 |
| Titel |
Evaluation of Different Performance Criteria for Calibrating Conceptual Hydrological Models |
| VerfasserIn |
Thomas Wöhling, Luis Samaniego, Rohini Kumar |
| Konferenz |
EGU General Assembly 2011
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 13 (2011) |
| Datensatznummer |
250051667
|
|
|
|
|
|
| Zusammenfassung |
| Different criteria can be used for the calibration and evaluation of conceptual hydrological
models with observed data. One of the most common performance criteria is the
Nash-Sutcliffe efficiency (NSE, Nash and Sutcliffe 1970). However, Gupta et al. (2009)
showed that there are systematic problems inherent with any optimization based on
formulations related to the NSE. In this study, 53 transfer parameters of the distributed
conceptual mHM model (Samaniego et al. 2010) for the upper part of the meso-scale
Goldersbach catchment (36.84 km²) in the Neckar basin (Germany) are calibrated using nine
years of observed daily runoff and the global search algorithm AMALGAM. The model
performance is evaluated using the following criteria: the NSE of discharge (NSEQ), the NSE
of log-transformed discharge (NSElnQ), the sum-squared error of discharge (SSEQ), the
sum-squared error of monthly discharge sums (SSEMQ), the King-Gupta efficiency (KGEQ,
Gupta et al. 2009), and three criteria obtained by a decomposition of the NSE. These criteria
are the linear correlation coefficient (R), a measure for relative variability (α), and a bias
component (β). The model calibration is posed in a multi-criteria context using three different
performance criteria. NSEQ, NSElnQ, and SSEMQ are used in the first optimization run
whereas R, α, and β are utilized in the second multi-criteria calibration run. In
each of the two AMALGAM runs, the trade-off between the different performance
criteria is determined by Pareto efficient parameter solutions. The extreme ends
of the bi-criterion Pareto fronts and an equally weighted compromise solution of
all three criteria are selected for further analysis. The results suggest that the 53
optimised transfer parameter values are very sensitive to the utilized performance
criterion. The parameter set that yields the largest NSEQ – value in the first run
tends to slightly favour peak flows but exhibits a very good overall performance
during the calibration period. On the other hand, the parameter set that yields the
largest NSElnQ – value (a common criterion to give emphasis to low flows) results in
relatively large volume balance errors and tends to underestimate variability. A very
good overall performance was achieved with the compromise solution of the first
optimization run that yielded better NSEQ, NSElnQ, SSEMQ, and SSEQ – values
than any of the solutions in the second AMALGAM run while maintaining R, α,
and β – values close to their ideal values. The results demonstrate the advantages
of multi-criterion calibration of conceptual hydrological models. To strengthen
confidence in the generality of the results, a number of other catchments will be tested. |
|
|
|
|
|
|