International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
2
= 4
x à
o 15 A à
: À
S 1 A
= A
o A
B asia à IR2-ch5
i A IR1—-ch4
= HE TE TET TE TE TET,
-05
0 20 40 60 80 100
Hurridity (52)
Figure 1. Influence of water vapour on differences in brightness
temperatures for the GMS-5/NOAA-14 split-window
channels (from Tokuno M., 1997)
Figure 1 shows the influence of water vapor by relating the
brightness temperature difference for /R/ vs. Ch4 and IR2 vs. Ch5
at increasing levels of humidity. The sensitivity of /R/ is almost
the same as that of Ch4, whereas the sensitivity of /R2 is greater
than CA5 for higher humidity levels as can be seen form the
increasing /R2-Ch5 values. This is possibly caused by instrument
differences between the two satellites.
1.0
0.5
0.5
6 8 10 12 14
Wavelength A (pm)
Instrument response curve for NOAA-14/AVHRR and
GMS-5/VISSR (after Yuichiroh, 2004)
Figure 2.
Figure 2 shows the normalized response curves (to a value of 1)
of the two instruments as a function of wavelength after
Yuichiroh (2004). Small, but apparently significant band-to-band
wavelength and response function differences call for a careful
selection of split-window algorithms for parallel land surface
temperature retrieval, inter alia crop canopy temperature, from the
two satellites.
2.2.4 Crop canopy temperature retrieval from GMS-5 and
NOAA-14: The retrieval of canopy temperatures from satellite
data is based on the Stephan-Boltzman black body emission
equation:
R = gc0T4 (Eq.17)
Where: :
R — is radiation emitted by the surface (W m")
216
07 5.672 x 10 * Wm"? K* (the Stephan-Boltzman constant)
£97 emissivity of the surface
T — surface temperature /K7
The emissivity term in the equation is a measure of the efficiency
with which the surface emits energy. A perfect emitter, the black
body, has an emissivity of 1. The black body is a theoretical
concept whose behaviour does not exist in nature. The emissivity
of most natural bodies lies between 0.91 and 0.98 in the thermal
wave region 8-14 jum (Qin and Karnieli, 1999). Actual surface
emissivity depends on surface characteristics such as the
vegetation and the surface wetness, so that its diurnal variation is
expected to be relatively small but day-to-day variation can be
significant. We estimated the surface emissivity from
NOAA/AVHRR visible channels by interpolating in-between days
when NOAA-14 observations were contaminated but GMS$-5
observations were cloud-free so temperature and emissivity
separation was still needed. The procedure is the same as Kerr et
al. (1992) and Sobrino et al. (2000), who estimated narrow-band
emissivity semi-empirically from a Normalized Difference
Vegetation Index (NDVI). The optical imagery from which this
NDVI is computed are atmospherically corrected based on the
SMAC-algorithm (Simplified Method for Atmospheric Correction
of Satellite Measurements in the Solar Spectrum) using standard
atmospheric conditions (Rahman, H., and G. Dedieu, 1994).
For observations from the polar-orbiting satellite the split-window
algorithm developed by Coll and Caselles (1997) was selected to
estimate maize crop canopy temperatures. The algorithm was
selected based on the notion that it accounts better for water
vapour then the split-window algorithm commonly applied, and
because it has been calibrated for data from the two satellites used
in this study. This in turn helps to improve to the consistency of
the inference. The algorithm was calibrated for GMS-5 by
Yuichiroh (2004) and used to estimate canopy temperatures in-
between days NOAA-14 observations were contaminated. This
algorithm takes the following form:
To - c (TH y. * e3TI HF es TILTI2 * c,(TI2)Y + Offset (Eq.18)
Where:
Ty = surface temperature [K]
T11, T12 = split-window brightness temperature [K]
3
The regression coefficient *c; corrects for atmospheric water
vapor; and the offset corrects for surface emissivity in bands 77/
and 772. For applications to geo-stationary satellites the split-
window algorithm takes the same form, except that the regression
coefficient ‘c;’ also accounts for the optical path length based on
the satellite zenith angle. In addition, the instantaneous amount of
precipitable water (PW) is required in order to select the
appropriate coefficients (c;) for a particular image scene. PW is
derived using the VISSR/IR data (range: 0 — 5, precision: 0.01 g
cm?), also utilizing a derivation of the split-window algorithm as
proposed by Chester et al. (1987), which for cloud-free conditions
takes the following form:
PW = (1/0.095)[(1/see O)In[(T11-T,)/(T12-T,)]-0.025](Eq. 19)
Where:
T, = air temperature (or 7//— 2.2)
a
To derive land surface temperature it is essential to detect cloud-
covered areas correctly because these equations are only available
when the satellite receives radiation from the surface, and not
from a cloud top. We utilized a cloud detection method which
utilizes a combination of a semi-automated threshold brightness
temperature T// filtering technique (unique for the two sensors)
Internati
EPIO
and a cas
GMS-5a
225 |
Temperz
relational
temperati
PS-n mc
variable.
data into
describes
imagery |
CONS
INITIAL
Figure 3.
Data and
processed
Monsoon
crop seas:
Satellite 1
(http://we:
For the s
NOAA Sa
same pix
could be «
methodolc
ranges of
Figure 3.
To avoid
region cha
characteris
flat terrain
soils. Loc
sq. km) w
up, within
of China
University
maize pro
productior
