493
Validation for Terrestrial Water Storage Variations
The MATSIRO-TRIP model simulations conducted here span over the period of
GRACE mission from 2002 to 2007 with a global 1° × 1° grid resolution. Atmospheric
forcing (precipitation, temperature, radiations, pressure, humidity, and wind speed)
was based on atmospheric reanalysis data provided by the Japanese Meteorological
Agency Climate Data Assimilation System (Onogi et al. 2007), and an altitude correction has been applied to temperature, pressure, and humidity (Ngo-Duc et al. 2005).
Ensemble simulations were conducted by using five observed global precipitation
datasets to reduce the uncertainties in forcing variables. Specifically, three groundbased observational products and two hybrid satellite products were retrieved from the
Global Precipitation Climatology Centre, Global Precipitation Climatology Project,
and Climate Prediction Center. All the reanalysis and observed precipitation datasets
were bilinearly interpolated or aggregated into the 1° × 1° grids. Observed daily or
monthly precipitation was disaggregated on the basis of the temporal distribution of
6-hourly reanalysis precipitation fields. Input land surface properties, including land
cover, soil texture, and soil and vegetation parameters, were specified according to the
Global Soil Wetness Project 2 (Dirmeyer et al. 2006). Additional model parameters in
MATSIRO followed the default values in the work of Takata et al. (2003).
In the MATSIRO-TRIP hydrologic simulations, the total TWS consists of three
main components: soil moisture, snow water, and river storage. Soil moisture and
snow water were calculated as the arithmetic mean of model ensemble simulations.
To obtain optimal river storage simulations for the realization of temporal variations
of effective velocity, the Bayesian model averaging (Duan et al. 2007) was applied
to 10 TRIP runs with perturbed effective velocities ranging from 0.1 to 1.0 m/s with
equal intervals. The average of ensemble simulations was optimized to maximize the
weight-averaged likelihood of ensemble members, and the observed Global Runoff
Data Center (GRDC) river discharge data were taken as the training data. The procedure was performed for each individual river basin in Figure 20.1, since effective
velocity is highly dependent on basin-specific topography and river morphology. The
simulated total TWSA was spatially averaged over each basin for the comparison
with the GRACE TWS. More details on the MATSIRO-TRIP ensemble simulations
can be found in the work of Kim et al. (2009).
20.4.4 data uSed in the coMBined land–atMoSPheRe
wateR Balance coMPutation
The data needed for the combined land–atmosphere water balance computation
(Equation 20.3) include the vertical profiles of atmospheric humidity, wind speed,
and river discharge. For the atmospheric data, the output from the NCEP–Department
of Energy Reanalysis 2 (Kanamitsu et al. 2002) was used here. It has a 6-hourly
temporal resolution and a 2.5° × 2.5° horizontal resolution at eight pressure levels
(1000, 925, 850, 700, 600, 500, 400, 300 mbar). Additional details about the reanalysis data used here can be found in the work of Yeh and Famiglietti (2008). For the
runoff, observed daily data provided by the GRDC for the selected 20 river basins
were used. Since the majority of GRDC data are for the periods of the twentieth
century and the length of data varies from basin to basin, only five basins with the
data period covering the GRACE data period (2003–2008) were used in the analysis.
Validation for Terrestrial Water Storage Variations
The MATSIRO-TRIP model simulations conducted here span over the period of
GRACE mission from 2002 to 2007 with a global 1° × 1° grid resolution. Atmospheric
forcing (precipitation, temperature, radiations, pressure, humidity, and wind speed)
was based on atmospheric reanalysis data provided by the Japanese Meteorological
Agency Climate Data Assimilation System (Onogi et al. 2007), and an altitude correction has been applied to temperature, pressure, and humidity (Ngo-Duc et al. 2005).
Ensemble simulations were conducted by using five observed global precipitation
datasets to reduce the uncertainties in forcing variables. Specifically, three groundbased observational products and two hybrid satellite products were retrieved from the
Global Precipitation Climatology Centre, Global Precipitation Climatology Project,
and Climate Prediction Center. All the reanalysis and observed precipitation datasets
were bilinearly interpolated or aggregated into the 1° × 1° grids. Observed daily or
monthly precipitation was disaggregated on the basis of the temporal distribution of
6-hourly reanalysis precipitation fields. Input land surface properties, including land
cover, soil texture, and soil and vegetation parameters, were specified according to the
Global Soil Wetness Project 2 (Dirmeyer et al. 2006). Additional model parameters in
MATSIRO followed the default values in the work of Takata et al. (2003).
In the MATSIRO-TRIP hydrologic simulations, the total TWS consists of three
main components: soil moisture, snow water, and river storage. Soil moisture and
snow water were calculated as the arithmetic mean of model ensemble simulations.
To obtain optimal river storage simulations for the realization of temporal variations
of effective velocity, the Bayesian model averaging (Duan et al. 2007) was applied
to 10 TRIP runs with perturbed effective velocities ranging from 0.1 to 1.0 m/s with
equal intervals. The average of ensemble simulations was optimized to maximize the
weight-averaged likelihood of ensemble members, and the observed Global Runoff
Data Center (GRDC) river discharge data were taken as the training data. The procedure was performed for each individual river basin in Figure 20.1, since effective
velocity is highly dependent on basin-specific topography and river morphology. The
simulated total TWSA was spatially averaged over each basin for the comparison
with the GRACE TWS. More details on the MATSIRO-TRIP ensemble simulations
can be found in the work of Kim et al. (2009).
20.4.4 data uSed in the coMBined land–atMoSPheRe
wateR Balance coMPutation
The data needed for the combined land–atmosphere water balance computation
(Equation 20.3) include the vertical profiles of atmospheric humidity, wind speed,
and river discharge. For the atmospheric data, the output from the NCEP–Department
of Energy Reanalysis 2 (Kanamitsu et al. 2002) was used here. It has a 6-hourly
temporal resolution and a 2.5° × 2.5° horizontal resolution at eight pressure levels
(1000, 925, 850, 700, 600, 500, 400, 300 mbar). Additional details about the reanalysis data used here can be found in the work of Yeh and Famiglietti (2008). For the
runoff, observed daily data provided by the GRDC for the selected 20 river basins
were used. Since the majority of GRDC data are for the periods of the twentieth
century and the length of data varies from basin to basin, only five basins with the
data period covering the GRACE data period (2003–2008) were used in the analysis.