From model divergence to robust decisions: Integrating structural uncertainty into local-scale forest management in Catalonia, Spain

Ecological modelling is increasingly used to inform forest management decisions under climate change. However, simulation experiments often rely on a single model and may therefore overlook structural uncertainty, i.e., the variability in projections arising from differences in underlying modelling assumptions. Although ensemble modelling can mitigate this issue, in practice managers often have access to only a limited number of locally calibrated models. In this study, we explored structural variability given this constraint and assessed its implications for forest management. Specifically, we addressed three questions: (i) Are current management recommendations viable under future climate conditions? (ii) What are the primary sources of variability in forest projections? and (iii) Are there robust decision thresholds for sustainable forest management based on simulation outputs? To answer these questions, we used two locally calibrated models for NE Spain and conducted long-term simulations of Pinus sylvestris stands across a climatic gradient. Simulations considered three climate change scenarios and three management alternatives, focusing on management-relevant indicators: timber provision, productivity, and sustainability ratio. Structural variability was quantified using variance partitioning to separate the contributions of site conditions, climate scenarios, and model choice to the overall variation in simulations. Further, to provide practical guidance, we define a decision window in which annual timber harvest is lower than the net annual increment (i.e., sustainability ratio < 1), and at the same time models show a low inter-model disagreement, such as their contribution to total variance is < 20%. Results suggest that forest model choice can outweigh climate change uncertainty in timber projections; yet both models agreed that sites characterised by high water availability (humid conditions) remain responsive to current guidelines under future climates. Decision thresholds were largely site-dependent, marking robust planning horizons between 60 and 100 years. Overall, forest management planning could benefit from inter-model agreement, provided the models are reliable. While we explored structural variability between the two available models, robust decision-making would further require adding newly calibrated forest models, integrating additional uncertainty sources, including multiple ecosystem services, and implementing adaptive rules that update management objectives under changing conditions.