467
Remote Sensing and Modeling of Global Evapotranspiration
in the tropics if other surface variables are accurate. Compared with the ECMWF
and NCEP zonal mean S↓ s from 2000 and 2001 (Figure 19.12e), GMAO S↓ s is relatively more accurate, except for much lower values in high latitudes of the northern
hemisphere.
Compared with the 30-year annual mean T avg in the CRU (Figure 19.12b), NCEP
tends to have lower values for almost all latitudes, with particularly large negative
biases (area-weighted bias of –1.43°C) from 20°S to 20°N. ECMWF generally agrees
well with the CRU but is somewhat higher in middle high latitudes and a little lower
in the tropics. For the 2000 and 2001 intercomparison (Figure 19.12f), NCEP tends
to have the lowest T avg , ECMWF has the highest, and GMAO is in the middle for
most latitudes. Although the bias in T avg is small, a small bias in temperature can
introduce relatively large errors in VPD and, consequently, in global ET because of
the nonlinear relationship between T avg and both VPD and ET.
A comparison of vapor pressure (e) shows that NCEP has higher values than
CRU in tropical and boreal latitudes, while ECMWF agrees well with CRU (Figure
19.12c). The area-weighted average biases at any given latitude are relatively small,
with values of 2.92 Pa (0.22%) and 64.04 Pa (4.89%) for CMWF and NCEP, respectively. The 2000 and 2001 reanalyses intercomparison (Figure 19.12g) shows that
GMAO e is generally good in the middle and high latitudes of the northern hemisphere relative to ECMWF. GMAO underestimates e in the tropics and southern
hemisphere (54.75°S to 8°N), with an area-weighted bias of 102.93 Pa (5.57%) with
respect to e in ECMWF.
NCEP has considerably lower VPD than both CRU and ECMWF (Figure 19.12d).
The area-weighted mean bias along latitude reaches –185 Pa (–25.66%) compared
with CRU. ECMWF generally agrees well with CRU with an area-weighted average bias of 2.55 Pa (0.35%). Comparison of reanalyses from 2000 and 2001 (Figure
19.12h) shows that GMAO is closer to ECMWF than NCEP, although GMAO VPD
tends to be lower from approximately 5°S to 40°N, and there are some discrepancies between GMAO and ECMWF at other latitudes. Both Figures 19.12d and
19.12h show that NCEP has lower VPD overall and it is much lower in tropical
areas.
Figure 19.13 is the spatial pattern of annual total MODIS ET driven by GMAO,
ECMWF (ERA40), and NCEP1, and Figure 19.14 is the corresponding zonal mean.
Obviously, ET by GMAO has more detailed spatial ET variations than the other
two, largely because first, GMAO has the finest resolution (1.0° × 1.25°) among the
three meteorological datasets, and second, overall, GMAO has the best quality at the
global scale, except for its low radiation in equatorial regions. Although ERA40 also
has high accuracy, its S↓ s is largely underestimated in tropical regions, resulting in
unreliable low ET estimates in the tropics.
In extratropical regions, both ETs by GMAO and ERA-40 have similar magnitudes, while coarser spatial resolution of ERA40 (2.5° × 2.5°) reduces the details of
spatial pattern in estimated ET relative to that by GMAO. Although NCEP1 has the
highest S↓ s in the tropics (Figure 19.12e), its lowest VPD (Figure 19.12h) counter19.12e), its lowest VPD (Figure 19.12h) counter.12e), its lowest VPD (Figure 19.12h) counter19.12h) counter.12h) counteracts high solar radiation and results in a low ET, as shown in both Figures 19.13 and
19.14. This is because VPD is not a dominant stomatal conductance control for the
tropical forests, but the low VPD term will underestimate ET in the P–M equation.
Remote Sensing and Modeling of Global Evapotranspiration
in the tropics if other surface variables are accurate. Compared with the ECMWF
and NCEP zonal mean S↓ s from 2000 and 2001 (Figure 19.12e), GMAO S↓ s is relatively more accurate, except for much lower values in high latitudes of the northern
hemisphere.
Compared with the 30-year annual mean T avg in the CRU (Figure 19.12b), NCEP
tends to have lower values for almost all latitudes, with particularly large negative
biases (area-weighted bias of –1.43°C) from 20°S to 20°N. ECMWF generally agrees
well with the CRU but is somewhat higher in middle high latitudes and a little lower
in the tropics. For the 2000 and 2001 intercomparison (Figure 19.12f), NCEP tends
to have the lowest T avg , ECMWF has the highest, and GMAO is in the middle for
most latitudes. Although the bias in T avg is small, a small bias in temperature can
introduce relatively large errors in VPD and, consequently, in global ET because of
the nonlinear relationship between T avg and both VPD and ET.
A comparison of vapor pressure (e) shows that NCEP has higher values than
CRU in tropical and boreal latitudes, while ECMWF agrees well with CRU (Figure
19.12c). The area-weighted average biases at any given latitude are relatively small,
with values of 2.92 Pa (0.22%) and 64.04 Pa (4.89%) for CMWF and NCEP, respectively. The 2000 and 2001 reanalyses intercomparison (Figure 19.12g) shows that
GMAO e is generally good in the middle and high latitudes of the northern hemisphere relative to ECMWF. GMAO underestimates e in the tropics and southern
hemisphere (54.75°S to 8°N), with an area-weighted bias of 102.93 Pa (5.57%) with
respect to e in ECMWF.
NCEP has considerably lower VPD than both CRU and ECMWF (Figure 19.12d).
The area-weighted mean bias along latitude reaches –185 Pa (–25.66%) compared
with CRU. ECMWF generally agrees well with CRU with an area-weighted average bias of 2.55 Pa (0.35%). Comparison of reanalyses from 2000 and 2001 (Figure
19.12h) shows that GMAO is closer to ECMWF than NCEP, although GMAO VPD
tends to be lower from approximately 5°S to 40°N, and there are some discrepancies between GMAO and ECMWF at other latitudes. Both Figures 19.12d and
19.12h show that NCEP has lower VPD overall and it is much lower in tropical
areas.
Figure 19.13 is the spatial pattern of annual total MODIS ET driven by GMAO,
ECMWF (ERA40), and NCEP1, and Figure 19.14 is the corresponding zonal mean.
Obviously, ET by GMAO has more detailed spatial ET variations than the other
two, largely because first, GMAO has the finest resolution (1.0° × 1.25°) among the
three meteorological datasets, and second, overall, GMAO has the best quality at the
global scale, except for its low radiation in equatorial regions. Although ERA40 also
has high accuracy, its S↓ s is largely underestimated in tropical regions, resulting in
unreliable low ET estimates in the tropics.
In extratropical regions, both ETs by GMAO and ERA-40 have similar magnitudes, while coarser spatial resolution of ERA40 (2.5° × 2.5°) reduces the details of
spatial pattern in estimated ET relative to that by GMAO. Although NCEP1 has the
highest S↓ s in the tropics (Figure 19.12e), its lowest VPD (Figure 19.12h) counter19.12e), its lowest VPD (Figure 19.12h) counter.12e), its lowest VPD (Figure 19.12h) counter19.12h) counter.12h) counteracts high solar radiation and results in a low ET, as shown in both Figures 19.13 and
19.14. This is because VPD is not a dominant stomatal conductance control for the
tropical forests, but the low VPD term will underestimate ET in the P–M equation.