COTSMod: A spatially explicit metacommunity model of outbreaks of crown-of-thorns starfish and coral recovery

Outbreaks of the Pacific crown-of-thorns starfish (COTS; Acanthaster cf. solaris) have been responsible for 40% of the decline in coral cover on the GBR over the last 35 years. With the intensity and frequency of bleaching and cyclonic disturbances increasing, effectively managing these outbreaks may allow reefs an opportunity to recover from these cumulative impacts. Significant research effort has been directed toward developing regional scale models for COTS outbreaks, but these have yet to be fit explicitly to long term time series at the scale of the entire GBR, nor do previous research efforts incorporate explicit estimates of cumulative disturbance history. We developed a stage-based metapopulation model for COTS at a 1 × 1 km resolution using long-term time series and modelled estimates of COTS larval connectivity, nutrient concentrations and important vital rates estimated from the literature. We coupled this metapopulation model to an existing spatially explicit model of coral cover growth, disturbance and recovery across the GBR from 1996 to 2017 to create a metacommunity model. Our results were validated against a spatially and temporally extensive dataset of COTS and coral cover across the GBR, predicting an average coral decline of 1.3% p.a. across the GBR, and accurately recreating coral cover trajectories (mean prediction error = 7.1%) and COTS outbreak classification (accuracy = 80%). Sensitivity analyses revealed that overall model accuracy was most sensitive to larval predation (boosted regression tree; relative importance = 46.7%) and two parameters defining juvenile density dependent mortality (21.5% and 17.5%). The COTS model underestimated peak COTS densities particularly in the Swains and Townsville sectors of the reef, while overestimating COTS density during non-outbreak years. A better understanding of inter-annual variability in larval connectivity, and regionally variable density dependence for adult COTS life stages may improve model fit during these extreme outbreak events. Our model provides a platform to develop upon, and with improvements to estimates of larval connectivity and larval predation could be used to simulate the effects of implementing varying combinations of COTS interventions. This research highlights the importance of the early life history stages of COTS as drivers of outbreak dynamics, emphasizing the need for further empirical research to estimate these parameters