Modelling irruptive species with hybrid machine learning: A case study of crown-of-thorns starfish outbreaks on the Great Barrier Reef

Owen Stewart, Samuel Alexander Matthews , David Warne, Michael Bode

Irruptions consist of rapid and severe increases in abundance, which for some species can result in significant economic and environmental damage. Forecasting the timing and location of these irruptions is crucial for effective management, however is difficult due to the complexity of the ecosystems in which they occur and the conditions under which they arise. Hybrid machine learning (ML) modelling has the potential to offer more accurate predictions than standard mechanistic, statistical, and ML approaches by combining the flexibility of ML with existing mechanistic knowledge. In this paper we develop a hybrid forecasting model for crown-of-thorns starfish (COTS) irruptions which contribute substantially to ecosystem degradation on the Great Barrier Reef (GBR). The hybrid components of our model include mechanistic models of COTS larval dispersal between reefs, whose predictive influence emphasises the central role of dispersal on COTS irruptions. Further analyses of variable importance support key hypotheses regarding COTS population dynamics, including the roles of chlorophyll and temperature, and the regional benefits of local management. When compared to existing COTS models, our hybrid approach produces more accurate short term forecasts that would improve managers' ability to target high-risk reefs. Along with producing accurate hindcasts, the model shows promise for predicting irruptions during critical periods at the initiation of GBR-scale irruptive events, which the model suggests are underway in 2026.