Many warehouses contain large quantities of combustible material stacked to a height of many metres. A fire that is started in a pallet at a low level may grow rapidly upwards. Unless the fire is tackled in its earliest stages, it may be very difficult to control and will spread to spread through the whole of the warehouse in less than one hour. The most important issues for a hazard assessment are: -
The features of a warehouse that have a significant influence on the magnitude of the hazard are: -
These are discussed in the following sections
Studies in the UK and US (1,2) deduced frequencies of approximately 10-2 per annum for the frequency of fire starts in a given warehouse - averaged across all industry sectors.
Most surveys of the causes of fires in warehouses have concluded that malicious ignition is the dominant contributor followed by electrical faults, then friction heat and sparks and careless disposal of smoking materials (3). Fire statistics tend to highlight several minor initiators, but surveys often fail to provide a complete picture of the causes of major fires. Hot working, including shrink wrapping, grinding, cutting and welding, which are not explicitly mentioned in some accident study reports, have been responsible for many fires. Other important initiators that should not be overlooked are: -
Nearly all fire initiators can be adequately controlled by relatively simple and inexpensive precautions, but it is difficult to protect premises against the determined arsonist, particularly if, as is often the case, he is an employee or ex-employee.
There are typically 100 major fires in warehouse premises in the UK each year. Usually the most significant factor in determining the rate of fire growth in a chemical warehouse is the flammability of the materials being stored. Some common categories in rough order of declining hazard are listed below
The containers in which material are stored are also crucial in determining the rate at which fires spread. Those which assist the spread of fire include: -
Video recordings and detailed test reports are available from HSL Fire Safety Section (Health and Safety Laboratory) for many of the commodity types described below:
Flammable liquids in plastic IBCs represent a very high risk - internal storage of this type is prohibited under the American NFPA Flammable Liquids code NFPA 30. The risk arises for two reasons:
This is another high risk commodity. A typical example in the context of COMAH assessments are pesticides formulated in flammable solvents for final sale in 1 or 5kg lots. The bottles of liquid are packed in cardboard boxes and shrink wrapped on pallets. Again the risk arises for two reasons:
The risks posed by this type of material were highlighted by a recent fatal fire involving finely ground rubber powder. Any combustible dusts or powders may burn very rapidly if bagged in paper or plastic and stored at elevation - for example in pallet stacks or on racks. As the powders run out of the burning bags they may become dispersed and burn violently as a dust cloud.
The extent to which this kind of risk can be successfully controlled with sprinklers is not yet clear. Water systems are likely to be appropriate.
Table 2: Burning rates and fire growth coefficients for commodities
|
No |
Commodity |
Heat release MW/m2 |
Coefficient (a) (KW/s2) |
|---|---|---|---|
|
1 |
Wood pallets, stack 0.46m high (5-12% moisture). |
1.4 |
0.009-0.02 |
|
2 |
Wood pallets, stack 1.5m high (6-12% moisture). |
5.2 |
0.025-0.123 |
|
3 |
Wood pallets, stack 3.1m high (5-12% moisture). |
10.6 |
0.064-0.156 |
|
4 |
Wood pallets, stack 4.9m high. |
17 |
0.064-0.178 |
|
5 |
Mail bags, filled, stored 1.9m high. |
0.4 |
0.028 |
|
6 |
Cartons, compartmented, stacked 4.6m high. |
1.7 |
0.28 |
|
7 |
Paper, vertical rolls stacked 6.1m high. |
1.276-3.906 |
|
|
8 |
Cotton (also PE, PE/Cot. Acrylic/Nylon/PE) garments in 3.67m high rack. |
0.567-2.268 |
|
|
9 |
"Ordinary combustibles" rack storage, 4.6-9.1m high. |
0.013-0.625 |
|
|
10 |
Paper products, densely packed cartons, rack storage, 6.1m high. |
0.004 |
|
|
11 |
Polyethylene letter trays, filled, stacked 1.5m high on cart. |
8.5 |
0.031 |
|
12 |
Polyethylene trash barrels in cartons, stacked 4.6m high. |
2 |
0.331 |
|
13 |
Polyethylene fibreglass shower stalls in cartons, stacked 4.6m high. |
1.4 |
0.138 |
|
14 |
Polyethylene bottles packed in cartons, compartmented, stacked 4.6m high. |
6.2 |
0.138 |
|
15 |
Polyethylene bottles packed. |
0.178 |
|
|
16 |
Polyethylene pallets stacked 0.9m high. |
