04
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B-YF. Istanbul 2004
band from the visible region, the near infrared and one from
mid-infrared can be selected for the best separibilty in the
vegetated areas using Landsat ETM+ image. As it is common to
use band 3 from the visible bands and 5 from the mid-infrared
bands the current study has also decided to do so. Therefore,
bands 3, 4 and 5 were finally selected for the interpretation
purpose.
3.2 Interpretation of Tropical Vegetation
Figure 2 shows a spectral library for the vegetation in
southeastern Bangladesh.
0,4
A
Qo NS
8 03 / Ne
© L ; NN *- Primary forest
o Ÿ fou N Secondary forest
o 7 7 Bamboo
5 0 2 Ÿ / Shrub
= ' U Rubber
= 7//4 * Teak/scattered trees
-
E t" Indigenous plantation
3 / / .]. [Acacia
a ff ,
o0 0,1 //
€ {7
ho dB
0
Band 1 Band 2 Band 3 Band 4 Band 5 Band 7
Landsat ETM+ bands
Figure 2. Comparison of spectral reflectance from different
vegetation types in south-eastern Bangladesh
Band 1, 2 and 3 of Landsat ETM+ represent to the visible light
of blue-green, green and red reflectance respectively. The
reflectance of vegetation in the visible spectrum dominates due
to the presence of leaf pigments. Shrubs usually contain a little
amount of biomass, which has a high reflectance in all the
visible bands. In contrast, bamboo, which is a monocotyledon,
appears with a high absorption in this region. Vegetation, which
contains a huge amount of green leaves and biomass (i.e.
bamboo, natural forest, Acacia plantation etc.) usually shows
more absorption in the visible green and red spectral region. On
the other hand, vegetation that has fewer amounts of those
components (i.e. shrub, rubber) exhibits relatively higher
reflectance in that spectral region. It should be noted that
vegetation having little amount of green biomass might contain
some reflectance from the soil underneath. It is interesting to
note that the reflectance in visible-red region is higher than that
of the green for vegetation containing little amount of biomass,
however, the relationship is opposite for those vegetation
having high amount of biomass, because red light is largely
absorbed by chloroplasts and used in photosynthesis. Therefore,
it can be concluded that the first derivative of visible red to
green reflectance might contain some useful information on the
green biomass content of tropical broad-leave vegetation. As
band 3 provides the highest variability among the visible
channels and consequently it makes sense to use this channel
for interpretation of vegetation (Figure 2).
Band 4 corresponds to the reflectance of near infrared region,
which is very high for vegetation and therefore, it is widely used
to separate vegetation from other type of landuse. Usually
reflectance in this band is transparent by chloroplasts, but
highly reflected by spongy mesophyl. Young secondary forest
159
shows the highest reflectance in this region and teak/scattered
trees have the lowest. Mature natural forest shows relative lower
reflectance than the young secondary forest. Though mature
forests have the similar species composition as the young
secondary forests, the difference might be the result of the
difference of the structure of mesophyl tissue. The mesophyl
tissue of the leaves of younger vegetation provides the stronger
reflectance than that of the mature vegetation leaves. In
addition, presence of more shadows on mature forest canopy
might have some influence on it. As band 4 shows a high
variability among vegetation classes it would be useful to use
this for vegetation interpretation.
Band 5 corresponds to the shortwave-infrared region, which is
quite sensitive to the amount of water present in plant leaves.
Rubber and shrub usually shows a higher reflectance in this
region. This might occur that those classes have a less amount
of water or reflectance was dominated by soil-background.
Acacia plantation exhibits the lowest reflectance. This
vegetation does not contain any true-leaf. Leaves usually shed at
the seedling stage and the phyllod become swollen and act the
function of leaf. Probably those modified phyllods contain large
amount of water than the leaves of other vegetation. This
spectral region has a high variability for different vegetation
classes. It indicates that the band could be useful for the
separation of different types of vegetation. All the spectral lines
of figure 2 have been crossed each other in between the near-
infrared and mid-infrared region. Consequently, the first
derivative between bands 5 to 4 might contain valuable
information for vegetation class separation in the tropics.
Band 7 also contains the information of shortwave-infrared
channel, which is also an indicator of the presence of water on
leaf. However, all the lines in figure 2 from bands 5 to 7 did
not cross each other. So, it contains only the redundant
information as band 5 does. Hence, if band 5 is already used
band 7 will not improve the interpretation capability of tropical
vegetation.
From the above discussion it can be concluded that the use of
bands 3, 4 and 5 could be optimal interpretation of vegetation.
As we can use maximum three channels as the basic colour
combinations. The next question arise which band should be
used on which colour combination. If 5 4 3 bands are assigned
to red, green and blue the vegetated area will appear as green. If
4 5 3 combination is the vegetation will be red. As human eye
can distinct red better than green the latter combination could
separate vegetation classes in a better way. However, working
continuously with red on computer screen is a stressful for eye,
the other combination is recommended for certain time.
3.3 Separation of Individual Class
As figure 2 shows only the mean of value of reflectance, further
graphs were developed to show the variability of reflectance in
each class (Figure 3). In this figure, high-low graph represents
the mean value at the centre with a plus and minus of 9594
confidence limit, which means that a chance at this probability
level the mean value of reflectance lies within this threshold.
