We selected 8 of the 10 reference sites from our previous study (Sluis and Tandarich,
2000) of shallow soil prairie ecosystems in the Midewin region for this experimental
flight (Figure 3). Over
all the sites there were 320 globally positioned (GPS) ground reference points
where the vegetation communities, soils and water tables were characterized.
Summary tables are given here for the soils mapped (Table
1), vegetation communities produced by TWINSPAN ordination (Table
2) and the edaphic analogues (soil and vegetation community correspondences
identified (Table 3)
in the ground reference study of Sluis and Tandarich (2000).
Prior to the flight, 2 m square Tyvek™ targets were placed at corner points
around the sites to frame them and GPS located. The target points served to
enable georeferencing of the sites to be done. The target and ground reference
points were mapped in geographic information system (GIS) coverage using ArcView™.
These ground reference points were analysis keys to the spectral responses on
the HSS imagery.
We chose a flight window of July 1 to July 15, when both Deschampsia caespitosia
(wet shallow-soil prairie indicator species) and Bouteloua curtipendula (dry
shallow-soil prairie indicator species) were flowering. We expect that the characterization
of their spectral responses and ground patterns will be most important for our
study. It will improve our understanding of these shallow soil ecosystems, as
well as deeper soil ecosystems, by accurately mapping their relative locations.
The methods investigated here may also be readily applied to other ecosystems
found in the Chicago region and beyond.
The flight was to be done by ITD -- Spectral Visions, a NASA subcontractor.
In consultation with ITD, four north-south flight lines were planned that would
cover the desired sites; this flight line configuration would render seven images
(Figure 4). The flight
took place on July 16. The sensor was the ITD Spectral Visions sensor of 120
channels.
Subsequent to the flight, ground reference data was collected at all points.
Percent cover of canopy species was measured. Water tables were measured. The
resulting data was compiled in a spreadsheet (Microsoft Excel™) and transformed
into a GIS ArcView database.
ITD Spectral Visions preprocessed ("shift-corrected’) four of the
seven HSS images to remove some distortion. However the shift-corrected images
needed further geometric correction (rectification) prior to analysis to the
UTM map projection.
Image georectification was accomplished using ERDAS Imagine™ software.
Both rubber sheeting and polynomial correction routines were used. The polynomial
correction appeared to provide a more satisfactory result. UTM coordinates were
derived from the GPS target points and from selected control points found on
US Geological Survey orthophotoquads of the area. The targets were located precisely
on the ITD imagery using MultiSpec™ software (Figure
5) by examination of the pixel spectral radiance values in the target area.
The highest radiance disclosed the target location. The resulting georectification
had the effect of "stretching" some pixels (Figure
6), but that is to be expected given the distortion in the original image.