presence of vegetation is to attenuate soil emission and to increase depolarization of the global land 
surface microwave emissivity (PD and PR decrease). So, depending on the specific conditions of the 
study, PD or PR could be related to the Leaf Area Index (Paloscia and Pampaloni, 1992), or to the 
vegetation cover density (Choudhury and Wang, 1990). Teng et al. (1993) used PR to parameterize the 
influence of the vegetation cover on the linear relationship between soil moisture and the brightness 
temperature T Bp . Yet, for surfaces with sparse vegetation, PD and PR are also strongly sensitive to soil 
moisture and to soil roughness conditions (Choudhury, 1989). Wigneron et al. (1993a), suggested that 
the use of PD or PD p (PD p =PD*TB p , p=v or h), is less sensitive to steep soil moisture variations than 
PR and improve vegetation growth monitoring. So, PR, PD or PD p can be most useful as qualitative 
indicator of vegetation density, but in most applications their use to actually achieve quantitative 
retrievals of vegetation cover is limited. 
3.2. Approaches based on forward model 
Several studies rely on parametric forward models to retrieve land surface features. The parameters 
may not be directly related to any geophysical parameters of interest, but they are necessary to fit the 
forward model to the actual microwave measurements. They may be calibrated prior to the retrievals 
using ground measurements, or may be retrieved simulteanously to the geophysical parameters of 
interest. 
Most models simulating vegetation microwave emission are based on radiative transfer equations (Kerr 
and Wigneron, 1993). At low frequency, the vegetation canopy can be modeled as a lossy dielectric 
medium with negligible scattering characterized by the opacity x (Matzler, 1990; Paloscia and 
Pampaloni, 1992) or include scattering effects using both parameters x and go (Mo et al, 1982; Kerr and 
Njoku, 1990). The vegetation opacity x is related to the vegetation water content W (Kg/m 2 ), using 
semi-empirical formulations: linear (Kirdyashev et al., 1979, Wegmiiller et al, 1993), logarithmic 
(Pampaloni and Paloscia, 1986) or parametric functions (Schmugge and Jackson, 1992). More simply, 
x can be related to W using a single parameter b, (Jackson et al., 1982; Matzler, 1990; Jackson and 
Schmugge, 1991): 
x= b.W (2) 
So, when using a very simple modeling approach for the vegetation canopy microwave emission 
(using the h.Q-model for soil emission, the x.co-model and equation (2) for vegetation emission), at 
least seven parameters must be taken into account: surface temperature, soil roughness parameters 1\ 
and Q, soil moisture m v , vegetation opacity x and synthetic parameters co and b. Different approaches, 
very specific to the conditions of the study, can be used to perform retrievals of geophysical features. 
When repeated observations are carried out over typical vegetation canopy, prior knowledge of b, 0), or 
other synthetic parameters can be obtained using ground measurements. Jackson and Schmugge (1991) 
found that in the L band range (18-21cm), a single value of b might be used regardless of cover type 
(b=0.15). At higher frequencies, they investigated the functional dependence of b on wavelength and 
on canopy type using published data. Estimates of the single scattering albedo co have also been done 
for several crop types (Brunfeldt and Ulaby, 1984; Pampaloni and Paloscia, 1986). In a recent study, 
Van de Griend and Owe (1993), used large scale knowledge of soil moisture to retrieve simulteanously 
co and x from SMMR data (figure (2)). Considering the size (100x100km) of the vegetated area which 
is observed by the spacebome microwave sensor, the relationship between retrieved values of (co,x) 
and actual vegetation characteristics is unknown. This important problem is yet to be adressed. 
Beyond 3-6GHz, scattering effects increase inside the vegetation canopy. So in radiative transfer 
formulations, the phase matrix P which model multiple scattering effects has an increasing importance.