501
Validation for Terrestrial Water Storage Variations
the quality of comparison in terms of both the timing and amplitude vary considerably
among basins. The comparisons are relatively good for the Columbia and Yukon basins,
but less satisfactorily in the Colorado, St. Lawrence, and Zambezi basins.
It should be indicated that most of previous analyses of GRACE TWS (for example,
see the work of Swenson et al. 2003, Wahr et al. 2004, Chen et al. 2005, Ramillien et al.
2005, and Seo et al. 2006) have focused on the monthly TWSAs rather than the TWSC
shown in Figures 20.8 and 20.9. As discussed by Yeh and Famiglietti (2008), reanalysis
data–based estimates of monthly TWS variations, as derived by taking integration from
the water balance–estimated TWSC, often contain an artificial multi year trend (“drift”)
due to the common problems such as observing system changes or soil moisture nudging in most climate model simulations. Therefore, without correction, TWS estimates
based on reanalysis data cannot be directly used to compare with other independent
estimates such as that from GRACE or from model simulations. Also, in most studies
of GRACE hydrologic applications (for example, see the work of Rodell and Famiglietti
1999, 2001, Syed et al. 2005, and Yeh et al. 2006), the interest has been in the total
TWSC (rather than TWS) because of its importance for water balance closure. Because
monthly TWSC is estimated from TWSA by taking the difference between monthly
water storage anomalies, small errors in TWSA will be amplified into larger discrepancies in the derived TWSC. Thus, the overall agreement in the seasonal cycle as shown in
Figures 20.8 and 20.9 is encouraging and indicates that, in addition to providing sound
estimates of monthly TWSA as shown previously (Figures 20.5 and 20.7), GRACE data
also have the potential for providing reasonable estimates of monthly TWSC at least for
the selected river basins in Figures 20.8 and 20.9.
20.6 CONCLUSIONS
In this chapter, GRACE TWS data are compared with other independent estimates
of TWS for selected regions and large river basins. The objective is to explore the
potential for GRACE to observe TWS variations and to develop a framework for
its validation. Long-term in situ measurements of soil moisture and groundwater
depth in Illinois are used to validate GRACE TWS estimates. Seasonal and interannual TWS variations at the continental scale are explored by comparing GRACE
TWS with the global-scale hydrological model simulation and estimated inferred
from combined land–atmosphere water balance computation for 20 selected large
river basins. In general, the comparisons yield encouraging results as can be judged
from the overall close agreement on the seasonal pattern of TWS variations between
GRACE TWS and other independent estimates (Figures 20.5 and 20.7 through 20.9).
The proposed GRACE validation framework is based on the combination of
comparing GRACE TWS data with direct observations, global-scale land surface
hydrological simulations constrained by observed streamflow, and combined land–
atmosphere water balance computation based on atmospheric reanalysis data. The
results obtained here can be expected to provide diagnostic information useful for
the GRACE validation over major river basins or large continental regions, as well
as to provide a benchmark for the LSM development and validation. Moreover, the
diagnostic study presented here is expected to shed light on the strengths and limitations of large-scale water balance computation and current atmospheric reanalysis
Validation for Terrestrial Water Storage Variations
the quality of comparison in terms of both the timing and amplitude vary considerably
among basins. The comparisons are relatively good for the Columbia and Yukon basins,
but less satisfactorily in the Colorado, St. Lawrence, and Zambezi basins.
It should be indicated that most of previous analyses of GRACE TWS (for example,
see the work of Swenson et al. 2003, Wahr et al. 2004, Chen et al. 2005, Ramillien et al.
2005, and Seo et al. 2006) have focused on the monthly TWSAs rather than the TWSC
shown in Figures 20.8 and 20.9. As discussed by Yeh and Famiglietti (2008), reanalysis
data–based estimates of monthly TWS variations, as derived by taking integration from
the water balance–estimated TWSC, often contain an artificial multi year trend (“drift”)
due to the common problems such as observing system changes or soil moisture nudging in most climate model simulations. Therefore, without correction, TWS estimates
based on reanalysis data cannot be directly used to compare with other independent
estimates such as that from GRACE or from model simulations. Also, in most studies
of GRACE hydrologic applications (for example, see the work of Rodell and Famiglietti
1999, 2001, Syed et al. 2005, and Yeh et al. 2006), the interest has been in the total
TWSC (rather than TWS) because of its importance for water balance closure. Because
monthly TWSC is estimated from TWSA by taking the difference between monthly
water storage anomalies, small errors in TWSA will be amplified into larger discrepancies in the derived TWSC. Thus, the overall agreement in the seasonal cycle as shown in
Figures 20.8 and 20.9 is encouraging and indicates that, in addition to providing sound
estimates of monthly TWSA as shown previously (Figures 20.5 and 20.7), GRACE data
also have the potential for providing reasonable estimates of monthly TWSC at least for
the selected river basins in Figures 20.8 and 20.9.
20.6 CONCLUSIONS
In this chapter, GRACE TWS data are compared with other independent estimates
of TWS for selected regions and large river basins. The objective is to explore the
potential for GRACE to observe TWS variations and to develop a framework for
its validation. Long-term in situ measurements of soil moisture and groundwater
depth in Illinois are used to validate GRACE TWS estimates. Seasonal and interannual TWS variations at the continental scale are explored by comparing GRACE
TWS with the global-scale hydrological model simulation and estimated inferred
from combined land–atmosphere water balance computation for 20 selected large
river basins. In general, the comparisons yield encouraging results as can be judged
from the overall close agreement on the seasonal pattern of TWS variations between
GRACE TWS and other independent estimates (Figures 20.5 and 20.7 through 20.9).
The proposed GRACE validation framework is based on the combination of
comparing GRACE TWS data with direct observations, global-scale land surface
hydrological simulations constrained by observed streamflow, and combined land–
atmosphere water balance computation based on atmospheric reanalysis data. The
results obtained here can be expected to provide diagnostic information useful for
the GRACE validation over major river basins or large continental regions, as well
as to provide a benchmark for the LSM development and validation. Moreover, the
diagnostic study presented here is expected to shed light on the strengths and limitations of large-scale water balance computation and current atmospheric reanalysis