ABSTRACT
This chapter discusses hyperspectral data was compared with multispectral data to evaluate the benefits of hyperspectral imagery for wetland vegetation species identification and mapping in the Kissimmee River Floodplain. Dominant wetland vegetation types such as broadleaf marsh, wet prairie, and wetland shrub were replaced by dry pasture and other upland vegetation types. Similar to the land cover-level mapping in the last chapter, object-based image analysis and machine learning classifiers were combined to map vegetation species using the HyMap data. Training, testing, and classification includes a data export procedure from ArcGIS to ASCII format, an import of ASCII format into Waikato Environment for Knowledge Analysis (WEKA) software package, and an import of outputs from WEKA into ArcGIS for mapping. Multispectral sensors such as Ikonos and QuickBird could achieve accuracy similar to hyperspectral sensors for mapping saltwater marshes. Fine spatial resolution hyperspectral imagery has a large data volume, which is a challenge in data processing for a normal computation and processing facility.
