445
Remote Sensing and Modeling of Global Evapotranspiration
research purposes (Guenther et al. 2002). The MODIS sensor provides higher quality data for monitoring terrestrial vegetation and other land processes than previous
sensors such as the Advanced Very High Resolution Radiometer (AVHRR), not only
because of its narrower spectral bands that enhance the information derived from
vegetation (Justice et al. 2002), onboard calibration to guarantee the consistent timeseries reflectance (Guenther et al. 2002), and orbit and altitude satellite maneuvers
to ensure subpixel geolocation accuracy (Wolfe et al. 2002) but also because leading
scientists are working as a team to improve the accuracy of the data from low-level
reflectance data to derived high-level land data.
This chapter concentrates on applying satellite data to terrestrial ET and water
cycle studies. Within this context, we reviewed the history of the development of
regional and global ET models, described the MODIS global terrestrial ET algorithm and its applications, and discussed the model uncertainties to the remote sensing data and meteorological data.
19.2 LITERATURE REVIEW
19.2.1 SuRface eneRgy Balance ModelS
Because remote sensing can provide land surface temperature (LST) information
through thermal spectral bands, energy balance–based models have been proposed
and widely used. In the early stage of energy balance–based models, most studies
used high-resolution remote sensing data; some data sources are even from airborne
sensors or sensor mounted above a site (e.g., Kalma and Jupp 1990; Norman et al.
1995; Bastiaanssen et al. 1998a,b; Su 2002). The energy balance models calculate the
ET through the residual of the surface absorbed energy as follows. The surface net
radiation (R n ) partitions into three components, including latent heat (λE), sensible
heat (H), and soil heat fluxes (G; Equation 19.1).
R n = λE + H + G.
(19.1)
G is calculated with some empirical relationship between R n and other biophysical variables, such as LST, surface reflectance, vegetation cover fraction estimated
with normalized difference vegetation index (NDVI), and LAI (Norman et al. 1995;
Bastiaanssen et al. 1998a; Bastiaanssen 2000; Su 2002; Nishida et al. 2003a,b). The
sensible heat term (H) is estimated by a function of the difference (ΔT) between surface dynamic temperature (T 0 ) and air temperature (T a ) at a given height.
H
c
T
r
c
T T
r
p
ah
p
a
a
=
=
−
ρ
ρ
∆
0
,
(19.2)
where ρ is the air density, c p is the specific heat of air at constant pressure, and r a
is the resistance to convective heat transfer and radiative heat transfer. Although
surface dynamic temperature is different from remotely sensed LST, all energy balance–based models use remotely sensed LST to replace T 0 in Equation 19.2. Some
models, such as the Surface Energy Balance Algorithm for Land (Bastiaanssen et al.
Remote Sensing and Modeling of Global Evapotranspiration
research purposes (Guenther et al. 2002). The MODIS sensor provides higher quality data for monitoring terrestrial vegetation and other land processes than previous
sensors such as the Advanced Very High Resolution Radiometer (AVHRR), not only
because of its narrower spectral bands that enhance the information derived from
vegetation (Justice et al. 2002), onboard calibration to guarantee the consistent timeseries reflectance (Guenther et al. 2002), and orbit and altitude satellite maneuvers
to ensure subpixel geolocation accuracy (Wolfe et al. 2002) but also because leading
scientists are working as a team to improve the accuracy of the data from low-level
reflectance data to derived high-level land data.
This chapter concentrates on applying satellite data to terrestrial ET and water
cycle studies. Within this context, we reviewed the history of the development of
regional and global ET models, described the MODIS global terrestrial ET algorithm and its applications, and discussed the model uncertainties to the remote sensing data and meteorological data.
19.2 LITERATURE REVIEW
19.2.1 SuRface eneRgy Balance ModelS
Because remote sensing can provide land surface temperature (LST) information
through thermal spectral bands, energy balance–based models have been proposed
and widely used. In the early stage of energy balance–based models, most studies
used high-resolution remote sensing data; some data sources are even from airborne
sensors or sensor mounted above a site (e.g., Kalma and Jupp 1990; Norman et al.
1995; Bastiaanssen et al. 1998a,b; Su 2002). The energy balance models calculate the
ET through the residual of the surface absorbed energy as follows. The surface net
radiation (R n ) partitions into three components, including latent heat (λE), sensible
heat (H), and soil heat fluxes (G; Equation 19.1).
R n = λE + H + G.
(19.1)
G is calculated with some empirical relationship between R n and other biophysical variables, such as LST, surface reflectance, vegetation cover fraction estimated
with normalized difference vegetation index (NDVI), and LAI (Norman et al. 1995;
Bastiaanssen et al. 1998a; Bastiaanssen 2000; Su 2002; Nishida et al. 2003a,b). The
sensible heat term (H) is estimated by a function of the difference (ΔT) between surface dynamic temperature (T 0 ) and air temperature (T a ) at a given height.
H
c
T
r
c
T T
r
p
ah
p
a
a
=
=
−
ρ
ρ
∆
0
,
(19.2)
where ρ is the air density, c p is the specific heat of air at constant pressure, and r a
is the resistance to convective heat transfer and radiative heat transfer. Although
surface dynamic temperature is different from remotely sensed LST, all energy balance–based models use remotely sensed LST to replace T 0 in Equation 19.2. Some
models, such as the Surface Energy Balance Algorithm for Land (Bastiaanssen et al.