0.044 |
|
|
17 |
Polyethylene pallets stacked 1.83-2.44m high. |
0.278-1.111 |
|
|
18 |
Polyurethane mattress single, horizontal. |
0.064 |
|
|
19 |
Polyurethane insulation board, rigid foam stacked 4.6m high. |
2 |
15.625 |
|
20 |
Polystyrene jars packed in cartons, compartmented, stacked 1.5m high. |
14 |
0.331 |
|
21 |
Polystyrene tubs nested in cartons, stacked 4.3m high. |
5.4 |
0.069 |
|
22 |
Polystyrene toy parts in cartons, stacked 4.6m high. |
2 |
0.064 |
|
23 |
Polystyrene insulation board, rigid foam, stacked 4.3m high. |
3.3 |
27.78 |
|
24 |
Polyvinylchloride bottles packed in cartons, compartmented, stacked 4.6m high. |
3.4 |
0.111 |
|
25 |
Polypropylene tubs packed in cartons, compartmented, stacked 4.6m high. |
4.4 |
0.1 |
|
26 |
Polypropylene and polyethylene film rolls, stacked 4.3m high. |
6.2 |
0.625 |
|
27 |
Distilled spirits in barrels, stacked 6.1m high. |
0.625-1.6 |
This type of storage is very common. A significant amount of fire exposure is required to trigger the failure of a steel drum even if it contains a volatile material such as acetone. As a rule of thumb for a full 210 litre drum full fire engulfment for around 5 minutes is necessary to cause drum failure. This clearly requires a relatively large and sustained fire. This risks associated with drum storage of flammable liquids therefore increase greatly if they are stored with other more readily ignited materials e.g. flammable liquids in any type of plastic container, timber pallets, packaging.
For less volatile liquids the burn through of pallets may cause falling and splitting of drums before they fail by internal pressurisation.
The failure of drums may cause a fireball (up to 30m diameter) a jet fire, a severe spreading pool fire or any combination of these. Once the first drum of flammable liquid spills its contents, fire spread through a drum stack or warehouse is highly likely.
This kind of storage should be protected with a foam sprinkler system. A properly specified system will significantly reduce the risk that an ignited leak will cause drum failure.
This commodity is again fairly easily ignited. Fire spread is initially slow with the development of spreading pool fire of molten plastic. In the later stages this fire can grow more rapidly especially if there are a three or more levels of storage in a stack or on racks.
These fires respond well to ceiling level sprinklers.
A large number of full scale tests have been carried out by Factory Mutual in the US (4). Many of the goods tested were packed in large cardboard boxes. Examples of results obtained are shown in Table 2. Fire growth results from these in these tests were fitted to quadratic curves i.e.
The best fit coefficient a gives a measure of the rate of fire growth in the commodity.
These are relatively hard to ignite but fire can spread quickly once it takes hold. The fully developed burning rates are very high - see Table 2.
HSL fire tests have shown that fire will spread through a large stack of HDPE barrels filled with water. The fire takes of order one hour to develop but if it is left all of the plastic barrels will be consumed and their contents spilled. This has obvious significance where toxic aqueous solutions are stored in this way. Fire detection in quiet hours is necessary.
The fire stops growing when: -
If a chemical warehouses contain significant quantities of flammable liquids it is likely that the fire will spread rapidly to the whole warehouse.
For relatively low risk warehouses there is some chance that the fire brigade may control the fire.
For most warehouses the amount of air required to burn the contents exceeds the initially enclosed air by several orders of magnitude. If the ventilation of the warehouse is limited and not increased through failure of elements of the cladding in the early stages of the fire, the concentration of oxygen will fall. Eventually the low oxygen levels will restrict the rate of heat release. The fire will typically continue to burn at a restricted rate that is determined by the available inflow of oxygen - this is known as ventilation controlled or ventilation limited burning. A rapidly growing fire in a relatively small store may reach the ventilation controlled regime in a few minutes. For fires that develop more slowly in large buildings the transition to ventilation controlled burning may take many tens of minutes. If the fully developed burning rate is low and the building is particularly leaky ventilation controlled burning may never occur.
