7.4.4 Scenarios 5 and 6: Sea surface pipeline leak, ESDV failure case

This scenario was the same as for the sea surface pipeline leak, scenarios 3 and 4, except that it considers the situation where the ESDV fails to close and the whole inventory of the CO2 pipeline is available to allow the establishment of steady state conditions. Figure 3.6 depicts this scenario.

Figure 3.6 Sea surface pipeline leak scenario: unconstrained inventory

In addition to the direct hazard to personnel resultant from a pipeline leak close to the platform, there exists the possibility that the cooling effect of the expanding liquid CO2 flashing off to gas and cooling as a result (see JT effect, 2.1.1.2), impacts personnel. The cool gas could potentially impact the legs of the platform or other structural components to the point at which their material properties change (see A.1). This would represent a 'worst case scenario', and was not considered further within the modelling exercise, but it is a phenomenon that should be addressed in actual hazard assessments.

3.7.4.5 Scenarios 7 and 8: subsea pipeline leak

Figure 3.7 shows the scenario. The conditions assumed are:

- A subsea pipeline with a 28" (711 mm) bore545556, 150 km long, completely level and straight, develops a leak from a 12" (300 mm) hole in the top of the pipe, 2 km from the landfall at a sea depth of 50 m.

- The subsea pipe contains pure dense phase CO2 at a temperature of 4 ˚C and a pressure of 150 barg, and has an internal roughness consistent with drawn steel, Ra = 40 microns (Rubert N6).

- Any ongoing flow in the pipeline, prior to rupture, can be ignored as it is negligible compared to the initial flow rates following the rupture.

- The inventory within the pipe is sufficient for steady state conditions to be reached.

- Ambient air temperature is 0 °C.

- Atmospheric stability is Pasquill category D.

- The relative humidity is 70 %.

- The CO2 pool size at the sea surface can be calculated using Guide to quantitative risk assessments for offshore installations57.

The liquid CO2 was further assumed to float toward the surface (since its density is less than that of seawater), absorbing heat from the water to form bubbles of gas, as described in 2.3.1. Although some of the CO2 will most probably dissolve in the seawater, some may form solids or hydrates. These effects were ignored, as the objective of the exercise was to demonstrate dispersion modelling, rather than provide an accurate worked example. The existence of the hydrate was also assumed to have no impact on the dispersion of the CO2.

Dispersion at two wind speeds, 1,5 m/s and 5 m/s, were modelled, and the surface roughness value chosen was that appropriate to the sea at the two wind conditions; the wind direction was not relevant, but it was to be clearly shown in the pictorial representations.

Figure 3.7 Subsea pipeline leak scenario

3.7.4.6 Scenario 9: Enclosed space release

The following configuration is assumed:

- The enclosed space is a 40' (12 m) container, with cross-section dimensions 2,4m x 2,4m.

- The leak is a full bore guillotine fracture of a 3/8" (10 mm) instrument line containing pure dense phase CO2 at a temperature of 4 ˚C and a pressure of 150 barg located in the centre of one end of the container.

- The instrument line internal roughness is assumed to be consistent with drawn steel, Ra = 40 microns, Rubert N6.

- Ambient air temperature taken is 15 °C.

- Relative humidity is 70 %.

- No isolation of the CO2 takes place so that steady state conditions will be reached.

3.7.5 Dispersion modelling results

The modelling was subcontracted to HSL, who carried it out using PHAST version 6.7. Their report included a commentary on the dispersion of the CO2 releases and of the visibility of the plume and any other aspects considered relevant as a result of the modelling. The commentary also included an assessment of the consequences of acute exposure in terms of impairment and survivability of persons (human vulnerability assessment) exposed to the CO2 with reference to the guidelines given in SPC/Tech/OSD/3058.

The report of the modelling work is included as Annex C; the following sections summarise the conclusions therein.

3.7.5.1 Scenarios 1 and 2, platform pipe leak

Scenario 1, which could be similar to a hole in a riser pipe, assumes that steady state conditions are reached. The dispersion modelling results show that there is the potential for fatalities on the platform deck, but since the maximum width of the cloud on deck up to 20 m away from the release point is approximately 1,3 m, the areas where fatal concentrations of CO2 are predicted (the area within the red line) are likely to be quite narrow (see Figure 3.8). There are unlikely to be fatalities at sea level. (see Figure 3.9).

Figure 3.8 Lethality plots on the platform deck for Scenario 1

Figure 3.9 Concentration plots at sea level for scenario 1

In this type of situation, a computational fluid dynamics (CFD) program might be a better method of accurately determining the concentration of the CO2, particularly since these have the ability to model obstructions, such as deck-mounted equipment that would perturb the movement of the cloud away from the release point. However, CFD modelling is slow and complex compared to integrated dispersion programs, and so it may be better to accept the more general predictions of the integrated dispersion program and design around them, rather than to follow the more complex route of a full CFD analysis.

The higher wind speed used in scenario 2, as might be expected, produces a longer, thinner and more elliptical shaped plume than in scenario 1, since low wind speeds tend to produce clouds which spread more laterally. Consequentially, the lethality footprint (indicated by the red line) obtained at deck height is slightly longer but narrower than that in scenario 1, as shown in Figure 3.10.

