462
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
ET estimates. Biophysical parameters, such as C L , VPD_close, and VPD_
open, used in the algorithm have the same values for a given biome type.
However, for different species within the same biome type, the differences
in these parameters can be large (Turner et al. 2003a,b). Water stress and air
temperature stress can greatly affect the ET estimates. Uncertainties from
inputs can introduce biases in ET estimates that are difficult to detect. For
example, when no water or air temperature stress was put on the stomatal
conductance, the tower-driven RMSE increased from 27.3 W ∙ m –2 (with
stress) to 43.2 W ∙ m –2 , and from 29.5 W ∙ m –2 (with stress) to 46.3 W ∙ m –2
when driven by GMAO data. The average LE bias of the tower-driven LE
estimates to the LE observations changed from –5.8 W ∙ m –2 (with stress) to
19.9 W ∙ m –2 and from –1.3 W ∙ m –2 (with stress) to 23.3 W ∙ m –2 when driven
by GMAO data. In addition, only little knowledge is obtained regarding
some parameters (e.g., boundary layer resistance for soil evaporation) and
the mechanisms involved. Therefore, further study is needed to improve the
ET algorithm for some ecosystems such as those in arid areas.
(3) Scaling from tower to landscape. The size of the flux tower footprint is
largely influenced by tower height and local environment conditions (Cohen
et al. 2003; Turner et al. 2003a,b). The direct comparison of observed ET
with the estimated ET from the 3 × 3 1-km 2 MODIS across all 19 sites may
introduce uncertainties due to the differences in tower footprints for different towers and under varying environmental conditions for a given tower.
In heterogeneous areas, the differing scales of the tower and MODIS ET
estimates should be performed via an upscaling process, such as that used
during the Bigfoot study (Cohen et al. 2003; Turner et al. 2003a,b). The
expense and intensity of such a study, however, limits our ability to perform
such comparisons.
19.6 APPLICATIONS OF THE RS-ET ALGORITHM AT
THE REGIONAL AND GLOBAL SCALES
19.6.1 Regional and gloBal validationS
In addition to the direct evaluation of ET through measurements at eddy flux towers
as shown above, ET can be indirectly validated with precipitation and MODIS GPP
at the regional and global scales.
The RS-ET algorithm was applied globally with the 0.05° resolution MODIS
input data and GMAO data. The 0.05° resolution global ET dataset during 2000–
2006 were used to do the global validation work. The spatial pattern of the mean
global annual total ET during 2000–2006 is reasonable, with a maximum ET of 1100
mm∙year –1 and an area-weighted average of 351 ± 221 mm∙year –1 over vegetated land
areas (Figure 19.8). As expected, tropical forests have the highest ET values, while
dry areas and areas with short growing seasons have low estimates of ET. The ET for
temperate and boreal forests lies between the two extremes (Figure 19.8). The mag19.8). The mag.8). The magnitudes and spatial patterns of global ET generally agree with estimates provided in
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
ET estimates. Biophysical parameters, such as C L , VPD_close, and VPD_
open, used in the algorithm have the same values for a given biome type.
However, for different species within the same biome type, the differences
in these parameters can be large (Turner et al. 2003a,b). Water stress and air
temperature stress can greatly affect the ET estimates. Uncertainties from
inputs can introduce biases in ET estimates that are difficult to detect. For
example, when no water or air temperature stress was put on the stomatal
conductance, the tower-driven RMSE increased from 27.3 W ∙ m –2 (with
stress) to 43.2 W ∙ m –2 , and from 29.5 W ∙ m –2 (with stress) to 46.3 W ∙ m –2
when driven by GMAO data. The average LE bias of the tower-driven LE
estimates to the LE observations changed from –5.8 W ∙ m –2 (with stress) to
19.9 W ∙ m –2 and from –1.3 W ∙ m –2 (with stress) to 23.3 W ∙ m –2 when driven
by GMAO data. In addition, only little knowledge is obtained regarding
some parameters (e.g., boundary layer resistance for soil evaporation) and
the mechanisms involved. Therefore, further study is needed to improve the
ET algorithm for some ecosystems such as those in arid areas.
(3) Scaling from tower to landscape. The size of the flux tower footprint is
largely influenced by tower height and local environment conditions (Cohen
et al. 2003; Turner et al. 2003a,b). The direct comparison of observed ET
with the estimated ET from the 3 × 3 1-km 2 MODIS across all 19 sites may
introduce uncertainties due to the differences in tower footprints for different towers and under varying environmental conditions for a given tower.
In heterogeneous areas, the differing scales of the tower and MODIS ET
estimates should be performed via an upscaling process, such as that used
during the Bigfoot study (Cohen et al. 2003; Turner et al. 2003a,b). The
expense and intensity of such a study, however, limits our ability to perform
such comparisons.
19.6 APPLICATIONS OF THE RS-ET ALGORITHM AT
THE REGIONAL AND GLOBAL SCALES
19.6.1 Regional and gloBal validationS
In addition to the direct evaluation of ET through measurements at eddy flux towers
as shown above, ET can be indirectly validated with precipitation and MODIS GPP
at the regional and global scales.
The RS-ET algorithm was applied globally with the 0.05° resolution MODIS
input data and GMAO data. The 0.05° resolution global ET dataset during 2000–
2006 were used to do the global validation work. The spatial pattern of the mean
global annual total ET during 2000–2006 is reasonable, with a maximum ET of 1100
mm∙year –1 and an area-weighted average of 351 ± 221 mm∙year –1 over vegetated land
areas (Figure 19.8). As expected, tropical forests have the highest ET values, while
dry areas and areas with short growing seasons have low estimates of ET. The ET for
temperate and boreal forests lies between the two extremes (Figure 19.8). The mag19.8). The mag.8). The magnitudes and spatial patterns of global ET generally agree with estimates provided in