Early detection of changes in health status of Norway spruce using hyperspectral data.

Abstract

Natural disturbances play a key role in forest ecosystems. However, they potentially cause severe economic damage to forest owners and may have an inappropriate ecological impact on the environment. In Central Europe, beetle infestations in Norway spruce (Picea abies, L.) forests increased step by step over the past two decades with some dramatically increase in consequence of dry years (e.g. years 2003/2004). Considering the climate change, it’s necessary to establish early detection and early warning systems with suitable automated monitoring methods. Insofar remote sensing techniques are considered effective for early detection of potentially infested areas. Numerous studies point out the capacity of remote sensing for detecting the (declining) health status of trees. While advanced stages of damage (so called "red-attack" and "grey-attack") can be easily detected, the early detection of an infestation (referred to as "green-attack") is still problematic. For this reason, research to detect infestation before visible discoloration (“green attackâ€) with help of remote sensing has to go on. In our study we analysed the suitability of very high spatial resolution multi- to hyperspectral remote sensing data for mapping spruce vitality in Central Europe. The test site in Bavaria (Germany) consists of two Norway spruce stands (60 and 100 years old). In each of the two stands, 70 trees were artificially weakened by girdling (done in May 2013). A 20 cm wide stripe of bark was circularly removed from the trunk. At the same time, unstressed control trees were identified. For monitoring of the trees, we used mainly airborne hyperspectral data with ground resolution between 0.5 m (VNIR) and 2.0 m (SWIR). The data were acquired during the vegetation period of 2013 and data collection continued throughout 2014. Additionally, WorldView-2 satellite data with 2.0 m spatial resolution and eight spectral bands from five acquisition times (2013 and 2014) were acquired. Object based methods were used to analyse individual tree crowns. The experiment was complemented by field based inspections of crown status as well as needle sample collection. The needles were measured with a field spectrometer and analysed chemically. With this unique multi-temporal data set we were able to locate the most sensitive wavelength regions for this phenomenon and to detect the extent where girdled trees still appear green but changes in reflectance spectra indicate a declining vitality. In autumn 2013 the test area was extended by another test site in Austria where the analysis methods were adopted.