Figure 3.10 Lethality plots on the platform deck for scenario 2

Again, there is the potential for ships to be in the vicinity of the platform, particularly during an evacuation. In this scenario, the area where the concentration of the CO2 is at a toxic lethality level of 1 (indicated by the red line) does not extend to sea level, thus personnel on the ship would be exposed to non-lethal levels of CO2. In the area in which ships (even rescue tenders) would be present, the CO2 concentration would be 3x104 ppm, or 3 %, which is below the SLOT value for exposure of workers (see Table 3.2). At sea level (see Figure 3.11) there is a small area (shown by the green contour line) where the CO2 concentration is 20 000 ppm (2 %) at which level there would be discomfort, but no permanent effects, even following prolonged exposure (see Table 3.1), and there is a larger area (shown by the blue contour), where the CO2 concentration is 5 000 ppm (0,5 %), at which point the physical effects of the CO2 would be barely noticeable. Loss of life at sea level would not be expected, based on these results.

Figure 3.11 Concentration plots at sea level for scenario 2

3.7.5.2 Scenarios 3 and 4: sea surface pipeline leak, ESDV operates (limited inventory)

Scenario 3 could represent a vessel colliding with an unprotected riser pipe. Since, at 1,5 m/s, the wind speed is quite low, the plume remains almost exactly vertical before momentum reduces and the plume begins to slump. The most hazardous concentrations are local to the source (see Figure 3.12). On the platform deck above the fracture the potential for serious injury or fatalities does exist, but the plume at this point is quite narrow (3 m to 5 m downwind, 1 m to 4 m upwind), so personnel would need to be in the immediate area to experience the effects (see Figure 3.13). The modelling has assumed there are no obstructions in the path of the plume. Therefore, in practice, the plume may be more diffuse than is suggested.

Figure 3.12 CO2 concentration plot (side view) for scenario 3, prior to closure of ESDV

Figure 3.13 Lethality footprint at platform level for scenario 3 at low wind speed

Away from the platform, by the time the CO2 has slumped to sea level, the concentration is such that it would not present a hazard to personnel. Once the release has stopped the concentration of CO2 at platform level quickly decays, and within about two minutes has fallen below the fatal level.

The higher wind speed modelled in scenario 4 produces a bending in the plume of CO2, so that it is angled from the vertical. The higher wind velocity also dilutes the cloud of CO2 before it starts to slump significantly. Again, the most hazardous concentrations are local to the source ,as shown in Figure 3.14 but away from the fracture point. The concentration is such that it would not present a hazard to personnel.

Figure 3.14 Lethality footprint for scenario 3 (at platform height) at higher wind speed

3.7.5.3 Scenarios 5 and 6: Sea surface pipeline leak, ESDV failure case (unconstrained inventory)

The unconstrained release modelled in scenarios 5 and 6 means that the CO2 inventory from the entire 150 km pipeline is available for dispersion, allowing the plume to reach steady state conditions. For scenario 5, within 3¾ seconds, following the start of the release, the shape of the CO2 cloud is the same as scenario 3 (see 3.7.5.2), because the release conditions are the same; steady state conditions are reached after 3,86 minutes. The lethality footprint at platform level then extends downwind just over 4,4 m and has a maximum width of approximately 3 m with no upwind extent.

Surprisingly, the concentration at sea level for scenario 5 is shown as only slightly larger than the maximum concentration footprint for scenario 3, and the dimensions are shown as slightly smaller than for scenario 3. This seems contradictory because operating the ESDV is supposed to be a form of mitigation. According to the results, the lethality when operating the ESDV is worse than if there were no ESDV. This is due to the fact that short releases are somewhat conservatively modelled as instantaneous clouds, and continuous releases as jets. This is a limitation of most integral models, and not just the program that was used by HSL.

The conclusions for scenario 6 are the same as for scenario 4. This is because both plumes reach steady state conditions before the release stops, or in the case of scenario 4, before the ESDV is assumed to operate.

3.7.5.4 Scenarios 7 and 8: subsea pipeline leak

Two different modelling steps are involved in these scenarios. The first step is to establish a source term, i.e. the manner in which the CO2 reaches the surface of the sea. The second step is to demonstrate how the CO2 subsequently disperses. For example, DNV's PHAST will only model the latter, therefore the source term was assessed separately59 as a pool 10 m in diameter.

The PHAST modelling of the low wind case (scenario 7) indicates lethality as being over a significant area, up to 4,8 km away downwind from the source of the leak and up to 8,6 km in width (see Figure 3.15). Personnel up to 8 km away may be expected to experience harmful effects.

Figure 3.15 Lethality footprint for scenario 7 at low wind speed

For the higher wind scenario (8) the lethality area is still large, as shown in Figure 3.15 (1,8 km downwind and 1,4 km wide), and the area over which harmful effects might be expected is up to 2,7 km downwind and 2,1 km wide.

