Comparison of two model calibration approaches and their influence on future projections under climate change in the Upper Indus Basin
Naz, Bibi S.
Wortmann, Michel
Disse, Markus
Bowling, Laura C.
Bogacki, Wolfgang
DOI: https://doi.org/10.1007/s10584-020-02902-3
Persistent URL: http://resolver.sub.uni-goettingen.de/purl?gldocs-11858/10667
Naz, Bibi S.; Research Centre Jülich, Institute of Bio and Geosciences: Agrosphere (IBG-3), Jülich, Germany
Wortmann, Michel; Potsdam Institute for Climate Impact Research, Potsdam, Germany
Disse, Markus; Department of Civil, Geo. and Environmental Engineering, Technical University of Munich, Munich, Germany
Bowling, Laura C.; Department of Agronomy, Purdue University, West Lafayette, USA
Bogacki, Wolfgang; Department of Civil Engineering, University of Applied Sciences Koblenz, Koblenz, Germany
Abstract
This study performs a comparison of two model calibration/validation approaches and their influence on future hydrological projections under climate change by employing two climate scenarios (RCP2.6 and 8.5) projected by four global climate models. Two hydrological models (HMs), snowmelt runoff model + glaciers and variable infiltration capacity model coupled with a glacier model, were used to simulate streamflow in the highly snow and glacier melt–driven Upper Indus Basin. In the first (conventional) calibration approach, the models were calibrated only at the basin outlet, while in the second (enhanced) approach intermediate gauges, different climate conditions and glacier mass balance were considered. Using the conventional and enhanced calibration approaches, the monthly Nash-Sutcliffe Efficiency (NSE) for both HMs ranged from 0.71 to 0.93 and 0.79 to 0.90 in the calibration, while 0.57–0.92 and 0.54–0.83 in the validation periods, respectively. For the future impact assessment, comparison of differences based on the two calibration/validation methods at the annual scale (i.e. 2011–2099) shows small to moderate differences of up to 10%, whereas differences at the monthly scale reached up to 19% in the cold months (i.e. October–March) for the far future period. Comparison of sources of uncertainty using analysis of variance showed that the contribution of HM parameter uncertainty to the overall uncertainty is becoming very small by the end of the century using the enhanced approach. This indicates that enhanced approach could potentially help to reduce uncertainties in the hydrological projections when compared to the conventional calibration approach.