Identifying (less) vulnerable forest structures to storms in different conifer-dominated European forest landscapes by simple spaceborne proxies

Wind is among the primary abiotic forest disturbance agents in Europe and globally. Remote sensing techniques have been efficiently used for disturbance detection and assessment but are still less exploited in vulnerability modeling and mapping. Building on a pilot study, we further explore spaceborne spectral reflectance data for deriving proxy predictors of wind disturbance occurrence, based on the Sentinel-2 imagery and the FORWIND database. Random forest regression models are built and tested on random samples from an equidistant network of virtual plots, tailored to the disturbance records of two landscape-scale case studies from the Alpine and lowland regions of Central Europe. The information content of the proxy variables is subsequently interpreted by aerial imagery on smaller subsamples. The results reveal the predictive power of the models, despite the modeling difficulties associated with orographic complexity or the meteorological field. Trend relationships are found between the proxy predictors and forest structure and spatial heterogeneity related attributes, less commonly and not uniformly addressed by previous modeling studies. The findings demonstrate the capacities of simple methods for an initial, landscape-scale assessment and identification of (less) vulnerable forest structures in an objective manner.