Identifying individual drivers of damage to oak during severe UK storms in winter 2021

• Published in: Agricultural and forest meteorology Vol. 373; p. 110797
• Main Authors: Halstead, Kate, Sanderson, Roy, Bonomo, Salvatore, Quine, Christopher, Suggitt, Andrew, Gaulton, Rachel
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 15.10.2025
• Subjects: Natural disturbance; Quercus; Random forest; Structural equation modelling; Terrestrial laser scanning; Wind risk modelling; Structural equation modelling; GLMM; TWI; LIDAR; Random forest; Natural disturbance; GNR; DBH; LOOCV; NDVI; Wind risk modelling; Quercus; Terrestrial laser scanning; SEM; TLS; AIC; DTM
• ISSN: 0168-1923
• DOI: 10.1016/j.agrformet.2025.110797

•Identification of individual drivers of damage to two species of oak native to Great Britain, following successive Storms Arwen and Barra in winter 2021.•Both Structural Equation Modelling (SEM) and Random Forest identified individual tree health and structural characteristics as key drivers of damage.•Higher incidence of structural defects and faster growth rates increased vulnerability to storm damage; disease symptoms were important at broader scales.•Individual tree-level traits, rather than topographic or site-scale factors, primarily explained patterns of storm damage in oak landscapes. There has been an increase in windstorm disturbance events in European forests over the past ∼50 years, exacerbated by anthropogenic climate change. In this study, we examined the factors influencing storm damage to oak tree species native to Great Britain (Quercus robur and Quercus petraea) following two successive and severe Storms, Arwen and Barra, in the UK in winter 2021. A combination of novel data collection methods, dendrochronology and remote sensing, and data analysis models, Structural Equation Modelling (SEM) and Random Forest, are used to evaluate storm impacts at both individual tree and site-wide scales. This research directly compares a well-established but data-driven machine-learning method, Random Forest, with a novel, untested approach for wind risk modelling, SEM, to determine whether SEM is a viable method for identifying predictors of storm damage. SEM is a hypothesis-driven method which allows testing of relationships based on prior ecological knowledge. This analysis investigates whether pre-existing health conditions, such as disease and structural defects, influence an oak tree’s vulnerability to storm damage. Both models indicated that individual tree characteristics, notably structural defects and growth rate, were stronger predictors of storm damage than topographic factors. Disease symptoms were also important at the site-wide scale. SEM enabled identification of indirect pathways - for example, showing that disease symptoms were associated with reduced growth, which in turn increased susceptibility to damage - relationships not detected in Random Forest outputs. Furthermore, oak trees with faster growth rates were more prone to storm impacts; across all sites, cumulative growth rates (2000–2015) of storm-damaged trees were 22.8% greater than those of undamaged trees. Our findings contribute to the understanding of wind risk, aiding the parameterisation of risk models for oak, whilst also providing site managers with insights to support conservation efforts. Identifying drivers of damage is crucial given increasing climate-induced storm risk.