 |
| Titel |
Global-scale regionalization of hydrological model parameters using streamflow data from many small catchments |
| VerfasserIn |
Hylke Beck, Ad de Roo, Albert van Dijk, Tim McVicar, Diego Miralles, Jaap Schellekens, Sampurno Bruijnzeel, Richard de Jeu |
| Konferenz |
EGU General Assembly 2015
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 17 (2015) |
| Datensatznummer |
250109457
|
| Publikation (Nr.) |
 EGU/EGU2015-9364.pdf |
|
|
|
|
|
| Zusammenfassung |
| Motivated by the lack of large-scale model parameter regionalization studies, a large set of
3328 small catchments (< 10000 km2) around the globe was used to set up and evaluate five
model parameterization schemes at global scale. The HBV-light model was chosen because
of its parsimony and flexibility to test the schemes. The catchments were calibrated against
observed streamflow (Q) using an objective function incorporating both behavioral and
goodness-of-fit measures, after which the catchment set was split into subsets of
1215 donor and 2113 evaluation catchments based on the calibration performance.
The donor catchments were subsequently used to derive parameter sets that were
transferred to similar grid cells based on a similarity measure incorporating climatic
and physiographic characteristics, thereby producing parameter maps with global
coverage. Overall, there was a lack of suitable donor catchments for mountainous and
tropical environments. The schemes with spatially-uniform parameter sets (EXP2 and
EXP3) achieved the worst Q estimation performance in the evaluation catchments,
emphasizing the importance of parameter regionalization. The direct transfer of
calibrated parameter sets from donor catchments to similar grid cells (scheme EXP1)
performed best, although there was still a large performance gap between EXP1
and HBV-light calibrated against observed Q. The schemes with parameter sets
obtained by simultaneously calibrating clusters of similar donor catchments (NC10 and
NC58) performed worse than EXP1. The relatively poor Q estimation performance
achieved by two (uncalibrated) macro-scale hydrological models suggests there is
considerable merit in regionalizing the parameters of such models. The global HBV-light
parameter maps and ancillary data are freely available via http://water.jrc.ec.europa.eu. |
|
|
|
|
|
|