Ventilation controlled fires are characterised by:
Eventually a major fire inside a warehouse will usually cause failure of a large area of the roof cladding and/or structural support. This cladding failure will generally end or significantly relax the ventilation control of burning rate. The size of the fire then increases dramatically typically producing a massive smoke plume with flames leaping to a height of many tens of metres. At this stage the off-site risks typically decline as plume lift-off becomes near complete. The most significant period for hazard assessment is usually during the earlier low buoyancy or ventilation controlled stage and a key issue in risk assessment is the time taken for significant failure of the building skin.
Simplified risk assessment of warehouse fires that consider a "worst case" scenario should normally focus on the early low buoyancy stage. It is not adequate to show that the plume lifts off completely once the building has completely collapsed. Even the simplest risk assessment must therefore include some consideration of the building ventilation and response to fire.
Most modern warehouses are constructed from plastic coated profiled steel sheets bolted onto a metal frame. Brick or concrete block walls separate fire compartments and often form the front wall of the warehouse. Roof lights are often double sheets of glass reinforced plastic (GRP) though glass or thermoplastics such as PVC and polycarbonate can be used.
Roofs of older buildings are commonly clad with asbestos cement. More modern buildings usually use have profiled steel, sometimes as a double layer enclosing a thickness of insulating material such as polyurethane foam or rockwool.
Some appropriate approximate rules of thumb that may be used in predicting building response include:
On many sites several large warehouses are joined together to form a single structure, raising the possibility of a fire in one spreading to the others. Primary considerations in this context are the ability of separating walls to withstand the fire, transmission of thermal radiation from the flame pillar through roof lights and possible effects on the dispersion of smoke.
Some sites operate high bay warehouses. These represent particularly high risks because of the high packing density, the vertical height of the combustible surfaces, the difficulty of carrying out effective fire-fighting and the "chimney effect" in which fire is drawn very rapidly upwards in the narrow gaps between stacks of goods.
Some specific comments on the effectiveness of sprinkler systems in controlling risks for a range of commodities are included in the case studies - Section 3.3.
The existence of a sprinkler system should be taken into account when judging the fire frequency assumptions in some COMAH safety reports. This is one factor along with the quality of site security, site location, ignitability of contents, level of process activity in storage areas etc. that may contribute to the judgement that the fire frequency at a site will be significantly higher or lower than the historical average.
No allowance for the sprinklers should be made for warehouses storing flammable liquids in plastic containers.
It is possible that COMAH safety cases may include mathematical models of fire in warehouses. These methods may be complex and liable to misuse or misinterpretation. If ALARP demonstrations rely critically on the results of models it may be appropriate to seek the advice of a topic specialist.
Risks vary considerably during the development of a fire. Some form of time dependent analysis may therefore be attempted in a safety case - in its simplest form this may be to set a time limit on the duration for which a worst case scenario is sustained.
The first stage of any form of time dependent assessment relies on an analysis of the rate at which fires grow. Models of fire spread should be empirically based. The assumed rate of fire development should be clearly identified and the early stages of (mainly vertical) fire growth should be broadly consistent with test results - e.g. Table 2. The later lateral stages of fire growth are very difficult to model or assess. Fires spread through radiative ignition of nearby commodities, roof level flame extension and flashover, collapse of burning stacks, flow of burning liquids and melts, burning brands and in many other ways. Some models may use the t2 fires define in NFPA 72. In any case final rate of heat release should be consistent with the fire load and any limit on fire size imposed by restriction of ventilation. If the time for a fire to reach its fully developed state is less than 10 minutes or more than 30 minutes this should be justified carefully.
