Forecasting potential bark beetle outbreaks based on spruce forest vitality using hyperspectral remote-sensing techniques at different scales

• Published in: Forest ecology and management Vol. 308; pp. 76 - 89
• Main Authors: Lausch, A., Heurich, M., Gordalla, D., Dobner, H.-J., Gwillym-Margianto, S., Salbach, C.
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier B.V 15.11.2013; Elsevier
• Subjects: Animal and plant ecology; Animal, plant and microbial ecology; Attack; Bark beetle; Biological and medical sciences; Forestry; Fundamental and applied biological sciences. Psychology; Hyperspectral remote sensing; Ips typographus (L.); Spruce forest; Synecology; Terrestrial ecosystems; Vitality; Bark beetle; Attack; Spruce forest; Ips typographus (L.); Vitality; Hyperspectral remote sensing; Forests; Coleoptera; Multispectral remote sensing; Insecta; Potential; Hyperspectral imaging sensor; Pest; Outbreak; Softwood forest tree; Gymnospermae; Ips typographus; Picea; Hyperspectral imagery; Measurement sensor; Forest ecology; Forecasting; Arthropoda; Forestry; Coniferales; Spermatophyta; Curculionidae; Technique; Invertebrata
• ISSN: 0378-1127; 1872-7042
• DOI: 10.1016/j.foreco.2013.07.043

•The study aimed to predict outbreak potential of bark beetle based on detection different vitality stages of spruce.•Hyperspectral data are able to detect changes in biochemical–biophysical vegetation characteristics in spruce forest.•Important spectral information for derivation vitality status of spruce are 450–890nm.•We found a classification accuracy of 64.04% between C3 – infestation 2010 and C5 – healthy.•Hyperspectral data with grain of 4m contain relevant information to estimate differences in vitality of spruce than 7m. The bark beetle (Ips typographus L.) is known for the detrimental impact it can have on Europe’s mature spruce forests with bark beetle outbreaks already having devastated thousands of hectares of spruce forests in Germany. This study analysed the hypothesis that the vitality of spruce vegetation is already susceptible from factors such as climate change or emissions to a certain extent before infestation, so that the role of the subsequent bark beetle infestation is only secondary. Hyperspectral remote-sensing techniques were used to detect changes in biochemical–biophysical vegetation characteristics in the spruce forest of the Bavarian Forest National Park, Germany. For this study, several spectral bands, vegetation indices and specific spectral band combinations of hyperspectral HyMAP remote-sensing data with a 4m and a 7m ground resolution were analysed and compared in terms of their classification accuracy, generating an ID3 decision tree. The vitality classes and thus also the attack stages of the spruce vegetation could be estimated with moderate to good accuracy using hyperspectral remote-sensing data. Clear spectral differences between the class with spruce trees that were still green but with reduced vitality (possibly the first stages of green-attack) and the class with healthy spruce trees could be ascertained. The best spectral characteristics, spectral indicators and spectral derivatives related to vitality classes and thus attack stages were typically based on wavebands related to prominent chlorophyll absorption features in the VI within the spectral range of 450–890nm. Only limited spectral information and derivatives could be found in the short-wave infrared region 1 (SWIR) within the spectral range of 1400–1800nm, which reflects the water content of the spruce needles. The class of spruce trees that were still green but with reduced vitality (possibly the first stages of green-attack) showed a trend towards detectability and differentiation with spectral indicators and index derivatives. However, the prediction of observed effects with 64% accuracy as observed here is regarded as insufficient in forestry practises. Hyperspectral data with a ground resolution of 4m were found to contain more information relevant to estimating the vitality class of spruce vegetation compared to hyperspectral data with a ground resolution of 7m.