BEA EEE
Fig. 2: Forest /non-forest classification (left) vs. ground truth (right)
6. Discussion and Conclusions
Examining the quality of satellite image and
then omission of TM band 7 in this study
confirm the results obtained by Ehrler (1993),
and stress again the necessity of quality control
before using these data.
The lack of a comprehensive definition of
forest led to the conclusion that sluices in forest
areas with low density are considered in both
possible forms (as forested and non-forested
areas), whilst the dominant spectral reflection
in the satellite image in such areas is closer to
non-forest. Releasing a comprehensive
definition of these different classes (by a
competent organisation or authority) is
therefore a necessary task. Such definitions
will help to increase the precision and
uniformity in such classification tasks.
The overall accuracy of 95% in such flat areas
in Iranian northern part confirms the high
potential assessing forested and non-forested
areas using Landsat TM data. Such a compiled
forest map with a high accuracy can be used in
executive works and in detailed studies. The
potential must also be evaluated in similar
investigations in other forests in the
northeastern and northwestern part of the
country (Golestan and Gilan Provinces) as well
as in the mountaineous parts of the northern
forests. This is necessary because their
topography has a serious impact on the
geometry and radiometry. Certainly the
different ecological conditions of the forests in
the western part of Iran require special
investigation.
7. REFERENCES
Borry F.C. and B.P. De Roover, 1990: Asessing the Value of Monotemporal SPOT1 Imagery for
Forestry Application Under Flemish Condition, PE&RS, 56 (8): 1147-1153
Darvishsefat A., 1994: Einsatz und Fusion von multisensoralen Satellitendaten zur Erfassung von
Waldinventuren, Dissertation, Geographisches Institut der Universitit Ziirich.
226 Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
