Exploring the potential to use in-between pixel variability for early detection of bark beetle attacked trees

Abstract. The European spruce bark beetle (Ips typographus L.) is a major disturbance agent in Norway spruce (Picea abies (L.) Karst) forests in Europe and it is estimated that a changing climate will result in more severe outbreaks in the future. To reduce the risk of large outbreaks it is important to have methods that enable early detection of bark beetle attacks to help forest managers to prevent population build-up, e.g by sanitary cutting. Several studies have been devoted to early detection of bark beetle attacks with Sentinel-2 data with a focus on spectral properties and vegetation indices for early detection with pixel-based methods. In this study we explore the potential to use changes in variability between pixels in windows of different sizes (3×3, 4×4 and 5×5 pixels). We compute the coefficient of variation for four vegetation indices (NDVI, NDWI, CCI and NDRS) in a time-series of Sentinel-2 data during a bark beetle outbreak in Sweden that was triggered by a drought in 2018. The results indicate that CCI is the most promising index for early detection and that the variability between pixels increase in windows with attacked trees from late July when the main swarming was the second week of May.