372
TABLE 1
Landsat Thematic Mapper Bands and Transformations Used in
Statistical Analysis
Band 1
Band 2
Band 3
Band 4
Band 5
Band 7
(TM1) Landsat TM Band
(TM2 )
(TM3)
(TM4)
(TM5 )
(TM7)
Band 4 - Band 1
Band 4 - Band 2
Band 4 - Band 3
Band 4 - Band 5
Band 4 - Band 7
(BD1) Band Difference
(BD2)
(BD3)
(BD5)
(BD7)
Band 4/Band 1
Band 4/Band 2
Band 4/Band 3
Band 4/Band 5
Band 4/Band 7
Band 2/Band 3
Band 3/Band 1
Band 5/Band 7
(R41) Simple Band Ratio
(R42)
(R43)
(R45)
(R47)
(R23)
(R31)
(R57)
(Band 4 - Band
(Band 4 - Band
(Band 4 - Band
(Band 4 - Band
(Band 4 - Band
0.3037(TM1 ) +
+0.5585(TM4) +
-0.2848(TM1) -
+0.7243(TM4) +
0.1509(TM1) +
+0.3406(TM4) -
1) /(Band 4 +
2) /(Band 4 +
3) /(Band 4 +
5)/(Band 4 +
7)/(Band 4 +
0.2793(TM2) +
0.5082(TM5) +
0.2435(TM2) -
0.0840(TM5) -
0.1973(TM2) +
0.7112(TM5) -
1) (ND1) Normalised Difference
(ND2)
(ND3)
(ND5 )
(ND7)
Band
Band 2)
Band 3)
Band 5)
Band 7)
0.4743(TM3)
0.1863(TM7)
(TMB)
0.5436(TM3)
0.1 800(TM7)
(TMG)
0.3279(TM3)
0.4572(TM7)
(TMW)
1
Brightness Index
1
Greenness Index
1
Wetness Index
First Principal Component (PC1 )
Second Principal Component (PC2)
Third Principal Component (PC3)
Source: 1. Crist (1983)
and r=-0.77 for the June and August Austrian
scene dates. These and all other regression
relationships that exhibit at least a "fair"
correlation (r > 0.50) are significant at
greater than the 99 per cent level of pro
bability .
Two methods were used to rank the effec
tiveness of the Thematic Mapper bands and
transformations in discriminating metal
stress in coniferous forests: the first ap
proach involved the simple ordering of the
Pearson product moment correlation coeffi
cients (r-values) for each of the seven soil
metal values and combinations (lead, zinc,
copper, lead + zinc, lead + copper, zinc +
copper, and lead + zinc + copper) and thirty-
one bands and transformations for each of
the four Landsat scene dates (a total of 28
data sets), whereas the second approach con
sisted of calculating the weighted frequency
sums of the thirty-one bands and transforma
tions for each of the 28 data sets to obtain
a "rank sum value" for each TM band and
transformation for each scene date. Both
methods yielded essentially identical re
sults in defining the individual rank posi
tions of the TM bands and transformations
and their groupings. TABLE 2 presents the
results of this analysis. The division of
the bands and transformations into four
groupings is arbitrary and based solely on
the similarity in the "rank sum values" of
the group members. The range of values re
presented by each of the four groups is
approximately the same for each correspond
ing scene date group. Only the top-ranked
14 or 15 TM bands and transformations for
each scene date are presented in TABLE 2, as
it is from these that the best ten bands and
transformations based on all four scene
dates are compiled.
The highest ranked TM bands and trans
formations for both the June and August scene
dates of Austria show a close correspondence
to each other, which is not the case with
respect to the January and August Spanish
scene dates.
The five TM bands and transformations
comprising Group I of the combined Spanish
and Austrian scene dates are ranked also
within the top 14 or 15 bands and transfor
mations of each of the four individual scene
dates; those listed in Group II are repre
sented in the upper ranks of only three of
the four scene dates. The two top-ranked
transformations in Group I (PC1 and BD1) are
essentially identical in their discriminat
ing power, as are the third and fourth ranked
TMB and TM5.
4. DISC
Tj.
and trc
spectir
tor be
needs 1
scene c
hazy oi
nation
as well
the fii
band dj
criterj
simplei
problen
princif
of its
which n
scenes
one (,s)
matic J 1
ever, i
charact
by allc
data se
sition
Th
Mapper
bands a
Spanish
water s
is to h
the hot
in this
water
acute a
Th
the top
of the
likely