Mapping peatland species fractions and condition using multi-temporal spaceborne imaging spectroscopy data

Rewetting drained peatlands is crucial in mitigating climate change, but special monitoring schemes are required to track rewetting effects. Species distribution and condition may give indications on water level and hence rewetting success. We mapped the fractions of seven peatland related and four non-peatland vegetation classes, as well as their dynamics in condition from five 30 m EnMAP images. We used neural network regression with synthetic training data augmentation to predict peatland vegetation fractions (e.g., Phragmites australis , Typha spp. and Carex spp.) from the hyperspectral EnMAP stack, as well as NPV (non-photosynthetically active vegetation), GV (green vegetation), and water fractions from the single-dates to derive intra-annual dynamics. Patterns of species fractions matched field observations at varying accuracies with the highest agreement for Typha and Phragmites (MAE = 16% and 22%, respectively). The changes in vegetation condition similarly matched field observations (GV: MAE = 12%, Water: MAE = 14%, NPV = 21%), and in combination with species fractions revealed class-specific trends, such as differences in timing of the green peak (maximum GV fractions). Intraspecific variations in green peak could be observed for the class Phragmites, indicating differences in water level. Further, dynamics in the extent of open water areas were apparent, highlighting the necessity of multi-temporal information. Increasing availability of spaceborne imaging spectroscopy will enable long-term monitoring of peatland rewetting, accounting for intra-annual variations in vegetation and water dynamics.