Identification of drought risk zones in current beech stands

Based on Sentinel-2 satellite images, vegetation indices are derived that provide information about the vitality of the trees. We have calculated and compared these indices for 2018 as well as for the previous years 2016 and 2017. If there are large differences to the previous years, this indicates premature discoloration of the trees in 2018. We have validated this classification with on-site observations, so-called ground truth data (Figure 1). Using the classified Sentinel-2 pixels (beech with and without drought symptoms) and machine learning algorithms, we conducted an analysis to identify relevant site factors that have led to drought symptoms and to predict areas with an increased risk of drought.

Through the comparison with ground observation data, we were able to show that the vegetation indices (e.g. Normalized Difference Vegetation Index (NDVI) and the Normalized Difference Water Index (NDWI)) can reliably detect prematurely discoloured beech trees (Figure 2). Threshold values derived from vegetation indices can then be used for mapping beech trees with drought symptoms (Figure 3).