Coupling habitat-specific temperature scenarios with tolerance landscape to predict the impacts of climate change on farmed bivalves

Due to climate change, heatwaves are likely to become more frequent, prolonged and characterized by higher peak values, compared with climatological averages. However, the thermal tolerance of organisms depends on the actual exposure, which can be modulated by environmental context and microhabitat characteristics. This study investigated the frequency of occurrence of mass mortality events in the next decades for two species of farmed bivalves, the mussel Mytilus galloprovincialis and the clam Ruditapes philippinarum, in a shallow coastal lagoon, characterised by marked diurnal oscillations of water temperature. The effect of heatwaves was estimated by means of tolerance landscape models, which predict the occurrence of 50% mortality based on the exposure intensity and duration. Scenarios of water temperature up to the year 2100 were modelled by combining two mechanistic components, namely: 1) monthly mean water temperatures, simulated using a hydrodynamic model including the heat budget; 2) daily oscillations, estimated from the harmonic analysis of a twenty year-long site-specific time series of water temperature. Scenarios of mean daily sediment temperature were estimated by means of a cross-correlation model, using as input the water temperature one: the model parameters were estimated based on a comprehensive set of site-specific water and sediment temperature observations. The results indicate that for both species the risk of mass mortality rapidly increases starting from the 2060s. Furthermore, the daily patterns of water temperature seemed to be relevant, as overnight it falls below the predicted mortality thresholds for a few hours. These findings suggest that further studies should address: 1) the improvement of tolerance landscape models, in order to take into account the integrated effect of repeated nonlethal stress events on mortality rate; 2) the prediction of environmental temperature in specific habitat, by means of both process-based and data driven models.