Kristina Blennow

Forest wind damage models are typically based on the assumption that windstorm damage results from the interaction between horizontal wind forces and forest stand properties. In complex terrain, mountain waves caused by stably stratified air flowing over mountains can generate standing waves and severe downslope windstorms on the leeward side. Using the windstorm of 19 November 2021 in a mountain valley in southeastern Norway as a case study, we tested two hypotheses:
1. Forest stand properties do not significantly contribute to explaining forest damage during a mountain wave event.
2. Meteorological variables related to atmospheric stratification, turbulence, and non-horizontal airflow significantly contribute to explaining forest damage during a mountain wave event.
To test these hypotheses, we combined forest damage observations with a high-resolution numerical weather prediction model and Random Forest modeling. We used SHapley Additive exPlanations (SHAP) values to quantify the contributions of individual model features. Incorporating forest stand variables did not significantly improve predictive performance, whereas potential temperature gradient, vertical airflow velocity, and wind gust speed, capturing turbulence, did. SHAP analysis showed that although wind gust speed helped explain damage, its influence was secondary to that of potential temperature gradient, which had the strongest explanatory power. The model demonstrated good discriminative power between damage and no damage in the test set. Our findings underscore the limitations of conventional models reliant on horizontal wind speed, highlighting the need for high-resolution numerical weather prediction models that resolve three-dimensional flow in complex terrain, especially during mountain wave events.