Pros and cons of various efficiency criteria for hydrological model performance evaluation

Confidence in hydrological predictions is linked to the model's performance in reproducing available observations. However, judgment of a model's quality is challenged by the differences which exist among the available efficiency criteria or objective functions. In this study, model output...

Full description

Saved in:
Bibliographic Details
Published inProceedings of the International Association of Hydrological Sciences Vol. 385; pp. 181 - 187
Main Author Onyutha, Charles
Format Journal Article Conference Proceeding
LanguageEnglish
Published Gottingen Copernicus GmbH 18.04.2024
Copernicus Publications
Subjects
Online AccessGet full text
ISSN2199-899X
2199-8981
2199-899X
DOI10.5194/piahs-385-181-2024

Cover

More Information
Summary:Confidence in hydrological predictions is linked to the model's performance in reproducing available observations. However, judgment of a model's quality is challenged by the differences which exist among the available efficiency criteria or objective functions. In this study, model outputs based on several objective functions were compared and found to differ with respect to various circumstances of variability, number of outliers, and model bias. Computational difficulty or speed of a model during calibration was shown to depend on the choice of the efficiency criterion. One source of uncertainty in hydrological modelling is the selection of a particular calibration method. However, this study showed that the choice of an objective function is another sub-source of calibration-related uncertainty. Thus, tackling the issue of uncertainties on model results should comprise combination of modelled series obtained based on (i) various objective functions separately applied to calibrate a model, (ii) different calibration methods, and (iii) several hydrological models. The pros and cons of many new and old efficiency criteria which can be found explored in this study highlight the need for modellers to understand the impact of various calibration-related sub-sources of uncertainties on model outputs.
Bibliography:ObjectType-Article-1
ObjectType-Feature-2
SourceType-Conference Papers & Proceedings-1
content type line 22
ISSN:2199-899X
2199-8981
2199-899X
DOI:10.5194/piahs-385-181-2024