Influence of three common calibration metrics on the diagnosis of climate change impacts on water resources

Seiller, G.; Roy, René; Anctil, François

doi:10.1016/j.jhydrol.2017.02.004

Cited by 37 publications

(25 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Kay et al 2009, Gosling et al 2011, but other studies have shown that hydrological model uncertainty can be greater for low flows than for mean or high flows (Vetter et al 2017, Giuntoli et al 2015. Similarly, Seiller et al (2017) showed that the choice of both the hydrological model and the objective function used for calibration can significantly alter the projected changes in low flow, thus the choice of hydrological model may be more important for drought analyses. Another factor not considered here is snow, due to limited availability of required temperature data.…”

Section: Discussionmentioning

confidence: 99%

National-scale analysis of low flow frequency: historical trends and potential future changes

et al. 2018

View full text Add to dashboard Cite

Article (refereed) -postprint Kay, A.L.; Bell, V.A.; Guillod, B.P.; Jones, R.G.; Rudd, A.C. 2018. Nationalscale analysis of low flow frequency: historical trends and potential future changes. Climatic Change, 147 (3-4). 585-599. AbstractThe potential impact of climate change on hydrological extremes is of increasing concern across the globe. Here a national-scale grid-based hydrological model is used to investigate historical trends and potential future changes in low flow frequency across Great Britain. The historical analyses use both observational data and ensemble data from a regional climate model . The results show relatively few significant trends in historical low flows (2-or 20-year return period), whether based on 7-day or 30-day annual minima. Significant negative trends seen in some limited parts of the country when using observational data are generally not seen when using climate model data. The future analyses use climate model ensemble data for both near-future and far-future timeperiods (2020-2049 and 2070-2099 respectively), which are compared to a baseline sub-period from the historical ensemble . The results show future reductions in low flows, which are generally larger in the south of the country, at the higher (20-year) return period, and for the later time-period. Reductions are more limited if the estimates of future potential evaporation include the effect of increased carbon dioxide concentrations on stomatal resistance. Such reductions in river flow could have significant impacts on the aquatic environment and on agriculture, and present a challenge for water managers, especially as reductions in water supply are likely to occur alongside increases in demand.

show abstract

Section: Discussionmentioning

confidence: 99%

National-scale analysis of low flow frequency: historical trends and potential future changes

et al. 2018

View full text Add to dashboard Cite

show abstract

“…In addition, much attention has been paid to parameter uncertainty assessment in hydroclimatic modeling recently [33,77]. In this study, both the yearly and multi-year calibration strategies with 1-, 2-and 4-periods were used for hydrological climate change impact assessment.…”

Section: Discussionmentioning

confidence: 99%

“…It would be more comprehensive to calibrate parameters over different periods, such as wet and dry climate periods [82], as well as by a multi-objective calibration approach [83]. In addition, Seiller et al, [77] showed that the diagnosis of the impacts of climate change on water resources is very much affected by the objective functions used for hydrological model calibration. Investigation of the parameters' nonstationarity in a changed climate could be another avenue for future studies [84].…”

Section: Discussionmentioning

confidence: 99%

Parameter Uncertainty of a Snowmelt Runoff Model and Its Impact on Future Projections of Snowmelt Runoff in a Data-Scarce Deglaciating River Basin

Xiang

Chen

et al. 2019

Water

View full text Add to dashboard Cite

The impacts of climate change on water resources in snow- and glacier-dominated basins are of great importance for water resource management. The Snowmelt Runoff Model (SRM) was developed to simulate and predict daily streamflow for high mountain basins where snowmelt runoff is a major contributor. However, there are many sources of uncertainty when using an SRM for hydrological simulations, such as low-quality input data, imperfect model structure and model parameters, and uncertainty from climate scenarios. Among these, the identification of model parameters is considered to be one of the major sources of uncertainty. This study evaluates the parameter uncertainty for SRM simulation based on different calibration strategies, as well as its impact on future hydrological projections in a data-scarce deglaciating river basin. The generalized likelihood uncertainty estimation (GLUE) method implemented by Monte Carlo sampling was used to estimate the model uncertainty arising from parameters calibrated by means of different strategies. Future snowmelt runoff projections under climate change impacts in the middle of the century and their uncertainty were assessed using average annual hydrographs, annual discharge and flow duration curves as the evaluation criteria. The results show that: (1) the strategy with a division of one or two sub-period(s) in a hydrological year is more appropriate for SRM calibration, and is also more rational for hydrological climate change impact assessment; (2) the multi-year calibration strategy is also more stable; and (3) the future runoff projection contains a large amount of uncertainty, among which parameter uncertainty plays a significant role. The projections also indicate that the onset of snowmelt runoff is likely to shift earlier in the year, and the discharge over the snowmelt season is projected to increase. Overall, this study emphasizes the importance of considering the parameter uncertainty of time-varying hydrological processes in hydrological modelling and climate change impact assessment.

show abstract

“…Squared error metrics, such as Nash Sutcliffe Efficiency (NSE) and Mean Square Error (MSE), have historically been thought to be useful to reduce simulation errors associated with high flow values (Oudin et al, 2006;Price et al, 2012;Seiller et al, 2017;de Boer-Euser et al, 2017). Although Gupta et al (2009) showed theoretically how and why the use of NSE and other MSE-based metrics for calibration results in the underestimation of peak flow events, our experience indicates that this notion continues to persist almost a decade later.…”

mentioning

confidence: 84%

On the choice of calibration metrics for high flow estimation using hydrologic models

Mizukami¹,

Rakovec²,

Newman³

et al. 2018

Preprint

View full text Add to dashboard Cite

Abstract. Calibration is an essential step for improving the accuracy of simulations generated using hydrologic models, and a key modeler decision is the selection of the performance metric to be optimized. It has been common to used squared error performance metrics, or normalized variants such as Nash-Sutcliffe Efficiency (NSE), based on the idea that their squarederror nature will emphasize the estimation of high flows. However, we find that NSE-based model calibrations actually result in poor reproduction of high flow events, such as the annual peak flows that are used for flood frequency estimation. Using 5 three different types of performance metrics, we calibrate two hydrological models, the "Variable Infiltration Capacity" model (VIC) and the "mesoscale Hydrologic Model" (mHM) and evaluate their ability to simulate high flow events for 492 basins throughout the contiguous United States. The metrics investigated are (1) NSE, (2) Kling-Gupta Efficiency (KGE) and variants, and (3) Annual Peak Flow Bias (APFB), where the latter is an application-specific "hydrologic signature" metric that focuses on annual peak flows. As expected, the application specific APFB metric produces the best annual peak flow estimates; however, 10 performance on other high flow related metrics is poor. In contrast, the use of NSE results in annual peak flow estimates that are more than 20% worse, primarily due to the tendency of NSE to result in underestimation of observed flow variability.Meanwhile, the use of KGE results in annual peak flow estimates that are better than from NSE, with only a slight degradation in performance with respect to other related metrics, particularly when a non-standard weighting of the components of KGE is used. Overall this work highlights the need for a fuller understanding of performance metric behavior and design in relation to 15 the desired goals of model calibration.

show abstract

Influence of three common calibration metrics on the diagnosis of climate change impacts on water resources

Cited by 37 publications

References 61 publications

National-scale analysis of low flow frequency: historical trends and potential future changes

National-scale analysis of low flow frequency: historical trends and potential future changes

Parameter Uncertainty of a Snowmelt Runoff Model and Its Impact on Future Projections of Snowmelt Runoff in a Data-Scarce Deglaciating River Basin

On the choice of calibration metrics for high flow estimation using hydrologic models

Contact Info

Product

Resources

About