High Spatial Resolution Digital Imagery
87
by the CAMIS frames, certainly would require an extensive network of field
plots, distributed over the entire 35,000 hectares, to produce a reliable thematic
product. However, remembering that the District is primarily interested in map-
ping the distribution of only two specific species, the number of field observa-
tions could be significantly reduced.
Perhaps as a general "rule-of-thumb," this investigation at Lake Okeechobee
suggests that the project budget for creating a reliable vegetation map from high
resolution multispectral imagery should include roughly equal dollars for image
acquisition and field data collection. Reliability is quantified for this remote
sensing application as an overall thematic accuracy of approximately 80%. Image
post-processing (if required) and image analysis costs, including supervised or
unsupervised classification, should be included as separate items.
8.4
Wetland Vegetation Mapping Over Blackwater Wildlife Refuge
The costs associated with this project started with a reasonable sum for image
acquisition. The vendor, 3Di, LLC captured and post-processed over 9 GB of
hyperspectral images, covering ~ 30,000 acres of Refuge property, for only
,000. However, as described above, a significant amount of additional post-
processing was required to create a marginally useful mosaic. The total additional
cost for developing this mosaic and producing the final class map was in excess
of 0,000. The contactor should be responsible for these tasks and should be
able to provide a detailed, and understandable, description of all acquisition and
post-processing techniques and algorithms as part of the required metadata
documentation.
A comprehensive set of ground truth sample sites would likely have cost an-
other ,000 to ,000. When compared with traditional aerial photography,
the digital imagery is initially cheaper to acquire and process. Furthermore, the
digital imagery will likely have a much quicker delivery as compared to analogue
photography. However, the questionable accuracy of the thematic map suggests
that for projects where very detailed vegetation classification is required (i.e.,
species alliances or associations), large scale true-color or false-color photogra-
phy, combined with intensive field data collection, may be the most cost effective
alternative. Digital aerial mapping camera technologies are also maturing rapidly
and may represent an alternative to hardcopy photography. While not specifically
designed or suited for computer-aided classification, very high-resolution digital
photos can provide an excellent source of interpretable images with very good
geometry (i.e., spatial accuracy) and with a very rapid turn-around time.
Future attempts to develop a useable base map for an area as large as the
Blackwater Wildlife Refuge should employ either medium or high-resolution
Previous Page
