Soil type: 11.1% Sand
61.7% Silt 27.
2% Clay
S
L
M
SI Field
2.37
160.21
0.0148
OD Field
5.46
64.55
0.085
SL Field
7.69
23.47
0.33
SU Field
18.73
44.80
0.42
Table II: Soil test
site characteristics
S rms height
L correlation length
M = S/L
RESULTS AND DISCUSSIONS
Due to the volume of the data set (from 1.4 to 90.0 GHz) and the lack of soil dielectric model for high
frequencies, we focalise our study on the three first frequencies (1.4, 5.05, 10.65 GHz). The following analysis
is oriented to study the soil moisture influence using smoothest field data SI (study 1), and surface roughness
influence using OD and SU data fields (study 2). Systematic inter-comparisons are made, on the first hand, for
the SI field between Wilheit, Fresnel simulated data and observed data and, on the other hand, between SPM,
PO, GO simulated data and observed data. The simulated brigthness temperatures (or the emissivities) are not
corrected from the atmospheric contribution. The dielectric properties of soil have been computed using the
Dobson semi-empirical model (Dobson et al. 1985), in considering averaged moisture and temperature profile,
for the Fresnel reflectivity, over soil layers approximately l/10th and l/5th of the wavelength respectively, and
soil thickness layers (0-0.5 cm, 0-1 cm, 0-2 cm, 0-5 cm) for the SPM, PO, GO. Emissivity of the soil medium
is obtained from one minus its reflectivity.
Study 1: radiative transfer model
A comparison between simulated TB and experimental TB is done and shown in figures 1 (a, b, c). It can be
seen from these figures that overall, there is a good agreement between experimental data and both Wilheit
and Fresnel models predictions. For both L and X bands, experimental data (except C band) and simulated
data, show the same angular behaviour, indicating the absence of roughness effects. So, this can be an
adequate situation to validate the theoritical model based on the assumption of smooth surface. For 5.05 GHz
the difference observed above 30 degrees of incidence angle may be attributed to the system performances.
The good agreement between the Wilheit model and the Fresnel model may be explained by the good
conditions and associated homogeneous temperature and moisture profiles. In order to study the general
behaviour of the model over a wide range of soil moisture, the TB calculated by the Fresnel model for average
soil moisture over different depths and the Wilheit model, are compared to measured data at 20 degrees of
incidence angle. The figures 2 (a, b, c) show a particular situation in which 2.0,1.0,0.5 cm depths average soil
moisture and temperature are considered, for the Fresnel model at L, C and X bands. The figures show the
sensitivity of the brightness temperature to soil moisture. Both models exhibit the same response in the full
range of soil moisture. However, it is very difficult, from this figure, to make any conclusion on the merits or
demerits of the Wilheit model over the Fresnel one. This could be attributed to the heterogeneity in the surface
soil moisture due to heavy rains for the experimental periods. Therefore, an error analysis of soil moisture
sampling, at different depths is performed, and the 95% of confidence interval obtained on soil moisture
(expressed in cm3/cm3), at the depths 0-0.5, 0-1.0 and 0-2.0 cm are, +/- 0.0445, +/- 0.036 and +/-0.031
respectively. This large error may explain the scatter in the brigthness temperature and soil moisture
relationships. Furthermore, errors in the determination of soil moisture affect the accuracy of the inputs of the
models, and consequently the quality of the simulated TB. This possibly masks the benefits which can be
expected from the Wilheit model in comparison with the Fresnel model. Hence, to bring out the advantage of
the Wilheit model, a regression analysis of the results from Wilheit and Fresnel models with experimental
data, is done. From statistics of table III, it can be concluded, that for L band, the Wilheit model predicts more
reasonable results, and the soil moisture gradient effect is more evident, also. In the case of X band, the
Fresnel model with 0.5 cm average depth gives better statistical results than the Wilheit model. This change
might have come from the extrapolation of surface soil moisture in the calculation of the radiative model and
the heterogeneity of surface soil moisture. In the case of 1.4 GHz, the Fresnel model gives the best statistiscal
results at 2.0 cm (which supports the earlier studies on sampling depths). Finally, the good performances of the
Wilheit model in L band confirm very clearly, the advantage of radiative transfer model which accounts for
soil moisture gradients in the TB calculation.
529