Application of feedforward artificial neural network in Muskingum flood routing: A black-box forecasting approach for a natural river system

Publikation: Beiträge in ZeitschriftenZeitschriftenaufsätzeForschungbegutachtet

Standard

Application of feedforward artificial neural network in Muskingum flood routing: A black-box forecasting approach for a natural river system. / Latt, Zaw Zaw.
in: Water Resources Management, Jahrgang 29, Nr. 14, a006, 11.2015, S. 4995-5014.

Publikation: Beiträge in ZeitschriftenZeitschriftenaufsätzeForschungbegutachtet

Harvard

APA

Vancouver

Bibtex

@article{38567b5c7eb9409d866049ba10cd1ac4,
title = "Application of feedforward artificial neural network in Muskingum flood routing: A black-box forecasting approach for a natural river system",
abstract = "Due to limited data sources, practical situations in most developing countries favor black-box models for real time flood forecasting. The Muskingum routing model, despite its limitations, is a widely used technique, and produces flood values and the time of the flood peak. This method has been extensively researched to find an ideal parameter estimation of its nonlinear forms, which require more parameters, and are not often adequate for flood routing in natural rivers with multiple peaks. This study examines the application of artificial neural network (ANN) approach based on the Muskingum equation, and compares the feedforward multilayer perceptron (FMLP) models to other reported methods that have tackled the parameter estimation of the nonlinear Muskingum model for benchmark data with a singlepeak hydrograph. Using such statistics as the sum of squared deviation, coefficient of efficiency, error of peak discharge and error of time to peak, the FMLP model showed a clear-cut superiority over other methods in flood routing of well-known benchmark data. Further, the FMLP routing model was also proven a promising model for routing real flood hydrographs with multiple peaks of the Chindwin River in northern Myanmar. Unlike other parameter estimation methods, the ANN models directly captured the routing relationship, based on the Muskingum equation and performed well in dealing with complex systems. Because ANN models avoid the complexity of physical processes, the study{\textquoteright}s results can contribute to the real time flood forecasting in developing countries, where catchment data are scarce.",
keywords = "Artificial neural network, Flood routing, Multilayer perceptron, Multiple-peaked hydrograph, Muskingum method, Nonlinear model, Ecosystems Research",
author = "Latt, {Zaw Zaw}",
year = "2015",
month = nov,
doi = "10.1007/s11269-015-1100-1",
language = "English",
volume = "29",
pages = "4995--5014",
journal = "Water Resources Management",
issn = "0920-4741",
publisher = "Springer Netherlands",
number = "14",

}

RIS

TY - JOUR

T1 - Application of feedforward artificial neural network in Muskingum flood routing

T2 - A black-box forecasting approach for a natural river system

AU - Latt, Zaw Zaw

PY - 2015/11

Y1 - 2015/11

N2 - Due to limited data sources, practical situations in most developing countries favor black-box models for real time flood forecasting. The Muskingum routing model, despite its limitations, is a widely used technique, and produces flood values and the time of the flood peak. This method has been extensively researched to find an ideal parameter estimation of its nonlinear forms, which require more parameters, and are not often adequate for flood routing in natural rivers with multiple peaks. This study examines the application of artificial neural network (ANN) approach based on the Muskingum equation, and compares the feedforward multilayer perceptron (FMLP) models to other reported methods that have tackled the parameter estimation of the nonlinear Muskingum model for benchmark data with a singlepeak hydrograph. Using such statistics as the sum of squared deviation, coefficient of efficiency, error of peak discharge and error of time to peak, the FMLP model showed a clear-cut superiority over other methods in flood routing of well-known benchmark data. Further, the FMLP routing model was also proven a promising model for routing real flood hydrographs with multiple peaks of the Chindwin River in northern Myanmar. Unlike other parameter estimation methods, the ANN models directly captured the routing relationship, based on the Muskingum equation and performed well in dealing with complex systems. Because ANN models avoid the complexity of physical processes, the study’s results can contribute to the real time flood forecasting in developing countries, where catchment data are scarce.

AB - Due to limited data sources, practical situations in most developing countries favor black-box models for real time flood forecasting. The Muskingum routing model, despite its limitations, is a widely used technique, and produces flood values and the time of the flood peak. This method has been extensively researched to find an ideal parameter estimation of its nonlinear forms, which require more parameters, and are not often adequate for flood routing in natural rivers with multiple peaks. This study examines the application of artificial neural network (ANN) approach based on the Muskingum equation, and compares the feedforward multilayer perceptron (FMLP) models to other reported methods that have tackled the parameter estimation of the nonlinear Muskingum model for benchmark data with a singlepeak hydrograph. Using such statistics as the sum of squared deviation, coefficient of efficiency, error of peak discharge and error of time to peak, the FMLP model showed a clear-cut superiority over other methods in flood routing of well-known benchmark data. Further, the FMLP routing model was also proven a promising model for routing real flood hydrographs with multiple peaks of the Chindwin River in northern Myanmar. Unlike other parameter estimation methods, the ANN models directly captured the routing relationship, based on the Muskingum equation and performed well in dealing with complex systems. Because ANN models avoid the complexity of physical processes, the study’s results can contribute to the real time flood forecasting in developing countries, where catchment data are scarce.

KW - Artificial neural network

KW - Flood routing

KW - Multilayer perceptron

KW - Multiple-peaked hydrograph

KW - Muskingum method

KW - Nonlinear model

KW - Ecosystems Research

UR - http://www.scopus.com/inward/record.url?scp=85027950645&partnerID=8YFLogxK

U2 - 10.1007/s11269-015-1100-1

DO - 10.1007/s11269-015-1100-1

M3 - Journal articles

AN - SCOPUS:85027950645

VL - 29

SP - 4995

EP - 5014

JO - Water Resources Management

JF - Water Resources Management

SN - 0920-4741

IS - 14

M1 - a006

ER -

DOI