A cross-European assessment on the pre-emergence detection of trees attacked by spruce bark beetle using UAV imagery

The European spruce bark beetle (Ips typographus) has driven unprecedented forest losses in Europe. Detecting infested trees before brood emergence is crucial for control. We present the first multi-country benchmark of pre- emergence detection using drone remote sensing. Data span Sweden, Finland, Italy, and Czechia (12 areas; 6 time series), with 26 multispectral and 16 hyperspectral image sets, covering 1798 infested and 13,020 healthy trees. We compare detection performance across remote sensing factors and tree-decline dynamics. Hyperspectral sensors performed best due to access to green-shoulder bands (~530 nm): the Green Shoulder Curvature Ratio (GSCRms) index detected up to twice as many infestations as the next-best spectral index (VI). Without these bands, hyperspectral and multispectral sensors were comparable using narrow- or broadband VIs. Within the red-edge, the 705–712 nm bands were 2–4 times more sensitive to stress than the 730–740 nm bands, guiding the choice of multispectral sensors. Normalizing red-edge VIs to pre-attack values markedly improved early detection and reduced false positives; green-shoulder VI was inherently stable and required no normalization. Outbreak timing was broadly similar among countries, but the decline progressed faster for trees attacked at outbreak peak. Decline was highly localized: adjacent plots often differed largely in decline rate (e.g., 50% vs. 17%). Localized decline rate dominated detectability, followed by outbreak phase, green-shoulder-band availability, red-edge wavelength, and VI normalization.