Explicitly integrating parameter, input, and structure uncertainties into Bayesian Neural Networks for probabilistic hydrologic forecasting

Xuesong Zhang, Faming Liang, Beibei Yu, Ziliang Zong

Research output: Contribution to journalArticlepeer-review

57 Scopus citations

Abstract

Estimating uncertainty of hydrologic forecasting is valuable to water resources and other relevant decision making processes. Recently, Bayesian Neural Networks (BNNs) have been proved powerful tools for quantifying uncertainty of streamflow forecasting. In this study, we propose a Markov Chain Monte Carlo (MCMC) framework (BNN-PIS) to incorporate the uncertainties associated with parameters, inputs, and structures into BNNs. This framework allows the structure of the neural networks to change by removing or adding connections between neurons and enables scaling of input data by using rainfall multipliers. The results show that the new BNNs outperform BNNs that only consider uncertainties associated with parameters and model structures. Critical evaluation of posterior distribution of neural network weights, number of effective connections, rainfall multipliers, and hyper-parameters shows that the assumptions held in our BNNs are not well supported. Further understanding of characteristics of and interactions among different uncertainty sources is expected to enhance the application of neural networks for uncertainty analysis of hydrologic forecasting. © 2011 Elsevier B.V.
Original languageEnglish (US)
Pages (from-to)696-709
Number of pages14
JournalJournal of Hydrology
Volume409
Issue number3-4
DOIs
StatePublished - Nov 2011
Externally publishedYes

Bibliographical note

KAUST Repository Item: Exported on 2020-10-01
Acknowledged KAUST grant number(s): KUS-C1-016-04
Acknowledgements: We sincerely appreciate the three anonymous reviewers for their valuable comments that help significantly improve the manuscript, especially those comments on critical posterior analysis, reorganization of the sections, and linkage and difference between the new BNNs and those reported in previous research. Dr. Xuesong Zhang is supported by the DOE Great Lakes Bioenergy Research Center (DOE BER Office of Science DE-FC02-07ER64494, DOE BER Office of Science KP1601050, DOE EERE OBP 2046919145). This research is partially supported by grants from the National Science Foundation (DMS-0607755 and CMMI-0926803) and the award (KUS-C1-016-04) made by King Abdullah University of Science and Technology (KAUST). We thank Mr. David Manowitz at the Joint Global Change Research Institute, Pacific Northwest National Laboratory and University of Maryland for professional editing.
This publication acknowledges KAUST support, but has no KAUST affiliated authors.

Fingerprint

Dive into the research topics of 'Explicitly integrating parameter, input, and structure uncertainties into Bayesian Neural Networks for probabilistic hydrologic forecasting'. Together they form a unique fingerprint.

Cite this