Sensitivity analysis for comparison, validation and physical-legitimacy of neural network-based hydrological models
This paper addresses the difficult question of how to perform meaningful comparisons between neural network-based hydrological models and alternative modelling approaches. Standard, goodness-of-fit metric approaches are limited since they only assess numerical performance and not physical legitimacy...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Published: |
IWA Publishing
2014
|
| Subjects: | |
| Online Access: | https://eprints.nottingham.ac.uk/28054/ |
| _version_ | 1848793496420352000 |
|---|---|
| author | Dawson, C.W. Mount, Nick J. Abrahart, R.J. Louis, J. |
| author_facet | Dawson, C.W. Mount, Nick J. Abrahart, R.J. Louis, J. |
| author_sort | Dawson, C.W. |
| building | Nottingham Research Data Repository |
| collection | Online Access |
| description | This paper addresses the difficult question of how to perform meaningful comparisons between neural network-based hydrological models and alternative modelling approaches. Standard, goodness-of-fit metric approaches are limited since they only assess numerical performance and not physical legitimacy of the means by which output is achieved. Consequently, the potential for general application or catchment transfer of such models is seldom understood. This paper presents a partial derivative, relative sensitivity analysis method as a consistent means by which the physical legitimacy of models can be evaluated. It is used to compare the behaviour and physical rationality of a generalised linear model and two neural network models for predicting median flood magnitude in rural catchments. The different models perform similarly in terms of goodness-of-fit statistics, but behave quite distinctly when the relative sensitivities of their inputs are evaluated. The neural solutions are seen to offer an encouraging degree of physical legitimacy in their behaviour, over that of a generalised linear modelling counterpart, particularly when overfitting is constrained. This indicates that neural models offer preferable solutions for transfer into ungauged catchments. Thus, the importance of understanding both model performance and physical legitimacy when comparing neural models with alternative modelling approaches is demonstrated. |
| first_indexed | 2025-11-14T19:01:13Z |
| format | Article |
| id | nottingham-28054 |
| institution | University of Nottingham Malaysia Campus |
| institution_category | Local University |
| last_indexed | 2025-11-14T19:01:13Z |
| publishDate | 2014 |
| publisher | IWA Publishing |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | nottingham-280542020-05-04T20:16:22Z https://eprints.nottingham.ac.uk/28054/ Sensitivity analysis for comparison, validation and physical-legitimacy of neural network-based hydrological models Dawson, C.W. Mount, Nick J. Abrahart, R.J. Louis, J. This paper addresses the difficult question of how to perform meaningful comparisons between neural network-based hydrological models and alternative modelling approaches. Standard, goodness-of-fit metric approaches are limited since they only assess numerical performance and not physical legitimacy of the means by which output is achieved. Consequently, the potential for general application or catchment transfer of such models is seldom understood. This paper presents a partial derivative, relative sensitivity analysis method as a consistent means by which the physical legitimacy of models can be evaluated. It is used to compare the behaviour and physical rationality of a generalised linear model and two neural network models for predicting median flood magnitude in rural catchments. The different models perform similarly in terms of goodness-of-fit statistics, but behave quite distinctly when the relative sensitivities of their inputs are evaluated. The neural solutions are seen to offer an encouraging degree of physical legitimacy in their behaviour, over that of a generalised linear modelling counterpart, particularly when overfitting is constrained. This indicates that neural models offer preferable solutions for transfer into ungauged catchments. Thus, the importance of understanding both model performance and physical legitimacy when comparing neural models with alternative modelling approaches is demonstrated. IWA Publishing 2014 Article PeerReviewed Dawson, C.W., Mount, Nick J., Abrahart, R.J. and Louis, J. (2014) Sensitivity analysis for comparison, validation and physical-legitimacy of neural network-based hydrological models. Journal of Hydroinformatics, 16 (2). pp. 407-424. ISSN 1464-7141 generalised linear model; index flood; neural network; partial derivative; physical legitimacy; sensitivity analysis; ungauged catchment http://www.iwaponline.com/jh/016/jh0160407.htm doi:10.2166/hydro.2013.222 doi:10.2166/hydro.2013.222 |
| spellingShingle | generalised linear model; index flood; neural network; partial derivative; physical legitimacy; sensitivity analysis; ungauged catchment Dawson, C.W. Mount, Nick J. Abrahart, R.J. Louis, J. Sensitivity analysis for comparison, validation and physical-legitimacy of neural network-based hydrological models |
| title | Sensitivity analysis for comparison, validation and physical-legitimacy of neural network-based hydrological models |
| title_full | Sensitivity analysis for comparison, validation and physical-legitimacy of neural network-based hydrological models |
| title_fullStr | Sensitivity analysis for comparison, validation and physical-legitimacy of neural network-based hydrological models |
| title_full_unstemmed | Sensitivity analysis for comparison, validation and physical-legitimacy of neural network-based hydrological models |
| title_short | Sensitivity analysis for comparison, validation and physical-legitimacy of neural network-based hydrological models |
| title_sort | sensitivity analysis for comparison, validation and physical-legitimacy of neural network-based hydrological models |
| topic | generalised linear model; index flood; neural network; partial derivative; physical legitimacy; sensitivity analysis; ungauged catchment |
| url | https://eprints.nottingham.ac.uk/28054/ https://eprints.nottingham.ac.uk/28054/ https://eprints.nottingham.ac.uk/28054/ |