9 Soil Moisture
205
ence issue is under a great deal of investigation, because relation 9.6 is violated
under certain conditions of roughnesses and incidence angles. Wang et al. (1980)
assumed the random roughness was independent of incidence angle and simplified
Eq. 9.6 to
R = Ro exp(- h)
(9.10)
More recent work by Promes et al. (1988) has shown that, for L-band, assuming
smooth field emissivity for most dry land agricultural conditions in which the row
height to row spacing is less than 2 will result in an error ofless than 3%.
Theis et al. (1986) have demonstrated the possibility of using a multisensor approach for improving the estimates of soil moisture under field conditions. In this
case, the effects of surface roughness were accounted for with scatterometer measurements. These were then used in a soil moisture equation which included terms
related to the emissivity measured by the radiometer and to the scatterometer
roughness term. Inclusion of the roughness term improved the r'2 values from 0.22
to 0.65 for C-band and from 0.69 to 0.95 for L-band.
Although roughness may not be a serious limitation for passive sensors, at least
for most natural surfaces, it is a major factor for radar. In many cases the effects of
roughness may be equal or greater than the effects of soil moisture on the backscatter. Thus the soil moisture problem becomes one of determining the roughness
effect independently so that a model can be inverted to yield a measure of soil
moisture.
The role of surface roughness in soil moisture estimation for the active case
needs to be understood through surface scattering processes. The theoretical work
on surface scattering can be divided into three categories: The small perturbation
model (SPM), the physical optics model (POM) and geometrical optics model
(GOM). In a broad sense, the geometrical optics model is best suited for a very
rough surface, the physical optics model is suitable for surfaces with intermediate
scales of roughness, and the small perturbation model is suitable for surfaces with
short correlation lengths. Figure 9.4 describes the region of validity for the three
models in terms of KI and Kcr, where K is the wave number, I is the correlation
length, and cr is the standard deviation of the surface roughness heights. The
mathematical expressions to calculate surface backscatter using these models and
their regions of validity in terms of RMS height, correlation length and wavelength
can be found in Ulaby et al. (1982). An examination of these surface backscattering expressions employing different scattering models shows that even though the
backscatter increases due to the increase of surface roughness, the soil moisture
sensitivity to backscatter diminishes due to sharp rate of decrease in the value of
reflectivity. As a result of two competing effects, the roughness effects overshadow
the soil moisture effects.
Unfortunately, even if roughness data are available, Oh et al. (1992) have shown
that the typical values of Kcr and KI found in the natural environment result fall in
the area outside of the various models regions of validity (Fig. 9.4). Consequently,
most people have had little success using these models.
However, based on the scattering behavior in limiting cases and experimental
data, Oh et al. (1992) have developed an empirical model in terms of the rms
205
ence issue is under a great deal of investigation, because relation 9.6 is violated
under certain conditions of roughnesses and incidence angles. Wang et al. (1980)
assumed the random roughness was independent of incidence angle and simplified
Eq. 9.6 to
R = Ro exp(- h)
(9.10)
More recent work by Promes et al. (1988) has shown that, for L-band, assuming
smooth field emissivity for most dry land agricultural conditions in which the row
height to row spacing is less than 2 will result in an error ofless than 3%.
Theis et al. (1986) have demonstrated the possibility of using a multisensor approach for improving the estimates of soil moisture under field conditions. In this
case, the effects of surface roughness were accounted for with scatterometer measurements. These were then used in a soil moisture equation which included terms
related to the emissivity measured by the radiometer and to the scatterometer
roughness term. Inclusion of the roughness term improved the r'2 values from 0.22
to 0.65 for C-band and from 0.69 to 0.95 for L-band.
Although roughness may not be a serious limitation for passive sensors, at least
for most natural surfaces, it is a major factor for radar. In many cases the effects of
roughness may be equal or greater than the effects of soil moisture on the backscatter. Thus the soil moisture problem becomes one of determining the roughness
effect independently so that a model can be inverted to yield a measure of soil
moisture.
The role of surface roughness in soil moisture estimation for the active case
needs to be understood through surface scattering processes. The theoretical work
on surface scattering can be divided into three categories: The small perturbation
model (SPM), the physical optics model (POM) and geometrical optics model
(GOM). In a broad sense, the geometrical optics model is best suited for a very
rough surface, the physical optics model is suitable for surfaces with intermediate
scales of roughness, and the small perturbation model is suitable for surfaces with
short correlation lengths. Figure 9.4 describes the region of validity for the three
models in terms of KI and Kcr, where K is the wave number, I is the correlation
length, and cr is the standard deviation of the surface roughness heights. The
mathematical expressions to calculate surface backscatter using these models and
their regions of validity in terms of RMS height, correlation length and wavelength
can be found in Ulaby et al. (1982). An examination of these surface backscattering expressions employing different scattering models shows that even though the
backscatter increases due to the increase of surface roughness, the soil moisture
sensitivity to backscatter diminishes due to sharp rate of decrease in the value of
reflectivity. As a result of two competing effects, the roughness effects overshadow
the soil moisture effects.
Unfortunately, even if roughness data are available, Oh et al. (1992) have shown
that the typical values of Kcr and KI found in the natural environment result fall in
the area outside of the various models regions of validity (Fig. 9.4). Consequently,
most people have had little success using these models.
However, based on the scattering behavior in limiting cases and experimental
data, Oh et al. (1992) have developed an empirical model in terms of the rms