Mapping and Assessing Seagrass Along the Western Coast of Florida Using Landsat TM and EO-1 ALI/Hyperion Imagery
Document Type
Article
Publication Date
12-10-2012
Keywords
image classification, water depth correction, submerged aquatic vegetation, leaf area index, remote sensing, seagrass
Digital Object Identifier (DOI)
https://doi.org/10.1016/j.ecss.2012.09.006
Abstract
Seagrass habitats provide a variety of ecosystem functions thus monitoring of seagrass habitat is a priority of coastal management. Remote sensing techniques can provide spatial and temporal information about seagrass habitats. Given the availability and accessibility of Landsat-5 Thematic Mapper (TM) and the advanced nature of Earth Observing-1 Advanced Land Imager (ALI) and Hyperion (HYP), we compared the capability of the three 30 m resolution satellite sensors and tested regression models based on two seagrass metrics [percent cover of submerged aquatic vegetation (%SAV) and leaf area index (LAI)] for mapping and assessing seagrass habitats within a shallow coastal area along the central western coast of FL, USA. We also evaluated a water depth correction approach to create water depth-invariant bands calculated from the three sensors' data. Then a maximum likelihood classifier was used to classify the %SAV cover into two classification schemes (3-class and 5-class). Based upon the two seagrass metrics measured in the field, six multiple regression models were developed and %SAV and LAI were estimated with spectral variables derived from the three sensors to assess the seagrass habitats in mapped units. Our results indicate that the HYP sensor produced the best seagrass cover maps in the two classification schemes: 3-class [overall accuracy (OA) = 95.9%] and 5-class (OA = 78.4%) and the best %SAV and LAI estimation models [R2 = 0.78 and 0.59, and cross-validation (CV) = 18.1% and 1.40 for %SAV and LAI, respectively] for assessing seagrass habitats. These results are likely due to the many narrow bands in the visible spectral range and rich subtle spectral information available in the HYP hyperspectral data. ALI outperformed TM (OA = 94.6% vs. 92.5% for the 3-class scheme, and OA = 77.8% vs. 66.0% for the 5-class scheme) for mapping %SAV likely due to its higher radiometric resolution. Our findings also demonstrate that the water depth correction approach was effective in mapping the detailed seagrass habitats with the data from the three sensors. The protocol developed and utilized here represents a new contribution to the existing set of tools used by researchers for documenting the amount of seagrass and which can guide future studies.
Was this content written or created while at USF?
Yes
Citation / Publisher Attribution
Estuarine, Coastal and Shelf Science, v. 115, p. 234-245
Scholar Commons Citation
Pu, Ruiliang; Bell, Susan; Meyer, Cynthia; Baggett, Lesley; and Zhao, Yongchao, "Mapping and Assessing Seagrass Along the Western Coast of Florida Using Landsat TM and EO-1 ALI/Hyperion Imagery" (2012). School of Geosciences Faculty and Staff Publications. 351.
https://digitalcommons.usf.edu/geo_facpub/351
