Unrealistic ‘worst case’ scenarios
On the 10th March 2005, Nottingham’s Eastcroft incinerator breached its dioxin limits. The spot measurement showed that emissions during the sampling period were 9 times higher than the emissions limit. Critically, given that dioxins are usually only measured every six months, the question arises were emissions 9 times higher over the entire six months? This could lead to the assumption that over the year dioxin emissions were much higher than the emissions limit.
A study showed that spot measurements do not give a representative indication of the actual emissions over a period [De Fré and Wevers, 1998]. Continuous monitoring over a period showed that actual emissions could be 30 to 50 times higher than spot measurements. Combining the two observations for a worst case scenario, the actual emissions could have been 450 (9 x 50) times higher than the emissions limit, over a six month period.
Given the toxicity and persistence of dioxins in the environment, this raises serious questions about the level of rigour applied to monitoring and enforcement [http://www.acrologic.co.uk/lib/public_EA_ConsultationSubmission.rtf].
Calculations used to declare incinerators ‘safe’ are not based on real data. Vyvian Howard coined the term “fact-free modelling” to describe this approach, because most or all of the input data are calculated or theoretical values [Howard C.V. The health impacts of incineration, with particular reference to the toxicological effects of ultrafine particulate aerosols, organo-chlorines and other emissions. Proof of Evidence submitted to East Sussex and Brighton and Hove Local Plan Public Inquiry, 2003].
Some key calculated parameters, such as partition coefficients, are then used to calculate other parameters, which themselves may then be used to calculate yet more parameters. Each time this is done, predictions errors increase, so that the final answers yielded by the model may bear little relation to reality.
Dearden et al  have assessed the performance of a number of software programs for the calculation of partition coefficients, and found that the best had a standard error (s) of 0.271 log unit (a factor of 1.9), whilst the worst had s = 0.654 (a factor of 4.5) [Dearden J.C., Netzeva T.I. and Bibby R. A comparison of commercially available software for the prediction of partition coefficient, in Ford M., Livingstone D., Dearden J. and van de Waterbeemd H. (Eds.), /Designing Drugs and Crop Protectants: Processes, Problems and Solutions/, Blackwell, Oxford, 2003, pp. 168-169].