Early Detection of Forest Bark Beetle Attack Using Time Series Spatial Variability of Spectral Indexes from Sentinel-2

Bark beetle infestations pose a significant threat to forest ecosystems, necessitating timely intervention for effective management and conservation. This study focuses on an 8,100 km² area in southern Sweden, heavily impacted by a bark beetle outbreak triggered by a severe drought in 2018. Using spectral indices derived from Sentinel-2 data, it explores the potential of a window-based approach for early detection of forest bark beetle attack, considering spatial variability between adjacent pixels. Four spectral indices (NDVI, NDWI, CCI, NDRS) are analyzed using a time series approach, and coefficient of variation (CV) between pixels in a window is employed to capture changes in vegetation health. The Chlorophyll Carotenoid Index (CCI) emerges as the most sensitive indicator for early detection. Detection algorithms, including DBEST, MLS, and CUSUM, pinpoint June 2018 as the month of bark beetle attack identification after a main swarming in May 2018, demonstrating superior performance compared to pixel-based frameworks. The results highlight the efficacy of integrating spatial properties with spectral indices in a time series analysis for enhanced early detection of bark beetle infestations.