EM 1110-2-2907
1 October 2003
Broadleaf marsh.
Miscellaneous wetland vegetation.
Upland forest.
Upland herbaceous.
Upland shrub.
Wetland forest.
Wetland shrub.
Wet prairie.
Vines
(2) Geological constrains did not aid in the identification of the vegetation classes.
Geologic constrains tend to be more useful in mapping plant communities in areas with a
more mature ecosystem or were there is significant variation in the substrate or soil.
Choosing a sensor capable of delineating healthy vegetation versus stressed vegetation was
another consideration that needed to be addressed by the researchers. This would allow land
use managers the opportunity to closely monitor the decline and rise of various species
throughout the duration of the wetland restoration.
c. Field Work.
(1) Airborne hyperspectral data were collected in conjunction with 146 ground-truth
data points (also known as training sites); this collection was made on-foot and by airboat.
Fieldwork was done and data collected during a flood by a botanist and a GIS specialist. In
the field, SAIC's hand held spectrometer was used to collect the spectral data associated
with mixed plant communities from within the Kissimmee River floodplain. These ground-
control points were then used to test the accuracy of the vegetation map developed from the
hyperspectral data.
(2) Problems arose using the plant classes defined by the 1996 field study. Classes
were subsequently altered to better suit the dechannelized ecology. A supervised classifica-
tion was applied to the data and two vegetation maps were produced denoting 68 vegetation
communities and 12 plant habitat types (Figure 5-25). The hyperspectral map was then
compared to the existing vegetation map produced in 1996.
d. Hyperspectral Sensor Selection. Researchers on this project had the opportunity to
choose between AVIRIS and HyMap. HyMap was eventually chosen for its accuracy,
spectral capabilities, and reasonable expense. HyMap, a hyperspectral sensor (HSI), was
placed on board a HyVista aircraft. HyMap maintains 126 bands across the 15- to 20-nm
range. The error in HyMap data was found to be at 3 m, equivalent to the accuracy of the
on board GPS unit. To learn more about HyMap and HyVista view
(1) For this project, the hyperspectral (HSI) data maintained clear advantages over
other sensor data. HSI's high spectral resolution allows for the distinction of spectrally
similar vegetation and had the potential to monitor vegetation health status. The shortwave
infrared (SWIR) wavelengths where found to be most sensitive to the non-photosynethic
6-3
Previous Page
