 |
| Titel |
Multi-criteria evaluation of hydrological models |
| VerfasserIn |
Oldrich Rakovec, Martyn Clark, Albrecht Weerts, Mary Hill, Ryan Teuling, Remko Uijlenhoet |
| Konferenz |
EGU General Assembly 2013
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 15 (2013) |
| Datensatznummer |
250074965
|
|
|
|
|
|
| Zusammenfassung |
| Over the last years, there is a tendency in the hydrological community to move from the
simple conceptual models towards more complex, physically/process-based hydrological
models. This is because conceptual models often fail to simulate the dynamics of the
observations. However, there is little agreement on how much complexity needs to
be considered within the complex process-based models. One way to proceed to
is to improve understanding of what is important and unimportant in the models
considered.
The aim of this ongoing study is to evaluate structural model adequacy using alternative
conceptual and process-based models of hydrological systems, with an emphasis on
understanding how model complexity relates to observed hydrological processes.
Some of the models require considerable execution time and the computationally
frugal sensitivity analysis, model calibration and uncertainty quantification methods
are well-suited to providing important insights for models with lengthy execution
times.
The current experiment evaluates two version of the Framework for Understanding
Structural Errors (FUSE), which both enable running model inter-comparison experiments.
One supports computationally efficient conceptual models, and the second supports
more-process-based models that tend to have longer execution times. The conceptual FUSE
combines components of 4 existing conceptual hydrological models. The process-based
framework consists of different forms of Richard’s equations, numerical solutions,
groundwater parameterizations and hydraulic conductivity distribution.
The hydrological analysis of the model processes has evolved from focusing only on
simulated runoff (final model output), to also including other criteria such as soil moisture
and groundwater levels. Parameter importance and associated structural importance are
evaluated using different types of sensitivity analyses techniques, making use of both robust
global methods (e.g. Sobol’) as well as several alternative local sensitivity analysis methods.
The latter methods can yield similar results, however they are much more computationally
frugal than the global methods and often are better suited to analysis of complex models.
Simple models are used to compare the global and local methods, and insights used to
interpret results for complex model for which the local methods are much more
convenient.
The analyses are carried out for a medium-sized catchment (200 km2) in the Belgian
Ardennes, for which meteorological, fluxnet data, in situ soil moisture and groundwater time
series are available. |
|
|
|
|
|
|