Site location 
DATA ACQUISITION 
The number of sites in a country depends on the 
proportion of the Community's arable land in that 
country. Since there are 50 sites, each site 
"represents", in one sense, 2% of the Community's 
arable land. At least one site falls in each of 
the Community's 12 member nations - with the 
exception of Belgium, the Netherlands and 
Luxembourg, which share a single site. 
The sites are located to avoid the necessity of 
special programming for SPOT, which is to say 
that they are normally located close to the track 
of SPOT and on the rows (J) of the SPOT grid. In 
the case in which, after acquisition of one site, 
it might be possible to re-orient the radiometer 
to acquire another, care was taken to ensure that 
the sites are sufficiently far apart in the 
along-track direction for the apparatus to re 
orient and stabilize. 
Suitability of data 
There is no way of determining whether TM or SPOT 
data will be available for a given site, and 
S.C.O.T. must be capable of interpreting either 
type of data. Both types have their particular 
advantages. SPOT data are be well adapted for 
defining field boundaries and for estimating 
areas under crops. TM data, resampled to 20 
metres, is suited (especially in the middle 
infra-red) to the discrimination of crop types, 
some of which may not be separable using SPOT's 3 
channels. 
MOS data have proved impossible to incorporate in 
the system up to now, since they apparently 
suffer from a number of technical problems. The 
sensor will be used when these problems have been 
solved. 
All sites are located so as to fall completely 
within the corresponding TM swath. The sites in 
the north of Europe are also, whenever possible, 
located in the si delap between neighbouring TM 
swaths. In this way, the opportunity for 
scanning the sites using TM is doubled. 
In as far as possible given these constraints, 
sites are placed in areas in which at least 10% 
of the land is cultivated. 
Although the Agriculture Project has planned to 
use MOS images as well, MOS orbital geometry was 
not taken into account in locating the sites. 
The sites are located independently of the 
administrative regions, and many therefore fall 
astride two or more such regions. 
Ground data and Segment location 
Areas for ground survey, called segments. are 
located at each site (Delineé 1988). Each 
segment measures 700 meters on a side. In most 
of the northern sites, sixteen segments are 
arranged in a regular 4 x 4 grid. Most of the 
sites in southern countries are stratified to 
exclude non-agricultural areas, and segments are 
only located in agricultural areas in the sites. 
A simple scheme involving the repetition of a 
regular grid results in an average of 16 segments 
in the stratified sites. 
Organisations under contract to the JRC send out 
field teams annually to identify all the crops in 
all fields in each of the segments in each site. 
Farmers exploiting the fields in the segments are 
also asked to provide data on the expected and 
final yield and quality of the crop. The 
contractors provide these ground data to the JRC 
at the end of the crop year. 
In 1988 ground data were collected at 17 sites; 
in 1989 30 sites were active, and in 1990 all 
fifty sites are being inventoried. 
As described in more detail below, the same 
segments investigated by ground teams are 
identified on the imagery and used by S.C.O.T. 
for the major effort put into the image 
interpretation. 
The ground data are collected entirely 
independently of the image analysis, and the 
image interpreters have no access to these data 
until the end of the year. The data on the 
images are therefore analysed without the benefit 
of current ground data. 
Data volume 
The Agriculture Project expects to base its 
"rapid estimates" on about 3 to 7 images per site 
per year, depending on cloud cover. The last two 
years ( 1989 and the start of 1990) have been 
exceptionally clear over much of Europe, and the 
project has had no trouble in filling this 
"quota" of image acquisitions. This number of 
images per year means that in the operational 
phase of the project the contractor will 
therefore interpret some 250 to 300 scenes per 
year. S.C.O.T. is building up its capacity over 
the years, with 10 active sites in the first 
lead-in period of 7 months (which ended in 
October 1989), then - currently - 20 active sites 
(the original 10 plus another 10) for the first 
full season lasting until the end of 1990, and 
finally 33 sites to round out the 3-year 
contract. The remaining 17 sites will be 
interpreted in the final 2 years of the JRC's 
pilot project. As explained below, it is also 
necessary to acquire imagery for sites that are 
not yet active, but that will become active in 
the following year. The JRC is therefore 
purchasing imagery for 33 sites in 1990. 
Scheduling 
Action 4 uses both TM and SPOT data, collected 
whenever cloud cover permits, with the 
restriction that once an image of a site is 
acquired, no further acquisitions for that site 
are programmed for a dead time whose length 
depends on the period of the year and the 
location of the site. Dead times are generally 
shorter in the cloudy north of Europe than in the 
clear skies of the south. 
Close links 
have been set 
up 
between 
the 
companies 
s e 
lling SPOT and 
Landsat da 
t a , 
S.C.O.T., 
and 
the JRC so that 
a l l 
parties 
are 
i n f o r m e d 
rapidly of a succès 
s f u l 
c l o u d - f 
ree 
acquisition. 
Using guidelines 
laid 
down by 
the 
JRC, S . C . 
0 . T . 
must then imm 
e d i a 
t e l y dec 
i d e 
whether or 
not 
to order the data 
184