 |
| Titel |
Getting a feel for parameters: using interactive parallel plots as a tool for parameter identification in the new rainfall-runoff model WALRUS |
| VerfasserIn |
Claudia Brauer, Paul Torfs, Ryan Teuling, Remko Uijlenhoet |
| Konferenz |
EGU General Assembly 2015
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 17 (2015) |
| Datensatznummer |
250111605
|
| Publikation (Nr.) |
 EGU/EGU2015-11741.pdf |
|
|
|
|
|
| Zusammenfassung |
| Recently, we developed the Wageningen Lowland Runoff Simulator (WALRUS) to fill the
gap between complex, spatially distributed models often used in lowland catchments and
simple, parametric models which have mostly been developed for mountainous catchments
(Brauer et al., 2014ab). This parametric rainfall-runoff model can be used all over the world
in both freely draining lowland catchments and polders with controlled water levels.
The open source model code is implemented in R and can be downloaded from
www.github.com/ClaudiaBrauer/WALRUS.
The structure and code of WALRUS are simple, which facilitates detailed investigation of
the effect of parameters on all model variables. WALRUS contains only four parameters
requiring calibration; they are intended to have a strong, qualitative relation with catchment
characteristics. Parameter estimation remains a challenge, however. The model
structure contains three main feedbacks: (1) between groundwater and surface water;
(2) between saturated and unsaturated zone; (3) between catchment wetness and
(quick/slow) flowroute division. These feedbacks represent essential rainfall-runoff
processes in lowland catchments, but increase the risk of parameter dependence and
equifinality.
Therefore, model performance should not only be judged based on a comparison between
modelled and observed discharges, but also based on the plausibility of the internal modelled
variables. Here, we present a method to analyse the effect of parameter values on internal
model states and fluxes in a qualitative and intuitive way using interactive parallel
plotting.
We applied WALRUS to ten Dutch catchments with different sizes, slopes and soil
types and both freely draining and polder areas. The model was run with a large
number of parameter sets, which were created using Latin Hypercube Sampling. The
model output was characterised in terms of several signatures, both measures of
goodness of fit and statistics of internal model variables (such as the percentage of rain
water travelling through the quickflow reservoir). End users can then eliminate
parameter combinations with unrealistic outcomes based on expert knowledge using
interactive parallel plots. In these plots, for instance, ranges can be selected for each
signature and only model runs which yield signature values in these ranges are
highlighted. The resulting selection of realistic parameter sets can be used for ensemble
simulations.
C.C. Brauer, A.J. Teuling, P.J.J.F. Torfs, R. Uijlenhoet (2014a): The Wageningen
Lowland Runoff Simulator (WALRUS): a lumped rainfall-runoff model for catchments
with shallow groundwater, Geoscientific Model Development, 7, 2313-2332,
www.geosci-model-dev.net/7/2313/2014/gmd-7-2313-2014.pdf
C.C. Brauer, P.J.J.F. Torfs, A.J. Teuling, R. Uijlenhoet (2014b): The Wageningen
Lowland Runoff Simulator (WALRUS): application to the Hupsel Brook catchment
and Cabauw polder, Hydrology and Earth System Sciences, 18, 4007-4028,
www.hydrol-earth-syst-sci.net/18/4007/2014/hess-18-4007-2014.pdf |
|
|
|
|
|
|