Application of parallel computing methods for improving efficiency of optimization in hydrologic and water quality modeling

Younggu Her, Fnu Cibin Raj, Indrajeet Chaubey

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Many heuristic algorithms for solving optimization problems have been proposed, and efforts have been made to improve their efficiency in finding optimal solutions. However, the sampling-based search strategies of the heuristic optimization algorithms still require considerable time to locate optima, particularly in the hyper-dimensional parameter space that are typical for distributed hydrologic models. This study evaluated the performance of parallel computing methods, 'spmd' and 'parfor,' of Matrix Laboratory (MATLAB) in improving computational efficiency of solving optimization problems using sampling-based algorithms in hydrologic and water quality modeling applications. The parallel computing methods were applied in calibrating 34 parameters for hydrologic simulation of the Soil and Water Assessment Tool (SWAT) model and in identifying watershed-scale optimum spatial distributions of corn stover removal using the AMALGAM optimization algorithm. In the applications, the SWAT model was calibrated with NSE of 0.86 and R2 of 0.93 at the watershed outlet for the calibration period of 1993 to 1999 using the parallel computing methods, and the calibrated model provided NSE of 0.84 and R2 of 0.92 for the validation period of 2000 to 2009. The results clearly demonstrated the effectiveness of the methods in reducing computational time of the parameter calibration and spatial optimization. The SWAT model code modification was not required, and changes in the source code of the original sequential optimization algorithm were minimal. Thus, the parallel computing methods of MATLAB are expected to provide simple and quick solutions for improving efficiency of solving the optimization problems in watershed model applications.

Original languageEnglish (US)
Pages (from-to)455-468
Number of pages14
JournalApplied Engineering in Agriculture
Volume31
Issue number3
DOIs
StatePublished - Jan 1 2015

Fingerprint

Parallel processing systems
Water quality
Watersheds
Soils
Calibration
Sampling
Water
Heuristic algorithms
Computational efficiency
Spatial distribution

All Science Journal Classification (ASJC) codes

  • Engineering(all)

Cite this

@article{ab1e315cde904455ac349434c80da86f,
title = "Application of parallel computing methods for improving efficiency of optimization in hydrologic and water quality modeling",
abstract = "Many heuristic algorithms for solving optimization problems have been proposed, and efforts have been made to improve their efficiency in finding optimal solutions. However, the sampling-based search strategies of the heuristic optimization algorithms still require considerable time to locate optima, particularly in the hyper-dimensional parameter space that are typical for distributed hydrologic models. This study evaluated the performance of parallel computing methods, 'spmd' and 'parfor,' of Matrix Laboratory (MATLAB) in improving computational efficiency of solving optimization problems using sampling-based algorithms in hydrologic and water quality modeling applications. The parallel computing methods were applied in calibrating 34 parameters for hydrologic simulation of the Soil and Water Assessment Tool (SWAT) model and in identifying watershed-scale optimum spatial distributions of corn stover removal using the AMALGAM optimization algorithm. In the applications, the SWAT model was calibrated with NSE of 0.86 and R2 of 0.93 at the watershed outlet for the calibration period of 1993 to 1999 using the parallel computing methods, and the calibrated model provided NSE of 0.84 and R2 of 0.92 for the validation period of 2000 to 2009. The results clearly demonstrated the effectiveness of the methods in reducing computational time of the parameter calibration and spatial optimization. The SWAT model code modification was not required, and changes in the source code of the original sequential optimization algorithm were minimal. Thus, the parallel computing methods of MATLAB are expected to provide simple and quick solutions for improving efficiency of solving the optimization problems in watershed model applications.",
author = "Younggu Her and {Cibin Raj}, Fnu and Indrajeet Chaubey",
year = "2015",
month = "1",
day = "1",
doi = "10.13031/aea.31.10905",
language = "English (US)",
volume = "31",
pages = "455--468",
journal = "Applied Engineering in Agriculture",
issn = "0883-8542",
publisher = "American Society of Agricultural and Biological Engineers",
number = "3",

}

Application of parallel computing methods for improving efficiency of optimization in hydrologic and water quality modeling. / Her, Younggu; Cibin Raj, Fnu; Chaubey, Indrajeet.

In: Applied Engineering in Agriculture, Vol. 31, No. 3, 01.01.2015, p. 455-468.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Application of parallel computing methods for improving efficiency of optimization in hydrologic and water quality modeling

AU - Her, Younggu

AU - Cibin Raj, Fnu

AU - Chaubey, Indrajeet

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Many heuristic algorithms for solving optimization problems have been proposed, and efforts have been made to improve their efficiency in finding optimal solutions. However, the sampling-based search strategies of the heuristic optimization algorithms still require considerable time to locate optima, particularly in the hyper-dimensional parameter space that are typical for distributed hydrologic models. This study evaluated the performance of parallel computing methods, 'spmd' and 'parfor,' of Matrix Laboratory (MATLAB) in improving computational efficiency of solving optimization problems using sampling-based algorithms in hydrologic and water quality modeling applications. The parallel computing methods were applied in calibrating 34 parameters for hydrologic simulation of the Soil and Water Assessment Tool (SWAT) model and in identifying watershed-scale optimum spatial distributions of corn stover removal using the AMALGAM optimization algorithm. In the applications, the SWAT model was calibrated with NSE of 0.86 and R2 of 0.93 at the watershed outlet for the calibration period of 1993 to 1999 using the parallel computing methods, and the calibrated model provided NSE of 0.84 and R2 of 0.92 for the validation period of 2000 to 2009. The results clearly demonstrated the effectiveness of the methods in reducing computational time of the parameter calibration and spatial optimization. The SWAT model code modification was not required, and changes in the source code of the original sequential optimization algorithm were minimal. Thus, the parallel computing methods of MATLAB are expected to provide simple and quick solutions for improving efficiency of solving the optimization problems in watershed model applications.

AB - Many heuristic algorithms for solving optimization problems have been proposed, and efforts have been made to improve their efficiency in finding optimal solutions. However, the sampling-based search strategies of the heuristic optimization algorithms still require considerable time to locate optima, particularly in the hyper-dimensional parameter space that are typical for distributed hydrologic models. This study evaluated the performance of parallel computing methods, 'spmd' and 'parfor,' of Matrix Laboratory (MATLAB) in improving computational efficiency of solving optimization problems using sampling-based algorithms in hydrologic and water quality modeling applications. The parallel computing methods were applied in calibrating 34 parameters for hydrologic simulation of the Soil and Water Assessment Tool (SWAT) model and in identifying watershed-scale optimum spatial distributions of corn stover removal using the AMALGAM optimization algorithm. In the applications, the SWAT model was calibrated with NSE of 0.86 and R2 of 0.93 at the watershed outlet for the calibration period of 1993 to 1999 using the parallel computing methods, and the calibrated model provided NSE of 0.84 and R2 of 0.92 for the validation period of 2000 to 2009. The results clearly demonstrated the effectiveness of the methods in reducing computational time of the parameter calibration and spatial optimization. The SWAT model code modification was not required, and changes in the source code of the original sequential optimization algorithm were minimal. Thus, the parallel computing methods of MATLAB are expected to provide simple and quick solutions for improving efficiency of solving the optimization problems in watershed model applications.

UR - http://www.scopus.com/inward/record.url?scp=84937207487&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84937207487&partnerID=8YFLogxK

U2 - 10.13031/aea.31.10905

DO - 10.13031/aea.31.10905

M3 - Article

VL - 31

SP - 455

EP - 468

JO - Applied Engineering in Agriculture

JF - Applied Engineering in Agriculture

SN - 0883-8542

IS - 3

ER -