The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
78
measured with a hand held GPS and added to georeferenced
SAR images as an attribute (Figure 2). Samples were collected
before or just after the seeding. In some fields cases it was
noticed that the farmers increased the density of the surface soil
by flattening with force in order to prevent soil moisture loss
before the planting season.
Soil samples were gathered using 100 cm3 metal cylinders, and
soil moisture levels were detected using a gravimetric method
in the laboratories (Black, 1965). Soil texture analyses were
made using the Hydrometer Method in order to correlate
reflection values with the soil moisture in collected soil samples
(Bouyoucous, 1951 and Soil Survey Staff, 1993). Fresh of soil
samples were weighed and noted. Then soil samples were oven
dried for 24 hours at 105°C temperature. Dried samples were
weighed once again for the dry weight. Gravimetric soil
moisture values were calculated for the 240 sample points (i.e.
72 points for PALSAR, 74 for ASAR and 94 for RADARSAT)
with the help of fresh weight and dry weight samples. Also clay
and sand contents of each sample were calculated. During the
field works the characteristics of the soil such as stoniness,
roughness and surface relief were noted carefully.
Sigma nought values for RADARSAT and ASAR images were
calculated using PCI Geomatica software. Sigma nought values
were taking the local incidence angle at that pixel position in
the range direction. Backscatter values for the PALSAR image
were used.
Wl .
The relation between the bare soil moisture and each type of
SAR data was investigated by calculating correlation
coefficients. For each sample point, instead of using the
corresponding pixel value of the actual ground point coordinate,
a 9 x 9 kernel window was used to calculate the average
backscattering value in dB (a°) or in DN.
5. RESULTS AND DISCUSSIONS
In this work, the areas showing similar roughness
characteristics, i.e. the plains, were chosen as the study area.
The contribution of SAR images to detection of soil moisture
was investigated by taking into account the interaction of SAR
images with surface relief. The stoniness did not exist in the
study area, which was basically constituted of plain grounds
that might affect the backscattering. It was assumed that only
the soil texture and the soil moisture could affect backscattering.
After analyzing the 240 soil samples from the study area, the
relation between moisture content (which was believed
affecting backscattering directly) and ratios of clay, silt, and
sand in soil was investigated. Although theoretically it is known
that clay texture has higher water holding capacity we still have
made an attempt to check the real situation in the research area,
(i.e. correlation calculations were performed in order to check
the coherence between the real situation in the study area and
the findings given in the literature). The correlation studies
indicated that the clay content was correlated with the moisture
at 0,72 level whereas the sand content was anti-correlated with
the soil moisture at a level of 0,70. It was also noticed that the
amount of the silt was not correlated with the moisture (Figure
3).
Figure 2. Georeferenced SAR images and the locations of the
sample points
Figure 3. The relations of soil moisture with the ratios of clay,
silt, and sand in soil samples
It is known that the moisture content of the soil varies
depending on the inorganic material and the clay content of the
soil. In the study area, the organic material content of the soil is
very low and homogeneous. The flat and nearly flat areas form
the physiological suture of the study area. In the area, clay
content of the soil is not homogenous. Thus soil moisture
content of the study area changes depending merely on the clay
content and the micro relief of the soil. In this study the relation
between the soil moisture and RADARSAT-1 backscattering,
