5.1.10.1 Auditing prediction in EIAs

Bisset examined 791 impact predictions from four projects in the UK, and found that only 77 of these predictions could be audited, and that 57 were probably accurate.

Henderson considered 122 predictions from two projects in Canada. Of these, 42 could not be audited because of lack of monitoring data, and 10 either because they were too vague or because of modifications to the project. Out of the 70 predictions audited, 54 were substantially correct, 13 partly correct or uncertain, and 3 were definitely wrong.

The fifth and most extensive environmental impact audit was performed by Culhane, who examined 239 impact predictions from 29 EIAs, intended to form a representative cross-section of EIAs prepared in the USA between 1974 and 1978. The projects were concerned principally with agriculture, forestry, infrastructure, waste management, and uranium processing. Culhane found that most of the impact predictions in these EIAs were very imprecise, with less than 25% being quantified. Few predictions were clearly wrong, but less than 30% were "unqualifiedly close to forecasts''.

More recently, 181 different predictions in EIAs were audited by Buckley in Australia as a nationwide exercise. The study showed that the average accuracy of quantified, critical, testable predictions in environmental impact statements in Australia to date is around 40-50%, and that predictions where actual impacts proved more severe than expected were on average 33% more than those where they proved as much or less severe (estimated to be around 54%). Table 5.2 shows some highlights of Buckely's study.