90
ERDC TR-05-1
The images were acquired near low tide during a leaf-off season. It is rec-
ommended that follow-on work acquire remotely sensed data during the growing
season and near high tide conditions. Furthermore, the smaller reference wet-
lands, while easily accessible from the ground, may not offer the variation in sur-
face features seen in the much larger Hell Hook Marsh. It is also recommended
that future remote sensing applications center on Hell Hook Marsh and other
large, diverse estuarine wetlands. Ground-truth data will be required to complete
a quantitative image classification and an objective land cover class map accu-
racy assessment.
9.2
Enhanced Levee Inspections
This application of high-resolution imagery to enhance levee inspections
showed that 1-m multispectral imagery was of only marginal utility. The Dis-
trict's evaluation clearly showed a need for even smaller pixels. It is recom-
mended that a single-CCD digital aerial camera system be employed to acquire
data with very high spatial resolution (e.g., 0.25 m). The estimated costs to re-
ceive a complete mosaic covering the 45 km (28 miles) of the Sny levee, includ-
ing an acceptable buffer region of 500 m on each side of the structure, would be
between ,000 and ,000.
9.3
Invasive Species Mapping at Lake Okeechobee, Florida
This project expended a significant amount of planning and execution to ac-
quire the airborne multispectral imagery. The satellite imagery offered a much
more cost effective data source for mapping the relatively large study area. The
District was periodically updated throughout the project. Their evaluation of the
thematic maps suggested that remotely sensed imagery could potentially offer a
cost effective source for monitoring hyacinth and lettuce distributions across all
of Lake Okeechobee. A rough estimate of the cost of acquiring 4-m IKONOS
imagery over the entire Lake was well over 0,000. A less expensive image
source was identified as the SPOT sensor. With 10-m multispectral and 2.5-m
panchromatic imagery, a pan-sharpened four-band product would cost between
,000 and ,000. The price range assumes that some post-processing will be
required to radiometrically balance (or equalize) the several SPOT scenes re-
quired to cover all of Lake Okeechobee. Image classification costs, including the
collection of the required minimum amount of field data, would be reduced by
employing an unsupervised clustering routine to identify hyacinth and lettuce
pixels. Field data collection could then be limited to verification of areas mapped
as hyacinth and lettuce. Depending on the available budget, SPOT imagery could
be acquired and processed seasonally (i.e., three to fours times per year), provid-