International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
in August, when ripening begins. An example of lodging is
represented in Figure 5, where an extensive area of lodging was
detected on 24 July 2003. On the upper part of Figure 5 there is
a multitemporal Envisat SAR image, which comprises of
images taken on 14 July and 11 August 2003. Nearly 8096 of
the crops were flattened in the middle part of the parcel as can
be seen in the photograph taken on 11 August 2003. The lodged
area in the SAR image somewhat corresponded to the area what
was mapped in the field surveys. Anyhow, the rest of the 8
lodged fields could not be detected from SAR images,
accordingly it seems that it is not possible to see lodging
reliably from Envisat SAR images. Most likely the reason for
this is, again, that the spatial resolution and radiometric
accuracy of Envisat SAR images is not enough compared to the
small size of parcels in Finland.
Figure 5. Yield damages caused by lodging in the test area in
2003.
3.3 Yield estimation using SAR images
Spring sown oats, barley and wheat are the major cereal crops
in the Seinäjoki test area. Although, the crop yields are
relatively low compared to the yields in the southern latitudes,
there is interest in the crop yield estimation in Finland.
Nowadays field agents of the Ministry of the Agriculture and
Forestry prepare the crop yield estimation on a subjective basis.
In Figure 3, in case of cereal crops, can be seen a slight increase
(3-4 dB) in SAR backscattering from the end of June till the end
of August when harvesting starts. At the same time soil surface
moisture in Figure 4 seems to be relatively constant. The
increase in the SAR backscattering seems to correspond to the
increase in the crop height, which might make crop yield
estimation indirectly possible. Problem is that the increase in
SAR backscattering has been obtained by averaging large areas
of same crops, and for example in Finland parcels are so small
that backscattering time series cannot be calculated for small
parcels using Envisat SAR images.
3.4 Summary of the suitability of Radarsat-1 and Envisat
SAR systems for agricultural monitoring in Finland
Table 2 summarizes our experiences in agricultural monitoring
using satellite SAR images. The statements in Table 2 are based
on our experiences of Radarsat-1 and Envisat SAR images in
2001 and 2003.
Table 2. Summary of the experiences of Radarsat-1 and Envisat
SAR images in agricultural monitoring in Finland.
Application | Radarsat-1 (HH) Envisat (V V/VH)
area
Cultivation Detection of | VV polarization might be used
practises sowed fields is | for detecting sowing date.
very good, which
could be used to
trigger the yield
prediction model
Yield Flooding in the | There were no flooding in 2003
damages fields can be | season, but presumably HH
detected very | polarization is better for
reliably. Damages | flooding detection. Lodging
caused by lodging | could not be detected. 30 meters
could not be | spatial resolution hinders the
detected. usability of the Envisat SAR
images.
Yield Not adequate for | Not adequate for individual
estimation individual parcels, | parcels, but when large areas are
but seems to be | averaged crop growth can be
usable when large | seen. Cross-polarization VH is
areas of same | probably best for the crop
crops are averaged | biomass estimation, but noise
equivalent 6° of Envisat SAR is
too high to see low vegetation.
4. CONCLUSIONS
Agricultural applications, such as the monitoring of cultivation
practises, estimation of crop yield and mapping of yield
damages, are among the most promising civilian application
areas of SAR images in the future. However, considerable
amount of research needs to be done to retrieve information
from SAR images. It is evident that several SAR images are
needed to cover the whole growing season from the sowing to
the harvesting. These multitemporal SAR images then provide a
time series of SAR backscattering for agricultural fields.
Multitemporal changes in SAR backscattering correspond to the
variations in physical parameters of the agricultural fields, such
as soil moisture, soil surface roughness, vegetation biomass and
vegetation moisture. According to our experience, an
approximate sowing date for individual parcel can be detected
reliable, because there is a big change in the soil surface
roughness from ploughed to sowed field. Crop growth or the
increase of the vegetation biomass can be seen in the
backscattering time series, but large areas of same crops must
be averaged because of the SAR speckle and poor spatial
resolution of satellite SAR systems. The change in the
backscattering is in order of 2 to 4 dB in C-band in all
polarizations. Crop yield damages caused by lodging were
detected only in very few cases and in this case the spatial
resolution must be in order of few meters before variations
within parcel can be seen.
Envisat SAR backscattering time series from 2003 will be
analysed in more detail. The agricultural research using the
Internatioi
ae
SAR imag
SAR imag
vegetation
still waits
images VW
available.
Blaes X:
Parameter
Coherence
pages 374
Brown Sa
Bennett,
Measurem
for Crop F
and Remo
1610.
ESA, 199
ERS-1. .
Division, |
ESA, 200
2002. htt
asarProdu:
Karjalaine
Kuittinen,
Resolutior
Productioi
Precision
Linden an
Skriver H
Thomsen,
Signatures
Remote Sc
2429.
Tike, 200
Informatic
in Finlar
linen281 11
keskuksitt
February
Ulaby Fav
Learned /
Second In
Physical |
ESTEC, 2
The authc
providing
announcer
to thank tl
and Fores
species in
operation |