The next stage of an assessment is the calculation of the rate of emission of hot gas. Field modelling (CFD) provides a very precise method of analysing the movement of smoke inside a warehouse and through apertures in the roof. This may be appropriate if the location and development of potential fires as well as the venting of the building are clearly defined. Normally a wide range of fire scenarios are possible and the building response is less certain A less sophisticated approach to calculating flows, such as zone modelling, is adequate.
As a rough rule of the thumb the flow velocity from small vents in warehouse during a serious fire is likely to be of order 10 m/s. Velocities up to 20 m/s are possible if the roof collapses . The temperature of the outflow is variable and obviously must be linked to the rate of burning. In the early stages of a fire a large proportion (>90%) of the heat released by the fire may be transferred into the fabric of the building. After roof collapse the proportion of heat lost from the plume is relatively small (<30%). In ventilation controlled fires heat transfer to (and eventually through) the building fabric may again be very efficient. The exact figure depends the type of fuel, extent of burning surfaces and ventilation area - 60-90% heat loss from the plume is typical.
Radiative heat transfer to the environment outside the warehouse is the main heat loss mechanism for large unconfined fires (5). The rate of heat loss from smoke depends on the size of the luminous flame pillar and the emissivity of the smoke. Most warehouse fires are characterised by an enormous plume of dense black smoke, which to a large extent, obscures the flame pillar. Close to the building the emissive power of the plume fluctuates widely as red/yellow flame blooms appear and disappear on a one or two second time scale, but further away the highly turbulent hot black smoke has an emissive power of less than 50 kW/m2. A precise calculation of the fraction of heat lost from the smoke plume by radiation is difficult and a figure close to 0.30 is usually assumed.
The key to a reliable assessment is the calculation of ventilation rates in the fire state. This ventilation may occur through
Vents are generally incorporated into warehouse roofs with the following objectives: -
Sprinklers and fire-fighting activities of the local fire brigade tend to result in large quantities of water on the floor of a warehouse. Large quantities of fire fighting water can be held in on-site lagoons and in naturally bunded areas of the site, but these measures are only effective if there are no surface water drains close to the warehouse, which discharge into a local stream or river.
Many modern chemical warehouses are designed to retain spills and fire fighting/sprinkler water by use of a ramp at the entrance and bunding of between 0.5 and 1m. In the event of a fire which engulfs the whole warehouse, most aqueous products and a large proportion of the organic liquids end up on the floor together with metal drums, fusible solids and a variety of other debris. Overtopping of a 0.5m bund is a possibility in warehouses that store agrochemicals in liquid form that are stacked to a height of several metres over more than 50% of the floor area.
COMAH sites by virtue of their classification store large quantities of very toxic, toxic, explosive and/or highly reactive substances. A fraction of these is released in the smoke plume during a fire, whilst the bulk of the contents of a warehouse is consumed by the flames and produce a variety of combustion products, some of which are toxic.
Many hundreds of different types of agrochemicals have been developed, but the combustion of only a few of them has been studied in the laboratory and then only under a restricted range of fire conditions. This lack of information presents problems for risk assessors and regulators wishing to predict the toxic effects of the smoke plume from a burning warehouse. Further complications arise from the wide variety of formulation methods and packing arrangements adopted by different manufacturers.
Experiments in which chemicals are burnt in flames or furnaces have been conducted in the UK and European laboratories with the aim of identifying the principal combustion products. There is reasonable agreement that chemicals with the appropriate heteroatoms can, under certain circumstances, release varying amounts of: -
The relative proportions of these compounds depends on the materials exposed and the conditions of the fire or test.
Despite the potential complexity suggested by the above list, almost all significant (acute) hazards arise from either:
Reference (6) reports measurements showing that the survival of medium molecular weight toxic organic compounds (e.g. pesticides) when burned in well ventilated pools is very low (<10-4). Unfortunately real warehouse fires are not usually simple pools. Both liquid and powdered materials are stored on racks well above the ground. If material is vaporised above the ground level in the intermittent section of a fire or carried up by the strong convection currents much larger survival fractions are observed (~30%). Overall survival fractions of around 10% are appropriate for high rack storage of materials where there is a large fire load and the fire is well ventilated (6).
