Hazmat 101 News - March 2004

An Analysis of 658 Incidents Involving Chlorine Compounds

By Cris Whetton

Introduction

The day before beginning this article I was in the kitchen, helping my wife unpack the groceries, when I noticed a rather expensive-looking, green-lettered box whose label included the words "sea salt - pure flaky crystals". I asked my wife if she had been indulging in any flaky crystals and, if not, what had possessed her to buy such a costly product. "Oh," she replied, "Edwina uses it in her organic restaurant. She told me it was much more economical than ordinary salt, because you need less of it. Look, it says so on the packet." Indeed it does. However, TV producers rarely have much of a scientific education, so I was forced to give my wife yet another brief chemistry lesson, after which I examined the packet in more detail, for something had caught my eye. In amongst the phrases peppered - or salted, if you prefer - with words such as "natural", "pure", "family", "healthy", was the obligatory list of ingredients. Each 100 grams of product was listed as containing no protein, no carbohydrates, no fats, no fibre - and 39 grams of sodium. End of list. And the remaining 61 grams? No mention of water and no mention of "The C-Word".

Which is why I have chosen to weary you with my domestic arrangements: you must not use the C-Word, especially if you are a manufacturer trying to promote a "green", healthy image. What do you think it would do to a salt company's sales if word got out that their product contained CHLORINE. There, I've done it. I have used the C-Word in public. But then, I do not sell salt.

Chlorine is a very useful element, many of its compounds equally so, but it has become one of the demons of the twenty-first century for some groups. What other element needs national and international organisations to support it against the somewhat greater numbers of organisations that exist to attack it? My own organisation has no axe to grind. We are not for or against any particular element or compound; our interest is in preventing accidents, whatever chemicals may be involved. As part of our work we maintain a database of hazardous incidents (currently about 16,000 entries) and publish a selection of these each month in the journal Hazards Intelligence (see www.saunalahti.fi/ility  for details). Some of this information is collected and analysed in special reports covering incidents in oil refineries, food and agriculture, transport and, of course, Chlorine Compound Incidents, the subject of this article. (We also track incidents in certain areas on behalf of our clients.)

About the Incidents

The current edition of Chlorine Compound Incidents (for brevity, these are referred to as 'chlorine incidents' throughout most of this note) analyses 658 incidents which occurred between January 1, 2000, and December 31, 2003, involving chlorine and its compounds. These incidents resulted in at least 50 deaths and over 750 persons receiving medical treatment, though often only as a precaution. Seventy-four percent of the incidents recorded occurred in the USA, with fourteen percent in the European Union (EU), and the remaining twelve percent occurring elsewhere. This must not be taken as indicative of the distribution of chlorine incidents throughout the world; in fact, this is simply the distribution of reports that we receive. The incidents are not only distributed geographically, but also in time and Figure 1 shows the aggregate monthly distribution of the 658 incidents recorded.

Figure 1: Aggregate monthly distribution of incidents

There is a suggestion in Figure 1 of a normal (Gaussian) distribution, centred on July. This distribution is more pronounced in Figure 2, which shows the distribution of only those incidents associated with swimming pools and similar recreational facilities.

Figure 2: Distribution of incidents occurring in swimming pools

Figure 2 shows only those incidents occurring in the northern hemisphere, north of the Tropic of Cancer. The curve correlates well with the monthly variation of temperature in these latitudes, and consequently with the expected levels of pool usage. This suggests - but does not prove - that the sample is statistically unbiased. Most of these incidents occurred in "professionally" managed facilities; we only record domestic incidents where there is injury or evacuation. It is also useful to examine how these incidents are distributed among the various 'things' (entities, in our usage) in which they occur.

Figure 3: Distribution of incidents by entity

Distribution by entity is shown in Figure 3. The four principal sources of incidents, together accounting for nearly two thirds of the total, are recreational users (i.e. pools, spas, water parks, etc.), road transport, industrial users (i.e. companies using chlorine or one of its compounds to make something other than another chemical), and water treatment (drinking water or sewage treatment). Incidents during the use of chlorine to manufacture other chemicals are significantly fewer. This suggests that incidents occurring in the "big four" may be the result of inexperience in the hazards associated with chlorine compounds. In the case of recreational use, there is further evidence for this.

