396
figure 7: Temporal evolution of MSAVI (o), o 0 (x) in db, B ( *) in db, A*50 (+) in db/dg. Scaling factor MSAVI*50-18.
Except for some responses due to rain events, the temporal evolution of these data are similar (figure 7).
The linear regression between cr 0 , B and MSAVI gave for the angle range [30°,40°]:
o 0 = 28.68*MSAVI-19.62 with r 2 = 0.97 and SEE = 0.514
B = 20.51 »MSAVI-14.65 with ^ = 0.7 and SEE =1.15
Where r 2 is the correlation coefficient and SEE the standard error of estimate. For the angle range [40°,50°] we
found:
ct 0 = 23.545*MSAVI-19.47 with r 2 = 0.88 and SEE = 0.54
The relationship with B is not good in that case. On the other hand it appears that, contrary to our expectation,
[30°,40°] range is better adapted to monitoring biomass than [40°,50°]. To understand this point a modeling
effort is necessary. Besides, the results obtained with B are less good. A possible reason may be the
assumptions made (constant roughness and negiigeable influence of soil moisture). Which means that they
should be taken into account
7. USE OF EMPIRICAL MODEL TO RETRIEVE SOIL PARAMETERS : SOIL MOISTURE AND
ROUGHNESS
The empirical model [16] is developed in term of ks (k is the wave number and s the rms height) and soil
moisture through the relative dielectric constant Its validity conditions include angular range above 20° for
smooth surface, soil moisture up to 31(%vol) and roughness parameters (ks) ranging from 0.1 to 6. The
empirical expression for bare soil is :
a 0 (6, e r ,ks) = g*cos 3 0*(7v(0)+Yh(O))/P w
Where g = 0.7*(l-exp(-0.65*(ks) lft )
p 1/2 = l-(2»0/7t) 1/3 * 1r »*exp(-ks)
0 = incidence angle
Ej = Relative dielectric constant [17]
Y 0 = Fresnel reflectivity of surface at nadir.
Y v and Yh, vertically and horizontally polarized Fresnel reflectivities of the surface.
By applying this expression on April May WSC data with ks = 1.2 (we assume homogeneous surface) and sm
(soil moisture) = 0.9 (%vol), we obtained figure 8. Because the model was developed for smooth surface with
0>2O°, we can not use it with 0 outside this range. This explains our simulated results below this angle.