Much higher survival fractions are possible for ventilation controlled fires. This is particularly true of relatively volatile, stable species that are spilled in large quantities in the early stages of the fire before the enclosed air is vitiated. The rate of consumption of the toxic material is greatly reduced by ventilation control of burning rate, so that the overall rate of toxic material emissions may be comparable in the ventilation controlled regime and following structural collapse. The former case is, of course, associated with far less heat and therefore constitutes a higher risk.
Elements that should appear in an analysis of a ventilation controlled fire, where this is a possibility, include:
One useful way of characterising combustion product formation is to determine the conversion efficiency with which the mass of elements such as chlorine, sulphur and nitrogen within agrochemicals is converted to HCl, SO2 NO2 and other toxic gases. These data enable the rate of production of low molecular weight toxic gases to be calculated if the rate of combustion of the parent agrochemical is known. A potential shortcoming of the approach is that some combustion products such as HCN and CO are themselves combustible and, when produced in a warehouse fire, might be consumed in the flames and fail to contribute to the off-site hazard. In spite of this problem, elemental conversion efficiencies are helpful indicators of the types and quantities of combustion products released by a fire and several laboratories world wide are engaged in their measurement.
Experiments carried out by HSL on the combustion of 10 nitrogen containing agrochemicals under well ventilated conditions found that the percentage of nitrogen converted to NO2 varied from 0.7 to 5.5. Other experiments on bagged ammonium nitrate fires on wooden pallets found that the location of the fire and the type of pallet has a significant effect on the quantity of nitrogen dioxide produced. The maximum conversion efficiencies of nitrogen to NO and NO2 were approximately 15% and 3.5% respectively.
Experimental work carried out at several laboratories has enabled agreement to be reached on the following:
Some laboratories quote the results of combustion experiments in terms of the fraction of elements converted to their oxide or chloride. A mass release rate of the combustion product in terms of the combustion rate of the parent compound is given by: -
where
Mf = mass conversion fraction.
Mwtp = molecular weight of the agrochemical.
Nc = number of atoms of the element forming the combustion
product.
Np = number of atoms of the element in the agrochemical.
Mwtc = molecular weight of the combustion product.
Conversion efficiency is defined as kgs of combustion product per kg of parent burnt.
For example if the conversion of nitrogen to NO2 when dimethametryn burns is given as 0.25, the mass of NO2 produced per kg of dimethametryn burnt is given by: -
Molecular weight of dimethametryn = 255.4
Molecular formula = C11H21N5S
Molecular weight of NO2 = 46
Elemental conversion efficiencies that should be assumed in warehouse fire analysis are given below. Operators who use smaller figures should justify them: -
Table 3: Elemental conversion efficiencies
|
Element |
Conversion fraction |
|---|---|
|
Chlorine to HCl |
0.95 |
|
Sulphur to SO2 |
1 |
|
Nitrogen to HCN |
0.05 |
|
Nitrogen to NO2 |
0.05 |
|
Carbon to CO |
0.05 |
Special consideration should be given to production of phosgene if rapid vapour leakage (into an external fire) from fire engulfed containers of chlorinated solvents is possible.
It would be sensible to include combustion products in a hazard assessment whenever they make a significant contribution to the hazard of the smoke plume, but this is not always easy to determine. There are situations where the hazard due to toxic smoke from the fire is completely dominated by vaporisation of highly toxic chemicals, (e.g. a warehouse filled with the pesticide phorate). But there are many other situations where the principal hazard is a result of the emission of toxic combustion products from an essentially non-toxic compound (e.g. an ammonium nitrate warehouse fire). In general the hazards from chemical warehouses are likely to be due in part to toxic combustion products formed in a fire.
3.10 Toxicity of combustion products
Most of the components of the smoke plume from a warehouse fire are hazardous to health, but only some are regarded as acutely toxic. In the main these are the above low molecular weight gases and vaporised very toxic agrochemicals (such as phorate), which have escaped combustion in the flame pillar. Each substance has an associated toxicity value attached to it, which describes the concentration needed to inflict a dangerous dose upon a human. The dangerous dose or Dangerous Toxic Load (DTL) usually applies to an exposure time of typically 30 minutes for a specific concentration (30minutes is assumed to be the average time required for countermeasures to be implemented). If a person receives the associated concentration for more than 30 minutes, then fatality is likely in some cases.
