197 
TABLE 2 
¡asa 
istics 
n and 
plant 
n to 
ec- 
ensity. 
bsorp- 
550 nm 
centra- 
mini- 
he 
tanges 
ab- 
.nds 
ated 
P. excelsa) 
ig at the 
:, mature 
lunts of fir, 
such as 
present as 
it understory 
:s , and small 
•ess are 
¡tand. 
idsat TM 
:a used in 
:om mid-Sep- 
1 scenes are 
, All scenes 
lave been 
ising the 
)d of Crane 
; form the 
2 study. A 
2 and station 
jround con- 
sles were 
lorizon and 
2 lower 
ree sampling 
sisted of ap- 
2dles and 
six year old 
Correlation Between 
Spruce Needle Metal 
Soil Metal (M ) and 
(M N ) Content S 
Correlation 
Coefficients 
(r-values) 
Pb 
s 
Zn 
s 
Cu 
s 
Pb N 
Zn N 
CU N 
Pb 
s 
1 .00 
0.91 
0.79 
0.15 
0.74 
0.34 
Zn 
s 
- 
1 .00 
0.82 
0.11 
0.73 
0.28 
Cu 
s 
- 
- 
1 .00 
0.26 
0.50 
0.16 
Pb N 
- 
- 
- 
1 .00 
0.21 
0.05 
Zn N 
- 
- 
- 
- 
1 .00 
0.25 
CU N 
- 
- 
- 
- 
- 
1 .00 
Level of significance: r>0.49, p>0.99 
0.30<r< 0.49, 0.95 < p< 0.99 
r < 0.30, p < 0.90 
A total of 44 soil and 44 tree samples 
collected from the 10.75 ha test site were 
analysed for total copper, lead, and zinc 
content using the atomic-absorption spectro 
photometry method. Soil lead values range 
from 10-10,000+ ppm, soil zinc values from 
60-6300 ppm, and soil copper values from 
20-940 ppm, whereas spruce needle lead values 
vary from 1-10 ppm, needle zinc values from 
30-340 ppm, and needle copper values from 
2-5 ppm. 
8. ANALYSIS OF GEOCHEMICAL AND LANDSAT DATA 
The approach followed in establishing 
the relationship between the soil, needle, 
and spectral data involved first merging the 
different data sets with one another and then 
applying a linear regression analysis to the 
various combined sets. Because of the common 
reference system employed in the collection 
of the soil and spruce needle samples, the 
spatial correspondence between these two 
data sets was already established. For the 
merging of the Landsat spectral data with 
the ground-collected data, the geographical 
correspondence between the Landsat MSS and 
TM pixel arrays of the four Landsat scenes 
had to be first determined and then the aver 
age soil and needle lead, zinc, and copper 
values calculated for each "pure" pixel re 
presenting 95 per cent or more forest cover. 
Soil and needle metal isopleth maps of each 
of the three metals provided the means of 
merging the ground information with the 
Landsat spectral reflectance information of 
the test site, using 100-point and 36-point 
dot grids scaled in accordance to the 76 by 
76 metre and 30 by 30 metre ground-projected 
instantaneous field of views of the Landsat 
MSS and TM sensor systems. 
The number of "pure" test site forest 
pixels contained in the September 1976 and 
1981 MSS scenes is 11 and 12, respectively, 
and in the June and July TM scenes 76. 
9. STATISTICAL ANALYSIS OF DATA 
Linear regression analysis and an ana 
lysis of variance formed the basis for 
establishing relationships between the soil, 
needle, and Landsat data sets. TABLE 2 lists 
the correlation coefficients (r-values) ob 
tained for the various pairwise combinations 
of the soil and needle metal data sets. The 
relationships between soil copper, lead, and 
zinc are very good to excellent, but are de 
cidedly poor between needle copper, lead, and 
zinc. The results of the regression of needle 
lead and copper content against their corres 
ponding soil lead and copper values are poor, 
but good with respect to needle zinc regres 
sed against soil zinc. 
TABLES 3 and 4 list the Landsat MSS and 
TM spectral bands and transformations regres 
sed against soil and needle copper, lead, and 
zinc pixel metal values. The results of the 
statistical analysis employing soil metal 
values have been published in previous papers 
(Banninger, 1985a, 1985b, 1985c, and 1986), 
and only the more significant findings from 
these studies will be presented here. 
For the MSS bands and transformations, 
the first principal component (PC1) and the 
Kauth-Thomas green vegetative (GVI) and soil 
brightness (SBI) indices show overall the 
highest correlations with soil metal content 
(r=-0.74 to r=-0.89), followed closely by 
Landsat bands 6 and 7, band differences BD6 
and BD7, and the perpendicular vegetation 
indices PVI6 and PVI7 (r-values ranging from 
-0.65 to -0.85). These values are significant 
at greater than the 98 per cent probability 
level. 
For the TM bands and transformations, 
the normalised differences ND1 and ND3, the 
band difference BD1, and the simple ratio 
R41 exhibit the highest correlation values 
with respect to soil metal content (r=-0.68 
to t--0.11), followed by the greenness index 
TMG, band difference BD3, and the first prin 
cipal component(PC1) (r-values from -0.6 3 to 
-0.70). These values are significant at 
greater than the 99 per cent probability 
level. 
The regression of the Landsat MSS and 
TM spectral bands and transformations against 
the spruce needle metal values produced for 
all Landsat scene dates (except the Septem 
ber 1976 MSS scene) at best only weak to fair 
correlations (r-values between -0.40 and 
-0.65, with the level of significance of the 
higher values only between 90 and 95 per 
cent). Relationships are, for the most part,