22
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
moisture values were observed. These discrepancies can be attributed to different
sensing depths and footprint areas and different sensitivities to soil surface roughness. For GPR, the effect of roughness was excluded by operating at low frequencies (0.2–0.8 GHz) that were not sensitive to the field surface roughness according
to Rayleigh’s criterion. The RMSE between volumetric soil moisture measured by
GPR and TDR was 0.038 m 3 /m 3 . For the radiometer, the RMSE decreased from
0.062 m 3 /m 3 (horizontal polarization) and 0.054 m 3 /m 3 (vertical polarization) to
0.020 m 3 /m 3 (both polarizations) after accounting for roughness using an empirical
model that required calibration with reference TDR measurements (see Jonard et al.
2011 for details). Relatively accurate soil moisture retrievals were possible with the
off-ground GPR and L-band radiometer, although accounting for surface roughness was essential for the L-band radiometer. Future improvements may focus on
the potential radiometer and GPR synergies for improving soil moisture estimates,
to be applied, for instance, in the upcoming NASA’s Soil Moisture Active Passive
mission.
2.4.4 Soil MoiStuRe PRofile chaRacteRization
For most hydrologic and agricultural applications, it is more relevant to characterize the root-zone soil moisture (0–30 cm) rather than shallow surface soil moisture
(0–10 cm; Vereecken et al. 2008). In addition, decoupling of surface and subsurface
soil moisture may occur under various specific conditions such as the case when
considering a wet soil subject to fast evaporation or the propagation of a wetting front
in a dry soil, especially in coarse materials (Capehart and Carlson 1997). In that
respect, the relatively low frequency of the GPR allows a larger penetration depth
than remote sensing instruments. Moreover, owing to the large frequency bandwidth
of the ultrawideband GPR system that we used, information over different depths can
be retrieved from the GPR data.
GPR data inversion accounting for two-layered and continuously varying soil
moisture profile was thus performed with GPR field data acquired over a layered
soil in an agricultural field in Walhain, Belgium (Minet et al. 2011) using the same
off-ground GPR approach as presented above. Following dry conditions, the shallow surface soil was crusted and drier than the subsurface soil. Figure 2.8 presents the two-layered and profile model inversion soil moisture maps. Surface and
subsurface soil moisture maps from two-layered and profile inversions showed, in
general, a coherent soil moisture profile with respect to terrain observations, that is,
soil moisture increases with depth. The subsurface (or second layer) soil moisture
was characterized by a lower spatial coherence, with a larger nugget effect, denoting that retrieved values may be more uncertain than the surface (or first layer) soil
moisture, as outlined by numerical experiments (not shown). The first-layer thicknesses retrieved in the two-layered model inversions were in good agreement with
the depths of inflexion points of soil moisture profiles in the profile model inversions
and were on average, about 4 cm. Except for some particular difference, the surface
soil moisture retrieved by the two-layered or profile model inversions appeared very
similar, whereas the subsurface (or second layer) soil moisture differed between the
two model inversions. When assuming a homogeneous soil medium, the retrieved
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
moisture values were observed. These discrepancies can be attributed to different
sensing depths and footprint areas and different sensitivities to soil surface roughness. For GPR, the effect of roughness was excluded by operating at low frequencies (0.2–0.8 GHz) that were not sensitive to the field surface roughness according
to Rayleigh’s criterion. The RMSE between volumetric soil moisture measured by
GPR and TDR was 0.038 m 3 /m 3 . For the radiometer, the RMSE decreased from
0.062 m 3 /m 3 (horizontal polarization) and 0.054 m 3 /m 3 (vertical polarization) to
0.020 m 3 /m 3 (both polarizations) after accounting for roughness using an empirical
model that required calibration with reference TDR measurements (see Jonard et al.
2011 for details). Relatively accurate soil moisture retrievals were possible with the
off-ground GPR and L-band radiometer, although accounting for surface roughness was essential for the L-band radiometer. Future improvements may focus on
the potential radiometer and GPR synergies for improving soil moisture estimates,
to be applied, for instance, in the upcoming NASA’s Soil Moisture Active Passive
mission.
2.4.4 Soil MoiStuRe PRofile chaRacteRization
For most hydrologic and agricultural applications, it is more relevant to characterize the root-zone soil moisture (0–30 cm) rather than shallow surface soil moisture
(0–10 cm; Vereecken et al. 2008). In addition, decoupling of surface and subsurface
soil moisture may occur under various specific conditions such as the case when
considering a wet soil subject to fast evaporation or the propagation of a wetting front
in a dry soil, especially in coarse materials (Capehart and Carlson 1997). In that
respect, the relatively low frequency of the GPR allows a larger penetration depth
than remote sensing instruments. Moreover, owing to the large frequency bandwidth
of the ultrawideband GPR system that we used, information over different depths can
be retrieved from the GPR data.
GPR data inversion accounting for two-layered and continuously varying soil
moisture profile was thus performed with GPR field data acquired over a layered
soil in an agricultural field in Walhain, Belgium (Minet et al. 2011) using the same
off-ground GPR approach as presented above. Following dry conditions, the shallow surface soil was crusted and drier than the subsurface soil. Figure 2.8 presents the two-layered and profile model inversion soil moisture maps. Surface and
subsurface soil moisture maps from two-layered and profile inversions showed, in
general, a coherent soil moisture profile with respect to terrain observations, that is,
soil moisture increases with depth. The subsurface (or second layer) soil moisture
was characterized by a lower spatial coherence, with a larger nugget effect, denoting that retrieved values may be more uncertain than the surface (or first layer) soil
moisture, as outlined by numerical experiments (not shown). The first-layer thicknesses retrieved in the two-layered model inversions were in good agreement with
the depths of inflexion points of soil moisture profiles in the profile model inversions
and were on average, about 4 cm. Except for some particular difference, the surface
soil moisture retrieved by the two-layered or profile model inversions appeared very
similar, whereas the subsurface (or second layer) soil moisture differed between the
two model inversions. When assuming a homogeneous soil medium, the retrieved