Hydrological Model Parameter (In)stability – Implications for the Assessment of Climate Change Impacts on Flood Seasonality

Thursday, 18 December 2014
Klaus Vormoor1, Deborah Lawrence2, Maik Heistermann1 and Axel Bronstert1, (1)University of Potsdam, Institute of Earth- and Environmental Science, Potsdam, Germany, (2)Norwegian Water Resources and Energy Directorate, Oslo, Norway
Using a multi-model/multi-parameter ensemble consisting of (i) eight combinations of global and regional climate models, (ii) two statistical downscaling methods, and (iii) the HBV hydrological model with 25 calibrated parameter sets, we simulated daily discharge for a control (1961-1990) and future period (2071-2099) to investigate the potential impacts of climate change on flood seasonality and flood generating processes (FGPs) in six catchments with mixed snowmelt-rainfall regimes in Norway. For the catchments in northern and south-eastern Norway, we found more frequent autumn and winter events (partly also of higher magnitude) leading to possible shifts in the current flood regime from spring and early summer to autumn and winter. The possible shifts in flood regimes correspond to an increasing importance of rainfall as a FGP in all catchments considered, while rainfall replaces snowmelt as the dominant FGP in those catchments showing the largest changes in flood seasonality. The analysis of the relative role of the single ensemble components in contributing to overall uncertainty show that hydrological model parameter uncertainty is highest in those catchments showing the largest shifts in flood seasonality and FGPs. This points to difficulties in the time-transferability of the calibrated hydrological parameter sets under changing hydrometeorological conditions and highlights the need of alternative calibration approaches.

In this study, we detect time periods in the observation data sets of catchments showing changes in observed hydrometeorological conditions and differing phases of predominant flood seasonality. The HBV model is calibrated for the detected time periods using the Dynamically Dimensioned Search (DDS) global optimization algorithm, and split sampling tests are applied to study the role of the calibrated hydrological parameter sets under changing conditions. Preliminary results show that the hydrological model parameters are sensitive to the different calibration periods. The approach allows for detecting hydrological parameter sets which are better suited to simulate discharge for future conditions and which may help to reduce hydrological parameter uncertainty in the assessment of climate change impacts on flood seasonality and its underlying generation processes.