It is natural to ask whether there are operations, such as maintenance, in which incidents are more likely to occur? Some clues are provided by Figure 4, which shows the state of the entity at the time the incident occurred. Two of the categories in this figure need some explanation. "Replenishment" refers to activities taking place within the plant boundary and includes mixing or preparation of chemicals before use. "Load/Unload" is the transfer of material across the plant boundary. Thus, if workers are changing a chlorine cylinder inside the plant, this is classed as replenishment. If, however, they are unloading chlorine cylinders from a truck and carrying them into the plant, this is loading/unloading. The distinction has been made because the activities are usually carried out by different workers. "Normal Operation" covers those cases where maintenance, replenishment or loading/ unloading were not being carried out. Not surprisingly, but rather unhelpfully, the majority of incidents occurred during normal operation.

Figure 4: Entity state at the time of the incident

The level of incidents "In Transit" (mostly road traffic accidents - RTA - and derailments) is consistent with Figure 3. The low level of unknown states is unusual, and we can only assume that incidents involving chlorine are better reported than most. Incidents in maintenance, replenishment, and loading/unloading occurred in roughly equal proportions, which is not surprising since they are similar activities involving human intervention and - presumably - prone to similar levels of human error. Causes, shown in Figure 5, are much more difficult to identify.

Figure 5: Immediate causes of incidents

The major surprise of Figure 5 is that the most common immediate cause of chlorine incidents is equipment failure. This rather contradicts accepted wisdom, which says that most incidents have "management failure" as their root cause. However, it should be noted that Figure 5 shows immediate causes, not root causes. The most interesting aspect of Figure 5 may be the difference between operator errors and maintenance errors. In, say, an oil refinery, one would not expect operators to be any less well trained or any less safety conscious than maintenance personnel. The same should apply to water treatment plants. However, there is evidence that this does not apply to pools and water parks, where many of the operators - who are also responsible for such tasks as replenishment - are seasonal employees, chosen for their sporting skills and with only minimal training.

What Were the Consequences

Given so much hostility to chlorine from certain quarters,  one might expect incidents involving chlorine and its compounds to be responsible for many deaths, on a similar scale, say, to domestic gas explosions or gasoline fires. The facts suggest otherwise: in almost 96% of incidents no deaths occurred. Of 66 deaths recorded in 658 incidents (one death per ten incidents) three were gorillas (counted because they are physiologically so similar to humans) and three occurred in "near misses" where a chlorine compound was being transported, but no release occurred, the deaths being due to other factors.

Figure 6: Distribution of injuries

Injuries are somewhat harder to assess. For the purposes of this study, "Injury" is recorded whenever a person received medical attention - even if that attention was no more than removal to hospital for observation. It is rarely possible to be more precise than that because in most countries the authorities refuse to give further details. Only if the incident is large enough or severe enough to warrant an official investigation is the extent and nature of the injuries accurately known. In 50% of incidents, no person received any medical attention. The general distribution of injuries is shown in Figure 6.

What Chemicals Were Involved

Figure 7: Chemicals involved

As Figure 7 shows, the most commonly occurring chemical is hydrochloric acid. Not surprisingly, chlorine figures largely, either as a raw material or as a reaction product. The split into hypochlorite (unspecified), sodium hypochlorite, and chlorine solution is a consequence of vagueness of many incident reports. Where a report refers to "chlorine powder" or "chlorine tablets" in the context of water treatment it has been assumed that hypochlorite was involved because this is the most common form. This has been recorded as "hypochlorite (unspecified)". Where sodium or calcium hypochlorite has been specifically mentioned, this has been recorded. Cases of "chlorine solution" are even more vague.

Types Of Incidents

In an attempt to categorise the different incidents prior to analysis, each incident is been assigned to one of the following types:

• Simple release [SR]: A chlorine compound (CC) is released with no reaction taking place.
• Reaction between chlorine compounds [CCCC]: Typically (but not always) producing a release of chlorine gas.
• Reaction between CC & ammonia compound [CCA].
• Reaction between CC & water [CCW].
• Reaction between CC & hydrocarbon [CCH].
• Reaction between CC & other chemical [CCO].
• Unknown reaction [UR]: We were unable to determine what the chlorine compound reacted with.
• Near miss [NM]: An incident occurred in which a chlorine compound was involved, but there was no reaction or release.
• False alarm [FA]: An incident was thought to have occurred, but subsequent investigation showed that it did not.
• Other: Anything which does not fit any of the above categories. Most such incidents involve the dumping of chlorine compounds, but with no release.

The distribution of these incident and reaction types is shown in Figure 8.

