. Istanbul 2004
'e and it can be
ch image file.
nage, but also
vere calculated
backscattering
| Identification
parcels were
Forestry. Some
ller parcels, so
ined one base
over | ha. This
average SAR
vas calculated.
cel was needed
ries of average
rop species in
scattering time
16-Sep-03
tt
Spring w heat
Autumn rye
- Barley
- Barley (malt)
Oats
— Potato
Turnip rape
Grass
Grass (silage)
CAP fallow
bH HSFPSR I
|i VV and VH
s in 2003 in
n compared to
r to have more
backscattering
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
2.2 Field surveys
In 2003 field surveys were organized similarly as the field
surveys in the Radarsat-1 study in 2001. Purpose of the field
surveys is naturally to have reference data for the analysis of
SAR backscattering time series. Due to the limited time
window, which in our case was +3 hour compared with actual
image acquisition, we had to select a small set of test parcels for
more detailed research. Thereby 24 parcels was chosen and
following information was surveyed for each of the parcel:
= Soil surface roughness
=» Seed row direction
= Soil surface moisture
= Crop height and growth stage
= Possible crop yield damages
Soil surface roughness was measured using Leica laser
distometer only once in the beginning of the growing season
and after that roughness was assumed to be constant during the
growing season. Soil surface moisture was measured using
ThetaKit TK2-BASIC soil moisture measuring device, which
measures volumetric moisture into the depth of 6 cm. Crop
height was simply measured using measuring tape. Crop yield
damages were mapped as they were found during the field
surveys. Time series of the soil surface moisture for the 24 test
parcels are given in Figure 4.
Soil Surface Moisture
e
e
Volumetric soil moisture %
RB
e
UN
IN
"X
20 Ke. m ESS Es |
SK 7 E a = i
10 Ew |
|
0 : : :
8-Jun-03 8-Jul-03 7-Aug-03 6-Sep-03
Date
Figure 4. Volumetric soil surface moisture for test parcels in
summer 2003.
3. RESULTS
3.1 Envisat SAR backscattering time series in 2003
In Figure 3 is represented the Envisat SAR backscattering time
series in VV and VH polarizations for all major cultivated
plants in the test area in summer 2003. Similar results were
obtained for Radarsat-1 SAR in HH polarization in summer
2001 (Karjalainen et.al., 2003). Naturally results from years
2001 and 2003 are not directly comparable due to the changing
weather conditions, but nevertheless, a general comparison of
the usability of Radarsat-1 and Envisat SAR images in
agricultural monitoring can be made.
In the beginning of the growing season, before sowing,
ploughed fields had relatively rough soil surface and
consequently backscattering values were higher than in sowed
fields. According to Radarsat-1 results in 2001, HH
backscattering for ploughed fields was approximately from —8
to even —4 dB. Backscattering variation was high in ploughed
fields due to the surface roughness and soil moisture variations.
After sowing, backscattering decreased, because sowing made
the soil surface smoother compared to the ploughed fields.
After sowing, VH backscattering was near -18 dB, but VV
backscattering was still as high as —10 dB, and reason for this
most likely is that VV polarization is more sensitive to the soil
surface moisture, which at same time is around 40% of
volumetric moisture. Soil surface moisture rapidly decreased by
the beginning of July as can be seen in Figure 4. and
simultaneously VV backscattering decreased to it minimum
around -14 dB for cereal crops.
It should be noticed that the noise equivalent o" for Envisat
alternating polarization images is from -19 to -22 dB
depending on the antenna look angle (ESA, 2002). From
backscattering time series in Figure 3, can be scen that in the
beginning of the growing season VH backscattering was close
to the noise equivalent o" and VH backscattering started to
increase only after the middle of July, which implies that the
early crop growth from emerging to 50 cm cannot be seen using
Envisat cross-polarization images.
For cereal crops both VV and VH backscattering gradually
started to increase in the middle of July when crop height was
approximately 50 cm. After this point of time soil surface
moisture stayed relatively constant from 10% to 30% of
volumetric moisture, thus the average increase in SAR
backscattering seemed to be caused by the increase of the
vegetation biomass. Similar results were obtained for Radarsat-
1 HH polarization SAR images in 2001. The increase of SAR
backscattering was from 2 to 4 dB depending on the crop
species and SAR polarization. According to these results it
seemed that cross-polarization i.e. VH polarization might be the
best choice for biomass detection in Finland for cereal crops.
but noise equivalent o" in Envisat SAR images is too high for
detecting crops with low biomass. Also, it should be
emphasized that averaging of relatively large areas of same
crops is needed since the radiometric accuracy of alternating
polarization images makes it hard to see backscattering changes
for individual parcels with small area compared to the spatial
resolution of 30 m.
The backscattering time series for cereal crops can be explained
fairly well, but for other species it is more difficult. For example
grass silage can be harvested many times per growing season
depending on the weather conditions. Also turnip rape and
potato have quite unique time series, which on the other hand
makes classification or identification of crop species possible.
For example potato fields have relatively high backscattering
throughout the growing season in VV polarization most likely
due to the much more rough soil surface than cereal crops have.
3.2 Detection of yield damages
In 2001, we concluded that crop yield damages caused by
lodging could not be detected using Radarsat-1 HH polarization
SAR images. Only one barley field out of totally 41 fields with
lodging was possible to be identified from Radarsat-1 SAR
images. We believed that HH polarization used in Radarsat-1
SAR images was not sensitive enough to see changes in the
crop vegetation biomass. When Envisat SAR images were used
in 2003, we detected 8 parcels where more than 50% of the
crops where flattened by lodging. The lodging usually happens