Figure 3.16 Lethality footprint for scenario 8 at higher wind speed

The lethality footprints shown in Figures 3.14 and 3.15 are at 0 m (sea level). Whilst there are local 'high spots', for instance immediately above the source of the leak, the majority of the lethal concentration is at a height of below 2 m. In reality, the larger ships which are likely to be in the deeper sea areas where CO2 pipelines are laid will have decks higher than 5 m, so the degree of hazard is probably less than what might appear to be the case.

In practice the areas over which the modelling indicates that lethal levels of CO2 might be reached may not be realised, because the factors that were ignored for simplicity (see 3.7.4.5) will have a significant impact in reducing the inventory of the CO2 reaching the surface:

- CO2 will dissolve in the seawater to form carbonic acid (see 2.2.1). This will not desorb to gaseous CO2, but increase the local seawater acidity

- As the pressure of the liquid CO2 drops at the leak point, it will flash off and its temperature will drop as a result of the JT effect (see 2.1.1.2). Some of the CO2 will absorb heat from the surrounding water, but some will form solids, which will float to the surface of the sea (because the density of solid CO2 is less than that of water, see 2.3.1). As solid, it will contain a volume equivalent of over 500 times (see 2.1.1.4), which would be released over a longer period of time as it absorbed heat from the surrounding seawater.

- CO2 hydrates will be formed, which will capture CO2. This CO2 will be released over a longer period of time as the hydrates absorb heat from the surrounding seawater.

The phenomena described are not well-understood, nor have they been quantified, and their impact in reducing the hazard associated with subsea leaks should be investigated more carefully. Whilst this is clearly a high consequence event, it should also be appreciated that this is expected to be a low frequency event, since subsea pipeline penetration events are statistically very unlikely, due to the preventative measures applied during the design (see 4.2.4), and by close inspection of incident data from North Sea operations over more than 3 decades.

3.7.5.5 Scenario 9: Enclosed space release

Here, 12 air changes per hour were assumed. Steady state conditions are reached when the concentration reaches about 13,3 % (see Figure 3.17), at which point personnel within the containment would probably be severely affected. The potential impact of the postulated leak may be related to the exposure reactions to pure CO2 that are given in Table 3.1, and reiterated in Table 3.13.

Figure 3.17 CO2 concentration vs. time graph for scenario 9

Table 3.13 Exposure reactions to scenario 9

The modelling predicts CO2 concentrations of 2 million ppmv, which is an apparent contradiction in arithmetic, but indicates a pressure build-up within the containment, should there not be any means of pressure relief.

There is guidance; for example, EIGA Safety Info 24/11/E60 provides guidance on the design of enclosed spaces where CO2 will be present. Good ventilation should be provided, and, when that is not enough, well specified and maintained gas detection. There should also be employee training on the hazards of CO2.

3.7.5.6 General observations on the use of Integral dispersion models

The intention of the modelling exercise was to demonstrate what was possible. The results from the scenarios considered have been used to provide guidance on which the hazards described in this publication have been based.

The modelling exercise in this document demonstrates the limitations of integral models; this information can be used in the development of future versions of PHAST and similar commercial integral models. The modelling exercise also demonstrates that integral models should be used together with other models, such as CFD-based models, in order to obtain a more accurate picture of potential hazards. Annex C further expands on these observations.

52 Guidelines for Use of Vapor Cloud Dispersion Models (2nd edition), S. R. Hanna, P. J. Drivas, and J. J. Chang , AIChE/ CCPS, 1996

53 An ESDV on a natural gas pipeline will typically initiate closure following detection of a 10bar pressure differential in the pipe and take about 20 seconds to close fully. More rapid closing increases the risk of damage resulting from fluid hammer effects (see 3.3.9.2). On a CO2 pipeline this initiation may take longer because of the pressure-loss initiated flashing effect within the pipe, which will mean that the pressure front that triggers the ESDV closure may take more time to travel down the pipeline.

54 This is an extremely pessimistic scenario as demonstrated by the robustness of the CATS gas pipeline when the anchor from the Young Lady snagged the pipeline 25 June 2007.

55 Full bore ruptures of large diameter pipelines are rare events according to the available failure rate data

56 A single pipeline diameter has been chosen because smaller pipelines are unlikely to be used offshore: the 28" size has been assumed so that it can be compared with work carried out by Mahgerefteh on onshore natural gas and carbon dioxide pipelines (Pressurised CO2 Pipeline Rupture, I.Chem.E Symposium series No. 154, 2008).

57 Guide to Quantitative Risk Assessment for Offshore Installations. New ISBN 9781870553360. Old ISBN 1870553365, published 1999

58 SPC/Tech/OSD/30, Indicative human vulnerability to the hazardous agents present offshore for application in risk assessment of major accidents, Health and Safety Executive, Version 3, issued 1 November 2010

59 As 20 % of the depth of the release point (50 m), based on work carried out in A guide to quantitative risk assessment for offshore installations, J. Spounge, DNV Technica, Publication 99/100a, Centre for Marine and Petroleum Technology, 1999.

60 Carbon Dioxide Physiological Hazards not just an asphyxiant! EIGA Safety Info 24/11/E, prepared by the Safety Advisory Council, 2011