Journal of Hydrology | Joseph et al. 
Assessing impacts of climate change on hydrology involves global scale climate projections by General Circulation Models (GCMs), downscaling of global scale projections to regional scale by statistical methods or regional climate models and then use of regional outputs in hydrological simulations. Hydrological simulations considers varying inputs starting with soil characteristics, land cover, vegetation types, control structures to social parameters such as human interventions, irrigation and water use. This makes the model highly parametrized and at the same time highly uncertain due to the non-availability of majority of input parameters. Here, we compare the contributions of uncertainty from hydrological parameterization in the hydrological projections of climate change to that generated from the use of multiple climate models. The Ganga River Basin in India was selected as the study region. For regional climate change projections, we use dynamic downscaling outputs from Coordinated Regional Climate Downscaling Experiment (CORDEX) and statistical downscaling outputs from a transfer function forced with 3 GCMs, Institut Pierre Simon Laplace (IPSL), European Consortium Earth System Model (EC-EARTH) and MPI (Max Plank Institut) ESM (Earth System Model). Monte-Carlo Simulations (MCS) are performed with 1000 generated sets of sensitive model parameters for each of the GCM-regional model combination. We find that the observed time series of river discharge is reproduced well but with bias in low-flow conditions. This is probably associated with human intervention and poor representation of baseflow in VIC due to the neglected groundwater storage which feed the surface water during low flow condition. The future projections show that the major uncertainty lies across climate models for all the four seasons (MAM, JJAS, ON and DJF) and for all the hydrological variables, soil moisture, evapotranspiration (ET), water yield and river discharge. The uncertainty resulting from the MCS is quite small as compared to the climate model uncertainty. We are unable to find any added value in hydrological simulations by rigorous hydrological calibration and parameterization in absence of many required data, when the forcing meteorological data has huge uncertainty. Our findings highlight the need of convergence of climate models before the studies on hydrological impacts assessment and subsequent development of adaptation strategies.
Full text can be found here.