Introduction 172A brief overview of Atlantis 172Biophysical 173Industry and socioeconomic 173Monitoring and assessment 174 Applications to date 174Lessons learned about EBM in practice 174 EBM in practice 174Monitoring in support of EBM 180Lessons learned about using models to inform EBM 181 Abstract Models are key tools for integrating a wide range of system information in a common framework. Attempts to model exploited marine ecosystems can increase understanding of system dynamics; identify major processes, drivers and responses; highlight major gaps in knowledge; and provide a mechanism to 'road test' management strategies before implementing them in reality. The Atlantis modelling framework has been used in these roles for a decade and is regularly being modified and applied to new questions (e.g. it is being coupled to climate, biophysical and economic models to help consider climate change impacts, monitoring schemes and multiple use management). This study describes some common lessons learned from its implementation, particularly in regard to when these tools are most effective and the likely form of best practices for ecosystem-based management (EBM). Most importantly, it highlighted that no single management lever is sufficient to address the many trade-offs associated with EBM and that the mix of measures needed to successfully implement EBM will differ between systems and will change through time. Although it is doubtful that any single management action will be based solely on Atlantis, this modelling approach continues to provide important insights for managers when making natural resource management decisions.

Humans are changing marine ecosystems worldwide, both directly through fishing and indirectly through climate change. One of the little explored outcomes of human-induced change involves the decreasing body sizes of fishes. We use a marine ecosystem model to explore how a slow (less than 0.1% per year) decrease in the length of five harvested species could affect species interactions, biomasses and yields. We find that even small decreases in fish sizes are amplified by positive feedback loops in the ecosystem and can lead to major changes in natural mortality. For some species, a total of 4 per cent decrease in length-at-age over 50 years resulted in 50 per cent increase in predation mortality. However, the magnitude and direction in predation mortality changes differed among species and one shrinking species even experienced reduced predation pressure. Nevertheless, 50 years of gradual decrease in body size resulted in 1-35% decrease in biomasses and catches of all shrinking species. Therefore, fisheries management practices that ignore contemporary life-history changes are likely to overestimate long-term yields and can lead to overfishing.

An important challenge for conservation is a quantitative understanding of how multiple human stressors will interact to mitigate or exacerbate global environmental change at a community or ecosystem level. We explored the interaction effects of fishing, ocean warming, and ocean acidification over time on 60 functional groups of species in the southeastern Australian marine ecosystem. We tracked changes in relative biomass within a coupled dynamic whole-ecosystem modeling framework that included the biophysical system, human effects, socioeconomics, and management evaluation. We estimated the individual, additive, and interactive effects on the ecosystem and for five community groups (top predators, fishes, benthic invertebrates, plankton, and primary producers). We calculated the size and direction of interaction effects with an additive null model and interpreted results as synergistic (amplified stress), additive (no additional stress), or antagonistic (reduced stress). Individually, only ocean acidification had a negative effect on total biomass. Fishing and ocean warming and ocean warming with ocean acidification had an additive effect on biomass. Adding fishing to ocean warming and ocean acidification significantly changed the direction and magnitude of the interaction effect to a synergistic response on biomass. The interaction effect depended on the response level examined (ecosystem vs. community). For communities, the size, direction, and type of interaction effect varied depending on the combination of stressors. Top predator and fish biomass had a synergistic response to the interaction of all three stressors, whereas biomass of benthic invertebrates responded antagonistically. With our approach, we were able to identify the regional effects of fishing on the size and direction of the interacting effects of ocean warming and ocean acidification.

Marine ecosystem management is increasingly expected to take into account a wide range of ecological and socio‐economic factors. Decision‐making is helped by end‐to‐end ecosystem models that allow exploration of alternative management scenarios given a complex range of interacting factors. We present Atlantis – a spatially structured largely deterministic end‐to‐end marine ecosystem model written in C, available for all major operating systems, based on dynamically interacting physics, biology, fisheries, management, assessment and economics submodels. A detailed installation guide and example application files are also provided. One of the main features of Atlantis is its modularity. At the simplest level Atlantis can have uniform forcing of oceanographic processes, a single primary producer and a consumer. At the most complex level, Atlantis can be used with a range of environmentally driven ecological responses, complex and habitat‐dependent food web, dynamic assessment, management and fishing effort driven by market forces and human behaviour. The combination chosen should be guided by the available data and the questions to be answered. Atlantis provides a large and customizable list of output files and summary statistics that can be analysed and plotted using a number of dedicated r packages. When applying the Atlantis package, the users should be aware of the caveats associated with complex models, such as parameter and structural model uncertainty and challenges interpreting interactions of multiple processes.

On the iconic Great Barrier Reef (GBR), the cumulative impacts of tropical cyclones, marine heatwaves and regular outbreaks of coral-eating crown-of-thorns starfish (CoTS) have severely depleted coral cover. Climate change will further exacerbate this situation over the coming decades unless effective interventions are implemented. Evaluating the efficacy of alternative interventions in a complex system experiencing major cumulative impacts can only be achieved through a systems modelling approach. We have evaluated combinations of interventions using a coral reef meta-community model. The model consisted of a dynamic network of 3753 reefs supporting communities of corals and CoTS connected through ocean larval dispersal, and exposed to changing regimes of tropical cyclones, flood plumes, marine heatwaves and ocean acidification. Interventions included reducing flood plume impacts, expanding control of CoTS populations, stabilizing coral rubble, managing solar radiation and introducing heat-tolerant coral strains. Without intervention, all climate scenarios resulted in precipitous declines in GBR coral cover over the next 50 years. The most effective strategies in delaying decline were combinations that protected coral from both predation (CoTS control) and thermal stress (solar radiation management) deployed at large scale. Successful implementation could expand opportunities for climate action, natural adaptation and socioeconomic adjustment by at least one to two decades.