Academica-e. Repositorio Institucional de la Universidad Pública de Navarra
Publicación
Artículo científico (JournalArticle)
Hybrid forest models in the age of abundant monitoring data and powerful data-driven algorithms: back to the future
Forests are entering an era in which future climate, disturbance regimes, and management conditions may have
no close historical analogues, while society increasingly demands the simultaneous provision of timber, carbon
sequestration, biodiversity conservation, and other ecosystem services. Supporting sustainable, climate-smart
forest management therefore requires predictive tools capable of anticipating ecosystem responses under
novel environmental conditions. Recent advances in artificial intelligence (AI), remote sensing, and large forest
databases have revolutionized forest monitoring and prediction. The rapid expansion of satellite observations,
sensor networks, forest inventories, and automated monitoring programs has generated unprecedented volumes
of data, allowing AI to greatly improve pattern recognition, forecasting, and decision support within historically
observed, data-rich domains. However, the principal limitation of purely data-driven models lies in their limited
transferability to environmental conditions beyond those represented in the training data. As forests experience
novel combinations of climate, disturbances, species interactions, and management practices, statistical relationships derived from historical observations may no longer remain valid, compromising extrapolative predictions. Alternatively, process-based models (PBMs) provide a complementary approach by explicitly
representing the physiological and ecological mechanisms governing forest dynamics, making them better suited
for exploring future scenarios outside the historical record. Therefore, integrating these mechanistic models with
AI tools and Internet of Things (IoT) technologies offers a promising pathway toward next-generation hybrid
models and forest Digital Twins that combine pattern-recognition capabilities with causal ecological understanding. Achieving this vision will require continued development of process-based models, improved hybrid
model architectures, standardized datasets, and rigorous uncertainty analysis to support robust, climate-smart,
multi-objective forest management., Funding for this research has been provided by the Spanish Ministry of Economy and Competitiveness. Open access support provided by the Library of the Public University of Navarre. Project grant TED2021-131081B-C21 funded by MICIU/AEI /10.13039/501100011033 and the European Union NextGenerationEU/ PRTR.