2.3. GRACE-based measurement

Data of RL05 spherical harmonic-based terrestrial water storage (TWS) were retrieved from the National Aeronautics and Space Administration (NASA) Jet Propulsion Laboratory (JPL) Tellus archive (Landerer and Swenson 2012). Gridded (1° × 1°) data files for 133 months were retrieved for the period 2003–2014. TWS data were used together in combination of three independent TWS solutions, from the Center for Space Research at the University of Texas at Austin, the NASA JPL and the German Space Agency (Geoforschungszentrum, GFZ). The detailed steps to obtain TWS from gravity measurements are available online¹. The Satellite Laser Ranging approach is used to replace degree 2 and order 0 coefficients in the RL05 spherical harmonics data (Cheng and Tapley 2004). Swenson et al. (2008) described a process to derive the degree 1 coefficients. The effect of subsurface elastic deformation related to post glacial rebound is accounted following A et al. (2013). A destriping filter is applied to remove correlated errors (Swenson et al., 2006). The data are processed with a Gaussian filter of 300 km width. In order to account for signal damping caused by this processing, scale factors are multiplied with the TWS data.

The RL05 mascon solutions were also used to derive TWS change (Watkins et al. 2015; Wiese et al. 2016). Similar techniques (as for the spherical harmonics – SH – products) are applied on the data for generating TWS information². The mascon (MS) approach is different from the SH approach in terms of post-processing filter application. For example, in JPL’s mascon approach, the entire globe is characterized as ~3° spherical mass concentration blocks with nearly equal area (Watkins et al. 2015). Use of a priori information facilitates correlated noise removal, which limits the use of post-processing filters (Watkins et al. 2015). Mascon products are not too much dependent on application of scale factors comparing the SH approach (Watkins et al. 2015). We applied scale factors with the TWS solutions.

Satellite-based groundwater storage anomaly signals can be disaggregated from the TWS anomalies (TWSA) after removing soil moisture (SMA) and surface water (SWA) anomalies:

Snow has been rarely observed in the northern-most part of the study region; hence, we ignore snow in this analysis. Continuous, ground-based measurements of soil moisture and surface water equivalents are very scarce in the region. We used Global Land Data Assimilation System (GLDAS) simulation outputs from NASA archives for estimating SMA and SWA (Rodell et al. 2004). An ensemble of three different models – the Community Land Model (CLM), Variable Infiltration Capacity (VIC), and Noah – is used to overcome the uncertainty (can appear as a function of different model physics) associated with any single model. Bhanja et al. (2016) showed better performance of the combination of models in contrast to any single model output comparing ground-based measurement in the study region. The satellite-based GWSA estimates in smaller basins are likely to have large errors because they are too small for GRACE to resolve and are only shown for comparison with the in situ observations.