APPLICATION OF THE MULTLLEVEL SAMPLING 
INTERPRETATION OF REMOTELY SENSED IMAGERY 
IN TERRAIN SURVEY AND MAPPING 
Ni Shaoxiang, Associate Professor 
Zhou Shenglu, Research Student 
Department of Geo and Ocean Sciences, Nanjing University, 
Nanjing, China 
Commission IV 
ABSTRACT: 
The conventional approach for interpretation of land types on remotely sensed imagery depends mostly on one kind of im- 
agery, usually on same scale, which has caused some problems to the interpretation. The present sutdy deals with the utility 
of multi-level sampling interpretation method for extracting land type information from remotely sensed imagery, taking 
Huangpuchuan River catchment in Jungar County of Inner-Mongolia, China as the test area. Based on this approach, land 
types at the different levels, namely, land system, land unit and land site were interpreted and mapped. The study demon- 
strated that this approach could make the interpretation more efficient and, therefore, much reducing time and costs in ter- 
rain survey and mapping. Nevertheless, the accuracy of resultant terrain maps depends greatly on appropriate selection of 
sampling areas. 
KEY WORDS: Remotely Sensed Imagery , Multi-level Sampling Interpretation, Land Type 
ner-Mongolia. In topography it is dominated by loess hills 
1. INTRODUCTION and sandstone hills along with sand-mantled loess or sand- 
stone hills. 
As one of the most foundamentary tasks in land studies, 
land classification and mapping can provide a significant Two kinds of imagery were used in this study, namely, the 
basis for land use planning. Remotely sensed imagery has false-colour Landsat TM image on 1:100, 000 scale, com- 
been used in land classification and mapping for several posited by bands 2 (blue). 3 (green) and 4 (red), acquired 
decades. However, the conventional approach for inter- in later August. 1986, and the colour infrared aerial pho- 
pretation of land types on remotely sensed imagery tographs on 1: 50, 000 and 1:10, 000 scales, obtained in 
depends mostly on one kind of imagery, usually on same early September 1987. 
scale, which has caused some probelems to the interpreta- 
tion. This is because of not only apparent difference in 3. PROCEDURES AND RESULTS 
overall characteristics of land types at different levels on a 
given kind of imagery, but also of great variation in apper- The multi-level sampling interpretation consists of several 
ance of land types even at same level but on different kinds stages (Fig. 1) in which the information of land type at 
of imagery. the different leverls were extracted based on interpretation 
of the imagery together with consulting thematic maps and 
Townshend (1981) has made a recommendation on using field survey. 
the multi-level sampling method for extracting terrain in- 
  
formation from remotely sensed imagery. Recently, this 3.1 Interpretation and Mapping of Land Systems 
approach was used and further developed in our study on 
the Loess Plateau in China. A land system is defined as "an area or group of areas, 
throughout which there is a recurring pattern of topogra- 
2. STUDY AREA AND IMAGERY phy. soil and vegetation" (Christian and Stewart, 1968). 
In order to identify and delineate the land systems on the 
Huangpuchuan River catchment as the study area with ap- TM image the study area was divided into five kinds of 
proximately 1847 sq. km is located in Jungar County of In- region based mainly on variation or spatial combination of 
178 
phot 
othe 
mor| 
tion. 
whit 
both 
be d 
yellc 
colo 
yellc 
ish 
3.2 
Lan 
Eacl 
corr 
land 
rela: 
disti 
area 
as a 
pho! 
thro 
phic 
with 
vati 
In 4 
land 
and