Better than human: can we make better predictions about site contamination through applied data analytics?

The assessment of site contamination is undertaken in a staged manner where each stage informs the need and effort for subsequent stages. The Detailed Site Investigation (DSI) and management/remediation stages are subject to significant quantitative research and regulation, however the Preliminary Site Investigation (PSI) is largely a subjective risk evaluation process. The PSI process is important, because it determines if there is a need for a DSI based on whether there is evidence or a suspicion of contamination. The PSI framework for assessing if there is evidence or a suspicion of contamination on a site is presented in the NEPM (NEPC, 2013). Consistent application of this approach by a 'population of practitioners' should theoretically lead to a predicable outcome. However, in practice, a PSI may be conducted by assessors with a wide range of experience, indicating that there's a potential for different assessors to draw different conclusions based on the same evidence. Research has demonstrated that expert judgement is often subject to biases, and that experience is not a reliable indicator of judgement accuracy among scientists (Davern et al., 2008; Burgman et al., 2011; McBride et al., 2012). The NEPM includes detailed information about the quality of data that is needed for decision making, but it only acknowledges errors that are caused by using inadequate measurement data (NEPC, 2013). (author)