Abstract: Anthropogenic stressors are causing increasing degradation of marine ecosystems worldwide. Currently, however, available approaches for mitigating the effects of these stressors are severely limited. Potentially, anthropogenic stressors might be timed such that their interaction with ecosystem dynamics mitigates their impact. A key measure of impact for conservation is resilience, the ability of a population, community or ecosystem to resist and recover from stress. However, finding the times when ecological resilience criteria are met is challenging because resilience typically emerges non-linearly from complex scenarios involving interactions of anthropogenic and natural stressors and local ecosystem dynamics. Here, we demonstrate the potential of stressor scheduling (aka ecological windows) to increase the realised resilience of complex marine ecosystems. We show for seagrass meadows distributed globally and subjected to dredging stress that ecological windows can achieve up to a fourfold reduction in recovery time and 35% reduction in extinction risk. These results were obtained by modelling cumulative impacts of dredging against baselines of ecological dynamics using a Dynamic Bayesian Network (DBN). The dynamics of 28 seagrass meadows worldwide impacted by coastal development were modelled. Ecological windows for these meadows were site-specific and varied most strongly in relation to life history and dredging duration. However, windows tended to occur during autumn and winter, and some windows were consistent across sites globally, despite variations in local environmental conditions. Our results contrast strongly with current windows-based approaches that typically focus on single ecological events (e.g. coral spawning) while ignoring stressor impacts on resilience. They also demonstrate that resilience is dynamic in space and time relative to specific stressor scenarios. Thus, resilience can be much more effectively managed using ecological windows for this and other ecosystems facing anthropogenic stresses amenable to scheduling.

Biography:  Dr Paul Wu is an early career researcher in statistics, receiving his PhD in 2009 from the Queensland University of Technology. He develops and applies Bayesian Networks (BN), Dynamic BNs, and other statistical (e.g. Bayesian hierarchical models), machine learning (e.g. trees) and simulation based tools (agent based models) to an array of environmental and industry problems, particularly risk modelling of complex ecological systems. He has been recognised with the QUT Vice Chancellors Performance Award for his research and engagement with industry. He has 14 publications in this area, including recent work on “Dynamic Bayesian Network Inferencing for Non-Homogeneous Complex Systems. Journal of the Royal Statistical Society, in review” and “Predicting the Temporal Response of Seagrass Meadows to Dredging using Dynamic Bayesian Networks. Proceedings, Modelling and Simulation Society of Australia and New Zealand, Gold Coast, Qld. MODSIM2015 21st International Congress on Modelling and Simulation”. He has a track record of interdisciplinary collaborations with government, industry and academia across a variety of complex systems projects including: marine ecosystems, healthy waterways, agricultural supply chains, airport business intelligence, and Defence and risk modelling, producing beneficial outputs for industry as well as academic outputs.