A key challenge in enhancing flood forecast relies in the difficulty of reducing predictive uncertainty. The precipitation-dependent hydrologic uncertainty processor (HUP) is a flexible model independent Bayesian processor that can be used with any hydrologic model to provide probabilistic forecast. This study investigates the use of HUP with different hydrologic models for hydrologic uncertainty quantification in a flood forecasting scheme for a semiurban watershed of southern Ontario, Canada. The purpose is to better understand predictive uncertainty and enhance flood forecasting system reliability in semiurban conditions. HUP is based on Bayes' theorem, and it updates the prior distribution given available information at the forecast time to obtain the posterior distribution that is close to future unknown actual value. In this study, the hydrological model (HYMOD) and the modèle du Génie Rural à 4 paramètres Horaire (GR4H) were selected to work with HUP, and the Bayesian processor was calibrated using a number of selected flood events from 2005 to 2014. The performance of the processor was assessed by graphical tools and performance metrics, like reliability plots, Nash-Sutcliffe efficiency (NSE), and continuous ranked probability score (CRPS). Results show that HUP provides a robust framework and a reliable analytic-numerical method for the quantification of hydrologic uncertainty, the actual values are well captured by the uncertainty bounds, the CRPS values are relatively small, and reliability curves lie close to the bisector. The comparison between the NSE calculated from the output of the sole deterministic hydrologic model (HYMOD/GR4H) and from the median of the predictive distribution produced by HUP-HYMOD/HUP-GR4H, demonstrates that HUP has the ability to improve the deterministic forecast. For low peak flow events, HUP combining with different hydrologic models presents similar predictive performance, while for high peak flow events, a well performed deterministic model is required in HUP to produce an accurate probabilistic forecast.

Assessing Hydrologic Uncertainty Processor Performance for Flood Forecasting in a Semiurban Watershed

Biondi D.
2019

Abstract

A key challenge in enhancing flood forecast relies in the difficulty of reducing predictive uncertainty. The precipitation-dependent hydrologic uncertainty processor (HUP) is a flexible model independent Bayesian processor that can be used with any hydrologic model to provide probabilistic forecast. This study investigates the use of HUP with different hydrologic models for hydrologic uncertainty quantification in a flood forecasting scheme for a semiurban watershed of southern Ontario, Canada. The purpose is to better understand predictive uncertainty and enhance flood forecasting system reliability in semiurban conditions. HUP is based on Bayes' theorem, and it updates the prior distribution given available information at the forecast time to obtain the posterior distribution that is close to future unknown actual value. In this study, the hydrological model (HYMOD) and the modèle du Génie Rural à 4 paramètres Horaire (GR4H) were selected to work with HUP, and the Bayesian processor was calibrated using a number of selected flood events from 2005 to 2014. The performance of the processor was assessed by graphical tools and performance metrics, like reliability plots, Nash-Sutcliffe efficiency (NSE), and continuous ranked probability score (CRPS). Results show that HUP provides a robust framework and a reliable analytic-numerical method for the quantification of hydrologic uncertainty, the actual values are well captured by the uncertainty bounds, the CRPS values are relatively small, and reliability curves lie close to the bisector. The comparison between the NSE calculated from the output of the sole deterministic hydrologic model (HYMOD/GR4H) and from the median of the predictive distribution produced by HUP-HYMOD/HUP-GR4H, demonstrates that HUP has the ability to improve the deterministic forecast. For low peak flow events, HUP combining with different hydrologic models presents similar predictive performance, while for high peak flow events, a well performed deterministic model is required in HUP to produce an accurate probabilistic forecast.
Bayes theorem; Flood forecasting; Hydrologic uncertainty processor; Posterior distribution; Uncertainty quantification
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.11770/294672
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 11
  • ???jsp.display-item.citation.isi??? ND
social impact