Identification of Submerged Aquatic Vegetation using Simulated Data from Sentinel-3
Watanabe, Fernanda Sayuri1; Imai, Nilton Nobuhiro1; Alcântara, Enner Herenio1; Faria Barbosa, Cláudio Clemente2; Rotta, Luiz Henrique3
1UNESP, BRAZIL; 2INPE, BRAZIL; 3UNESO, BRAZIL
There is a great expectation for the launch of Sentinel-3 by the European Space Agency (ESA), scheduled for 2014, as well as the possible applications of its data. Sentinel-3 presents an advance related to the number of bands, particularly in the region of the visible spectrum. Moreover, the penetration of the radiation into the water column is greater on that spectral region, thus it is expected that Sentinel-3 images enable the mapping of submerged targets. Submerged Aquatic Vegetation (SAV) stands out among the submerged targets due to its important role in the balance of aquatic ecosystems. However, anthropogenic changes on these environments are causing problems of high colonization by these plants. Therefore, the mapping of SAV is essential for deployment of management plans for weed control. This article aims to evaluate, on the spectral point of view, the performance of simulated data of Sentinel-3 in the identification of SAV. The field data were collected in stream 'Ferreira', located in the Nova Avanhandava reservoir, lower course of the Tietê River (Sao Paulo State, Brazil). The field survey was carried out from 27th to 29th September 2011. We collected 20 samples of hemispherical-conical reflectance factor (HCRF), using a ASD FielSpec ® portable spectroradiometer, model HandHeld UV / NIR, adopting a field of view of 25°. In addition to the radiometric measurements, data about height of the water column over the vegetation canopy and turbidity were collected. HCRF samples include points with and without SAV, varying the depth (1-20 m) and height of the water column over the vegetation canopy (0.01 to 3 m). Using HCRF data was simulated the reflectance factor of 21 channels planned for the Sentinel-3 sensor. From the simulated data was applied a supervised classification using Spectral Angle Mapper (SAM). Also different band index were tested to distinguish station with and without SAV. From the results obtained using simulated data of Sentinel-3 is expected that the adopted processing and sensor spectral data are suitable for mapping of SAV. However, further research must be made, since the sensor is designed to acquire images with spatial resolution of 300 m. And even using modeling more robust, it may not be possible to detect the signal of the vegetation. The signal-to-noise-ratio (SNR) is another issue to be investigated due to the low signal of submerged targets.