Research Article: A preliminary approach to quantifying the overall environmental risks posed by development projects during environmental impact assessment

Date Published: July 7, 2017

Publisher: Public Library of Science

Author(s): Sam Nicol, Iadine Chadès, Yong Deng.


Environmental impact assessment (EIA) is used globally to manage the impacts of development projects on the environment, so there is an imperative to demonstrate that it can effectively identify risky projects. However, despite the widespread use of quantitative predictive risk models in areas such as toxicology, ecosystem modelling and water quality, the use of predictive risk tools to assess the overall expected environmental impacts of major construction and development proposals is comparatively rare. A risk-based approach has many potential advantages, including improved prediction and attribution of cause and effect; sensitivity analysis; continual learning; and optimal resource allocation. In this paper we investigate the feasibility of using a Bayesian belief network (BBN) to quantify the likelihood and consequence of non-compliance of new projects based on the occurrence probabilities of a set of expert-defined features. The BBN incorporates expert knowledge and continually improves its predictions based on new data as it is collected. We use simulation to explore the trade-off between the number of data points and the prediction accuracy of the BBN, and find that the BBN could predict risk with 90% accuracy using approximately 1000 data points. Although a further pilot test with real project data is required, our results suggest that a BBN is a promising method to monitor overall risks posed by development within an existing EIA process given a modest investment in data collection.

Partial Text

In many countries, environmental impact assessment (EIA) is the main vehicle to protect the environment [1]. For the purposes of this paper, ‘EIA’ refers specifically to the assessment process that considers the potential impacts posed by proposed construction and development projects that will impact environmental features that are legally protected by environmental law. Where impacts of development projects will occur, the main opportunity for prevention and mitigation occurs during EIA, and the conditions that define a future breach of the law are set during EIA. It is therefore critical that the EIA process is able to correctly predict the overall risk (herafter ‘risk’) posed by development, i.e., the likelihood and consequence that a project will have an impact on the environment.

Although overall risk-based approaches are well-developed for a number of related environmental fields, to our knowledge, there are few formal quantitative systems to learn the risks posed by construction and development projects requiring EIA. This is surprising, given that EIA is often the primary method for controlling the impacts of development projects on the environment. EIA assessments often consider potentially severe or irreversible consequences where the impacts of development projects are highly uncertain, so there is a demand for objective, data-driven information. A risk-based approach has many potential advantages, e.g. improved prediction and attribution of risk; sensitivity analysis; continual learning; and optimal resource allocation. Here we demonstrate that established data mining techniques provide the necessary tools to represent overall risk for EIA, and that our approach is feasible given sufficient data. Although further testing with real data is desirable to provide certainty about the effectiveness of a BBN to predict overall EIA risk, other disciplines are already using BBNs to manage analogous problems, so there is a strong precedent for applying these techniques to an EIA context. In many cases the only technical barrier to entry is data availability, so we suggest that regulators who may anticipate using an overall risk-based approach in future should begin to collect data as early as possible.




0 0 vote
Article Rating
Notify of
Inline Feedbacks
View all comments