Characterizing forest disturbance and recovery with thermal trajectories derived from Landsat time series data

The increasing frequency of forest disturbances caused by climate change has highlighted the importance of understanding the entire process of disturbance, from its onset to forest recovery. Many previous studies have used multispectral Landsat time-series data describing forest dynamics. However, there is a lack of studies using thermal imagery, which may provide information about bio-climatic changes and energy balance during the forest disturbance and recovery. Our objective in this research was to detect the main features of insect forest disturbance and subsequent forest recovery (disturbance duration, disturbance severity, recovery duration) using thermal Landsat imagery. We also aimed to determine the relationship between the identified thermal features and topography. The study area was a Norway spruce (Picea abies [L.] Karst.) dominated forest located in a Central European border region between the Czech Republic (Sumava Mountains) and Germany (Bavarian Forest). For more than three decades, forests in this area have experienced widespread defoliation due to the spruce bark beetle (Ips typographus L.). We determined the forest's surface temperature (ST) between 1985 and 2015 from thermal Landsat time-series data and then normalized it (STn) to be comparable across the years in question. We plotted thermal trajectories for this period and then clustered them based on their variances. The resulting thermal trajectories well described both disturbance and recovery, showing a rise in the STn during the onset of the disturbance, the severity, and a decrease during recovery. The mean disturbance duration was 8.44 years (SD = +/- 3.775), with a maximum severity as indicated by the STn of 4.92 (SD = +/- 1.5) and a duration of 12.84 years (SD = +/- 3.38) until 50% recovery and 17.85 years (SD = +/- 1.74) until 80% recovery. Clustering of the thermal trajectories revealed clusters with similar year-groups of bark beetle attack and a spatially aggregated pattern closely related to topography. Severity showed a positive correlation with altitude, whereas the spatially aggregated patterns of disturbance and recovery duration can be attributed to more complex topographical characteristics. Our study demonstrates the ability of thermal infrared imagery to provide highly relevant data for assessing not only the main features of forest disturbance and recovery, but also bio-climatic canopy functions in terms of topography and other environmental variables.