 |
| Titel |
Evaluation of a hydrological model based on Bidirectional Reach (BReach) |
| VerfasserIn |
Katrien Van Eerdenbrugh, Stijn Van Hoey, Niko E. C. Verhoest |
| Konferenz |
EGU General Assembly 2016
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 18 (2016) |
| Datensatznummer |
250126054
|
| Publikation (Nr.) |
 EGU/EGU2016-5731.pdf |
|
|
|
|
|
| Zusammenfassung |
| Evaluation and discrimination of model structures is crucial to ensure an appropriate use of
hydrological models. When evaluating model results by aggregating their quality in (a subset
of) individual observations, overall results of this analysis sometimes conceal important
detailed information about model structural deficiencies. Analyzing model results within their
local (time) context can uncover this detailed information. In this research, a methodology
called Bidirectional Reach (BReach) is proposed to evaluate and analyze results of a
hydrological model by assessing the maximum left and right reach in each observation point
that is used for model evaluation. These maximum reaches express the capability of the
model to describe a subset of the evaluation data both in the direction of the previous (left)
and of the following data (right). This capability is evaluated on two levels. First, on the level
of individual observations, the combination of a parameter set and an observation
is classified as non-acceptable if the deviation between the accompanying model
result and the measurement exceeds observational uncertainty. Second, the behavior
in a sequence of observations is evaluated by means of a tolerance degree. This
tolerance degree expresses the condition for satisfactory model behavior in a data series
and is defined by the percentage of observations within this series that can have
non-acceptable model results. Based on both criteria, the maximum left and right
reaches of a model in an observation represent the data points in the direction of the
previous respectively the following observations beyond which none of the sampled
parameter sets both are satisfactory and result in an acceptable deviation. After
assessing these reaches for a variety of tolerance degrees, results can be plotted in
a combined BReach plot that show temporal changes in the behavior of model
results.
The methodology is applied on a Probability Distributed Model (PDM) of the river Grote
Nete upstream of Geel-Zammel with 1 106 randomly sampled parameter sets for three
separate years. Acceptable model results must fit in the 95 % uncertainty bounds of
observed discharges and tolerance degrees of 0 %, 5 %, 10 %, 20 % and 40 % are
applied. An evaluation of BReach results with regard to other variables, such as the
magnitude and the rate of change of the observed discharges enables to detect recurring
patterns in model errors. This results in an augmented understanding of the model’s
structural deficiencies, revealing the incapability of the PDM model to simulate both
high and low flow simulations with a single parameter set for this catchment. As
the methodology can be applied for different hydrological model structures, it is a
useful tool to gain understanding of the difference in behavior of competing models. |
|
|
|
|
|
|