Water Resources Research

Supporting Information for

Global analysis of approaches for deriving total water storage changes from

GRACE satellites

Di Long1,2, Laurent Longuevergne3, and Bridget R. Scanlon1

1. Bureau of Economic Geology, Jackson School of Geosciences, University of Texas at Austin, Austin, Texas, USA

2. State Key Laboratory of Hydroscience and Engineering, Department of Hydraulic Engineering, Tsinghua University, Beijing, China

3. Geosciences Rennes, UMR CNRS 6118, Université de Rennes 1, Rennes, France

Contents of this file

Text S1

Figures S1 to S12

Tables S1 to S3

Additional Supporting Information (Files uploaded separately)

Captions for Tables S1 to S3

Introduction

This Supporting Information includes:

(1) Descriptions of physics of land surface models (LSM) used in this study, i.e., Noah2.7, Mosaic, VIC, and CLM2.0 in GLDAS-1 (1.1), NCAR’s CLM4.0 (1.2), and the WaterGAP Global Hydrological Model (WGHM2.2) (1.3). These models were used to generate scaling factors following the processing of GRACE data, i.e., truncation of the maximum degree and order of 60, destriping, and filtering using a 300 km Gaussian filter. In addition, three approaches of deriving total water storage (TWS) anomalies, i.e., the scaling factor approach, additive correction approach, and multiplicative correction approach were compared with TWS anomalies from WGHM2.2. Data availability can be found in the acknowledgment section of the context.

(2) Tables in the Supporting Information contain statistics of the comparison of three approaches of deriving TWS anomalies with WGHM output, i.e., correlation coefficient (Table S1), root mean square difference (RMSD, Table S2), and mean absolute difference (MAB, Table S3);

(3) Supplementary figures 1-11; and

(4) References in the Supporting Information.

Text S1.

Descriptions of LSMs used in this study

1.1 GLDAS-1 Land Surface Models

A Global Land Data Assimilation System (GLDAS) is a global, high-resolution, offline (uncoupled to the atmosphere) terrestrial modeling system that incorporates satellite- and ground-based observations in order to produce optimal fields of land surface states and fluxes in near-real time that are valuable for predicting climate change, weather, biological and agricultural productivity, and flooding, and for performing a wide range of studies in the broader biogeosciences [Rodell et al., 2004].

GLDAS-1 makes use of the new generation of ground- and space-based observation systems, which provide data to constrain the modeled land surface states. Constraints are applied in two ways: (1) by forcing the LSMs with observation-based meteorological fields, biases in atmospheric model-based forcing can be reduced, and (2) by employing data assimilation techniques, observations of land surface states can be used to curb unrealistic model states. There are four LSMs involved in GLDAS-1, including Noah2.7, Mosaic, VIC, and CLM2.0 that provide valuable hydrological state and flux variables of the land surface at a temporal resolution of 3 hour (or monthly) and a spatial resolution of 1°× 1° (0.25°×0.25° for the Noah model) for the period 1979-present. These LSMs do not consider groundwater storage changes and reservoir and lake storage changes. The simplification may, therefore, result in uncertain simulations of TWS changes over areas with extensive irrigation.

1.2 NCAR’s CLM4.0

CLM4.0 is the terrestrial component of the Community Earth System Model (CESM) and the Community Atmosphere Model (CAM) [Gent et al., 2011], a collaborative project between scientists in the Terrestrial Sciences Section (TSS) and the Climate and Global Dynamics Division (CGD) at the National Center for Atmospheric Research (NCAR) and the CESM Land Model Working Group. The model examines the physical, chemical, and biological processes by which terrestrial ecosystems affect and are affected by climate across a variety of spatial and temporal scales. The hydrologic cycle over land in this model includes interception of water by plant foliage and wood, throughfall and stemflow, infiltration, runoff, soil water, and snow. Components of terrestrial water storage output by the model include soil moisture, snow, vegetation canopy storage, channel storage in rivers, as well as unconfined aquifer storage, in which water storage changes below the model’s 10 layer soil column are tracked. The bulk aquifer layer receives recharge from the soil column when the water table is deeper than the lowest soil layer, and discharges baseflow to the river routing model [Lawrence et al., 2011].

