487
Validation for Terrestrial Water Storage Variations
observed by GRACE include all mass redistribution processes such as hydrologic redistribution in surface and subsurface, ocean water movements and tides, variations of atmospheric masses, and cryospheric changes. Therefore, appropriate corrections are required
to separate hydrological signals, and background models are often used to remove other
nonhydrological components. For example, the European Center for Medium-Range
Weather Forecasts pressure field is used to remove the effect of atmospheric mass redistribution. Modeled wind- and pressure-driven ocean motions are used to remove the
tide and ocean water movement effects. Glacial isostatic adjustment effect is corrected
by using postglacial rebound models. The errors in the short wavelengths, which appear
as a north to south–oriented long stripe pattern, are dominant because those signals are
attenuated more along the increasing latitudes. Swenson and Wahr (2006) investigated
the spectral signature of the correlated errors in Stokes coefficients and introduced a
filter to remove the “striping” problem. The version of the “dpc200711” dataset that was
used in this study was destriped by using a modified algorithm proposed by Chambers
(2006). The destriped filter is applied up to spherical harmonic degree 40 and truncates
at the order; thus, the data only contain wavelengths longer than approximately 1000
km. Afterward, the equivalent Gaussian smoother is applied with different half-widths
of smoothing radius. In this chapter, destriping and 0-, 300-, 500-km smoothing applied
monthly gravity fields from three major data centers are used to compare with TWS
estimates derived from other independent approaches.
Previous studies on the evaluation of GRACE accuracy (e.g., Rodell and Famiglietti
1999, 2001; Swenson and Wahr 2002; Swenson et al. 2003, 2006; Yeh et al. 2006)
have indicated that TWS variations would likely be detectable, depending on the size
of the region and the magnitude of the variations themselves. The accuracy of GRACE
estimates of water storage variability within a region depends on the GRACE measurement errors, and the degree to which the gravity signal from the water storage can
be separated from other time-variable gravity signals (e.g., atmospheric mass redistribution). In general, uncertainties in the GRACE-derived water storage variations
decrease with increasing spatial and temporal scales. GRACE is capable of estimating monthly changes in TWS to accuracies of better than 1 cm of water depth for
areas of >200,000 km 2 . Swenson et al. (2006) found that the GRACE TWS anomaly
estimates agree well with in situ measurements in Illinois averaged over an area of
~280,000 km 2 . Accuracy can be improved to better than a few millimeters for areas
of >1,000,000 km 2 . This achievement indicates the potential for GRACE to provide
direct measurements of seasonal TWS variations from individual river basins to continental water balance analyses, which are unprecedented in the history of hydrology.
However, GRACE signal separation can be a severe issue. GRACE gravity measurement made in space provides no information about the vertical distribution of mass. With
time-variable gravity signals alone, there is no way of telling whether a time-variable
gravity signal is caused by mass variability at the earth’s surface, in the atmosphere,
or deep within the mantle. Thus, GRACE data can be used to constrain only the vertically integrated water storage variability (assuming other geophysical contributions; for
example, atmospheric and solid earth mass changes are known or negligible) and cannot
separate soil moisture from surface water or from water deeper underground.
Six-year (2003–2008) GRACE monthly TWS anomaly (TWSA) data for the 20
selected river basins in Figure 20.1 are plotted in Figure 20.2. All of the nine GRACE
Validation for Terrestrial Water Storage Variations
observed by GRACE include all mass redistribution processes such as hydrologic redistribution in surface and subsurface, ocean water movements and tides, variations of atmospheric masses, and cryospheric changes. Therefore, appropriate corrections are required
to separate hydrological signals, and background models are often used to remove other
nonhydrological components. For example, the European Center for Medium-Range
Weather Forecasts pressure field is used to remove the effect of atmospheric mass redistribution. Modeled wind- and pressure-driven ocean motions are used to remove the
tide and ocean water movement effects. Glacial isostatic adjustment effect is corrected
by using postglacial rebound models. The errors in the short wavelengths, which appear
as a north to south–oriented long stripe pattern, are dominant because those signals are
attenuated more along the increasing latitudes. Swenson and Wahr (2006) investigated
the spectral signature of the correlated errors in Stokes coefficients and introduced a
filter to remove the “striping” problem. The version of the “dpc200711” dataset that was
used in this study was destriped by using a modified algorithm proposed by Chambers
(2006). The destriped filter is applied up to spherical harmonic degree 40 and truncates
at the order; thus, the data only contain wavelengths longer than approximately 1000
km. Afterward, the equivalent Gaussian smoother is applied with different half-widths
of smoothing radius. In this chapter, destriping and 0-, 300-, 500-km smoothing applied
monthly gravity fields from three major data centers are used to compare with TWS
estimates derived from other independent approaches.
Previous studies on the evaluation of GRACE accuracy (e.g., Rodell and Famiglietti
1999, 2001; Swenson and Wahr 2002; Swenson et al. 2003, 2006; Yeh et al. 2006)
have indicated that TWS variations would likely be detectable, depending on the size
of the region and the magnitude of the variations themselves. The accuracy of GRACE
estimates of water storage variability within a region depends on the GRACE measurement errors, and the degree to which the gravity signal from the water storage can
be separated from other time-variable gravity signals (e.g., atmospheric mass redistribution). In general, uncertainties in the GRACE-derived water storage variations
decrease with increasing spatial and temporal scales. GRACE is capable of estimating monthly changes in TWS to accuracies of better than 1 cm of water depth for
areas of >200,000 km 2 . Swenson et al. (2006) found that the GRACE TWS anomaly
estimates agree well with in situ measurements in Illinois averaged over an area of
~280,000 km 2 . Accuracy can be improved to better than a few millimeters for areas
of >1,000,000 km 2 . This achievement indicates the potential for GRACE to provide
direct measurements of seasonal TWS variations from individual river basins to continental water balance analyses, which are unprecedented in the history of hydrology.
However, GRACE signal separation can be a severe issue. GRACE gravity measurement made in space provides no information about the vertical distribution of mass. With
time-variable gravity signals alone, there is no way of telling whether a time-variable
gravity signal is caused by mass variability at the earth’s surface, in the atmosphere,
or deep within the mantle. Thus, GRACE data can be used to constrain only the vertically integrated water storage variability (assuming other geophysical contributions; for
example, atmospheric and solid earth mass changes are known or negligible) and cannot
separate soil moisture from surface water or from water deeper underground.
Six-year (2003–2008) GRACE monthly TWS anomaly (TWSA) data for the 20
selected river basins in Figure 20.1 are plotted in Figure 20.2. All of the nine GRACE