In this study we used daily run-off fields (0.5º x 0.5º) from the global hydrological model PCR-GLOBWB to calculate yearly water availability (van Beek et al., 2011; Wada et al., 2014). PCR-GLOBWB was forced using bias-corrected meteorological data (0.5º x 0.5º) as provided by the ISI-MIP project (Hempel et al. 2013) using five Global Climate Models (GCMs): GFDL-ESM2M, HadGEM2-ES, IPSL-CM5A-LR, MIROC-ESM-CHEM, NorESM1-M. Climate projections used in this study are based on three representative concentration pathways (RCPs), namely RCP2.6, RCP6.0 and RCP8.5 (Taylor et al., 2012; Van Vuuren et al., 2011). The resulting daily run-off values per climate projections were aggregated into yearly totals per water province using hydrological years, necessary in this assessment since the statistical analysis requires the time-series of annual water availability to be statistically independent. Within this study we follow the approach of Ward et al (2014) using long-term mean (LTM) maximum water availability as a proxy for the distinction between two types of hydrological years: 1) standard basins (October – September); 2) basins for which the mode month of long-term monthly maximum water availability falls in September, October or November (July – June). Four time periods were selected for further analyses, namely: historic (1975-2004); 2030 (2015-2044); 2050 (2035-2064); and 2080 (2065-2094). Per water province and for each climate projection and GCM a Gamma distribution using Maximum Likelihood Estimators (MLSEs) was fitted through each of the 30-year time-series of annual water availability, whilst excluding years with zero water availability. To test the accuracy of the estimated Gamma shape and scale parameters in approximating its original water availability distribution we applied the Kolmogorov-Smirnov test, in this application also known as the Lilliefors test. P-values of ≥0.001 were used in this study as acceptable goodness-of-fit. We used the resulting ‘accurate’ Gamma parameters to estimate annual water availability, under all GCM-RCP combinations, for nine return periods, ranging from 2 years up to 1000 years, while accounting for the probability of exceedance of zero water availability. We repeated these steps to estimate GCM ensemble-mean annual water availability per water province for the different return periods, covering the four time-periods and including all climate projections.