RELIABILITY OF SIMULATED DISCHARGES FOR DIFFERENT GAUGE LOCATIONS IN A SEMI DISTRIBUTED RAINFALL RUNOFF MODEL

摘要

In terms of flood forecasting in alpine environments, predictions at different gauges as well as sites with exposed infrastructure within the catchment are required. The used semi-distributed hydrological model HQsim combines runoff formation and surface runoff routines with an implemented channel routing for river reaches. This allows the estimation of discharges at selected channel segments. As a case study a large alpine catchment with a size of 890 km2 is used. The uncertainty in the discharge prediction is investigated at three discharge gauges located along the main river. The basis of our experimental set-up are 15,000 samples describing the prior parameter distribution obtained by means of a Latin Hypercube sampling. Out of this, we calculated a Generalized Likelihood Uncertainty Estimation (GLUE) for the flood discharge at each gauging station. As informal likelihood a combination of different Nash Sutcliffe Efficiencies (NSE) is used covering summer season as well as flood periods containing peak discharges. Based on the behavioral parameter settings for each individual gauge, the model prediction distribution and their means for the remaining gauging stations are computed and analyzed.