Crop 
April 
May 
June 
July 
Figure 3. Crop-discrimination scheme obtained by the combined use of greenness (GR) and TVI; crops included are 
the main crops cultivated in the irrigation district Grande Bonifica Ferrarese in Italy 
characteristic of each crop. 
This multi-index multi-temporal method has been ap 
plied to a set of MSS-images in 1980 and compared 
with agricultural statistics of the Po plain. The 
result is given in Table 3 and, although some devia 
tions are still rather large, e.g. for alfalfa, these 
results seem very promising. 
control plots, belong to the same normal parent dis 
tribution. 
Under the assumption that pairs of samples (S^, S2) 
are extracted from a single normal distribution a 
Student's t-test can be applied to the difference be 
tween the two means. The variable t is obtained as 
(Kenney and Keeping 1959): 
4 STATISTICAL SIGNIFICANCE OF DIFFERENCES IN TRANS 
FORMED VEGETATION INDEX VALUES 
t 
N 1 N 2 
31/2 
N 1 + N 2 
(2) 
The main problem with the multi-index multi-temporal 
method is whether the observed differences are 
statistically significant. 
Mean vegetation indices value and variance have 
been calculated for a number of ground control plots 
in the two irrigation districts of the Po valley and 
of Mendoza. The significance assessment is focussed 
on TVI because numerically it is the least sensitive 
to variations in reflectance and because of its wide 
spread use. 
Two procedures have been applied to test the sig 
nificance of differences in mean TVI values. The first 
procedure is an approximate procedure. In this proce 
dure it is assumed that image samples, i.e. the ground 
where x-| , ^ are the means, N-| , N2 the number of 
elements of S-| respectively S2 and 82 is an estimate 
of , the variance of the parent distribution given 
by: 
. V’ 
+ N„o 
2 2 
N, + N- - 2 
(3) 
2 2 
where , 02 are the observed variances of S-j , S2. 
The distribution of t, as defined by eq. (2), has 
n = N^ + N2 - 2 degrees of freedom.