Hyperspectral remote sensing protocol development for submerged aquatic vegetation in shallow waters

Abstract

Submerged aquatic vegetation (SAV) is an important indicator of freshwater and marine water quality in almost all shallow water aquatic environments. Throughout the world the diversity of submerged aquatic vegetation appears to be in decline, although sufficient historical data, of sufficient quantitative quality is lacking. Hyperspectral remote sensing technology, available from low altitude aircraft sensors, may provide a basis to improve upon existing photographic regional assessments and monitoring concerned with the aerial extent and coverage of SAV. In addition, modern low altitude remote sensing may also help in the development of environmental satellite requirements for future satellite payloads. This paper documents several important spectral reflectance signature features which may be useful in developing a protocol for remote sensing of SAV, and which is transferable to other shallow water aquatic habitats around the world. Specifically, we show that the shape or curvature of the spectral reflectance absorption feature centered near the chlorophyll absorption region of ˜ 675 nm is strongly influenced not only by the relative backscatter region between 530-560 nm, but by a "submerged vegetation red edge" that appears in the 695 to 700 nm region in extremely high density vegetative areas in very shallow waters (= 0.5m depth). This "aquatic biomass red edge" is also observable in deeper waters where there is a shallow subsurface algal boom as demonstrated in this paper. Use of this submerged aquatic red edge feature will become an important component of SAV remote sensing in shallow aquatic habitats, as well as in phytoplankton-related water quality remote sensing applications of surface phytoplankton blooms .

Collecting in situ remote sensing reflectances of submersed macrophytes to build up a spectral library for lake monitoring

“…Within a pre-processing chain the plant signal must be separated form external influences. Above the water body the main attenuation sources are cloud coverage, illumination and viewing angles of the system and roughness of water surface [Mertes et al, 1993;Bostater et al, 2004;Morel and Belanger, 2006]. Below the water surface, effects of the overall water column [Mumby et al, 1998;Silva et al, 2008] have to be considered, as water itself or suspended and dissolved materials affect radiative transfer [Mobley, 1994].…”

Implementation of a ground truth process for development of a submerged aquatic vegetation (SAV) mapping protocol using hyperspectral imagery

“…Absorption features are present in all spectra at about 550 nm and 605 nm indicating a potential response to the phycocyanin and phycoerythrin pigments in the drift algae and epihytes. 9 Reflectance in the 710 nm region increases consistently with increasing SAV density and reduce depths to canopy. …”

“…6,8 A basic principle in remote sensing is that features of interest reflect or emit light in uniquely different ways and these differences can be detected and recorded by the remote sensing system. 9,10,11 Our experience has shown that one of the major issues hindering the scientific utility and application of hyperspectral data to resource mapping is the uncertainty or error that exists when attempting to link ground truth data with spectra extracted from pixels in the image cube. Hyperspectral images collected from aircraft using line scanning systems such as PHILLS or AISA often contain significant spatial error as a result of aircraft pitch, roll and yaw making direct comparisons between ground truth data and individual pixels difficult or even impossible.…”

Plant pigment types, distributions, and influences on shallow water submerged aquatic vegetation mapping

“…Absorption by these three pigments is generally considered the major factor producing the spectral shape common to vascular plants (see Figure 5a). 15 Image processing and reflectance spectra extraction were conducted with ENVI. Average reflectance values were generated for six target areas, three over sand and 3 over SAV.…”