The impact of climate change on the hydro-climatology of the Indian subcontinent is investigated by comparing statistics of current and projected future fluxes resulting from three RCP scenarios (RCP2.6, RCP4.5, and RCP8.5). Climate projections from the CORDEX-South Asia framework have been bias-corrected using the Distribution-Based Scaling (DBS) method and used to force the HYPE hydrological model to generate projections of evapotranspiration, runoff, soil moisture deficit, snow depth, and

more »

... lied irrigation water to soil. We also assess the changes in the annual cycles in three major rivers located in different hydro-climatic regions. Results show that conclusions can be influenced by uncertainty in the RCP scenarios. Future scenarios project a gradual increase in temperature (up to 7˝C on average), whilst changes (both increase and decrease) in the long-term average precipitation and evapotranspiration are more severe at the end of the century. The potential change (increase and decrease) in runoff could reach 100% depending on the region and time horizon. Analysis of annual cycles for three selected regions showed that changes in discharge and evapotranspiration due to climate change vary between seasons, whereas the magnitude of change is dependent on the region's hydro-climatic gradient. Irrigation needs and the snow depth in the Himalayas are also affected. Assessment of future climate change impacts on water resources commonly involves climate variables (i.e. precipitation, temperature) from global circulation models (GCMs) in combination with hydrological models [10, 11] . GCMs demonstrate significant skill at the continental and hemispheric spatial scales and incorporate a large proportion of the complexity of the global system; however, they are inherently unable to represent local basin-scale features and dynamics [12] . To narrow the gap between GCMs' abilities and hydrological needs, regional climate models (RCMs) have been developed to downscale the GCM output and, thus, provide high-resolution meteorological inputs to hydrological models. To improve the confidence in regional trends of hydro-climatic key variables and increase robustness in hydrological long term predictions, the World Climate Research Programme (WCRP) has recently launched a framework, called COordinated Regional climate Downscaling EXperiment (CORDEX), to generate and evaluate fine-scale ensembles of regional climate projections for all continents globally [13] . CORDEX has several domains that are defined as regions for which the regional downscaling is taking place. In particular, the efforts in the South Asia (SA) domain aim to translate regionally-downscaled climate data into meaningful sustainable development information in the monsoon South Asia area [14] . CORDEX-SA was initiated in 2012 and the RCM outputs have only recently become available. While RCMs transfer the large-scale information from GCMs to scales which are closer to the basin scale (10-50 km), the output often shows large bias in the magnitude and spatial distribution of precipitation and, to a lesser extent, temperature [15] . RCM data are, therefore, not considered to be directly useful for assessing hydrological impacts at the regional and/or local scale [16] . A way to tackle the problem of RCM misrepresentation is to bias correct the RCM data to make them reproduce historical observed statistics to the degree possible [17] . Different approaches to bias correction have been made, with various complexity [18] . Simpler methods include shifting long-term annual or seasonal means to agree with observations whereas more advanced methods include adjustment of the full frequency distribution. A distribution-based approach is attractive not least for precipitation, for which both bias and future change are generally found to depend on the intensity level [19] . Bias correction often includes an implicit downscaling component, in that higher-resolution reference observations are used when fitting the RCM mapping functions. Bias correction generally preserves the variability described by different climatic conditions generated by RCM projections [20] ; however, the RCM may perform differently depending on the season or governing atmospheric circulation. For instance, a typically wet weather regime (e.g., pattern or season) can have a different precipitation distribution in time and space than a dry regime. Projected hydrologic information is prone to considerable uncertainty/errors at various steps of the modelling chain, i.e., climate projection, bias correction and downscaling techniques, and hydrological simulation [21] [22] [23] . These errors can propagate in a very complex way (e.g., magnitude of error could vary both in space and time) which could be misinformative for management decisions [24, 25] . A major source of uncertainty, among others, concerns the future emission scenarios, described by the representative concentration pathways (RCP), which further results in different climate projections. [26] showed that towards the end of the 21st century, the emission scenarios (here RCPs) are the dominant source of uncertainty in climate projections. The spatiotemporal variability of water fluxes differs between RCPs, particularly in areas with unique weather systems, i.e., monsoon [27, 28] . However, the choice of GCM and RCM may also have a large impact on the results and generally an ensemble of projections-encompassing different GCMs, RCMs, and emission scenarios-is recommended in hydrological climate change impact assessments [29, 30] .