Figure 8: Incident types

Specific Industries

The data is sufficient to allow detailed examination of specific industries to see whether there are common characteristics which can be identified and used to suggest ways of preventing accidents. What follows is no more than the briefest summary of the findings.

Recreational Facilities 
The majority of incidents - about 12% more than overall, see Figure 4 - occurred in normal operation. The proportion of incidents which occurred during maintenance and loading/unloading is the same as for overall. What may be significant is the large proportion of incidents occurring during 'Replenishment'. This includes such tasks as filling automatic dispensers with chemicals already on site and the mixing of such chemicals before adding them to a dispenser or subsidiary storage tank.

The proportion of operator error is almost double that for overall. This tends to confirm - though it does not prove - the theory that many incidents in this sector are the result of inexperienced, inadequately trained staff. The most common operator error was a mistake in mixing chemicals.

Water Treatment 
With few exceptions, the majority of the incidents occurred in either drinking water or sewage treatment plants and during normal operation. On water treatment plants - theoretically staffed by well-trained professionals - more incidents occurred in normal operation, apparently as a result of equipment failure, than in recreational systems. The major difference is evident in replenishment, which accounted for only 7% of incidents in water treatment, but 23% in recreational facilities. Water treatment plants have a permanent staff.

For water treatment plants, equipment failure was the dominant identified immediate cause of incidents, with human error (operational plus maintenance) less common. Loss of containment - mostly simple leaks - was the dominant equipment failure mode.

Industrial Users 
While the types of users are quite diverse it must be remembered that they are users of chemicals, not processors, and do not always have the same levels of expertise as do their colleagues in the processing industries. The proportion of incidents occurring in normal operation is comparable to the overall proportion, shown in Figure 4, but the proportion of incidents occurring during maintenance is far higher than for any other group. This suggests that industrial users may have less appreciation of the risks involved in maintenance than workers in the chemicals processing industries. Industrial users had a higher proportion of incidents resulting from human error (operator error + maintenance error) than any other sector except transportation.

Transportation 
The largest number of incidents occurred in one of two modes of transportation: road or rail. Of these, over 90% occurred 'In transit', with the remainder occurring during loading, unloading, maintenance or standing idle.

Surprisingly, there were as many incidents caused by equipment failure as by traffic accidents. The majority of identifiable failures were simple leaks in the bulk container. Some of these may have been caused by outside damage, such as fork lift trucks, but most seem to be failures of the container through corrosion, cracking or similar mechanisms. In one spectacular incident a road tanker split along its ventral seam.

Conclusion

Although it can be dangerous to generalise, especially as the incident reports which we collect are biased towards certain industries, the data we have suggests that chlorine and its compounds do not figure excessively in accidents. In the number of chemicals occurring in our incident database (and this should not be taken as a global figure), the most common chemical is fuel oil (diesel, kerosene, bunker fuel, etc.), followed closely by gasoline. Chlorine--most often as a reaction product--occurs one quarter as often. Hydrochloric acid, occurs one sixth as often.

The principal findings of this report are:

• The largest source of chlorine incidents was recreational facilities such as pools and spas, followed by road transport, industrial users, and water treatment. The safe use of chlorine compounds in such situations could be greatly improved by better training.
• The majority of incidents took place in normal operation, regardless of the entity involved.
• Overall, the predominant immediate cause of incidents was equipment failure. However, in the case of recreational use the major causes of incidents were operating error and equipment failure, both of which occurred in similar proportions.
• While chlorine incidents were responsible for few deaths, incidents in recreational facilities caused many to seek medical treatment.
• The misuse of chlorine compounds in cleaning could be avoided with better training.
• There is a lot of confusion over chlorine compounds, many of which have names which sound alike.
• Reactions between pool sanitation compounds and chemicals such as water and hydrocarbons might be avoided by the use of better packaging.

About the Author

Cris Whetton is CEO and Technical Director of ility Engineering, a consultancy specialising in the reliability and safety of process plants. (ility is the last word in Reliability, Availability, and Maintainability.) He is the publisher and managing editor of Hazards Intelligence and also responsible for two of the modules of the Master of Science engineering course in Process Safety and Loss Prevention at the University of Sheffield, U.K. Contact: ility Engineering, Näsilinnankatu 30, B34, 33200 Tampere, FINLAND, Tel: +358-3-214-4550 E-mail: ility@sci.fi  URL: www.saunalahti.fi/ility