Remote Sensing Parameterization of Meso-scale ...
415
of the transfer equation for sensible heat can be applied which excludes the necessity to collect
oT:: ix from scarse density observation networks in complex landscapes:
(15.28)
It should be stressed that r;:';x in Eq. (15.28) should be based on as best as possible estimates
of kB-1. For specific pixels where >'Emix ~ 0, Hmix may be determined from {Q. - Go}mix
without involving Eqs. (15.1) to (15.7). The areal maximum value for oT::ix will in this way
be obtained. Also the other extreme, viz. Hmix = 0, provides information on the magnitude
of oT::iX(oT::ix ~ 0). A new approach by combining (T;:..iJ,oT;:..ij) and (T;:..iti,oT:::!ti) where
j represents a large group of pixels with dry land surfaces (>..Emix ~ 0) and jj the class
with wet surfaces (Hmix ~ 0), provides the opportunity to internally calibrate C3 and C4 on a
regional scale for the moment of image acquisition without needing in-situ T::ix observations or
radiosoundings. This approach should be conceived as an essential element in the improvement
for estimating H from thermal remote sensing at data sparse terrain. In this way, the high
absolute accuracy of Toix required in sensible heat flux estimations (error:::; 0.5 K: Brutsaert et
al., 1993) is not longer needed: oT::ix is solved from Hmix of specific surfaces according to Eq.
(15.28), rather than from Toix and T;:..if seperated. Although Tomix is still needed to spatially
interpolate oT::ix, the standard error of Toix becomes to a certain extent irrelevant (it only
affect C3 and C4, not the dynamic range of H).
15.8 Classes of Remote Sensing Flux Algorithms
The impact of the selection of Zsur and spatial variability of T::ix for the different Hschematizations addressed at Table 15.2 is briefly examined by means of an example with
EFEDA data. Five different land use types (A to E) which differ in surface radiative properties are selected from a Thematic Mapper image acquired on June 12, 1991 (Pelgrum and
Bastiaanssen, 1996). Table 15.5 shows the primary remote sensing measurements on Toix,
r(j'ix and NDVlmix. The remaining SVAT- parameters in Table 15.5, are derived from semiempirical relationships, i.e. t(j'ix, z(j'':'x, oT::ix and u::,ix and empirically determined using the
equations proposed by Bastiaanssen (1995).
The air temperature was in this case considered at 10 m above surface level. The air temperatur~
at the blending height (in this case an arbitrary value of 100 m. has been considered) was 22.8°C
while the wind speed at the same height is 5.7 m S-l. The air heat capacity was taken as 1230
J m- 3 K- 1 and kB- 1 = 2.3 has been taken as an arbitrary value. To compare the H-fluxes
computed with the various remote sensing algorithm classes distinguished in Table 15.3, areaeffective values of u::,ix, z(j'':'x, Toix and T::ix(zsur) are required as input for algorithm class 1
to 4. The following aggregation rules have been applied to obtain u: 1 I, z~!!, T;I I for class 1
to 4. The value for T:ff has been quantified from T;ff using Eq. (15.27).
u ell
{~ L:PaU:;,iX2}0.5p;1
. n
(m S-l)
(15.29)
ell
1
.
(15.30)
ZOm
exp{;;- L: In z;;'~X} (m)
T;ff
{( ~ L: t;;,ix aT;,ix4)( < to > a)} 0.25
n
(K)
(15.31 )
The impact of the selection of the algorithm on local Hmix-fluxes for land use types A to E
and on the area- averaged flux, < H >, is presented in Fig. 15.7. The area-averaged flux was
415
of the transfer equation for sensible heat can be applied which excludes the necessity to collect
oT:: ix from scarse density observation networks in complex landscapes:
(15.28)
It should be stressed that r;:';x in Eq. (15.28) should be based on as best as possible estimates
of kB-1. For specific pixels where >'Emix ~ 0, Hmix may be determined from {Q. - Go}mix
without involving Eqs. (15.1) to (15.7). The areal maximum value for oT::ix will in this way
be obtained. Also the other extreme, viz. Hmix = 0, provides information on the magnitude
of oT::iX(oT::ix ~ 0). A new approach by combining (T;:..iJ,oT;:..ij) and (T;:..iti,oT:::!ti) where
j represents a large group of pixels with dry land surfaces (>..Emix ~ 0) and jj the class
with wet surfaces (Hmix ~ 0), provides the opportunity to internally calibrate C3 and C4 on a
regional scale for the moment of image acquisition without needing in-situ T::ix observations or
radiosoundings. This approach should be conceived as an essential element in the improvement
for estimating H from thermal remote sensing at data sparse terrain. In this way, the high
absolute accuracy of Toix required in sensible heat flux estimations (error:::; 0.5 K: Brutsaert et
al., 1993) is not longer needed: oT::ix is solved from Hmix of specific surfaces according to Eq.
(15.28), rather than from Toix and T;:..if seperated. Although Tomix is still needed to spatially
interpolate oT::ix, the standard error of Toix becomes to a certain extent irrelevant (it only
affect C3 and C4, not the dynamic range of H).
15.8 Classes of Remote Sensing Flux Algorithms
The impact of the selection of Zsur and spatial variability of T::ix for the different Hschematizations addressed at Table 15.2 is briefly examined by means of an example with
EFEDA data. Five different land use types (A to E) which differ in surface radiative properties are selected from a Thematic Mapper image acquired on June 12, 1991 (Pelgrum and
Bastiaanssen, 1996). Table 15.5 shows the primary remote sensing measurements on Toix,
r(j'ix and NDVlmix. The remaining SVAT- parameters in Table 15.5, are derived from semiempirical relationships, i.e. t(j'ix, z(j'':'x, oT::ix and u::,ix and empirically determined using the
equations proposed by Bastiaanssen (1995).
The air temperature was in this case considered at 10 m above surface level. The air temperatur~
at the blending height (in this case an arbitrary value of 100 m. has been considered) was 22.8°C
while the wind speed at the same height is 5.7 m S-l. The air heat capacity was taken as 1230
J m- 3 K- 1 and kB- 1 = 2.3 has been taken as an arbitrary value. To compare the H-fluxes
computed with the various remote sensing algorithm classes distinguished in Table 15.3, areaeffective values of u::,ix, z(j'':'x, Toix and T::ix(zsur) are required as input for algorithm class 1
to 4. The following aggregation rules have been applied to obtain u: 1 I, z~!!, T;I I for class 1
to 4. The value for T:ff has been quantified from T;ff using Eq. (15.27).
u ell
{~ L:PaU:;,iX2}0.5p;1
. n
(m S-l)
(15.29)
ell
1
.
(15.30)
ZOm
exp{;;- L: In z;;'~X} (m)
T;ff
{( ~ L: t;;,ix aT;,ix4)( < to > a)} 0.25
n
(K)
(15.31 )
The impact of the selection of the algorithm on local Hmix-fluxes for land use types A to E
and on the area- averaged flux, < H >, is presented in Fig. 15.7. The area-averaged flux was