Ecological risks of pesticides under future climate and land-use scenarios: A Bayesian network approach

Abstract

In the future, pesticide transport and exposure in aquatic environments are expected to be altered by land-use and climate changes. These changes may lead to higher exposure concentrations and changes in agricultural management practices, which will vary depending on the climate region and affected processes. Traditional risk assessment often applies deterministic approaches, using single-value estimates and assessment factors to account for uncertainty, and usually does not take into account future changes in climate and land-use. To overcome some of these deficiencies, this PhD focused on exploring the application of Bayesian networks (BNs) to facilitate probabilistic risk assessment. The developed BNs are meta-models which use various types of information and data sources to assess exposure and effect of pesticides, such as toxicity tests, monitoring data, and process-based or case-based prediction models. The BNs predicted distributed risk quotients for two northern European case studies. For a southern European case study, the probability of an effect on various biological endpoints, endpoint groups, and communities was predicted. This research showed how BNs can be applied to cover some of the shortcomings of traditional risk assessment by better accounting and communicating uncertainty in all model compartments thereby aiding risk management decisions.