Symposium on Remote Sensing for Resources Development and Environmental Management / Enschede / August 1986
Optimal Thematic Mapper bands and transformations for discerning
metal stress in coniferous tree canopies
C.Banninger
Institute for Image Processing and Computer Graphics, Research Centre Joanneum, Graz, Austria
ABSTRACT: A statistical analysis of multitemporal Landsat Thematic Mapper (TM) data acquired
from a pine and a spruce tree stand growing in copper-lead-zinc enriched soils helped define
the TM spectral bands and transformations best suited for detecting metal stress in conifer
ous tree canopies. Based on four Landsat TM scene dates, the first principal component (PC1)
and the band difference BD1 rank the highest of thirty-one single and transformed TM bands
examined, followed by the TM brightness index (TMB) and Thematic Mapper bands 5 and 7.
1 . INTRODUCTION
To effectively employ geobotanical re
mote sensing techniques in mineral explora
tion, a simple and expedient approach is
required that will help define suitable tar
get areas for follow-up examination by ground-
based exploration methods. Plants growing in
soils containing toxic concentrations of
heavy metals normally undergo changes that
alter both their physiological make-up and
their external appearance. These changes to
a plant are detectable by remote sensing
means, and in particular by the Thematic
Mapper sensor system on board Landsat 4 and
5. Both single and transformed TM bands have
been employed to monitor the state of vege
tation growth and the condition of its
health, although these have been applied in
most studies on a singular basis and may not
have been the most suitable band or trans
formation for the purpose or task at hand.
The question still remains as to which The
matic Mapper bands and transformations are
best at defining metal-related stress in
vegetation stands.
To help answer this question - at least
with respect to coniferous forests - thirty-
one single and transformed Thematic Mapper
spectral bands (TABLE 1) for each of four
Landsat scene dates of two mineralised forest-
covered area's in southwestern Spain and south
eastern Austria were statistically evaluated
to determine the most effective TM bands and
transformations for metal stress detection.
The transformed TM bands employed in the
analysis are configured to enhance or better
define important plant-energy relationships
associated with plant growth and health.
A brief description of the test sites,
ground and Thematic Mapper data utilised in
the study, and the analytical and statisti
cal approaches employed in the evaluation
of these data sets is presented in the fol
lowing sections. A more comprehensive des
cription of these topics can be found in
Banninger 1984, 1985a, b, c, and 1986.
2. DESCRIPTION OF TEST SITES, GROUND DATA
COLLECTION, AND THEMATIC MAPPER SCENES
A mature, densely wooded pine and spruce
forest comprise the Spanish and Austrian test
sites. High concentrations of copper, lead,
and zinc occur in the soils at both test sites,
and soil metal isopleth maps derived from the
soil sample data provided the means to relate
the soil geochemistry to the overlying tree
canopies, and subsequently to the Thematic
Mapper canopy radiance values. Soil metal
values for the Spanish test site range from
26-153 ppm for copper, 30-325 ppm for lead,
and 39-205 ppm for zinc; for the Austrian
test site, copper values range from 20-940
ppm, lead values from 10-10,000+ ppm, and
zinc values from 60-6300 ppm. No overt mani
festations of stress are present in the two
test site tree stands.
Thematic Mapper canopy radiance values
employed in the study consisted of January
1983 and August 1984 acquired scene data for
the Spanish test site and June and August 1984
acquired scene data for the Austrian test
site. Atmospheric haze corrections were ap
plied to all scene radiance values before
their computational use.
3. STATISTICAL ANALYSIS AND RESULTS
A linear regression analysis of the two
ground and four Landsat scene data sets pro
vided the basis for evaluating the thirty-
one TM bands and transformations. Landsat
canopy pixel radiance values were regressed
against Landsat pixel metal values, after
first transforming the soil metal data con
tained in the isopleth maps to their corre
sponding TM test site pixels. A dot grid
scaled to the dimensions of the instantane
ous field of view of the Thematic Mapper
sensor system was used to obtain the average
copper, lead, and zinc values for each pixel
position.
Pearson product moment correlation coef
ficients (r-values) for the higher ranking
TM bands and transformations of each of the
four Landsat scene dates extend to a maximum
of r=-0.80 and r=-0.68 for the January and
August Spanish scene dates, and up to r=-0.73
