A criterion of efficiency for rainfall-runoff models

Garrick, M.; Cunnane, C.; Nash, J.E.

doi:10.1016/0022-1694(78)90155-5

Cited by 136 publications

(77 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…It has been argued by previous authors that the utility of the Nash-Sutcliffe efficiency as a performance measure may be limited by bias in its evaluation (Garrick et al, 1978;Ma et al, 1998;Weglarczyk, 1998;Sauquet & Leblois, 2001), especially because σ i 2 is not necessarily smaller than 2 obs σ , when 2 obs σ is small. When behavioural models under the GLUE framework were selected with Nash-Sutcliffe efficiency >0.5, for example, those models having a good match between observed and modelled high flow (with much higher observation variance) were more likely to be chosen; while those having a good match between observed and modelled low flow (even under the same model variance but with lower observation variance) had less chances to be chosen (with lower Nash-Sutcliffe efficiency).…”

Section: Discussionmentioning

confidence: 99%

Parameter conditioning and prediction uncertainties of the LISFLOOD-WB distributed hydrological model

Pappenberger

Beven

et al. 2006

Hydrological Sciences Journal

View full text Add to dashboard Cite

Distributed hydrological models are considered to be a promising tool for predicting the impacts of global change on the hydrological processes at the basin scale. However, distributed models typically require values of many parameters to be specified or calibrated, which exacerbates model prediction uncertainty. This study uses the generalized likelihood uncertainty estimation (GLUE) technique to analyse the parameter sensitivities of a distributed hydrological model, LISFLOOD-WB. Discharge time series and event volume data of the Luo River at upstream and downstream sites, Lingkou and Lushi, are used to analyse parameter uncertainty. Eight key parameters in the model are selected for conditioning and sampled using the Monte Carlo method under assumed prior distributions. The results show that maximum efficiency of model performance is lower and the number of behavioural parameter sets giving acceptable performance is fewer in the Lingkou sub-basin than in the Lushi sub-basin with the same criteria of acceptability. For both sub-basins the distribution shape parameter B in the fast runoff generation scheme is the most sensitive in predicting both discharge time series and event volume at the outlet. It is also shown that the value of parameter B at which the highest efficiency is derived is shifted from a high value for Lushi to a low value for Lingkou, consistent with past experience of model calibration that the larger the basin, the larger the B value is. The channel Manning coefficient N c shows some sensitivity in the prediction of discharge time series, but less in the prediction of event volumes. The other key parameters show little sensitivity and good simulations are found across the full range of parameter values sampled. The uncertainty bounds of predicted discharges at the Lushi sub-basin are broad in the peak and narrow in the recession. The normalized difference between the upper and lower uncertainty bounds for both discharge and evapotranspiration are broad in summer and narrow in winter and that of recharge is the opposite.Key words distributed hydrological model; GLUE; parameter calibration; uncertainty Conditionnement de paramétrage et incertitudes de prévision du modèle hydrologique distribué LISFLOOD-WBRésumé Les modèles hydrologiques distribués sont considérés comme un outil prometteur pour la prévision des impacts du changement global sur les processus hydrologiques à l'échelle du bassin versant. Cependant, les modèles distribués nécessitent généralement la spécification ou le calage des valeurs de nombreux paramètres, ce qui aggrave l'incertitude de prévision de la modélisation. Cette étude utilise la technique GLUE (generalized likelihood uncertainty estimation) pour analyser les sensibilités de paramétrage du modèle hydrologique distribué LISFLOOD-WB. Des séries temporelles de débit et des données de volume événementiel de la Rivière Luo, au niveau des sites amont et aval de Lingkou et Lushi, sont utilisées pour analyser l'incertitude de paramétrage. Huit paramètres clefs du m...

show abstract

Section: Discussionmentioning

confidence: 99%

Parameter conditioning and prediction uncertainties of the LISFLOOD-WB distributed hydrological model

Pappenberger

Beven

et al. 2006

Hydrological Sciences Journal

View full text Add to dashboard Cite

show abstract

“…Indices of agreement are popular in the literature and generally involve the comparison of a deviation measure between simulated and observed time series with some reference measure (Nash and Sutcliffe, 1970;Garrick et al, 1978;Willmott, 1981;Legates and McCabe, 1999;Willmott et al, 2012). Common reference measures include deviation measures between the observed data points and the mean of observations, or deviation measures between the observed data points and a baseline or naive model of the variable being simulated.…”

Section: Evaluation Methodsmentioning

confidence: 99%

“…However, that variance encompasses the trend line behavior. Therefore, we also normalize RMSE for variable i in region j by the observed variance about the trend line, following the convention of comparing deviation measures to a selected baseline to provide more targeted information about model performance (Garrick et al, 1978;Willmott, 1984;Legates and McCabe, 1999).…”

