re for 
rio 
forest 
vice, 
r, 
st 
ry 
m on 
SS, 
VII, 
awa , 
iniature 
- 1779 - 
TIMBER AND SOIL CLASSIFICATION USING NUMERICAL INTERPRETATION 
OF LANDSAT-2- AND AIRCRAFT SCANNER DATA 
Sipi Jaakkola, Esa Franssila and Pekka Saukkola 
Technical Research Centre of Finland, Laboratory of Land Use 
Espoo, Finland 
ABSTRACT 
More than two thirds of the land area of Finland is covered by 
forest which, predominantly, comprises managed conifer stands. . 
This fact is a justification for the research and development of 
remote sensing oriented timber and soil mapping methods. For 
this study, aircraft scanner data was acquired in 1977 and used 
in detailed classification of timber characteristics. Comparisons 
were made between aircraft and satellite data. The emphasis, 
however, was in supervised classification of numerical Landsat-2 
data by tree species and rough soil classes. A model relating 
the timber volume to the spectral reflectances of Landsat bands 
was developed and tested in a multistage sampling experiment. 
In the paper, the research objectives and methods, involving tim- 
ber and soil classification as well as volume estimation in two 
test sites, are discussed. 
Le INTRODUCTION 
Although the Landsat-satellite program has now been carried on 
for six years, the number of operational applications of Landsat 
data still remains relatively small. Remembering that the first 
satellites of the program were announced to be experimental, the 
international user community, nevertheless, seems to appreciate 
better spatial and spectral resolution for most earth resources 
inventory applications. Now that Landsat-D, in 1981, is likely 
to meet adequate resolution requirements, new expectations are 
being directed towards meaningful applications of the data. An 
overall objective of the remote sensing research program at the 
Technical Research Centre of Finland (VTT) is to prepare for a 
full utilization of the improved European and American satellite 
data in early 1980's. Major technical advances have to be made, 
within a few years, in order to meet the objective. Among those 
are, e.g., 1) improvements in numerical interpretation of vege- 
tation and soil from multispectral data, 2) improvements in geo- 
metric transformations of satellite data for map referencing pur- 
poses, 3) improvements in linking the satellite data to geo- 
graphic information systems. 
In this study, procedures are being developed for forest tree 
Species identification, as well as timber area and volume esti- 
mation, applying numerical Landsat and aircraft scanner data