Figure 3 shows the sampling area in a sub-region of land
system B (loess-covered sandstone hill). The sampling
area consists chiefly of two kinds of land unit, namely,
cultivated and gently sloping hillside and rocky gully
slope. They together account for approximately 85 per
cent of the total sampling area, indicating a close correla-
tion between the sampling area depicted by photomorphic
characteristics on the TM image and the real distribution
of the land units. Using this approach, not only all land u-
nits in the sampling area were recognized and mapped, but
also all land units in the land system were deduced based
on the area proportion and real distribution of the land u-
nits in the sampling area, and the area proportion of the
sampling areas in the sub-region or even in the land sys-
tem.
3.3 Interpretation and Mapping of Land Sites
A land site as part of the land unit is, for all practical pur-
poses, uniform throughout its extent in landform, soil and
vegetation. In general, the interpretation and mapping of
land sites in a land unit followed the approach similar to
that for land unit. First, the land unit was divided into a
number of small areas with more homogeneous photomor-
phic characteristics on 1: 50, 000 scale photograph. Sec-
ond, a sampling plot was chosen for each of small areas
and its boundary was as accurately as possible located on
1:10, 000 scale photograph. Third, elementary relief units
and the associated soils and land cover/use types in the
plot were identified. Final, all land sites in the sampling
plot were interpreted and delimited based on comprehen-
sive analysis on photomorphic characteristics and consult-
ing thematic maps as well as field studies.
Figure 4 illustrates a rocky gully slope plot. In this plot,
the land site of gully slope (581) accounts for approxi-
mately 68 per cent in area, indicating a good correlation
between the sampling plot depicted by photomorphic char-
acteristics on 1:50, 000 scale for land site interpretation
and mapping. Based on the area proportion and spatial
distribution of land sites in the sampling plot as well as the
area proportion of the sampling plot in the land unit, all
land sites in the unit could be deduced and mapped.
4. CONCLUSSIONS
The multi-level sampling interpretation of remotely sensed
imagery in terrain survey and mapping conducts the inter-
pretation successively from entirety to part and from high-
er level to lower level of land type. This stage-by-stage
approach is much beneficial to improving quality of the in-
180
terpretation. In addition, time and costs in the survey and
mapping can be substantially reduced as the approach is
based on the statistical sampling theory and land type in-
formation is extracted step-by-step from sampling areas
with relatively homogeneous photomorphic characteristics
on imagery. However, it should be noted that the accura-
cy of resultant land type maps produced based on the ap-
proach is largely dependent on appropriate selection of
Therefore,
much attention should be paid to field survey and inves-
sampling areas. while using this approach
tigation to ensure the accuracy of the interpretation and
mapping.
5. REFERENCES
Christian, C. S. and Stewart, G. A.. 1968. Methodology
of Integrated Surveys. In: Unesco Conference on Princi-
ples and Methods of Integrating Aerial Survey Studies of
Natural Resources for Potential Development. Thoulouse,
Unesco, Paris, pp. 233-280.
Townshend, John, R.G., 1981. Regionalization of Ter-
rain and Remotely Sensed Data. In: Terrain Analysis and
Remote Sensing edited by John, R. G. Townshend.
George Allen & Unwin, London, pp. 109-132.
