Influence of canopy structure and illumination geometry on spectral anisotropy of aquatic vegetation in ultra-high resolution hyperspectral imagery

Advanced remote sensing technologies, including the use of unmanned aerial vehicles (UAVs) equipped with hyperspectral sensors, have revolutionized wetland monitoring providing ultra-high spatial and spectral resolutions data. However, aquatic vegetation canopy structure and substrate, as well as solar and viewing geometry, exert a substantial influence on the quantity of reflected light within the sensor's field of view, resulting into potentially biased reflectance measures; such bias could in turn affect spectral indices derived, as well as the estimation of biophysical and biochemical vegetation parameters. Nonetheless, few studies have focused so far on analysing vegetation anisotropy in aquatic environments. This study aims to quantitatively investigate radiometric variability of aquatic vegetation (at leaf and canopy scale) under different solar and viewing angular configurations, using data acquired from a hyperspectral push-broom camera mounted on a UAV. For this purpose, we collected hyperspectral imagery at centimetric resolution with varying solar elevation over a stand of Nuphar lutea, a floating-leaved aquatic plant, in Lake Pusiano (Italy), adjusting the UAV flight plans so that camera scan direction was either parallel to the solar principal plane (PP), orthogonal to it (OP), or set at approximately 45° to the PP (diagonal plane; DP), to cover a range of sun-target-sensor configurations. The acquisitions were carried out along one day, with four solar zenith angles (SZA), spanning from 45.57° to 71.80°. Within each image, Region of Interest (ROIs) were delineated to capture a portion of the adaxial leaf side (i.e. leaf-level ROIs) or a mixture of different plant parts and water background (canopy-level ROIs). ROIs were drawn as evenly distributed over the target stand to cover the range of across-track camera viewing zenith angles (VZA), spanning from -11.15° to +11.15°. Our findings revealed a strong wavelength dependency of floating aquatic vegetation anisotropy, with more pronounced effect in the visible (VIS, 400-700 nm) than in the near-infrared (NIR, 700-1000 nm) domain, especially within the Red (630-700 nm) and the Blue (450-500 nm) ranges. Specifically, angular configurations with the highest solar elevation (i.e. low SZA) and when scan direction approximately align with the PP, led to an increasing reflectance in the lateral zone of the image where VZA tends towards the sun (i.e. forward direction). This effect is mirrored by a decreasing reflectance in the backward direction with respect to the nadiral reflectance, both at leaf and canopy scale (even if attenuated in magnitude). Increment in reflectance in the forward view direction tends to decrease when approaching DP and OP, and with higher SZA. These results revealed a link between water, canopy structure (leaf orientation) and aquatic vegetation anisotropy, as the increasing reflectance in forward direction is possibly driven by the presence of water film or drops above floating and horizontal N. lutea leaves at leaf scale, and by water background pseudo specularity at canopy scale. To further explore how the influence of water and canopy structure could affect aquatic vegetation spectral anisotropy, similar analyses will be carried out on a stand of Phragmites australis, a tall, emergent (up to 3-4 meters above water) riparian plant species with growth form and spectral response characteristics very different from floating-leaved N. lutea.