Assessment of physiological status and spatial distribution of emergent macrophytes based on imaging spectroscopy

Wetlands are ecosystems encountered at the land-water ecotone and hence inheriting rich complexity and biodiversity. Emergent macrophytes are a prime example of this variability manifested by the co-occurrence of vegetation associations at a very fine spatial level. From 1960s onwards an abrupt deterioration of reed beds in Europe has been systematically observed and denoted as the ‘reed die-back’. Since then, earth observation has been utilized mainly to map the extent of reed beds based on multispectral information. Hyperspectral remote sensing has frequently been employed in vegetation related studies, however the spectral information content of macrophytes has not been widely investigated. This study focuses on assessing the potential of imaging spectroscopy for assessing the ecophysiology of lake shore vegetation at leaf level and mapping macrophytes species associations from airborne imagery. Concurrently acquired spectroscopic, chlorophyll fluorescence and chlorophyll content information from field samples around Lake Balaton, Hungary are employed to identify spectral regions and propose narrowband indices which can aid the identification of reed ecophysiological status based solely on vegetation spectral characteristics. Macrophyte species as well as Phragmites in different phenological states have concretely separate spectral responses, however stable and die-back reed are not crucially different. Bathymetry regulates consistently the spectral response of Phragmites. Narrow band ratio 493/478 (0.65***) correlates with the ETR, the latter being an indication of the photosynthetic activity of the plant, and hence the vegetation physiological status. Most indices correlating with fluorometric parameters are located in the optical domain. Based on R2 graphs, the ratios 699/527 and 612/516 can be used to estimate Y(II) from AISA hyperspectral data. Estimation of the photophysiological parameters of a reed bed is possible based solely on airborne hyperspectral imagery. Simultaneously acquired airborne AISA Eagle, Hawk and discrete return LiDAR data are combined in order to stress out the potential of each dataset in classifying the reed bed in terms of species associations. An application of SVM on noise-reduced Eagle data, at the chlorophyll and near infrared absorption spectral regions, provides the most concrete results in terms of overall accuracy (89%). SVM outperforms ML and infrared sensors as well as LiDAR data do not improve the categorization of macrophyte classes. While airborne data inherit a superior spectral and spatial resolution, they are difficult to acquire in an operational context. Upcoming satellites will provide imagery with progressively higher spatial and spectral capabilities. A simulation of Sentinel-2 image over a reed bed in a nature protected area indicates the potential of satellite imagery in mapping macrophytes. Main classes can be distinguished, despite the fact that inter-class separability is becoming vague. Given the very large swath of Sentinel-2 (290km) an operational categorization of main macrophytes is foreseen achievable.