Florida Lake Management Society Annual Conference, Naples, Florida, June 4 – 7, 2007
USING HYPERSPECTRAL IMAGERY TO DETECT SUBMERGED AQUATIC VEGETATION IN THE ST. JOHNS RIVER
Courtney Hart Idea Integration Jacksonville, Florida Dean Dobberfuhl St. Johns River Water Management District Palatka, FL Jonathan Jordan UF, Agricultural and Biological Engineering Dept. Gainesville, FL
Airborne hyperspectral imagery identify submerged aquatic vegetation
worldwide to successfully detect and study describes the image-processing
methodology that was developed for 2003 and 2006 airborne hyperspectral imagery to the extent of SAV in the lower basin of the St. Johns River. Results from the 2003 preliminary results from the 2006 imagery will be discussed.
determine study and
The St. Johns River is a 500-km low-gradient, black water river in northeast Florida. The Lower St. Johns River basin (LSJRB) encompasses the final 170 km of the river and is subject to tidal exchange and reverse flow events. In the lower basin, the St. Johns is wide (ca. 3-5 km) and shallow (< 2 m), and there is a broad littoral shelf frequently supporting meadows of SAV. Imagery collected in both 2003 and 2006 covered the shoreline of the St. Johns River from downtown Jacksonville to south Palatka.
Five hyperspectral images and their associated ground-truth data, along with reflectance data collected in the field, were used in the creation of a spectral library that was applied to the full collection of hyperspectral images acquired in 2003. This approach was utilized as a basin- wide SAV mapping tool that successfully mapped SAV. Classification results were comparable to SAV estimates made from independently collected transect data. Results from the 2003 data analysis provided guidance for selecting finer spatial resolution (pixel size) and more specific spectral resolution (narrower bandwidths) for the 2006 data collection. Hyperspectral data were again acquired in the spring of 2006 with the goal of not only identifying SAV, but potentially distinguishing specific SAV species. Substantially more groundtruthing was conducted for the 2006 imagery and produced a more accurate and diverse spectral library. Over 80 ground truthing plots were identified and delineated in the field using GPS to serve as training polygons for the subsequent image analysis. Groundtruthing was performed within two weeks of the image acquisition. The improved spectral library was applied to the 2006 images that coincided with the areas used in the creation of the 2003 spectral library.
Session 7A – Page 2