6. MERGING IMAGE ANALYSIS RESULTS
Classification of differences shown on E-TM images may be
problematic due to the difficulty to differentiate categories of
cover classes; often a continuum in vegetation from low to high
cover is observed and class boundaries cannot be properly
identified. A solution is available by converting the image to an
NDVI map. Cover classes that are distinctly visible on the
image, but that looses that characteristic when converted to
NDVI data form a complication. Following a dualistic
interpretation approach solves both problems.
An example is taken from part of the ETM-150/36 image, taken
in August 1999 (Figure 4). An NDVI map was prepared, and
based on comparing a mosaic of images, the NDVI scale was
adjusted by the formula [0.70 (NDVI — 110) + 110], to reduce
the level of greenness to NDVI's expected by end-Oct/early-
Nov. (Figure 4). The NDVI map is based on a gliding scale; no
attempt is made to link levels to cover classes.
Comparing the NDVI map with the original TM image shows
that the dark-reddish areas do not represent active-green
vegetation. Consequently information loss took place. The dark
areas however comprise of pine trees and shade; pine trees have
a low chlorophyll activity. Both are very distinct classes and
must be re-introduced onto the NDVI map.
The ETM was subsequently classified through conventional
routines to map trees and shade (Figure 4). Equally other cover
types occurring in the area like snow, ice and clouds, were
mapped. Superimposed on the NDVI image the classified image
produced the merged product (Figure 4).
Ro ; 4 à
LE dent Ré au A i ws
115 "bare 127 vegetated 451
MRR TF T GS BRE
IAPRS & SIS, Vol.34, Past 7, “Resource and Environmental Monitoring", Hyderabad, India,2002
Classified
trees and
Figure 4. Merging Distinct Cover Classes with a Continuum
NDVI Interpretation.
After the product is prepared, GIS features as available on topo-
sheets and on other higher-resolution images (e.g. 15 m Aster
images), provides a map that is ready for local-level survey and
land use planning exercises. An example of a village map
prepared on the above lines is shown in Figure 5.
Village boundary
uL Streams
Ex Villages
== Ridges
m Contours
[I sf
ab E |
Very Bare to 50% Bare
Poorly vegetated
Somewhat vegetated
Well vegetated
Pine Trees
Figure 5. Adding Features to an Interpreted ETM Product.
[Miandam, located in the Swat valley, has an abundance of pine trees
>
especially on northern slopes. The main hillside, till the crest, has many
hamlets, is less vegetated and is mainly used for grazing (cyan color on
the image). Dispersed, many fields with annual crops were found
(yellow); orchards are situated close to the streams (light green color).]
t UN -—— UN ee 0A (5 oJ
Se ti p
es
a a CAELO EE Sn