208 
As a result of background research programmes exe 
cuted during the NIWARS project from 1972-1977 in 
The Netherlands, for land observation three spectral 
bands have been selected positioned at center wave 
length values of 550, 670 and 870 nm respectively. 
Using one CCD only, the off-nadir view-angle was 
kept below 15°. Studies on the non- Lambertian beha 
viour of vegetation canopies have shown that simulta 
neous acquisition of multispectral data under diffe 
rent angles of view increases discrimination between 
vegetation types. For this reason a forward-looking 
module with a centre off-nadir view angle of 54°, 
observing in the same set of three spectral bands, 
can be used. The optical system should be dimen 
sioned in such a way, that corresponding with the 
integration time required to obtain a ground reso 
lution in flight direction of 0.5 m at nominal 
flight altitude, a radiometric resolution of a noise 
equivalent value for a reflectance difference of 
0,5% would be achieved during the growth season be 
tween 10 a.m. and 4 p.m. local time. 
For the detection of the spectral distribution of 
the upwelling radiance from (sea)water, the user 
requirements differ considerably. Instead of three 
spectral bands, eight bands within the range between 
400 and 1100 nm are required. Averaging over conse 
cutive pixels and a longer integration time are 
needed in order to reduce signal variation due to 
local variations of specular diffuse sky reflectance 
caused by the moving water surface. 
The dynamic range of the radiance contribution from 
the water is relatively small compared to the case 
of land observation. In order to detect small inten 
sity differences, a noise equivalent value for the 
reflectance difference of 0.05% is required. It was 
decided to select the same set of spectral bands 
within the available spectral region as defined for 
the Ocean Colour Monitor, a candidate optical in 
strument for a future ESA remote sensing satellite. 
In particular for sea observation absolute calibra 
tion of the sensor output signals is required. By 
using two spectral bands in the near infrared part, 
the radiance contribution due to atmospheric path 
radiance and surface sky glint is measured for each 
pixel. Since the scattered surface and path radiance 
is hardly determined by the spectral distribution of 
the upwelling radiance, relations between the total 
scattering term for the spectral bands in the visible 
and near infrared part can be applied to assess the 
upwelling radiance in the visible bands. Such rela 
tions between scattered radiance in the visible and 
near infrared bands can be derived from radiative 
transfer models in which the relative optical depth 
at flight altitude, the phase and observation angle 
and the assumed aerosol type are incorporated. 
The spectral bands in the visible part of the spec 
trum are positioned in such a way that an improved 
discrimination may be achieved between chlorophyll 
present in phytoplankton, the so-called yellow stuff 
and other anorganic matter and pollutants. The scan 
plane should also be tiltable up to an off-nadir 
angle between 10° and 20° in order to avoid detec 
tion of direct sunglint. 
The user requirements for land and sea observation 
are summarized in table 1. 
The position of the spectral bands selected for land 
and sea observation is presented in figure 1 to 
gether with examples of the spectral reflectance of 
green vegetation, dry bare soil and turbid water. 
CONFIGURATION OF THE SENSOR SYSTEM 
During the different trade-off studies for the con 
figurations derived from the required sets of spec 
tral bands, it was concluded that the optimum choice 
would be a cluster of 4 cameras. Each camera should 
measure in three spectral channels. The measurements 
in 3 different spectral bands could be realised by 
means of field separation and the insertion of spec 
tral filters mounted in front of each of the 3 
CCD's. By aligning three cameras in nadir direction, 
for land observation a high accuracy of the interband 
registration could be obtained by using for instance 
the central CCD's. For sea observation the combina 
tion of the 3 cameras provided the required 9 bands 
for which the triplets of forward, nadir and aft 
looking channels are already geometrically regis 
tered. The system software should be applied off-line 
for averaging by increasing the pixel size and for 
interband registration of all channels. The module 
containing the three cameras could also be tilted 
along the same axis. 
Another advantage of this choice was that a compact 
configuration could be obtained, sized to the total 
field of view provided by the optical window in the 
bottom of the fuselage of the NLR laboratory air 
craft. 
A cut-away view of the design of the CAESAR camera 
module is shown in figure 2. The field separation for 
the forward and aft looking channels is minimized by 
means of reflection prisms which also avoid the intro 
duction of beam polarization. The pushbroom scan prin 
ciple and the consecutive scanning in the object 
space by means of three CCD's per camera for sea ob 
servation are illustrated in figure 3. 
SENSOR DIMENSIONING AND CCD SELECTION 
For the dimensioning of the sensor system, the dy 
namic range of the measured signal should be deter 
mined. This depends on the solar irradiance, the 
directional reflectance of the object observed, the 
atmospheric transmittance, the aperture of the ima 
ging optics, the transmittance of the optics and the 
spectral filter, the spectral sensitivity, band width, 
band position and integration time. In addition the 
detector noise also plays an important role. For land 
observation under marginal conditions, the required 
intensity resolution for a spatial resolution of 0.5 
m could be obtained for the value F/2.6. 
As a result of the expected low level of the mea 
sured radiance from water surfaces, a temperature 
correction applied to the dark current calibration in 
combination with improved temperature control would 
be needed in order to meet the requirements for the 
intensity resolution in all bands for the same 
F-number. 
Table 1 
Land observation 
Sea observation 
Band 1: 535-565 nm 
Band 2: 655-685 nm 
Band 3: 845-895 nm 
Instead of three discrete 
bands, spectral correla- 
lation filters for soil, 
green vegetation and clear 
deep water can be used. 
Forward-looking module: 
centre off-nadir angle: 54° 
Same bands as for nadir 
looking 
Band 1: 400-420 nm 
Band 2: 435-455 nm 
Band 3: 510-530 nm 
Band 4: 555-575 nm 
Band 5: 620-640 nm 
Band 6: 675-695 nm 
Band 7: 770-800 nm 
Band 8: 990-1050 nm 
(measured twice) 
NE A p S 0,5% for all bands NE A p S 0,05% for 
all bands 
Spatial resolution: 0,5x0,5 m between 10 and 20 m 
(minimal) 
Total field of view (nadir looking): 25.7°. 
Instantaneous field of view: 0.26 mrad.