298
Table 7. Significance of differences in TVI-values,
as obtained with LANDSAT-MSS 7, MSS 5 measurements;
Mendoza, Argentina; N = vineyard, W = olive trees,
W+N = olive trees plus vineyard, N+P = vineyard plus
vegetable crops; ns = not significant, s = significant
at P = 0.05, hs = highly significant at P = 0.01
Crop
28-8-1984
27-2-1985
15-3-1985
N W W+N
N+P
N W W+N
N+P
N W W+N
N+P
N
-■ s ns
hs
- hs hs
hs
hs ns
hs
W
ns
hs
ns
ns
hs
ns
W+N
-
-
-
s
-
ns
N+P
-
-
as t* for (Nj-1) respectively (N2-1) degrees of free
dom. The different number of degrees of freedom is
the main difference between the test of hypothesis
a] = 02 respectively op 4 02 •
The procedures just described have been applied to
a number of image samples. An approximate indication
of the normality of an observed distribution is ob
tained by considering the criterions skewness and
kurtosis. Large values of both criterions suggest
non-normality. A x^-test has been applied to a se
lection of 'worst cases'. In Table 4, as well as in
the next few ones, codes have been assigned to ground
control plots according to the format:
CROP IRRIGATION DISTRICT PLOT NUMBER
and irrigation districts
being:
A = East Sesia
B = Grande Bonifica Ferra-
rese
On the basis of the results in Table 4 we can con
clude that even large values of kurtosis, e.g. for
samples ВАЗ, CA4 and IA2 do not imply that the x^~
test will indicate significant deviations from nor
mality. As for example plots CA2 and 0B2 show, one
should apply a x^-test when considering samples with
many elements, 50 or more say.
Notwithstanding the important theoretical differ
ences between procedure 1 and II, the outcome of a t-
test can be essentially the same, as shown by the
results in Table 5. The plots have been chosen by
taking the largest observed phenological difference
for each crop. As expected because of the fewer
degrees of freedom applying to procedure II,differ
ences assessed as not-significant according to pro
cedure i, are significant according to procedure II,
The capability of measuring by LANDSAT significant
differences in TVI is confirmed by Table 6. Discrim
ination of phenological differences, such as con
sidered in Table 6 is of obvious relevance to crop
monitoring. It is, however, clear that this capability
makes crop discrimination more difficult. This is one
additional reason to translate phenological vari
ability within an area into quantitative terms, e.g.
as done in Section 2 by means of the function A(t).
The need for discrimination of different intercrop
ping schemes has been explained in Section 2. The re
sults in Table 7 demonstrate that the differences
noted between the graphs in Figure 2 are actually
significant on at least one overpass date. Discrimina
tion of these intercropping patterns is, therefore,
feasible.
The Thematic Mapper under any count is a big im
provement on the Multi Spectral Scanner from an in
strumental point of view. A more difficult question
is whether the TM improves our capability to dis
criminate crops. Namely the much improved radiometric
sensitivity and spatial resolution do allow for ob
with crop codes being:
Code Crop
B rice
C pasture
I corn
0 soybeans
Table 8. Significance of differences in TVI-values;
East Sesia, Po valley, Italy; for explanation of plot
coding see text; ns = not significant, s = signifi
cant, hs = highly significant
A: LANDSAT MSS 7 and MSS 5 measurements; samples
include between 3 and 10 pixels
B: LANDSAT TM 4 and TM 3 measurements; samples in
clude between 4 and 10 pixels
C: LANDSAT TM 4 and TM 3 measurements; samples in
clude between 15 and 43 pixels
Plots
САЗ
СА4
ВАЗ
BA4
IA3
IA4
A
CA3
-
ns
hs
hs
hs
hs
CA4
-
hs
hs
hs
hs
ВАЗ
-
ns
hs
hs
BA4
-
s
hs
IA3
-
ns
IA4
-
В
САЗ
-
ns
hs
hs
hs
hs
CA4
-
hs
hs
hs
hs
ВАЗ
-
ns
hs
hs
ВА4
-
ns
ns
IA3
-
ns
IA4
-
С
САЗ
-
ns
hs
hs
hs
hs
СА4
-
hs
hs
hs
hs
ВАЗ
-
ns
hs
hs
ВА4
-
hs
hs
IA3
-
ns
IA4
-
servation of within-field variability. From our point
of view, therefore, the most sensible way of comparing
the TM with the MSS sensor is to assess which one
gives the higher significance of differences in TVI
values.
If we compare Table 8-A with Table 8-B we see that
at equal number of elements per field plot, the MSS
fares marginally better than the TM. The difference
between BA4 and IA3 is significant with MSS measure
ments and not significant with TM ones. It is only
the larger number of TM elements for a given field
plot (Table 8-C) which gives a slightly better sig
nificance: the difference BA4 versus IA3 is 'highly
significant' according to Table 8-C.
These results can be looked at from different points
of view. From one side we have found that the per
formance of MSS, from our particular point of view,
is at least comparable with TM. It is, however, true
that by working with five TM pixels only, one can
discriminate plots of 4500 m^ versus the 23,305 m^ of
5 MSS pixels. Moreover the TM 5, TM 6 and TM 7
measurements allow for detection of crop-specific
effects, after crop mapping is accomplished.
5 RADIOMETRIC ACCURACY VERSUS SIZE OF SIGNIFICANT
DIFFERENCES
To draw a conclusion on the size of significant dif
ferences in TVI-values we have calculated the means
of (xi - X2)/[(x^ + X2)/2] for each one of a large
number of significance tests, as done with the MSS
and TM images considered in this paper.
The result reads: not significant = 4%, significant
= 10% and highly significant = 10%. These percentages
give beforehand an indication of how large relative
differences in TVI-values should be, for these differ
ences to result as either not significant or signifi -
cant aft
Let us
tion of
assessed
lish whe
The nomi
is 0.5%
nominal
of the f
results
We can
ciple me
with sat
Actual
quite di
take the
flectanc
ity grea
fore be
It mus
analysis
digital
values (
Furtherrr
have to
least nc
the imag
6 CONCLU
We conci
patterns
can be о
0.5 ha a
and MSS
that a ir
scheme c
agricult
satellit
feasible
as prese
REFERENC
Andersor
earth
Azzali,
proacf
An app
East S
1611 .
sear cl
Azzali,
app lie
rarese
Manage
Azzali,
data t
for Le
geninp
Crist, I
mental
opment
3-13.
Hinzman,
Growtl
wheat
Jackson,
crimii
varioe
atmosp
208.
Kenney,
statif
pp.
Kuipers
in th<
LAND Si
