258 
22nd JULY 
7th AUGUST 
17 th SEPTEMBER 
Figure 2. Mean values, in digital counts (DC) for the three 
classes of cover types. 
Table 1. Different cover types defined as training parcels. 
Cover Type 
n Q pixels 
Agriculture 
Rice (Bahía) 
214 
Rice (Sequial) 
273 
Rice (Lido) 
339 
Rice (Pierina) 
207 
Rice (Rubino) 
51 
Citrics 
158 
Rural 
225 
Table 2. Dates of the 1985 images used and development stages 
of rice plants. Figures between brackets mean dates since 
sowing. 
Data image Soil conditions Phenological Stage 
22nd July Just flooded 
7th August Under water 
17th September Just drined 
Maturation Phase (70-80) 
Maturation Phase (85-95) 
Harvest Phase (125-135) 
Three images from Landsat 5 (TM) corresponding to 22nd 
July, 7th August and 17th September, 1985 have been used to 
carry out this study. Some help has been obtained out of 1: 
50.000 cartographies maps. Table 2 shows phenomenological 
data of the dates of the images. 
3. CLASIFICATION OF THE INDIVIDUAL SCENES. 
The results of the measurements carried out in three of the 
reference parcels placed in the Sueca zone are shown in figure 2. 
Reflectivities are expresed in digital counts (DC). A clear 
difference in the DC is shown for the three cover types of the 
images of July and August. In the September image rice and 
citricS appear confussed and only band 4 is slightly significant. 
In spite that the images of July and August give reflectivities 
so different in the three types of parcels, the overall classification, 
of the zone presents difficulties. The high number of bordering 
pixels introduces a noticeable degree of incertitude in the 
classification due to the scattering of values within categories. 
This incertitude is greater between the classes urban and citric - 
vegetables, because this cover type corresponds to the smaller 
parcels. 
For the overall classification of the zone the bands chosen as 
significant have been bands 3 and 4. Any other band 
combination does no introduce better achievements. The 
processing of the bands has been made by the evaluation of the 
vegetation index (VI): 
B4-B3 
VI = 
B4 + B3 
Some authors apply different types of filtering prior to per 
point classification achieving less scattering in each class values 
(Atkinson, 1985). However, this degradation technique of the 
image is not so useful in an area of great parcelation. But the 
application of a low pass averaging filter to image VI instead of 
doing it to the bands individually has permited a better 
classification under particular conditions. 
Figure 3 shows the histograms which come out of the 
1985 SUECA 
0 20 40 60 80 100 120 140 160 180 200 
VEGETATION INDEX 
Figure 3. Histograms of the smoothed vegetation index images. 
120 T 
80 
60 
40 
20 
0 
B 
Figure 4. S] 
rice varietie 
procedure a 
considered. 
August. Ho 
where the s: 
citrics, figu: 
eliminates it 
rice and citri 
Test areas 
with the sm 
pixels, of 
distinguisha 
confuse the 
short vegeta 
The over, 
rice, 1910 
urbanized ai 
The imag 
greater distil 
image of S 
completely i 
area. Final!; 
advanced p 
Figure 4 sho 
counts (DC) 
The most 
varieties is 
figure 5 is a 
of August 
separated. It 
varieties but 
A first es 
results of c 
supplied by 
obscured by 
We unde 
measuremer 
1984]) in an 
give a separi 
trying now t 
not exist) I 
components 
4. REFERE1 
Atkinson, P 
1985. Imi 
using filte 
- 961. 
Berg, A. & ] 
a remote 
(Agreste 1 
Commissio 
Etablisse