Sensitivity analysis of the Green Shoulder indices in pre-emergence detection of single trees attacked by European spruce bark beetle

Abstract

The application of hyperspectral cameras for forest health monitoring enables precise detection of stress-related changes in vegetation, such as those caused by spruce bark beetle infestations. In a previous study, Green Shoulder Indices (GSIs) were proposed, which exhibited high capacities to indicate trees with decreasing vitality caused by spruce bark beetle infestation using hyperspectral drone images. However, the detection accuracy of these indices may be influenced by the selection of various parameters. This study conducts a sensitivity analysis of the indices and aims to assess how the detection accuracy is impacted by different parameter choices. The detectability obtained from the GSIs was calculated and compared when (1) using the brightest or centremost pixels from the crown segments with different thresholds, (2) smoothing the spectral curves with different levels, and (3) using bands with varying bandwidths. The results showed that the GSIs were not sensitive to whether the brightest or centremost pixels were used for detection. Stronger smoothing caused the derivative peak at 545 nm to shift towards smaller wavelengths when a tree was under increasing stress, but the detectability obtained using GSIs did not decrease with stronger smoothing. The simplified GSIs using three wide spectral bands centred at 490 nm, 530 nm, and 550 nm (MS GSIs) slightly decreased the detection accuracy compared to narrowband MS GSIs, but the differences were minor, e.g. decreased from 0.86 to 0.80 using the index with the highest detectability (𝛿⁢𝐺𝑆𝐶𝑅⁢1𝑀𝑆). This study highlights the robustness of GSIs against tested factors and implies their potential for forest stress monitoring and damage control.

Keywords

European spruce bark beetle; early detection; remote sensing; hyperspectral imagery; drone imagery

Published in

International Journal of Remote Sensing
2025

SLU Authors

• Huo, Langning

  • Department of Forest Resource Management, Swedish University of Agricultural Sciences
    

Associated SLU-program

SLU Forest Damage Center

UKÄ Subject classification

Forest Science

Publication identifier

• DOI: https://doi.org/10.1080/01431161.2025.2482747

Permanent link to this page (URI)

https://res.slu.se/id/publ/141356