Skógræktarritið - 15.05.2001, Qupperneq 154

Fig. I. A: The study area on Southern
and Northern sides of the Lake
Tornetrásk. The test site consisted of
plots grouped into clusters (white
squares) with 500 m or I km between
each other. B: Most clusters consisted
of 9 circular plots (solid linesl.'but a
few had four extra plots inside the
cluster (dotted line) and two plots 500
m to the east and west from the cen-
tral plot (not visible).
three Landsat TM scenes. On the
other hand they found significant
relationships when the satellite
data were stratified into vegeta-
tion index classes and related to
an average biomass from ground
data.
In many large area studies
Landsat TM data are commonly
used. However, nowadays there
exists a range of optical satellite
data, which can potentially be
very useful. For example, Tiwari
(1994) used all bands from ÍRS
LISS (Indian Remote Sensing
Program, Linear Imaging Self-
Scanning Sensor) data to classify
different crown cover classes.
Allometric functions were esti-
mated between crown cover and
biomass using a log-linear model
with R2 values around 0.97. The
accuracy obtained with this
method is dependent on the
crown cover classification and
the allometric model between
crown cover and biomass.
One of the problems of using
remote sensing in mountainous
and high latitude areas is that
the topography and low sun
angle will cause differences in
illumination of slopes in differ-
ent directions. The effects of
topography on classification and
in extracting estimates from
satellite images are well docu-
mented and several topographic
correction models have been
suggested (Teillet et al. 1982,
Parlow, 1996 Gu and Gillespie
1998, Gu et al. 1999). In most
cases the cosine of the incidence
angle (cos(i)) is used as a correc-
tion factor to reduce the effect of
different illumination of slopes
in different directions (aspects).
Another more empirical approach
is to include the product of the
sine of slope with the sine of
aspect (sin(slope) x sin(aspect))
as an interaction term in the
regression function. This would
correct for the extra variation due
to the relationship between
topography and satellite data.
Other problems are that for
periods there are no good satel-
lite images from any optical
satellite available. This relates to
the short vegetation period in
the Scandinavian Mountains,
which is typically only about two
months, which often are cloudy.
The aim of this study was to
establish the possibility to use
satellite data for estimating bio-
mass of mountain vegetation, in
an area in Northern Sweden. For
Fig. 2. The forest on the south side of
Lake Tornetrásk.
this attempt, regression func-
tions were estimated using IRS
LISS 111 data and ground data
from a test site. The topographic
variables sin(slope), sin(aspect),
the interaction term (sin(slope) x
sin(aspect)), and the elevation
were also tested in the regres-
sion. The IRS data were chosen
because it was a cloud free scene
that covered the whole test site
during the short vegetation
period.
Materials and Methods
Study area
The study area was located in a
mountainous area in northern
Sweden (Latitude 68°20' N,
Longitude 18°50' E) on the
Southern and Northern sides of
Lake Tornetrask (Figure 1A). The
hills at the southern area.pre-
dominantly slope to the north,
whereas the steeper hills at the
northern area slope to the south
and west. The vegetation on both
sides of the lake was predomi-
nantly heath and open mountain
birch forest (Betula pubescent ssp.
aerepanovii) (Figure 2), but on the
Northern side the mountain
birch forest was richer and con-
sisted of tall herb meadows with
a few relatively large birch trees
(Figure 3). The tree line was
152
SKÓGRÆKTARRITIÐ 2001 l.tbl