1.3 WGHM

GRACE TWS changes from the scaling factor approach using CLM4.0 (the gridded product), and additive correction and multiplicative correction approaches were compared with TWS changes from the WaterGAP WGHM2.2 model which comprehensively considers SWS, SMS, GWS changes and human impact on water storage changes. WaterGAP [Alcamo et al., 2003] consists of both the WGHM model [Doll et al., 2003] and five water use models for the following sectors: irrigation [Doll and Siebert, 2002] and livestock, household, manufacturing and cooling of thermal power plants [Vassolo and Doll, 2005]. WGHM computes time series of rapid-surface and subsurface runoff, groundwater recharge and river discharge as well as water storage variations in canopy, snow, soil, groundwater, lakes, man-made reservoirs, wetlands and rivers as a function of climate, soil, land cover, relief and observed river discharge. Location and size of lakes, reservoirs, and wetlands are defined using the global lakes and wetland database (GLWD) [Lehner and Doll, 2004], with an addition of more than 6000 man-made reservoirs [Doll et al., 2003]. Groundwater storage is affected by diffuse groundwater recharge via the soil, which is modeled as a function of total runoff, relief, soil texture, hydrogeology and the existence of permafrost or glaciers. Focused groundwater recharge from rivers, lakes, and wetlands is not simulated in WGHM. This type of recharge may be important, in particular in semi-arid and arid regions.

Sectorial water uses for irrigation livestock, households, manufacturing and cooling of thermal power plants are computed using separate models. The irrigation model GIM [Doll and Siebert, 2002] computes consumptive water use, i.e., the part of the withdrawn water that evaporates during use. For all other sectors, both water withdrawal and consumptive water use are quantified by the water use models. Taking into account information on the water source, and making assumptions about irrigation water use efficiencies and return flows, the sub-model GWSWUSE computes net abstractions from groundwater and from surface water [Doll et al., 2012]. Net abstractions are computed as the difference between water withdrawals from the specific source and the return flows from water use to the source. Net abstractions are negative if abstractions (withdrawals) are less than return flows, which can only occur in the case of irrigation from surface water. In this comparison, we assume that WGHM TWS changes represent reality to compare the three approaches to restore GRACE signal because the WGHM model comprehensively considers water storage changes in surface water, soil layers, and aquifers, and especially accommodates human impacts on SWS and GWS changes.


Figure S1. A global map of aridity index (AI). Climate is classified as humid (AI > 0.65), sub-humid (AI ≤ 0.65 and > 0.5), semi-arid (AI ≤ 0.5 and > 0.2), arid (AI ≤ 0.2 and > 0.05), and hyperarid (AI ≤ 0.05).

Figure S2. A schematic showing GRACE data processing based on the scaling factor approach. Trapezoids denote inputs, diamonds denote processing, and rectangles denote intermediate variables. Outputs (rectangles with blue background) are GRACE TWS anomalies based on CLM scaling factors and Noah scaling factors.

Figure S3. Concepts related to approaches of correcting TWSA from GRACE signals (a). The red rectangle represents the basin function. The curve represents the effective basin function that is filtered using the same filter as GRACE data. The difference between the basin function and the effective basin function within an area of interest represents bias. The difference between the basin function and the effective basin function outside the study area of interest represents leakage. Note that bias and leakage can be estimated by synthetic data (e.g., Noah in GLDAS-1 in this study). GRACE signal for the area of interest has been filtered (b), resulting in the bias and leakage effects that could be corrected by synthetic data.

Figure S4. A schematic showing GRACE data processing based on the additive correction approach. Trapezoids denote inputs, diamonds denote processing, and rectangles denote intermediate variables. Output (the rectangle with blue background) is GRACE TWS anomalies from the, additive approaches.

Figure S5. A schematic showing GRACE data processing based on the multiplicative approach. Trapezoids denote inputs, diamonds denote processing, and rectangles denote intermediate variables. Output (the rectangle with blue background) is GRACE TWS anomalies from the multiplicative approaches.

Figure S6. TWS anomalies from the scaling factor approaches for the period Jan 2003-Jul 2013 using different LSMs for some basins of the dry basins, with showing monthly precipitation anomalies for the corresponding period.

