1692

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 42, NO. 8, AUGUST 2004

(plain curves). We first see that when considering

firm that the phase difference is a function of the dielectric

increases with the

very high values of rms height (rough interface), multiple re-

contrast as previously mentioned:

moisture content of the paleosoil. We can also see in Fig. 14(c)

flections occur and lead to an isotropic distribution of the radar

then decreases to

occurs for different

and (d) that the maximum value of

backscattering: the phase difference

sand thicknesses when the rms height  and the permittivity of

zero.

increases with respect to the permittivity

[Fig. 14(c)], the

the paleosoil change. For

As expected,

when considering incidence angles of 20 and 30 . For higher

maximum of the phase difference corresponds to a sand layer

incidence angles of 40 and 50 , the value of

seems to

cm and for

thickness of 2.1 m for

[Fig. 14(d)], the sand thickness is about 2.6 m for

cm.

be quite independent of the paleosoil moisture content. For an

incidence angle of 40 corresponding to the RAMSES acquisi-

The latter value is close to the 2.9-m sand layer thickness

is close to 17 for a

tion, FDTD simulations show that

observed for the Pyla dune in Fig. 4.

paleosoil roughness

cm, indicating that a phase differ-

Differences in the value (17 versus 23 ) and location (2.6 m

ence between both H and V polarizations at the sand­paleosoil

versus 2.9 m) of the maximum of the copolarized phase differ-

between our model and observations could be

interface occurs and must be taken into account.

ence

obtained from FDTD

due to vegetation remains inside the paleosoil layer that could

We then have to take into account

simulations in the two-layer IEM we proposed in Section V-B:

produce a permittivity gradient and/or moisture heterogeneity

it allows us to add the multiple-scattering component of the

effects. According to the WBK model [37], a single reflection on

sand­paleosoil interface to our single-scattering IEM model.

a permittivity gradient changes the phase of the incident wave.

for the horizontal polar-

We shall then consider a value

Thus, the two-layer single-scattering IEM model reproduces

ization obtained from (10) and the value

for the vertical

the observed phase behavior fairly well when combined to

polarization is obtained using

FDTD simulations that provide the multiple-scattering contri-

bution of the buried paleosoil layer. It also confirms that the

(26)

phase difference observed on L-band SAR images of the Pyla

dune is mainly due to the wet sand­paleosoil interface.

The new copolarized phase difference is then given by

VII. CONCLUSION

The objective of this work was to investigate and model the

capability of L-band SAR systems to penetrate soils to retrieve

(27)

information about subsurface moisture. The Pyla sand dune was

chosen as an experimental test site as it allows high radar pene-

Results are presented in Fig. 14 where new values of the

tration and presents large subsurface wet structures (paleosoils)

were computed for

copolarized phase difference

at varying depths. By analyzing RAMSES L-band (1.6 GHz)

two permittivity values and for rms height values in the range

SAR data of the Pyla dune, we established that a phase signal

corre-

[1.6­4.0 cm] at a local incidence angle of

is correlated to the buried wet paleosoils: a phase difference be-

sponding to the RAMSES data acquisition. We assumed that

tween HH and VV channels reaching 23 was clearly observed.

value is close to zero at the point where the paleosoil

the

This phase signature also allows detection of paleosoils down

outcrops (it is there dry) and then linearly increases up to the

to a larger depth than when only considering HH and HV am-

FDTD computed value for a covering sand layer of thickness

plitude signals. In order to confirm this result, field measure-

distribution for two

1 m. Fig. 14(a) and(b) presents the

ments were performed, which led to the same observed phase

[Fig. 14(a)]

values of paleosoil permittivity:

difference. We could fit our observations to the semiempirical

[Fig. 14(b)] and for a paleosoil rough-

and

model proposed by Oh and Sarabandi and reproduce the ob-

cm. Compared to Fig. 10(a) and (b), we can see

ness of

served phenomenon using a two-layer single-scattering IEM

that the phase difference spectrum is wider when considering

model that was completed by the results of FDTD simulations.

values and that the maximum value for

is also

Our model has shown that the overlying sand layer weakly con-

profile for each

higher. We computed the average

tribute to the phase difference and confirmed that the soil mois-

permittivity [smooth curves in Fig. 14(a) and (b)]: we obtain a

ture significantly influences the radar response in terms of the

maximum phase difference of 17 for

and

phase difference between the copolarized modes. The proposed

18.4 for

. These results indicate that the

two-layer model reproduces the observed phase difference fairly

multiple scattering that occurs at the sand­paleosoil interface

well when combined to FDTD simulations, which provide the

should be taken into account in the values of the backscattering

multiple-scattering component of the sand­paleosoil interface.

and

in (15) and (18), which could then

coefficients

This phase signature could be used as a new tool to map subsur-

significantly modify the copolarized phase difference obtained

face moisture in arid regions.

from (19).

also increases

The copolarized phase difference

ACKNOWLEDGMENT

with the rms height of the paleosoil surface as shown in

The authors would like to acknowledge the French National

Fig. 14(c)­(e), while FDTD simulations showed that the phase

from the paleosoil decreases as the paleosoil

Program for Remote Sensing (INSU/PNTS) for financial sup-

difference

roughness increases. Results presented in Fig. 14(e) also con-

port, ONERA/DEMR for providing RAMSES data, and H.