EM 1110-2-2907
1 October 2003
late evening hours to reduce the effects of atmospheric haze. The digital hemispherical
photographs were later analyzed using a specialized forestry software, which measures both
LAI and canopy leaf structure. LAI calculations based on the computed hemispherical digi-
tal images compared favorably with the LAI measurements from the meter instrument.
d. Sensor System.
(1) Satellite data were collected with the use of Landsat MTI, Hyperion, and ALI
(Advanced Land Imager) on 25 August 2001. The ALI sensor has nine spectral bandwidths
plus a panchromatic band. Three bands where analyzed for this study 773.31 nm, 651.28
nm, and 508.91 nm. The forested areas appear bright red, urban areas are gray-blue, and the
water is depicted by the dark blue regions.
(2) The sensor radiance was converted with the use of 6S, an atmospheric corrections
model that converts sensor radiance to estimated surface reflectance. The differences and
consistencies in the two sensors were then easily compared with the spectral data collected
in the field. Then, a more detailed study of the forest site was made, using measured geo-
metric and optical parameters as input to the SAIL multi-layer canopy reflectance model.
The ETM+ and ALI data were then compared with the SAIL (Scattering by Arbitrarily In-
clined Leaf) reflectance model and the high resolution Hyperion, a hyperspectral imaging
instrument (see http://eo1.usgs.gov/instru/hyperion.asp for details).
e. Study Results. A comparison of the panchromatic ETM+ and ALI data show dramatic
differences. The ALI data provided better definition of the marina and pier area as well as
natural water features (urban and water targets). Relative to the ETM+ images the ALI data
maintained a reduced DN value for all forest pixels, increasing the contrast in the forest re-
gion. The authors suggested the higher resolution and the narrow bandwidths accounted for
the dramatic contrasts between the image data sets.
(1) Spectral plot comparisons of the multispectral bands for different ground targets
(grass, water, urban features, and forest) illustrating the relationship between reflectance and
wavelength indicated a close match between the two sensors. The spectral plots were cre-
ated by the selection of training pixels for each target group. ALI spectral values were closer
in value than those seen in the ETM+ data; again, this is a result of the narrow bandwidths
and higher resolution. The only notable difference in the spectral response between the two
sensors was evident in band 5 for grass and urban features. These targets had up to 20%
variation in signal response between the sensors. Specifically, the ALI band 5 with a reflec-
tance of 0.35 m is ~20% higher than the ETM+ value of 0.29 m.
(2) The combined spectral plot of data from ETM+, ALI, Hyperion, and the empiri-
cally derived SAIL show overall an excellent agreement. The three satellite data sets closely
match one another, with slightly different values recorded in the SAIL model data. SAIL
values best matched those of the sensors in the visible portion of the spectrum.
f. Conclusions. The authors of this study were able to establish a simple, multi-layer
canopy reflectance model using measured parameters from the site to compare the ETM+
and ALI spectra. Hyperspectral data were also compared against the satellite and ground
data. Additional work is needed to establish the relationship between leaf area index (LAI)
6-11
Previous Page
