Alternative Approaches to a Calibration of Rainfall- Runoff Models for a Flood Frequency Analysis

Abstract

Abstract One of the tools which are currently being used in flood frequency analysis (FFA) is rainfall-runoff (RR) modelling. Its use in FFA often confronts the problem of how to correctly calibrate RR models to extreme flows. Since FFA only deals with extreme flows, traditional calibration techniques using simple objective functions such as the Nash-Sutcliffe model’s efficiency criterion are not sufficient. In this paper we have focused on proposing alternative approaches for calibration techniques of RR models in order to enhance the description of extreme flows. We have selected the HBV type conceptual, lumped model HRON as an RR model. We have suggested two alternative calibration approaches: 1) the use of a new optimization function that compares only values higher than the 95th percentile of observed flows and 2) using two sets of parameters to separately simulate low and high flows. Each of these improvements has enhanced the simulation of extreme flows, which has been demonstrated in the empirical cumulative distribution function calculated for the simulated and observed annual maximum series of flows. The results of this paper show that improvement can be obtained by both approaches, which give good agreement between observed and simulated extreme flows, while preserving a good simulation of low and medium flows