Mapping and Assessing Seagrass Abundance Using Landsat TM and EO-1 ALI/Hyperion Images

Seagrass habitats are characteristic features of shallow waters worldwide and provide a variety of ecosystem functions. Remote sensing techniques can provide spatial and temporal information about seagrass resources. In this study, we evaluate a protocol that utilizes image optimization algorithms followed by atmospheric and sunglint corrections to the three satellite sensors [Landsat 5 Thematic Mapper (TM), Earth Observing -1 (EO-1) Advanced Land Imager (ALI) and Hyperion (HYP)] and a fuzzy synthetic evaluation technique to map and assess seagrass abundance in Pinellas County, FL, USA. After image preprocessed with image optimization and atmospheric and sunglint correction approaches, the three sensors' data were used to classify the submerged aquatic vegetation cover (%SAV cover) into 5 classes with a maximum likelihood classifier. Based on three biological metrics [%SAV, leaf area index (LAI), and Biomass] measured from the field, nine multiple regression models were developed for estimating the three bio-metrics with spectral variables derived from the three sensors. Then, five membership maps were created with the three bio-metrics along with two environmental factors (water depth and distance-to-shoreline). And finally seagrass abundance maps were produced by using a fuzzy synthetic evaluation technique and membership maps. The experimental results indicate that the HYP sensor produced the best mapping results of 5-class classification of %SAV cover (overall accuracy = 87% and Kappa = 0.83 vs. 82% and 0.77 by ALI and 79% and 0.73 by TM) and better multiple regression models for estimating the three biological metrics (R2 = 0.66, 0.62 and 0.61 for %SAV, LAI and Biomass vs. 0.62, 0.61 and 0.55 by ALI and 0.58, 0.56 and 0.52 by TM) for creating seagress abundance maps along with two environmental factors. Our results also demonstrate that the image optimization algorithms and the fuzzy synthetic evaluation technique were effective in mapping the detailed seagrass habitats and assessing seagrass abundance with the 30-m resolution data collected by the three sensors.