Page 268 - Digital Analysis of Remotely Sensed Imagery

230 Cha pte r S i x

spectrum. Subtraction of the two bands leads to a near-zero differ-
ence in the resultant image for them. This effectively exaggerates the
spectral disparity between vegetation and nonvegetative covers,
making vegetation more visible in the outcome band. In other words,
the resultant image maximizes the vegetative signal and suppresses
the visibility of soil and other background covers.
VIs indices are related closely to the amount of vegetative
cover present on the ground, and its greenness or biomass. Highly
correlated to green-leaf density, they can be viewed as a proxy for
aboveground biomass (Bannari et al., 1995; Rasmussen, 1998).
Vegetation indexing using a multitude of spectral bands is an
effective means of taking advantage of the rich spectral informa-
tion of multispectral satellite data. These indices can serve as good
surrogate measures of vegetative cover if calculated properly from
the right combination of spectral bands. In the multispectral
domain, which band should be the near infrared and which should
be the red band depends upon their wavelength range, which in
turn is governed by the spectral reflectance curve. For Landsat
Multispectral Scanner (MSS) data they are bands 7 and 5 [Eq.
(6.11)], but channels 1 and 2 for AVHRR data [Eq. (6.12)].

VI = MSS7 − MSS5 (6.11)

VI = ch 2 – ch 1 (6.12)

All VIs must be radiometrically standardized to account for the
atmospheric effect, differences in solar elevation, and differences in
instrument calibration from one image to another if they are to be
compared with one another directly, as in a longitudinal study. Sea-
sonal and geographic variations can be taken into account by stan-
dardizing the derived VIs through division by the sum of the same
two bands [Eq. (6.13)], and the resultant index is the normalized dif-
ference vegetation index (NDVI). In this way, all indices obtained from
images of different seasons across the globe are directly comparable
to one another.
R − R
NDVI = NIR red (6.13)
R + R
NIR red
where R and R represent spectral reflectance at the near-infrared
NIR red
(0.73 − 1.10 μm) and red (0.58 − 0.68 μm) wavelengths, respectively
(Holben, 1986). Again, the actual spectral bands corresponding to
R and R vary with the sensor, such as bands 7 and 5 for Landsat
NIR red
MSS data [Eq. (6.14)].
MSS7 − MSS5
NDVI = (6.14)
MSS7 + MSS5