GRACE (Gravity Recovery and Climate Experiment) satellite data monitor large-scale changes in total terrestrial water storage (TWS), providing an invaluable tool where in situ observations are limited. Substantial uncertainty remains, however, in the amplitude of GRACE gravity signals and the disaggregation of TWS into individual terrestrial water stores (e.g. groundwater storage). Here, we test the phase and amplitude of three GRACE TWS signals from five commonly used gridded products (i.e. NASA's GRCTellus: CSR, JPL, GFZ; JPL-Mascons; GRGS GRACE) using in situ data and modelled soil moisture from the Global Land Data Assimilation System (GLDAS) in two sub-basins (LVB: Lake Victoria Basin; LKB: Lake Kyoga Basin) of the Upper Nile Basin. The analysis extends from January 2003 to December 2012, but focuses on a large and accurately observed reduction in TWS of 83 km 3 from 2003 to 2006 in the Lake Victoria Basin. We reveal substantial variability in current GRACE products to quantify the reduction of TWS in Lake Victoria that ranges from 80 km 3 (JPL-Mascons) to 69 and 31 km 3 for GRGS and GRCTellus respectively. Representation of the phase in TWS in the Upper Nile Basin by GRACE products varies but is generally robust with GRGS, JPL-Mascons, and GRCTellus (ensemble mean of CSR, JPL, and GFZ time-series data), explaining 90, 84, and 75 % of the variance respectively in "in situ" or "bottom-up" TWS in the LVB. Resolution of changes in groundwater storage (GWS) from GRACE TWS is greatly constrained by both uncertainty in changes in soil-moisture storage (SMS) modelled by GLDAS LSMs (CLM, NOAH, VIC) and the low annual amplitudes in GWS (e.g. 1.8-4.9 cm) observed in deeply weathered crystalline rocks underlying the Upper Nile Basin. Our study highlights the substantial uncertainty in the amplitude of TWS that can result from different data-processing strategies in commonly used, gridded GRACE products; this uncertainty is disregarded in analyses of TWS and individual stores applying a single GRACE product. 1 Introduction Satellite measurements under the Gravity Recovery and Climate Experiment (GRACE) mission have, since March 2002 (Tapley et al., 2004), enabled remote monitoring of large-scale (i.e. GRACE footprint: ∼ 200 000 km 2), spatiotemporal changes in total terrestrial water storage (TWS) at 10-day to monthly timescales (Longuevergne et al., 2013; Humphrey et al., 2016). Over the last 15 years, studies in basins around the world (Rodell and Famigli