 |
| Titel |
Calibration and Evaluation Techniques for Large-scale Hydrological Models |
| VerfasserIn |
Chantal Donnelly, Göran Lindström, Joel Dahné, Johan Strömqvist |
| Konferenz |
EGU General Assembly 2010
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 12 (2010) |
| Datensatznummer |
250032600
|
|
|
|
|
|
| Zusammenfassung |
| Widespread availability of regional and global databases as well as increases in computer processing speeds introduce the potential to setup and use traditional high-resolution hydrological models over multi-basin scales. In a multi-basin hydrological model, the spatial distribution of topography, soils and vegetation is used to predict spatially-varying catchment behaviour over the model domain, which may include many river basins, crossing regional and international boundaries, and a number of different geophysical and climatic zones. Important issues when setting up such a hydrological model are (1) how to evaluate model performance for both calibration and validation when simultaneously compared to many runoff gauging stations over the model domain, and (2) how to calibrate the model to optimise this model performance. The HYPE hydrological model was set up for two multi-basin applications, a pan-Swedish application and and an application for the Baltic Sea drainage basin, and a similar approach for model calibration and validation, based on representative hydrological response units (HRUs), tested on both applications. At these scales, larger gains in model performance were made by improving input data from regional and global databases, than by parameter tuning. Following sufficient improvements to input databases, parameters were tuned to optimise the performance of small subsets of gauged basins with dominant representations of each of the land and soil classes. This was shown to improve the performance of most of the basins in the catchment where measurements were available. This method was tested both in calibration of representative dominant gauged basins, then validated on both independent, individual gauged basins, as well as on all available data in the model domain. The study therefore presents a method for tuning parameters in a large-scale hydrological model and for validating the parameters for predictions in ungauged basins. The results indicate that this calibration approach can produce good overall model performance in a simultaneously calibrated model over a large, multi-basin model domain. |
|
|
|
|
|
|