 |
| Titel |
Reducing equifinality of hydrological models by integrating Functional Streamflow Disaggregation |
| VerfasserIn |
Stefan Lüdtke, Heiko Apel, Manuela Nied, Peter Carl, Bruno Merz |
| Konferenz |
EGU General Assembly 2014
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 16 (2014) |
| Datensatznummer |
250099805
|
| Publikation (Nr.) |
 EGU/EGU2014-15631.pdf |
|
|
|
|
|
| Zusammenfassung |
| A universal problem of the calibration of hydrological models is the equifinality of
different parameter sets derived from the calibration of models against total runoff
values. This is an intrinsic problem stemming from the quality of the calibration
data and the simplified process representation by the model. However, discharge
data contains additional information which can be extracted by signal processing
methods. An analysis specifically developed for the disaggregation of runoff time series
into flow components is the Functional Streamflow Disaggregation (FSD; Carl &
Behrendt, 2008). This method is used in the calibration of an implementation of the
hydrological model SWIM in a medium sized watershed in Thailand. FSD is applied
to disaggregate the discharge time series into three flow components which are
interpreted as base flow, inter-flow and surface runoff. In addition to total runoff, the
model is calibrated against these three components in a modified GLUE analysis,
with the aim to identify structural model deficiencies, assess the internal process
representation and to tackle equifinality. We developed a model dependent (MDA) approach
calibrating the model runoff components against the FSD components, and a model
independent (MIA) approach comparing the FSD of the model results and the FSD of
calibration data. The results indicate, that the decomposition provides valuable
information for the calibration. Particularly MDA highlights and discards a number of
standard GLUE behavioural models underestimating the contribution of soil water to
river discharge. Both, MDA and MIA yield to a reduction of the parameter ranges
by a factor up to 3 in comparison to standard GLUE. Based on these results, we
conclude that the developed calibration approach is able to reduce the equifinality of
hydrological model parameterizations. The effect on the uncertainty of the model
predictions is strongest by applying MDA and shows only minor reductions for MIA.
Besides further validation of FSD, the next steps include an extension of the study
to different catchments and other hydrological models with a similar structure. |
|
|
|
|
|
|