Understanding the sources of uncertainty to reduce the risks of undesirable outcomes in large-scale freshwater ecosystem restoration projects: An example from the Murray–Darling Basin, Australia

•We model uncertainty in achieving ecological outcomes under the Basin Plan.•We compare a simplistic ecosystem condition proxy and an ecosystem state metric.•Both metrics report improvements in the ecosystem health of the Coorong.•Managing to the simplistic metric may not fully account for ecosystem complexity.•Incorporating risk analysis information could be helpful for decision-making.

There are a growing number of large-scale freshwater ecological restoration projects world-wide. Assessments of the benefits and costs of restoration often exclude an analysis of uncertainty in the modelled outcomes. To address this shortcoming we explicitly model the uncertainties associated with measures of ecosystem health in the estuary of the Murray–Darling Basin, Australia and how those measures may change with the implementation of a Basin-wide Plan to recover water to improve ecosystem health. Specifically, we compare two metrics – one simple and one more complex – to manage end-of-system flow requirements for one ecosystem asset in the Basin, the internationally important Coorong saline wetlands. Our risk assessment confirms that the ecological conditions in the Coorong are likely to improve with implementation of the Basin Plan; however, there are risks of a Type III error (where the correct answer is found for the wrong question) associated with using the simple metric for adaptive management.