Hyperspectral drone images indicate that green shoulder indices are robust in pre-emergence detection of spruce bark beetle infestation across spatial and temporal scales

Abstract. Bark beetle (Ips typographus L.) outbreaks are one of the main threats to forest health in northern Europe, with recent events causing extensive damage to spruce forests. While management for population control relies on detecting infested trees before the emergence of the filial generation, identifying robust spectral indicators remains a major challenge. In this study, we evaluate the performance of vegetation indices (VIs) derived from hyperspectral drone imagery for detecting bark beetle infestations in southern Sweden. We calculated detection rates based on the cases where VI values for infested trees deviated from the value range observed in healthy trees. We tested different scenarios for defining the range of healthy values to assess spatial and temporal consistency of VI performance. Green shoulder VIs, particularly GSCR1MS and GSCR2MS, consistently showed the highest detection rates. Their performance was stable across different weeks and forest stands, indicating stronger generalizability and higher potential for pre-emergence detection. In contrast, red edge VIs showed limited temporal consistency and strong dependence on normalization. SWIR-based VIs presented low detection rates in all scenarios, therefore showing limited potential.