Developing predictive models of wind damage in Austrian forests

International audienceAbstract• ContextAmong natural disturbances, wind storms cause the greatest damage to forests in Austria.• AimThe aim of this study is to quantify the effects of site, stand and meteorological attributes on the wind disturbance regime at the operational scale of forest stands.• MethodsWe used binomial generalized linear mixed models (GLMMs) to quantify the probability of damage events and linear mixed models (LMMs) to explain the damage intensity at the forest stand level in four management units with a total forest area of approximately 28,800 ha.• ResultsTimber stock volume, stand age, elevation, previous disturbances, wind gust speed and frozen state of soil contributed in explaining probability of wind damage. While the model of disturbance probability correctly classified 90 % of all cases in the data set (specificity 95 %, sensitivity 26 %), the model for damage intensity explained only low percentages of the variation in the observed damage data (full model R2 = 0.38, fixed effects-only model R2 = 0.09; cross-validation in the four forest management units yielded similar R2 values).• ConclusionThe developed models indicated that decreasing the proportion of Norway spruce (Picea abies [L.] Karst), limiting stand age and reducing the timber stock in course of tending treatments in stands exposed to wind disturbance can mitigate the risk and the expected damage intensity. High gust speeds and salvage cuts after earlier damage increase the probability of further wind disturbance events