Abstract

Storms affect forests in many parts of the world with varying degrees of damage. With climate change, their frequencies, intensities and effects are expected to change in certain regions, yet we still lack quantifications of their current occurrence across the world’s forests. The talk will focus on remote sensing-based methods deployed to map annual storm-related forest disturbances across the world for the period 2002-2014. Harnessing widely distributed reference sites, meaningful structural and environmental covariates, machine learning and high-performance computing systems, the model built achieved an overall accuracy exceeding 85%. The global map produced will provide quantitative insights into storm relative importance in forest disturbance regimes and will provide an empirical basis for better constrained simulations of forest dynamics and carbon fluxes.