Tuesday, September 9, 2008


Hazard assessment is often used to predict what risks there are from a chemical or chemicals. One method of hazard assessment is to use a model which determines the concentration of a chemical in the air or water within a given area. This could either be within a facility or in the environment near the facility. These models are then used for making decisions on equipment, operating procedures, etc.

But from my observations, from both within industry/consulting and from outside, there are severe problems with these assessment models. Note that the concept itself is not the target of my criticism. Rather it is how it is applied and what models are used to make the assessment.

The most important of these problems is that managers do not know how to use the results from hazard assessment. Generally, managers take an all or nothing approach to the assessment. For example, say an emission standard is 500 ppm. If the predicted concentration in an area is 450 ppm, they will say "the level is below the standard therefore we do not need to do anything".

Let me interject at this point one common misconception. People often think that if a pollutant's level is below the standard then it has been scientifically considered safe. Not true. A scientifically determined safe level is called a criteria. Standards may be based on criteria, but they also take into account such things as technology, and most importantly, politics. Regulatory agencies and legislators are often heavily lobbied by the chemical and other industries. And in some countries the standard making body is also responsible for promoting the same industry. Bottom line: standards are about as much politics as they are science.

So back to our manager. Rather than saying "let us reduce the hazard to the best degree available", he will nothing. Or worse he may remove existing or planned safeguards. What should he do? He should be using the results as guides as to determine where to best use his resources, the best technology to use, etc.

The second problem with the hazard assessment models I have seen is something called sensitivity. Sensitivity in models is the change in the results due to small changes in the input values. A model is called sensitive if a small change in the input produces large changes in the output.

The hazard assessment models I have used are very sensitive. For example, in one case I changed the value of density in one program by only 5 percent and the result was a doubling of the output value. If models are too sensitive, then there output values are unreliable.

The last problem is the use of models on either only one part of the plant or on only one plant within a given area. For example, in Thailand near an industrial estate, many people complain about air pollution. But the environmental impact assessments for the individual plants all say there should be air pollution problem. What has happened is that the models used considered only the specific plant and not other nearby sites. But the sum of those emissions is a problem.