IN
IMAGES
artinen,
ssessment. The
pecified times
nising civilian
area is one of
The main food
itoring project
parison of the
it the research,
framework of
vere organized
rements of soil
sed by drought
| as well as a
e of the crop
that at the end
dominated by
et al. (2003)
nce of ithe
measure for
we also been
nterferometric
ation biomass
suitable SAR
et.al., 2003).
sensing is an
that recorded
ical properties
ss, vegetation
slope and seed
would like to
sion from the
ed since other
r hand, direct
backscattering
simulation), is
(wavelength,
on. Simulation
lex scattering
jn.
ied research in
> beginning oi
Crop Growth
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
Monitoring System by Joint Research Centre) in the FGI were
based on weather data, historical crop yield statistics and optical
low-resolution satellite images, and it turned out that cloudiness
almost completely hampered the use of optical satellite images
in Finland. The launch of ERS-1 satellite in 1991 activated
vivid research on agricultural remote sensing using satellite
SAR images (ESA, 1995). The excellent radiometric stability of
the ERS-1 and ERS-2 SAR images enabled a long-term
continuous monitoring of the agricultural fields. In 1997
agricultural SAR research was started in the FGI, firstly for crop
species classification purposes, but later also for crop damage
assessments and yield estimation purposes.
In 2001 FGI started a study in the test area near the city of
Seindjoki in Finland. The objective was to evaluate the
potential of high-resolution high-repetition multitemporal SAR
images in agricultural monitoring including aspects of the crop
yield estimation and assessment of the crop yield damages. For
the growing season in 2001 Radarsat-1 Fine beam satellite SAR
images were used. The image delivery from the Tromsg
Satellite Station took only few days, so near real-time
agricultural monitoring was possible. Field surveys were made
simultaneously with each of the image acquisition. The results
were promising in some cases, for example flooded areas and
cultivation practises were detected well, but the crop growth
caused only weak change to the SAR backscattering and yield
damages caused by lodging were detected only in very rare
cases. Conclusion was that the HH polarization of the Radarsat-
| is not sensitive to the changes of biomass of small leaved
crops, but on the other hand, it is sensitive to variation of the
surface roughness and soil surface moisture. Example of the
detection of the flooded area caused by snowmelt in the end of
April is represented in Figure |.
Figure 1. Snowmelt caused a minor flooding in April 2001.
Original data © distributed by — Radarsat
International/TSS/Novosat Ltd. Processed by FGI.
In 2003 the study continued using Envisat SAR images, which
were obtained in the framework of ESA’s Envisat
announcement of opportunity (AOE-488). We choose to have
Envisat SAR images in VV/VH alternating polarization mode,
which means that both VV and VH images are acquired
simultaneously with some loss in the radiometric accuracy
compared to one polarization imaging mode. Although two
polarizations give more information about the research subject,
a drawback with Envisat SAR in comparison with Radarsat-1
Fine beam SAR images is significantly worse spatial resolution.
Envisat SAR has 30 m and Radarsat-1 SAR has 8 m spatial
resolution. This paper describes our experiences of using
satellite SAR images in the remote sensing of agricultural fields
in Finland.
2. TEST AREA AND DATA
2.1 The test area and satellite images
The test area near the city of Seinäjoki in Finland is located at
the latitude of 63 degrees North and it is one of the
northernmost consistent agricultural areas in the world.
Percentages of different crop species in the arable land in 2003
were: oats 25%, barley 25%, grass silage 18%, fallow land
10%, turnip rape 5%, wheat 3%, rye 1%, potato 3% and the rest
10% comprises of sugar beet, grassland, pasture, garden etc
(Tike, 2003). Cereal crops are mostly spring-sown varieties,
except small amount of rye and autumn wheat. In 2003 the
average crop yields in the Seinäjoki region were: oats 3.6 t/ha,
barley 3.8 t/ha, wheat 3.5 t/ha and rye 2.8 t/ha (Tike, 2003).
Crop yields are low compared with the yields in the southern
regions of Europe. The river of Kyrójoki, which has been
embanked in the last centuries, divides the test area, and
nowadays there are no regular flooding in the springtime, but in
some places floodwater from melting snow can stay on the low-
lying fields for few weeks. The growing season in the test area
lasts from 150 to 160 days a year. Sowing date 1s usually in the
end of May or at the latest in the beginning of June. The
harvesting of food crops occurs at the end of August or early
September.
In 2001 altogether 20 Radarsat-1 SAR Fine beam images were
used in the agricultural monitoring study. The description of the
Radarsat-1 SAR images is given in Karjalainen et al. (2003).
In 2003 Envisat alternating polarization SAR were used.
Altogether 16 SAR images were ordered in early spring 2003,
but due to the problems in Envisat satellite and conflicts
between commercial image orders, we were able to have 12
SAR images. All images were acquired from descending orbit,
thus the local time was around 12:00 at the time of the
acquisition. The SAR incidence angle varies from 23 degrees to
41 degrees. On the average, the time interval between image
acquisitions was approximately two weeks, but in the most
intensive growing period in the beginning of July there were
even two image acquisitions in a week. The list of Envisat SAR
images used is represented in Table 1.
Table 1. The list of Envisat alternating polarization (VV/VH)
SAR images in 2003.
Image # Date PAF Swath
] 15 June 2003 D-PAC 4
2 18 June 2003 I-PAC 3
3 21 June 2003 D-PAC 2
4 28 June 2003 I-PAC 6
5 04 July 2003 I-PAC
6 07 July 2003 UK-PAC 3
7 14 July 2003 I-PAC 6
8 23 July 2003 I-PAC 3