A Bayesian overhaul of thermal tolerance landscape models: Predicting ectotherm lethality buildup and survival amid heatwaves

Abstract

In the face of escalating heatwaves, accurately forecasting ectotherm population mortality is a pressing ecological challenge. Current Thermal Tolerance Landscape (TTL) models, while surpassing single-threshold metrics by incorporating individual survival times, are constrained by frequentist regression parametrization reliant on constant-temperature experiments, omitting probabilistic outcomes. This study addresses these limitations by pioneering the application of Approximate Bayesian Computation-Sequential Monte Carlo (ABC-SMC) to analyze survival data from Baltic Mytilus mussels subjected to both microcosm (constant temperature) and mesocosm (dynamic temperature) heatwave regimes. The ABC-SMC yields probabilistic predictions of individual lethality buildup and population survival trajectories, closely aligned with observed survival data across both experimental conditions. Informed by more realistic dynamic data, the TTL model predicts local mussel resilience against the most extreme summer heatwaves projected for this century, albeit with considerations for sublethal impacts and potential recruitment declines. Our approach can enhance the predictive accuracy concerning the sensitivity of key marine populations amidst intensifying heatwaves, addressing the urgent need for accurate modeling tools to inform conservation practices and ecosystem management, ultimately aiding in the preservation of marine biodiversity.