the
ean
nding of
erranean
work, we
erranean
d ERS-2
1978) to
le by the
ly good
S. In the
sulted in
effective
regions.
ion with
ests; and
erranean
bjectives
climatic
ve to the
biomass
context.
Tal Svoray
Two specific study sites were selected to represent the extreme conditions of the climatic gradient in the study area:
Avisur site representing the northern and humid zone and Lehavim site representing the more southern and arid zone.
The remote sensing data used in this study include three ERS-2.SAR.PRI C-band images, acquired at February, April
and May 1997. These dates represent the vegetation status in the study area at the beginning, peak, and the end of the
winter season and thus provide a variety of phenological phases of annual herbaceous vegetation. In addition to the
remote sensing data, a detailed field campaign was carried out simultaneously with the satellite overpass.
Measurements of gravimetric soil moisture (at the top 5 cm) using the double weight technique were taken and
herbaceous vegetation green biomass, was measured randomly along 100 meter transects using the harvest and
assessment method (Tadmor et al. 1975).
3 METHODS
The remote sensing methodology proposed here includes a calibration of the ERS-2.SAR.PRI image DN to sigma
naught (decibels and power units) considering both viewing geometry and local topography. In addition, correction of
the image’s file coordinates to Real World coordinates (Israel New Grid) was carried out based on 250 ground control
points and a first order algorithm with a root mean square error of less than one pixel. The derivation of backscatter
(6^) from the DN value of ERS-2 SAR PRI image (equation 1) was executed based on the method of Laur et al.
(1997):
c 9 = x DN i [ee sin c (1)
N j-i k SIN OL er
where N is the number of pixels within an area of interest; DN is the ERS-2 SAR image digital number and the
average in the square parenthesis is calculated following the application of a mean filter with a 3x3 pixel window size
to reduce speckle effects; K is the processing center specific calibration constant; C accounts for updating the gain due
to the elevation antenna pattern implemented in the processing of the ERS SAR PRI data products and o, and os are
the mid-range and reference incidence angle respectively. Since the study area is hilly, backscatter was adjusted for
variations in the local angle of incidence, derived from a digital elevation model (Shoshany et al. 1998).
The image preprocessing enabled to derive ERS-2 SAR backscattering coefficient values from surface covered by the
four dominant Mediterranean vegetation formations. The image sampling procedure was executed at 50, relatively
homogenous, plots per vegetation formation per date. The understanding of the backscattering mechanism from the
Mediterranean vegetation formations has lead us to apply the semi-empirical water cloud model (Attema and Ulaby
1978). This model was applied earlier for vegetated surfaces based on ERS-2 SAR data but was mainly used for
agricultural crops (Xu et al. 1996). The general form of the water cloud model is described in equation 2:
c°= Af(L)cos 6(1-exp(-2BL/cos 0))+(C+ Dm,)exp(-2BL/cos 0) (2)
where 6° is the radar backscatter [m°/m°]; cos 6 is the cosine of the incidence angle; L is a canopy descriptor such as
LAI or biomass; and (C+Dm,) is the soil contribution which can be derived from the linear soil moisture model and
converted to power units [m"/m^]; A and B are the canopy coefficients and should be adapted empirically to each
vegetation layer. The determination of the canopy coefficients was executed by a non-linear least square regression
code based on the E04FDF procedure of the NAG libraries. The initial points were set to 0.005 for both coefficients
based on previous works that determined the canopy coefficients for agricultural crops such as wheat (Prevot et al.
1993) and sugar beet (Xu et al. 1996).
This form of the water cloud model could be used for homogenous areas but is not suitable for the heterogeneous
environment that characterize the Mediterranean region. In order to adapt the water-cloud model for Mediterranean
environments we propose to include in the model the unmixing method of Shoshany and Svoray (2000) which is
based on a multi-temporal multi-spectral approach. This adaptation has enabled us to receive the contribution of each
of the vegetation formations to each pixel. Our suggestion for the application of the water-cloud model in
heterogeneous areas is to incorporate the cover fraction assessments applied on Landsat TM images with the general
form of the water cloud model. Such a an incorporated model can use the form described in equation 3:
0° = Aficos 0(1-exp(-2B,GVDy/cos 0)) + Agf;cos O( l-exp(-2B4GVDycos 0)) + (3)
Anfncos 60(1-exp(-2B,GVDy/cos 6)) +f,(C+ Dm,)exp(-2B,GVDy/cos 6)
where different A and B canopy coefficients are related to the different three vegetation formations — shrubs, dwarf
shrubs and herbaceous vegetation; different f are the cover fractions of the three vegetation formations and different
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part Bl. Amsterdam 2000. 323