Retrieving Macrophyte Pigments From Spectral Reflectance

Plant traits play a key role in the delineation and quantitative description of vegetation functional features within and across ecosystems. Under a data-driven approach, biological and ecological models rely more and more on the availability, completeness and quality of plant traits used as their input. During the last decade, hyperspectral imaging sensors have demonstrated their potential for quantifying plant traits in terrestrial ecosystems, especially forests and grasslands. Despite their crucial biogeochemical interactions within aquatic ecosystems, and the important contribution to diversity and function of such environments, only few works have worked on using remote sensing for assessing quantitative traits for aquatic plants, or macrophytes. Leaf pigments are among the most studied traits using remote and proximal sensing-based spectroscopy, with a specific attention on chlorophylls (a and b), as their content can be linked to nitrogen (N) content and photosynthetic performance. The main aim of this work is to test and calibrate spectral reflectance-based models for estimating pigments content for floating and emergent macrophyte species common in temperate and continental regions, focusing on chlorophyll a (Ca) and b (Cb), their ratio (Ca/Cb), as well as total carotenoids (Cxc). The spectroscopy-based pigments models were implemented using high-resolution reflectance (at leaf and canopy scales) data and plant pigment content collected in situ. Best performing macrophyte pigments models were applied to hyperspectral imaging datasets acquired with the APEX sensor for mapping macrophyte pigments. APEX data were collected during airborne campaigns sponsored by EUFAR through Transnational Access action in connection with EU FP7 INFORM project, covering three shallow European water bodies in 2014 and 2016: Lake Hídvégi (Hungary), Mantua lakes system (Italy), and Nemunas Delta (Lithuania). Leaf Cxc concentration scores measured in our dataset were highly correlated with Ca concentration (R2 > 0.8), thus not permitting an independent estimation of the two parameters. Ca and Cb content were instead retrieved with good reliability (R2 > 0.6) using two-band and three-band spectral indices (SIs). The optimal spectral ranges were different for models at leaf and canopy scale. At leaf scale, best performing models were based on SIs including spectral reflectance around 495-508 nm and 690 nm for Ca, and around 485-489 nm and 661 nm for Cb. At canopy scale, best performing models included SIs exploiting spectral reflectance around 730 nm and 810 nm for Ca, and around 517-525 nm and 600 nm for Cb. The ratio of two chlorophylls (Ca/Cb) was retrieved with lower but still acceptable accuracy (R2 > 0.5), combining spectral reflectance in the ranges around 687-695 nm and 605 nm or 650 nm for leaf and canopy scale, respectively. Maps of macrophyte Ca content, Cb content and their ratio were derived from APEX data for the three study sites, at canopy scale (3-5 m pixel size). These maps were used for assessing the variability of macrophyte pigments content at species and ecosystem levels, taking into account both within-site and across-site variability factors, thus providing a first high-resolution spatialized insight into aquatic vegetation functional traits assessment.