hide
Free keywords:
Bias correction; Discharge simulation; Hydrological modelling; Statistical downscaling
Abstract:
In the current study, two regional climate models (MM5 and REMO) driven by different global boundary conditions (the ERA40 reanalysis and the ECHAM5 model) are one-way coupled to the uncalibrated hydrological process model PROMET to analyze the impact of global boundary conditions, dynamical regionalization and subsequent statistical downscaling (bilinear interpolation, correction of subgrid-scale variability and combined correction of subgrid-scale variability and bias) on river discharge simulation. The results of 12 one-way coupled model runs, set up for the catchment of the Upper Danube (Central Europe) over the historical period 1971-2000, prove the expectation that the global boundaries applied to force the RCMs strongly influence the accuracy of simulated river discharge. It is, however, noteworthy that all efficiency criteria in case of bias corrected MM5 simulations indicate better performance under ERA40 boundaries, whereas REMO-driven hydrological simulations better correspond to measured discharge under ECHAM5 boundaries. Comparing the hydrological results achievable with MM5 and REMO, the application of bias-corrected MM5 simulations turned out to allow for a more accurate simulation of discharge, while the variance in simulated discharge in most cases was better reflected in case of REMO forcings. The correction of subgrid-scale variability within the downscaling of RCM simulations compared to a bilinear interpolation allows for a more accurate simulation of discharge for all model configurations and all discharge criteria considered (mean monthly discharge, mean monthly low-flow and peak-flow discharge). Further improvements in the hydrological simulations could be achieved by eliminating the biases (in terms of deviations from observed meteorological conditions) inherent in the driving RCM simulations, regardless of the global boundary conditions or the RCM applied. In spite of all downscaling and bias correction efforts described, the RCM-driven hydrological simulations remain less accurate than those achievable with spatially distributed meteorological observations. © 2013 Springer-Verlag Berlin Heidelberg.
