The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
79
soil moisture and ASAR backscattering, and soil moisture and
PALSAR backscattering is compared. While processing the
SAR data, the resolution of the images was resampled to 8 m
for RADARSAT-1 and PALSAR data and to 30m for ASAR
data. SAR data has different resolutions. The resolution of
RADARSAT-1 and PALSAR data is three times better than the
ASAR data. There is a pretty high correlation (R 2 = 0.86)
between the soil moisture content and the PALSAR
backscattering, which is the best among all the SAR images
(Figure 4a). RADARSAT-1 gives better result (R 2 = 0.81) than
ASAR (Figure 4b). ASAR is the worse (R 2 = 0.76) among the
all (Figure 4c). This result is related not only with soil moisture
content but also with soil texture. In addition, due to the fact
that PALSAR has the biggest wavelength causing to penetrate
deeper, gives higher correlation with the moisture content of the
soil. The significance of linear association was tested for all
SAR backscattering results from different satellites using
Fisher’s F test for a = 0.05 significance level, and all were
found significant (for the PALSAR F=431,5706952 and
significance F=2,09369E-31, for the RADARSAT
F=380,7293985 , significance F=2,81845E-34, for the ASAR
F=244,5656022, significance F=l,07252E-24).
In investigating the soil moisture change, the spatial resolutions
of the each C band SAR data affect the average soil moisture
value of the represented area. However, the polarisation does
not affect the result significantly since both RADARAT-1 (HH
polarised) and ASAR (VV polarised) give a good (i.e.
acceptable) correlation with the soil moisture. With regard to
the band difference; despite the resolutions of both images were
the same, PALSAR (L) band gave better correlations than that
of RADARSAT-1 (C band) for the same soil characters.
Comparing the graphs, linear association from PALSAR
analysis is about 5 % better than RADARSAT-1 analysis. It is
also better than ENVISAT analysis by about 10 %.
ASAR (C Band - W| 08 June 06 a
Backscattering ln dB (cT°)
RADARSAT (C Band-HH) 28 May 06 fc>
Backscattering in dB (O 0 )
PALSAR (L Band-HHj -10 June 06 C
Figure 4. Variation of backscattering coefficient with
gravimetric soil moisture a) for ASAR b) for RADARSAT-1
and c) for PALSAR
6. CONCLUSIONS
In this study the potential of RADARSAT, ASAR, and
PALSAR data was investigated for estimating soil moistures
over bare soils in Menemen Plane of Western Turkey.
RADARSAT, ASAR and PALSAR images are collected on 28
May 2006, 8 June 2006 and 10 May 2006 respectively. In most
of the studies comparisons were made between the multi
polarisations or multi incidence angles of the same sensor or
comparisons made between the ASRAR and RADARSAT-1
data. In this study, the estimates of soil moisture obtained from
the SAR data of three different satellites for HH and VV
polarisations and for C and L bands are compared to find the
best sensor for measuring the bare soil moisture in agricultural
areas as in Menemen Plain.
It has been found out that for the bare soils having a relief of
less than 1% and having no stoniness, the correlations between
the soil moisture content and backscattering of ASAR,
RADARSAT-l,and PALSAR images were 76%, 81% and 86%
respectively. The RADARSAT-1 Fine Beam image has the
same resolution (6.25m x 6.25m) with the resolution of
PALSAR image (6.25m x 6.25m). Although they both were
resampled to 8m, PALSAR gave better correlation than
RADARSAT-1 image. Although the resolution of
RADARSAT-1 and PALSAR images is far higher than that of
the ASAR image (30mx30m), the significance of the results
produced is almost similar in such a flat area. Despite the
spatial resolution difference and polarization differences of
RADARSAT-1 and ASAR images (8m-HH versus 30m-VV),
the estimated soil moistures show high correlation with
backscatter values for the both image types. From the point of
view of monitoring and mapping the soil moisture content of
agricultural fields; for the areas having larger fields, both SAR
images can be utilized almost equivalently whereas for the
areas having smaller fields RADARSAT-1 image gives better
results. Regarding to the band difference, RADARSAT-1 (C
band) and PALSAR (L band) images having the same
resolution are compared, and as expected PALSAR data gave %
5 better estimation than the RADARSAT-1 data. This means L
band resulted in the best correlation between the ground soil
moisture and the estimated soil moisture.
Baghdadi et al. (2006) found out in their study that ASAR
sensor does not seem to offer any advantage compared to the
mono- polarization and multi incidence RADARSAT -1 sensors.
They studied HH and HV polarizations for different incidence
angles and recommended to study VV polarisation to see the
potential of ASAR versus RADARSAT in estimating soil
moisture. In this study it has been revealed that VV polarisation
of ASAR sensor does not prove any better to RADARSAT-1
sensor. If the price differences per scene are taken in to account,
one should consider purchasing ASAR (VV) for the vast
acreages. In this sense, 5% difference in R 2 values of ASAR
and RADARSAT-1 correlations do not show any significance.
Among all the sensors, the PALSAR is the most cost effective
data and gives the best moisture estimation in bare soil. For the
soils having the same characteristics as with the Menemen Plain,
PALSAR images can be preferred for moisture estimations.
This study revealed that the PALSAR which is designed to