Tal Svoray
GVD are the green leaf volumetric densities of the three vegetation formations as formulated in the study of Svoray et
al. (2000). The use of the green leaf biomass volumetric density rather than the green leaf absolute biomass provides a
better representation of the backscatter contributors in the different vegetation canopies. Green leaf biomass [kg m?]
values of the GVD were received from the above mentioned field campaign. The soil moisture coefficients and the
local incidence angle were considered as described in the general form of the water cloud model. The mixed proposed
model can be more clearly illustrated and explained using figure 2.
The difficulties involved with modeling the backscatter
from a mixed pixel of herbaceous and woody formations
are mainly related to different scattering mechanisms of the
two canopies. Due to the low penetration depth at C-band,
the interaction of the radar beam with the surface includes
the soil moisture effect in the case of herbaceous vegetation
but involves only the uppermost layer of the canopy in the
case of the woody formations. Thus, contribution of soil
moisture to backscatter of a mixed pixel is only from the
parts of the pixel that are covered with herbaceous
vegetation. The other parts of the pixel, covered with
woody formations, contribute only from the upper part of
the canopy. The solution provided by us is to include the
soil contribution only in the areas covered by the Figure2. An Illustration of backscatter from a
herbaceous vegetation. mixed pixel.
4 RESULTS AND DISCUSSION
This section describes the linear relationship that were found between ERS-2 SAR backscatter and volumetric soil
moisture in the study area; multi-temporal ERS-2 SAR backscatter curves of the dominant vegetation formations in
the study area; and finally results of applying the water cloud model in homogenous and heterogeneous plots are
discussed and the predicted values of the inversion models are being assessed.
Examination of the relationship between volumetric soil moisture and ERS-2 SAR backscatter have shown a strong
linear correlation between volumetric soil moisture and ERS-2 SAR backscatter. Statistical tests have proven
(0=0.05) that this relationship did not vary along the climatic gradient. This similarity between the linear soil moisture
models of a humid and arid sites and other models that were applied to different environments around the world, have
lead us to conclude that a unified model (figure 3) could be applied for the entire climatic gradient region.
The results in the unified model illustrate high level
of scattering in points at the lowest levels of
volumetric soil moisture (between 0% to 10%) and
at the highest levels of moisture (between 25% to
40%). The points in between these ranges are
characterized by much more closeness to the
regression line. This result may imply on the
sensitivity of the ERS-2 SAR backscatter and the
linear soil moisture model to changes in the soil
moisture concentrations. Thus, although the results
in the upper and lower parts of the plot are
reasonable and the total coefficient of determination
Backscatter [dB]
is high (0.9), it appears that the best range to be 0 M 20 an
represented by the soil moisture model is the Volumetric Soil Moisture [%]
between 10% to 25% of volumetric soil moisture
concentration. This range, although is relatively Figure 3. The relationship between the ERS-2
short, is very meaningful for the analysis of water SAR backscatter and the volumetric soil
Stress in semi-arid areas. moisture along the climatic gradient.
A study of ERS-2 SAR backscatter from canopies of the dominant vegetation formations in the study area (trees,
shrubs, dwarf shrubs and herbaceous vegetation) prove a clear difference between the multi-temporal curves of
backscatter from woody formations (trees, shrubs and dwarf shrubs) and the multi-temporal curves of herbaceous
vegetation (grasses and wheat) canopies. In general, the woody formations provide relatively stable multi-temporal
324 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part Bl. Amsterdam 2000.
CU
th
th
rac
pk
vel
for
lov
mu
rel
the
the
wa
hay
are
WO
shr
the
inv
me