In hot water: Uncertainties in projecting marine heatwaves impacts on seagrass meadows.

Seagrass ecosystems, vital as primary producer habitats for maintaining high biodiversity and delivering numerous ecosystem services, face increasing threats from climate change, particularly marine heatwaves. This study introduces a pioneering methodology that integrates Dynamic Bayesian Networks of ecosystem resilience with climate projections, aiming to enhance our understanding of seagrass responses to extreme climate events. We developed cutting-edge metrics for measuring shoot density and biomass in terms of population and site extinction, presented as annual ratios relative to their respective baselines. These metrics include associated uncertainties and projected recovery times. This innovative approach was applied in a case study focusing on Zostera muelleri in Gladstone Harbour, Australia. Utilising five downscaled climate models with a 10 km resolution, our study encompasses a range of Shared Socioeconomic Pathways and emissions trajectories, offering a comprehensive perspective on potential future scenarios. Our findings reveal significant variations in seagrass resilience and recovery times across different climate scenarios, accompanied by varying degrees of uncertainty. For instance, under the optimistic SSP1-1.9 scenario, seagrass demonstrated a capacity for recovery heat stress, with shoot density ratios improving from 0.2 (90% Prediction Interval 0.219, 0.221) in 2041 to 0.5 (90% PI 0.198, 1.076) by 2044. However, this scenario also highlighted potential site extinction risks, with recovery gaps spanning 12 to 18 years. In contrast, the more pessimistic SSP5-8.5 scenario revealed a significant decline in seagrass health, with shoot density ratios decreasing from 0.42 (90% PI 0.226, 0.455) in 2041 to just 0.2 (90% PI 0.211, 0.221) in 2048, and no recovery observed after 2038. This study, through its novel integration of climate models, Dynamic Bayesian Networks, and Monte Carlo methods, offers a groundbreaking approach to ecological forecasting, significantly enhancing seagrass resilience assessment and supporting climate adaptation strategies under changing climatic conditions. This methodology holds great potential for application across various sites and future climate scenarios, offering a versatile tool for integrating Dynamic Bayesian Networks ecosystem models.