Exploring the spatial heterogeneity of bark beetle infestation risk factors with geographically weighted regression and random forest

Abstract

Listen

Translate article

ABSTRACT This study investigated the spatial dynamics of bark beetle infestations and their relationships with risk factors in south-eastern Swedish forests during the years 2018–2020, highlighting the importance of spatial heterogeneity of risk factors, particularly under climate change. First, global Random Forest (RF) models were developed to examine the relationships between bark beetle infestations and risk factors during normal and drought periods. Second, spatial heterogeneity of the relationships was explored in local RF models by integrating Geographically Weighted Regression (GWR) with RF. The global RF models achieved accuracies of 0.89 and 0.84 for the normal and drought periods, respectively. In contrast, the local RF models performed better in many areas, capturing spatial variations in infestation drivers. Local models successfully identified the varying importance of risk factors, such as tree species composition, stand age, and local climate conditions, especially during drought. These findings underscore the necessity of spatially adaptive forest management strategies. The essential targeted interventions should consider local conditions, particularly during droughts, to mitigate infestation damage, instead of applying a “one-size-fits-all” strategy. The research highlights the need for monitoring and interventions at local scales, offering a more effective approach to managing bark beetle outbreaks in vulnerable forest ecosystems.