Coupling machine learning and forest simulations to promote the applicability of long-term forest projections under climate change perspectives
No hay miniatura disponible
Fecha
Título de la revista
ISSN de la revista
Título del volumen
Editor
FAO ;
Resumen
Descripción
Projecting forest dynamics is the foundation for sound decision support in adaptive forest management. However, due to their complexity, many forest modeling techniques addressing global changes in terrestrial ecosystems are limited to scientific applications. Integrating conventional research and artificial intelligence technologies has the potential to bridge research and practical use. In this study, we propose a Machine Learning (ML) framework that facilitates the implementation of long-term forest projections under climate change scenarios. Our approach combines ML and forest simulations based on process-based models to project forest dynamics. The goal is to leverage the complementary strength of process-based and state-of-the-art ML models to improve predictions at a reduced computational cost. We use environmental data and periodic field measurements at a national scale to train ML models to predict forest growth. By integrating process-based simulations we investigate how the additional variables can improve the prediction accuracy. The proposed hybrid ML framework identifies forest dynamics processes and drivers across spatial and temporal scales, contributing at many levels to the climate change adaptation: from increasing awareness of the climate-induced hazards to enhancing education and assisting in sustainable natural resource management and planning.
Keywords: adaptive forest management, climate change, forest growth modelling, machine learning
ID: 3623078