A safety report should consider all off-site and on-site initiators of major accidents and discuss in qualitative terms their likelihood and consequences. It should quantify the consequences of the most probable accidents to people and the local environment. The accidents that can have a serious effect on nearby populations and the environment primarily involve fires. Spills into surface water drains generally have a small or insignificant effect on the local environment because eco-toxic substances are not stored in bulk and small spills should not leave the site.
The consequences of major fires, which need to be determined include: -
At most warehouse sites a large number of hazardous substances are stored together for varying periods of time. Accident consequence analysis based on the inventory on a particular day is not necessarily appropriate for a COMAH safety report and for full compliance with the Regulations the hazards arising from annual worst case inventories, taking account of the maximum quantity of each substance that is likely to be stored should be analysed.
When processing of chemicals is not carried out and there are no bulk liquid tanks on a site, the probability of a major fire is around 0.01/year. The events which can initiate a fire may be classified as external or on-site and include: -
External Events
The safety report should discuss all of these accident initiators and put forward reasoned arguments as why the risk is not significantly greater than for other similar sites in the UK. Some sites will in fact be subject to a higher risk than the average UK site from one or more of these initiators because they border a busy railway line, they are close to an airport, or the area is known to be prone to subsidence. In such cases the occupier should estimate the frequency of a major accident and attempt to show that it is tolerable.
There is little an occupier can do about the probability of accidents initiated by off-site events such as aircraft impact and earthquake. However, all reasonably practicable precautions should be taken to eliminate the possibility of adverse environmental events such as lightening strike, severe storms and high winds setting off a chain of events which could result in a fire. Assessors should expect to see some or all of the following measures mentioned: -
The probability of flooding leading to ground/river contamination should be low, and if the site is on the banks of a river, the safety report should describe the measures to prevent flooding. These should include:-
At sites located in an area liable to subsidence, bulk storage tanks should be supported on deep piles. Speed and other restrictions should be applied to on-site vehicles to ensure that the probability of one of them initiating a warehouse fire is extremely low. If a major highway bound the site, barriers should be in place to prevent an out-of-control vehicle crashing on to the site and causing a major accident. A chain link fence supported by steel posts may not be sufficient to halt high-speed vehicles and there should be at least a wall, a building or a large open area between warehouses and the perimeter fence. If a busy railway line passes through, or borders, the site, warehouses should be situated well away from the path of a crashing train.
The on-site accident initiators a safety report should address include: -
In principle, a spill of a highly volatile flammable liquid can result in a vapour cloud explosion but in order for this to be sufficiently severe to be counted as a major accident, a number of drums would have to spill their contents simultaneously. While this is conceivable in the event of an earthquake, the probability of such an event is generally considered to be low enough for it to be neglected. However, the safety report should address all of the other hazards mentioned above and describe the measures and procedures that are in force to reduce the risk to a tolerable level.
Spills of highly eco-toxic liquids during HGV loading or unloading of drums and IBCs can result in several tens of litres entering the site drainage system and escaping off-site before an isolation valve can be closed. Assessors should expect Operators to describe the measures they have taken to reduce the probability of this type of accident.
Assessors are likely to see a variety of different approaches used to predict the consequences of major fires at a mixed chemical warehouse and it is sometimes difficult to determine if the assessment satisfies the requirements of the COMAH Regulations. Many Operators do not have access to computer programs to calculate accident consequences and carry out very simple hand calculations, based on published information. They may employ a consultant in which case the hazard analysis is likely to be more detailed. Irrespective of the approach adopted, the CA should demand that a minimum standard is attained and as a general rule the simpler the approach, the more pessimistic the assumptions should be.
As far as the hazards from the smoke plume are concerned, Assessors are likely to see quite wide variations in the assumptions made about five main topic areas: -
The following paragraphs discuss some of the more popular assumptions and comment on their level of conservatism.
One of the principal aims of a safety assessment for a chemical warehouse is to predict the extent of the consequences to people and the environment of the worst accident(s). Unfortunately is not obvious what size of fire and building response will give rise to the greatest toxic hazard range and if a site has several warehouses, the fire which produces the greatest hazard to people is unlikely to be the same as that which maximises the environmental impact. A fire in the warehouse with the largest toxic inventory when the wind speed is high is likely to present the greatest hazard to people. However, a fire in the warehouse containing the largest quantity of ecotoxic substances and one that is brought under control by copious quantities of fire-fighting water, which then escape off-site will have the greatest impact on the environment.
If a mathematical model is used to predict the consequences of fires, it is reasonable to expect that sensitivity studies are carried out to determine the effect on nearby populations of varying the: -
A variety substances contribute to the toxicity of the smoke plume from a warehouse, but those which probably dominate the hazard are: -
FIREPEST 3 calculates the response of a warehouse to fire and determines a time dependent rate of release of toxic vapour and combustion products, but most computer programs used in the production of COMAH safety reports for warehouses do not have this level of sophistication. They usually assume that 5-10% of the toxic substances are released into the smoke plume over a period of 1-2 hours. These assumptions should be regarded as acceptable provided there is conservatism in other aspects of the modelling. If however, an Occupier takes a minimalist approach in all aspects of the consequence analysis (minimum toxic inventory, 5% release over 2 hours, a rising smoke plume etc.) then the results should be challenged.
