nierons ) of some 
for vegetation 
.6 
.7 
.8 
.1 
.52 
.60 
.69 
.90 
.75 
.5 
.35 
.59 
.69 
.89 
.73 
capture of the data have undergone a tremendously 
rapid development during the 1970s and 1980s. Since 
the launch of the first Landsat satellite in 1972, 
remote sensing has become a freguently used tool for 
resource inventories of different kinds. One of the 
more important developments during this period, is 
the introduction of new sensing systems. The 
resolution has become higher, the number of spectral 
bands has increased and as a consequence, the data 
volumes have become enormous, see Tab. 3. One 
consequent of the increasing data volumes, is that 
the demands on the analysis equipment have risen, and 
the gap between the rich and the poor users has 
widened. 
Table 3. Spatial resolution, data volumes, expressed 
for the 
repeat 
cycle 
16 
16 
1-5 
1-5 
1 
as Landsat MSS equivalents and repeat cycle, 
most commonly used satellite systems. 
Sensing system: spatial 
resolution (m) 
data 
volume 
Landsat MSS 
79 
100 
% 
Landsat TM 
30 
825 
% 
Spot (multispectral mode) 
20 
917 
% 
Spot (panchromatic mode) 
10 
1222 
% 
NOAA 6 and 7 AVHRR * 
1100 
5 
% 
* Local area coverage (LAC) data 
.90 
.10 
.93 
.50 
.68 
.10 
.93 
.5 
.5 * 
sn multispectral 
t five wavelength 
rran 1980, Tucker 
Relationship to 
vegetation amount 
Strong negative 
Weak positive 
Strong negative 
Weak negative 
Strong positive 
In spite of the intensive research and the many pilot 
and feasibility studies, the number of truly opera 
tional applications of remote sensing is limited. A 
major constraint has been, and still is, the rela 
tively long interval between image recordings. The 
chances of obtaining a cloudfree image over a certain 
region at a certain time of the year are often very 
small. Justice et.al. (1985) reported that only one 
or two scenes, during wet season, since the first 
launch of the Landsat system, were available over 
parts of West African Sahel. The situation is even 
worse over the wet tropics. One way of overcoming 
this problem has been to use low-resolution data from 
weather satellites, e.g. the NOAA-series, with daily 
recordings (Hellden 1985, Tucker et.al 1985a, Tucker 
et. al. 1985b, Justice et.al. 1985). Several com 
parisons between Landsat MSS and NOAA AVHRR have been 
carried out, e.g. Nelson & Holben 1986, concerning 
information content and classification accuracy. 
Townshend & Tucker (1984) reported that AVHRR data 
represented approximately 70 % of the variation in 
bands 5 and 7 of Landsat MSS. 
The ability to direct the sensors of the SPOT- 
satellite will be another, very promising, way to 
solve some of the problems with cloud cover. The 
repeat cycle of SPOT-nadir images is 26 days, but 
using the off-nadir viewing (maximum 27 degrees) the 
average repeat cycle over the equator is 4 days, 
maximum 5 days. 
The use of sensors recording data at large angular 
views, e.g. NOAA AVHRR (56 ) and SPOT (27 ), involves 
however several new problems. The problems arise 
mainly from the following factors: 
near infrared 
STSIDERATIONS 
1 Varying atmospheric effects due to different path 
length 
2 Vegetation has often anisotropic reflectance 
properties, and will not follow a Lambertian model. 
3 The topographic influence will be much more 
pronounced 
4 Distorted pixel size and image geometry 
forms has been an 
two decades, and 
sensed data has 
. Both techniques 
ta as well as for 
A comprehensive investigation of off-nadir effects 
on SPOT and NOAA AVHRR is presented by Singh & 
Cracknell (1986) . To reduce these effects many 
authors have used only a fraction of the image data 
(NOAA AVHRR). Tucker et.al. (1983 & 1984) restricted 
the analysis to areas within 20 viewing. Duggin 
et.al. (1982) reported that simulated NOAA AVHRR 
single-band radiance values varied over 10 % across 
the swath if a scan angle of 14 was used, and sug 
gested that only the central 25 % of the scene should 
be utilized for generation of vegetation indices. 
Investigations of off-nadir effects due to aniso 
tropic reflectance properties are presented by 
Bartlett et. al. (1986) and Hugli & Frei (1983). 
In the future we can most likely also count on all- 
weather-systems, based on active remote sensing tech 
niques. The European remote sensing satellite, ERS-1, 
is planned to be launched in late 1989 (Guignard 
1985). Its primary use is within the field of ocean 
monitoring (waves and winds), though the C-band SAR 
may be useful also for applications on land. The 
Canadian Radarsat is scheduled for launch in 1991, 
and will carry a C-band SAR as well as a high- 
resolution sensor in the visible to near infrared 
part of spectrum. 
4. SATELLITE REMOTE SENSING, PREPROCESSING 
4.1 Radiometric correction 
When working with single date imagery, the require 
ments of preprocessing are much less than in the 
multi-date case. Radiometric correction becomes 
necessary only when absolute comparisons between 
different dates are carried out. Radiometric correc 
tion is needed to: 
1) correct for the difference in sun elevation (i.e. 
illumination differences due to time of the year 
of recording and the time of the day), 
2) correct for different sensor characteristics 
3) correct for atmospheric effects. 
A sound scientific principle is to convert the 
satellite data into a more tangible physical unit 
than just digital grey level (DGL). Then the data 
becomes comparable with most other types of remotely 
sensed data, and not only with other images from the 
same sensing system. The Landsat MSS-system records 
data as radiances, while the Thematic Mapper data 
(TM) is recorded in spectral radiances, i.e. the band 
width is taken into consideration. In order to make 
MSS and TM data comparable the MSS data must be 
converted into spectral radiances, expressed in the 
unit (rriW/cm2 sr microns) . Absolute radiance values 
are calculated according to the following formulas. 
Landsat TM: Rs = (Rmax-Rmin)/S * DGL + Rmin 
Landsat MSS: Rs = ((Rmax-Rmin)/S * DGL + Rmin)/BW 
where: Rs = spectral radiance (mW/sr cm2 micron) 
DGL = digital grey level 
S = saturation, maximum DGL 
Rmax = maximum radiance * 
Rmin = minimum radiance * 
BW = band width (microns) 
* = Landsat TM: spectral radiance (mW/sr cm2 micron) 
Values of Rmax and Rmin are found in Table 4. 
Calibration data for transformation of NOAA AVHRR 
data to either spectral albedo or radiance are given 
in each scan line in the tapes provided by NOAA/ 
NESDIS. The formulae for the transformation are 
described in NOAA Polar Orbiter Data, Users Guide 
(Kidwell 1985) and by Singh & Cracknell (1986). 
A problem when working with radiometric correction 
of data is that different receiving stations use 
different methods for the system correction. In some 
cases it can be impossible to aquire the proper 
information needed for radiometric calibration, which 
1043