Remotely sensed quantification of non-fire-related disturbances and their contribution to global forest dynamics

Stand-replacing disturbances vary in their type, frequency and severity across regions, with consequent heterogeneity in their effects on forest stand age, structure and carbon cycling. These effects can be assessed using dynamic global vegetation models (DGVMs), but representations of disturbances in these models are generally limited to wildfire, with other major natural agents, such as windthrow and biotic outbreaks, typically neglected. Furthermore, human actions, such as harvest and forest management, may intensify or dampen disturbance regimes and modulate forest susceptibility to certain agents. Here, we assess the relative importance of non-fire-related disturbances in canopy turnover, as opposed to fire-related, accounting for human influence across the world. Using the Landsat-derived Global Forest Change product, we delineate contiguous patches of tree cover loss at 30m resolution for the period 2001-2016 and combine them with different global satellite-based data products to identify fire disturbances, land conversion and exposure to human activities. We then calculate disturbance rates by forest ecoregion, excluding land use change and differentiating by fire association and forest intactness. We find that the rate of non-fire-related disturbances in intact forests, mainly windthrow and biotic outbreaks, exceeds the rate of fire in most tropical regions, as well as in British Columbia’s coastal and mountain temperate conifer forests and West Scandinavian and Russian taiga, whereas wildfire dominates in remaining boreal regions. Outside intact forests, the rate of non-fire-related disturbances, with the addition of anthropogenic disturbances, mainly harvest, exceeds the rate of fire in most regions of the world, apart from the fire-prone boreal Siberia, Mediterranean Australia, boreal North America and West American mountain temperate conifer and Mediterranean forests. The dominant disturbance agent shifts from fire in intact forests to non-fire outside intact forests, mainly in West Siberian taiga, East Australian temperate broadleaf and mixed forests, and some North American boreal forests. This analysis shows the potential of leveraging satellite remote sensing for assessing disturbance regimes and provides a globally consistent dataset that can help achieve more realistic disturbance simulations in DGVMs and reduce uncertainties in projections appraising the future role of forests in global carbon cycling.