Section: Metrics For Model Evaluationmentioning

confidence: 99%

Evaluation of integrated assessment model hindcast experiments: a case study of the GCAM 3.0 land use module

2017

View full text Add to dashboard Cite

Abstract. Hindcasting experiments (conducting a model forecast for a time period in which observational data are available) are being undertaken increasingly often by the integrated assessment model (IAM) community, across many scales of models. When they are undertaken, the results are often evaluated using global aggregates or otherwise highly aggregated skill scores that mask deficiencies. We select a set of deviation-based measures that can be applied on different spatial scales (regional versus global) to make evaluating the large number of variable-region combinations in IAMs more tractable. We also identify performance benchmarks for these measures, based on the statistics of the observational dataset, that allow a model to be evaluated in absolute terms rather than relative to the performance of other models at similar tasks. An ideal evaluation method for hindcast experiments in IAMs would feature both absolute measures for evaluation of a single experiment for a single model and relative measures to compare the results of multiple experiments for a single model or the same experiment repeated across multiple models, such as in community intercomparison studies. The performance benchmarks highlight the use of this scheme for model evaluation in absolute terms, providing information about the reasons a model may perform poorly on a given measure and therefore identifying opportunities for improvement. To demonstrate the use of and types of results possible with the evaluation method, the measures are applied to the results of a past hindcast experiment focusing on land allocation in the Global Change Assessment Model (GCAM) version 3.0. The question of how to more holistically evaluate models as complex as IAMs is an area for future research. We find quantitative evidence that global aggregates alone are not sufficient for evaluating IAMs that require global supply to equal global demand at each time period, such as GCAM. The results of this work indicate it is unlikely that a single evaluation measure for all variables in an IAM exists, and therefore sector-by-sector evaluation may be necessary.

show abstract

“…These values are very similar, even through the shape of the hydrograph for the simulation without storage is dissimilar to the observed hydrograph. This relates to the weakness of the Nash-Sutcliffe index noted by Garrick et al (1978), where model efficiency does not improve much with much better fitting models. The RMSE value for the model with full storage is ±0.28m, which is well within the performance of other comparable hydraulic models in the literature.…”

Section: Model Developmentmentioning

confidence: 99%

Rural land management impacts on catchment scale flood risk

Pattison¹

2010

Role of Hydrology in Managing Consequences of a Changing Global Environment

View full text Add to dashboard Cite

The full-text may be used and/or reproduced, and given to third parties in any format or medium, without prior permission or charge, for personal research or study, educational, or not-for-pro t purposes provided that:• a full bibliographic reference is made to the original source• a link is made to the metadata record in Durham E-Theses• the full-text is not changed in any way The full-text must not be sold in any format or medium without the formal permission of the copyright holders.Please consult the full Durham E-Theses policy for further details. Most approaches to this problem aim to upscale from individual grid cells to whole catchments, something that restricts the complexity of possible process representation, produces models that may not be parsimonious with the data needed to calibrate them and, faced with data uncertainties, provides computational limitations on the extent to which model uncertainty can be fully explored. Rather than upscaling to problems of concern, this thesis seeks to downscale from locations of known flood risk, as a means of identifying where land use management changes might be beneficial and then uses numerical modelling to identify the kinds of management changes required in those downscaled locations. Thus, the aim of this thesis is to test an approach to understanding the impacts of rural land management upon flood risk based upon catchment-to-source downscaling.This thesis uses the case study of the River Eden catchment (2400 km 2 ) as a test (1) it was shown to have a significant impact on downstream flood risk; and (2) it had range of data and information needed for modelling land use changes.The second part of this thesis explored the land management scenarios that could be used to reduce flood risk at the catchment scale. The scenarios to be tested were determined through a stakeholder participation approach, whereby workshops were held to brainstorm and prioritise land management options, and then to identify specific locations within the Eamont sub-catchment where they could tested. There were two main types of land management scenarios chosen: (1) landscape-scale changes, including afforestation and compaction; and (2) corresponding to a 5.8% decrease in peak discharge. A key conclusion is that land management practices have been shown to have an effect on catchment scale flooding, even for extreme flood events. However, the effect of land management scenarios are both spatially and temporally dependent i.e. the same land management practice has different effects depending on where it is implemented, and when implemented in the same location has different effects on different flood events.

show abstract

A criterion of efficiency for rainfall-runoff models

Cited by 136 publications

References 1 publication

Parameter conditioning and prediction uncertainties of the LISFLOOD-WB distributed hydrological model

Parameter conditioning and prediction uncertainties of the LISFLOOD-WB distributed hydrological model

Evaluation of integrated assessment model hindcast experiments: a case study of the GCAM 3.0 land use module

Rural land management impacts on catchment scale flood risk

Contact Info

Product

Resources

About