One Spa-
ls in the
value to
pixel in
Ss present
of looks
g image,
S, With a
lata, but
ring the
lata. The
T tO con-
istics on
for very
vhich are
ese scal-
accuracy
sing step
ature re-
speckle
tures de-
filtering
eckle re-
looks. It
xceeding,
)geneous
POT-XS
ed using
he DEM
g a set of
:stimated
.e. about
had been
.
[TH
bility of
onditions
al., 1993,
taken by
est areas
yrrelation
d mean
, agricul-
cipitation
days be-
haracter-
uld allow
tions on
es.
| to react
f correlà-
| whereas
relation
| the tim
e, as Can
- average
precipitation of 2 or 4 days compared to the grassland and the
forested areas.
Meanof | 2Days | 4Days | 10 Days | 20Days
: 0.20 0.39 0.64 0.70
-0.05 0.46 0.62 0.69
0.36 0.58 0.69 0.83
-0.18 0.00 0.38 0.32
Table 1: Correlation Coefficients Between Backscatter and Pre-
cipitation Averaged on Various Time Frames Before Each ERS-
1 Acquisition.
Table 2 summarises the results from linear correlations between
temperature at various times of the day and mean backscatter
values of deciduous and coniferous forest stands. The ERS-1
backscatter of the three coniferous stands exhibit strong correla-
tions to temperature, with maximum correlation coefficients to
the last temperature measurement (09h00 GMT) before the ERS-
] acquisition (10h20 GMT). On the other hand, the mixed de-
ciduous forest stand shows no significant correlation of ERS-1
backscatter to temperatures.
A theoretical study (Wegmiiller et al., 1994) shows that C-band
backscatter in ERS-1 configuration of coniferous trees decreases
with decreasing temperature, leading to the high positive correla-
tions in Table 2. This study also allows the backscatter behaviour
of deciduous trees to be explained as the result of two opposite
mechanisms. In late autumn and winter an increase of backscatter
due to leaffall might be balanced by the decrease of backscatter
due to change in trees' dielectric properties with decreasing tem-
peratures. As a result of these opposite effects, no significant
correlations are found in Table 2 for deciduous trees.
OOUh | I2h | 15h
0.28 | 0.24 | 020
0.87 | 0.35 0:82
0.82 | 0.80 | 0.77
0.77 | O75 | 071
Table 2: Correlation Coefficients Between Backscatter of a
Mixed Deciduous, two Scot Pine and a Douglas Fir Stand and
Temperatures at 6 h, 9 h, 12 h and 15 h GMT.
3. VEGETATION AND FOREST DYNAMICS USING
ERS-1
Grassland, agriculture and forest, vegetated areas characterising
a landscape, were used to study phenological and seasonal influ-
ences on ERS-1 backscatter evolution from which typical exam-
ples are presented. As can be seen from the temporal ERS-1 sig-
natures in Fig. 1, discrimination of grassland (in this case natural
vegetation) and agriculture is possible from October onwards.
From the knowledge of local agricultural practice, the increase of
backscatter for the agriculture area can be attributed to harvest-
ing and soil treatment, which result in bare soil conditions during
the winter period.
The same is also valid for the differentiation of forest and agri-
culture (Fig. 2), even if test areas are considered with a mixing of
Sighatures during the summer period.
c
=
-6
=
9° -8
o
z
9 -10
E
=
o
—16
211: 229 232 235 262 274: 292-304 310 (319 53473401123 (1/58
Day of Year
Fig. 1: Temporal Sequence of ERS-1 Backscatter for Agriculture
and Grassland Test Areas.
Sigma Nought [dB]
211 2295232 235 1262 271 292,304 510+ 319-334 340 125 38
Day of Year
Fig. 2: Temporal Sequence of ERS-1 Backscatter for Forest and
Agriculture Test Areas.
[dB]
Sigma Nought
=15
211 229 232.235 262 271 292 304 310-319. 334 340x125 © 38
Day of Year
Fig. 3: Temporal Sequence of ERS-1 Backscatter for Deciduous
and Coniferous Forest Test Areas.
Fig. 3 shows the comparison of the temporal ERS-1 signature of
a mixed deciduous and a coniferous (spruce) stand. There is no
clear indication of any effect of leaf fall on backscatter of de-
333
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