It is probably overly pessimistic to assume that the whole of the warehouse is on fire, the roof has failed and the smoke plume is neutrally buoyant, because a fire in a large warehouse can generate several Gigawatts of heat. In practice once a fire has caused failure of the roof of a warehouse the smoke rises high into the atmosphere and disperses safely. The fire only poses a hazard in its early stages, therefore it is reasonable for a hazard analysis to assume that only a fraction of the inventory is released over a period of about 1 hour. One acceptable method of determining this fraction is to use the lift-off criterion: -
where
Q = heat content of the smoke plume (MW)
H = height of the building (m)
U = wind speed (m/s)
This allows Q to be determined and related to the area of warehouse on fire if a heat generation rate per square metre of fire can be determined. However, this calculation relies on knowledge of:
Example: Well ventilated burning of formulated agrochemicals - three pallets high - building intact.
A pool fire and a stack fire for this several metres high generate of the order of 5MW/m2. If together the loss processes are conservatively assumed to account for 90% of the heat generated, then the heat in the smoke plume (Q) is around 0.5MW/m2. The maximum area of warehouse on fire that produces a smoke plume, which does not lift-off, is therefore given by: -
This implies that if the height of the warehouse is 10m and the wind speed is assumed to be 10m/s, then the maximum area on fire that will produce a smoke plume which does not lift-off is approximately 400m2. The toxic source term should therefore be calculated as either : -
mtoxic= release of toxic substances (kg).
Mtotal= total mass of toxics in warehouse (kg).
A = palletted floor area of warehouse (m2).
MP = most toxic mass in a palletted area of 400m2.
R = release fraction (5%-10%).
The time of the release should generally be assumed to be in the range 1 -2 hours.
Clearly there is considerable scope for slightly different assumptions and Assessors should recognise that the uncertainty attached to the rate of release of any particular substance is at least a factor of 2.
COMAH safety reports for warehouses may use a Gaussian dispersion model to predict down wind concentrations of hazardous substances. Such an approach should be considered acceptable if: -
The concentration that a person standing under the centre line of a neutrally buoyant plume is exposed to is approximately: -
m = rate of release of combustion products (kg/s).
szsy = standard deviations in the Y and Z directions.
U = wind speed at a height of 10 m (m/s).
H = height of the plume centre line (m).
Numerous equations for the standard deviations of a Gaussian plume in different weather conditions and over different terrain have been proposed, but the spread they produce in concentration profile is generally less than the uncertainty resulting from other sources. For the purpose of warehouse hazard assessment it does not matter much, which equations are used.
Adaptations of Gaussian models to allow for buoyancy by simply elevating the source should be fully explained and justified.
The consequences of wind speeds up to 15 m/s should generally be examined.
Most agrochemical warehouses contain formulations with a toxic label therefore in the event of a major fire, the smoke plume will contain some toxic substances and their combustion products. A hazard range should be calculated assuming that the dose fractions of all dangerous substances are additive. For example if a warehouse contains two toxic substances and a large quantity of non toxic substances as indicated in Table 7, the hazard range from the smoke plume is determined by calculating dangerous dose fractions for each toxic component as shown in Table 7. The result is 1.04 indicating that the hazard range is >100m.
Table 7: Typical analysis for an agricultural warehouse
|
Substance |
Quantity in warehouse |
Rate of release to smoke plume |
DDose |
DDose fraction at 100m during first 30minutes |
|
X |
Mx |
mx |
DDx |
0.2 |
|
Y |
My |
my |
DDy |
0.3 |
|
PVC (chlorine containing) |
1,000 |
mHCl |
DDHCl |
0.17 |
|
Isoprothiolane (sulphur containing) |
700 |
ms02 |
DDSO2 |
0.22 |
|
Metamitron (nitrogen containing) |
600 |
mNO2 |
DDNO2 |
0.15 |
Assessors are likely to see several other approaches used in COMAH safety reports including:-
All of these approaches introduce an acceptable amount of conservatism into the hazard range calculation, which Assessors need to balance against optimism incorporated in other parts of the consequence analysis.
There is a school of thought which believes that it will never be possible to build mathematical models to accurately predict the hazards from a burning pesticide warehouse that contains a large number of substances in different chemical and physical form and packaging arrangements. This school therefore recommends that the hazard analysis is based on a hypothetical "average" agrochemical which releases a variety of low molecular weight toxic gases when burnt. The composition of the average agrochemical is determined from the molecular composition and masses of different substances in the warehouses. The example shown in the table below illustrates how to calculate release rates of the different gases.
Table 8 shows how the composition of the average agrochemical can be determined from the inventory of three agrochemicals is: -
C 8.8 H 10.9 C1 1.54 S 1.5 N 1 O 2.1
The average molecular weight is 267 and if the burning rate of material in the warehouse was 20kg/s, then 74.8mols/s would be combusted. From this the rate of release of different products is:-
HC1 4.2kg/s
SO2 7.2
NO2 0.17
HCN 0.1
CO 19.7
A hazard range based of the addition of fractional dangerous doses is then likely to be conservative unless the warehouse contains very large quantities of an extremely toxic substance such as phorate.
Table 8: Calculation of a representative substance for an agrochemical warehouse
|
Quantity |
Mol. weight |
Composition |
No. of kg atoms carbon |
No. of kg atoms hydrogen |
No. of kg atoms chlorine |
No. of kg atoms sulphur |
No. of kg atoms nitrogen |
No. of kg atoms oxygen |
|---|---|---|---|---|---|---|---|---|
|
50,000 |
322.5 |
C7H7Cl3NO3PS |
1,085 |
1,085 |
465 |
310 |
155 |
465 |
|
20,000 |
296.6 |
C9H4Cl3NO2S |
606.9 |
2,698 |
202 |
134.8 |
67.5 |
138 |
|
40,000 |
380.5 |
C20H32N2O3S |
2,102.5 |
3,364 |
0 |
210 |
210 |
315.5 |
|
3,794.4 |
4,719 |
667 |
655 |
432.5 |
915 |
|||
|
8.8 |
10.9 |
1.54 |
1.51 |
1 |
2.1 |