Figure S7. TWS anomalies from the scaling factor approaches for the period Jan 2003-Jul 2013 using different LSMs for the Irrawaddy basin, with showing monthly precipitation anomalies for the corresponding period.

Figure S8. TWS anomalies from the scaling factor approaches for the period Jan 2003-Jul 2013 using different LSMs for the Yukon basin, with showing monthly precipitation anomalies for the corresponding period.

Figure S9. Comparison of scaling factors using the destriping filter in our study and those in Landerer and Swenson (2012) for 46 basins. Shading areas indicate uncertainties in scaling factors due to selection of different LSMs (CV).

Figure S10. Basins with relatively high CVs of monthly scaling factor time series and its temporally constant scaling factors for the period Jan 2003-Dec 2009 derived from GLDAS-1 Noah2.7.

Figure S11. TWS anomalies from the scaling factor approach with CLM4.0 scaling factors and CSR SH coefficients, the additive correction approach, and multiplicative correction approach, and from WGHM2.2 for some humid basins for the period Jan 2003-Dec 2009, with showing monthly precipitation anomalies for the corresponding period.

Figure S12. TWS anomalies from the scaling factor approach with CLM4.0 scaling factors and CSR SH coefficients, the additive correction approach, and multiplicative correction approach, and from WGHM2.2 for some high-latitude basins for the period Jan 2003-Dec 2009, with showing monthly precipitation anomalies for the corresponding period.

Table S1. Correlation coefficient from comparison of GRACE TWS and WGHM anomalies for the 60 basins. GRACE TWS anomalies include gridded TWS products from three processing centers with CLM4.0 scaling factors, the additive correction approach, and the multiplicative correction approach. Subscripts 1-5 represent TWS anomalies from CSR gridded, JPL gridded, GFZ gridded, additive correction approach, and multiplicative correction approaches, respectively.

Table S2. Root mean square difference (RMSD, mm) from comparison of GRACE TWS and WGHM anomalies for the 60 basins. GRACE TWS anomalies include gridded TWS products from three processing centers with CLM4.0 scaling factors, the additive correction approach, and the multiplicative correction approach. Subscripts 1-5 represent TWS anomalies from CSR gridded, JPL gridded, GFZ gridded, additive correction approach, and multiplicative correction approaches, respectively.

Table S3. Mean absolute bias (MAB, mm) from comparison of GRACE TWS and WGHM anomalies for the 60 basins. GRACE TWS anomalies include gridded TWS products from three processing centers with CLM4.0 scaling factors, the additive correction approach, and the multiplicative correction approach. Subscripts 1-5 represent TWS anomalies from CSR gridded, JPL gridded, GFZ gridded, additive correction approach, and multiplicative correction approaches, respectively.

References

Alcamo, J., P. Doll, T. Henrichs, F. Kaspar, B. Lehner, T. Rosch, and S. Siebert (2003), Development and testing of the WaterGAP 2 global model of water use and availability, Hydrol. Sci. J.-J. Sci. Hydrol., 48(3), 317-337.

Doll, P., and S. Siebert (2002), Global modeling of irrigation water requirements, Water Resour Res, 38(4).

Doll, P., F. Kaspar, and B. Lehner (2003), A global hydrological model for deriving water availability indicators: model tuning and validation, J. Hydrol., 270(1-2), 105-134.

Doll, P., H. Hoffmann-Dobrev, F. T. Portmann, S. Siebert, A. Eicker, M. Rodell, G. Strassberg, and B. R. Scanlon (2012), Impact of water withdrawals from groundwater and surface water on continental water storage variations, J Geodyn, 59, 143-156.

Gent, P. R., et al. (2011), The Community Climate System Model Version 4, J Climate, 24(19), 4973-4991.

Lawrence, D. M., et al. (2011), Parameterization Improvements and Functional and Structural Advances in Version 4 of the Community Land Model, J Adv Model Earth Sy, 3.

Lehner, B., and P. Doll (2004), Development and validation of a global database of lakes, reservoirs and wetlands, J Hydrol, 296(1-4), 1-22.

Rodell, M., et al. (2004), The Global Land Data Assimilation System, B Am Meteorol Soc, 85(3), 381-394.

Vassolo, S., and P. Doll (2005), Global-scale gridded estimates of thermoelectric power and manufacturing water use, Water Resour Res, 41(4).

10