|Abstract or Summary
- The relationships between spectral reflectance in the Landsat
Thematic Mapper (TM) bands and grass canopy variables were evaluated
using in situ remote sensing techniques. Reflectance data were
collected from experimental plots of annual ryegrass (Lolium multiflorum)
and tall fescue (Festuca arundinacea) using a Barnes Modular
Multiband Radiometer (MMR). The canopy variables used were canopy
height, canopy cover, total wet biomass, total dry biomass, aboveground
plant water, and leaf area index.
Statistically significant relationships were found between the
spectral bands and the canopy variables. Inverse relationships in
the visible (TM1, TM2, TM3) and middle infrared (TM5, TM7) regions
were related to spectral absorption by plant pigments (visible) and
moisture within plant tissue (middle infrared). Direct relationships
in the near infrared (TM4, MMR5) were attributed to enhanced reflectance
resulting from spectral scattering. Overall, no one spectral band was found to be superior in all situations, but TM5 consistently
showed the lowest correlations with the canopy variables.
Data sets were collected during three annual ryegrass phenological
stages: early stem extension (June), anthesis (July), and
senescence (August). The most significant correlations between reflectance
and the canopy variables were found for the June data.
High levels of biomass in July and plant senescence in August adversely
affected the spectral reflectance/canopy relationships.
Data from the tall fescue plots were obtained from a wide range
of total wet biomass levels (16.5 - 1677.9 g/m²). The asymptotic
limits, or the biomass range for which the reflectance could be used
to predict changes in the canopy variables, were studied. The
reflection asymptotes were nearly twice as high for the near infrared
(TM4) as for the visible and middle infrared bands (TM1, TM2, TM3,
TM7). The use of band ratios and normalized difference transformations
did not consistantly increase correlations of spectral reflectance
with the grass canopy variables. Logarithmic transformations of both
the spectral bands and the canopy variables were successfully used to
linearize the spectral reflectance/canopy regression functions. Redundancy
was found among the absorption bands (TM1, TM2, TM3, TM7)
and between the near infrared bands (TM4, MMR5). Principal component
transformations were utilized to eliminate these spectral band redundancies.
The seven spectral bands were reduced to two principal
components, while maintaining nearly all of the variability found in
the original bands.