The relationship between concentration and dose can be linear or non-linear, depending on which substance is in question. Toxicity relationships can be represented in the form Cnt =k, where C is the concentration, t is the exposure time in minutes and k denotes the dose (units depend on power of n). Data is available for the light gases and it is possible to derive corresponding data for most agrochemicals using methods recommended by HSE. Human toxicity data for specific agrochemicals are sparse, but Dangerous Toxic Loads (dangerous dose) corresponding to the Land Use Planning SLOT criteria, can be derived using the following guidelines:-
Table 4 : Dangerous toxic load for some combustion products
|
Substance |
Dangerous toxic load (ppmn.min) |
'n' value |
|---|---|---|
|
Hydrogen chloride |
23,730 |
1 |
|
Phosgene |
300 |
1 |
|
Chlorine |
108,000 |
2 |
|
Sulphur dioxide |
4,655,000 |
2 |
|
Hydrogen cyanide |
300,000 |
2 |
|
Nitrogen dioxide |
96,000 |
2 |
|
Carbon monoxide |
3600 |
0.7 |
In general, if animal inhalation data are available then:-
If only oral data are available then:-
The LD1-5 for humans (i.e. the dose causing 1-5% mortality) is assumed to be ¼ of the LD50 multiplied by 70 (the assumed weight in kg of an average individual). The air concentration required to produce a dangerous dose if inhaled for a period of 30 minutes is given by:-
Note 0.625m3 is the approximate volume of air inhaled in 30 minutes by an adult male weighing 70kg).
3.11 Mixtures of toxic substances
There is very little information available on the health effects resulting from exposure to mixtures of toxic substances, and there is a lack of a validated method for predicting whether such effects are additive, antagonistic, or synergistic in nature. For this reason, any approach developed to estimate the hazard range for a warehouse smoke plume containing a mixture of combustion products and their surviving parent compounds, such as agrochemicals, must be based on pragmatic default assumptions and will need to be refined as and when new data becomes available.
As far as COMAH safety reports are concerned Assessors should expect the consequence analysis to be based on the following pessimistic assumptions:-
The dose fraction referred to above is the fraction of a specified toxic load such as a dangerous dose. Thus when adding responses from different gases and vapours the following approach is recommended: -
Although the above rules determine how the dose to individuals exposed to the smoke plume from a burning warehouse containing agrochemicals should be calculated, their application presents problems for practical situations. A typical warehouse is likely to contain hundreds of different products formulated and packaged in different ways. As the fire sweeps through the warehouse the toxic substances in the smoke plume are numerous and change continually.
A further problem is that many of the substances in a warehouse, which have toxic and consequently need only be considered if the accident involves "COMAH substances". Added to this is the possibility that a warehouse may contain a mix brand name agrochemicals all containing the same active ingredient but in different concentrations - see Table 5 for Lindane.
Table 5: Examples of formula products containing lindane
|
Brand name |
AI concentration |
Formulation |
|---|---|---|
|
Atlas Steward |
560g/l |
suspension concentrate |
|
Gamma-Col |
800g/l |
suspension concentrate |
|
Fumite Lindane 10 |
20.8 g a.i |
smoke generator |
|
Fumite Lindane 40 |
84.4 g a.i |
smoke generator |
|
Gammasan |
30% w/w |
Flowable concentrate |
|
Kotol FS |
125 g/l |
Flowable concentrate |
|
Unicrop Leatherjacket pellets |
1.8% w/w |
ganular bait |
It is possible that one formulation will carry a toxic label because of the high concentration of active ingredient while another with only marginally less active ingredient could have an irritating label. A warehouse could contain very much more of the second formulation than the first and if it is neglected in the hazard assessment, the predictions could be grossly optimistic.
The question arises, which substances should be included in the risk assessment? The answer is all formulations in the warehouse where dangerous substances are stored, but this makes derivation of the source term difficult. The problem is compounded by the fact that the location of the seat of the fire and the way the different compounds are distributed throughout the warehouse has a significant influence on the toxic load received by an individual down wind.