1169 
(SP) 0«wB|« 
where <DN 2 > is the average value of squared digital numbers DN of the site, K a calibration constant given by 
the ERS-1 tape header and p the incidence correction over the scene, depending on the satellite incidence angle 
for the center of the image (a re ^=23°) and for the central pixel of the plot (a mc ). 
3. USE OF ERS-1 SAR DATA FOR SOUL MOISTURE MONITORING 
Most of the radar studies conducted till today within hydrological framework focused on observation over 
agricultural fields in controlled conditions with either ground based sensor (Bertuzzi & Bruckler, 1991), 
airborne scatterometer (AGRISCATT'88 campaign: Prevot et al., 1993b) or more recently spacebom SAR 
(NAIZIN watershed pilot experiment: Loumagne et al., 1994; Le Toan et al. 1993). These studies 
demonstrated the strong dependence of o° upon soil surface properties, mainly moisture and roughness. But 
they demonstrated too that vegetation layer can significantly affect the signal, as expected from general radar 
backscatter expression derived from radiative transfer theory (Attema & Ulaby, 1978): 
a° = a° v + t 2 o° s + a° sv (3) 
where o° v represents the scattering contribution by vegetation volume, o° s the direct soil surface contribution, 
r 2 the two-way attenuation through the vegetation layer and a 0 ^ interactions between soil surface and 
vegetation volume. From above studies, the observed vegetation effects were mainly due to attenuation (r 2 
term) (Le Toan et al. 1993), even in the case of grass covered areas as biomass approaches and exceeds 
1 kg/m2 (Dobson et al., 1992). The specific volume scattering (o°v term) was observed only at higher level of 
biomass (dense crops). Consequently we can expect that, concerning semi-arid conditions, the low amount of 
biomass (< 1 kg/m2, Tab. 1 ) is a great advantage to monitor soil moisture as attenuation should be the only 
vegetation effect to take into account, if significant. 
3.1. o Q sensitivity to soil moisture and roughness and to vegetation properties 
Fig.] shows that the temporal trend of radar backscattering o° on four selected MF sites seems to follow quite 
well rainfall events and the mean surface soil moisture of the watershed. However, when considering a° 
dependence upon soil moisture site by site (Fig. 2), the relation appears to be weaker than expected. 
Fig.l: temporal trend of SAR ERS-1 o° on MF sites. Fig. 2: radar backscattering vs volumetric soil 
Dots stand for rainfall events (mm) and solid moisture, for 4 different dates. Numbers 
line for mean volumetric surface soil moisture correspond to met/lux sites. 
(0-5cm) observed in the waterhed (% sm). 
To explain this behavior, we must stress first of all the small range of soil moisture compared to 
values encountered under temperate climate (like those related in the studies mentioned above) and which 
makes difficult the detection of o° evolution with soil moisture. In fact these low values do not result directly 
from the particular semi-arid climate but from the typical sandy-loam soils of these regions (about 20% loam 
and 70% sand here) for which volumetric moisture can hardly exceed 20-25 % sm (field capacity) whatever the 
rainfalls amount.