ul 2004
—
ficiency
ie black
oretical
lissivity
thermal
surface
as the
ation is
can be
from
en days
GMS-5
issivity
Kerr et
w-band
ference
ich this
on the
rrection
tandard
).
window
ected to
Im was
r water
ed, and
es used
ency of
S-5 by
ures in-
d. This
(Eq. 18)
> water
ids 711
e split-
"ression
ased on
iount of
ect the
PW is
0.01 g
ithm as
ditions
y. 19)
: cloud-
vailable
ind not
| which
ghtness
;:ensors)
I
I
|
|
|
|
I
|
|
|
|
I
|
|
|
|
|
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
and a case-by-case analysis of the visible band, and in the case of
GMS-5 also the water vapour band (77, 6.7um).
2.2.5 Integration of Remotely Sensed Crop Canopy
Temperature into the PS-n Model: Figure 3 below presents a
relational diagram of the methodology for deriving canopy
temperatures from satellite imagery and integrating them in the
PS-n model by updating the temperature difference forcing
variable. The flow diagram shows two parallel processes that feed
data into the PS-n model. The right side of the flow diagram
describes the canopy temperature retrieval process from satellite
imagery using the split-window technique.
VIS NIR
IS,
CONSTANTS / SYNOPTIC DAILY
INITIAL VALUES DATA
Instantaneou |
Offset maps
AVHRR/VISSR
T11, T12
Instantaneou )
Tcanopy map
split- window
coefficients
Figure 3. Relational diagram of the integration of satellite-derived
crop canopy temperatures in the PS-n model
Data and study area: Satellite data for GMS-5 were routinely
processed and archived under the GAME/Tibet (GEWEX Asian
Monsoon Experiment) project (Koike et al., 1999) and for the
crop season of 1999 imagery was obtained through the Weather
Satellite Image Archive as published by Kochi University, Japan
(http://Weather.is.kochi-u.ac.jp/archive-e.html).
For the same period NOAA/AVHRR images, obtained from the
NOAA Satellite Active Archive WWW site, were aggregated to the
same pixel resolution of the GMS-5/VISSR TIR bands so they
could be combined for our proposed multi-sensor crop production
methodology. The table below shows the precision and value
ranges of the satellite products, some of which are also depicted in
Figure 3.
Map Range Precision
NDVI -1-1 0.001
€ 0-1 0.001
PW 0-5 0.01
To 250 - 350 0.1
Table 1. Data precision and value ranges
To avoid mixed-land cover observations in one image pixel, a
region characterized by homogeneous land cover and uniform soil
characteristics was identified. The North China Plain consists of
flat terrain at 40 m.a.s.l with uniform, re-washed loess (loam)
soils. Located in these plains uniform Land Use Systems (>250
sq. km) were selected where experimental maize fields were set-
up, within the administrative district Quzhou, People’s Republic
of China. Here, researchers from the China Agricultural
University, Beijing routinely conduct the field trials, inter alia on
maize production potentials. They kindly provided experimental
production and yield data and correlated weather data recorded
217
from an automatic recording station within the experimental site.
In addition, planted areas and yields of surrounding administrative
counties were provided for validation and calibration of the PS-n
simulations.
3. RESULTS AND DISCUSSION
The results from the retrieval of canopy temperatures from
satellite data from GMS-5 and NOAA-14 showed good internal
agreement with a RMSE of 0.85, a BIAS of 0.92, and a STDV of
6.26 (Kelvin), as is also confirmed by the scatter plot of 37
observations as depicted in Figure 4.
325.0 ee eee
320.0
X 3150
a i
A i
9 3100 |
2 |
9 Í
= 3050 |
© i
o
^ 3000
295.0
290.0
N N w C w w œ
«o © © d — N N
o a a o a o e
eo e e e e © ©
Ww
S
e
T, NOAA-14/AVHRR (K)
Figure 4. Scatter plot of estimated canopy temperature for
NOAA-14/AVHRR and GMS-5/VISSR
As ‘cloud-free’ AVHRR imagery that could be obtained for the
crop season of 1999 was not entirely free of cloud on all dates,
selected pixels with no contamination were identified for further
analysis. The PS-n model was run using canopy temperature data
obtained from these selected pixels to update the “TEMPDIFF’
(AT) forcing variable including the pixel containing the Quzhou
maize research site. As only 24 cloud-free AVHRR observations
could obtained for the 1999 crop-season, it was necessary to fill in
the days when canopy temperature data were unavailable. A
linear interpolation procedure was applied conform the
computational steps as detailed in section 2.2 of this paper (Eq.6
and 7) so as to obtain proxies for missing days. The upper part of
Figure 5 shows the output curves of simulated (PS-n) structural
plant matter development based on NOAA-14/AVHRR data alone.
Crop stress indicated by the grey line (c/H20) shows that the crop
suffered from water shortage on multiple occasions. Specifically
stress period 1a (JD: 178 — 184) and 2a (JD: 262 — 265) indicated
by the vertical, light grey lines could very well be erroneous since
the interpolation technique has to rely for its guess on relatively
few observations. For its guess of the first stress period (la) the
technique relies on a single NOAA-14 observation (JD: 182) of
(temporal) canopy heating during a period of 21 days of no
observations; only two observation of no crop stress preceded and
proceeded (JD: 177 — 199). For estimating the second stress
period (2a) the interpolation technique can rely on more NOAA-14
observation (JD: 263, 264, 265). However, before the onset of this
particular stress period (JD 262) there are again few observations
available, with the last cloud-free satellite overpass occurring on
JD 250. Hypothetically, the duration of crop stress could have
* Julian day
