Water Resources Management

, Volume 27, Issue 3, pp 715–729

Use of Gene-Expression Programming to Estimate Manning’s Roughness Coefficient for High Gradient Streams


DOI: 10.1007/s11269-012-0211-1

Cite this article as:
Azamathulla, H.M. & Jarrett, R.D. Water Resour Manage (2013) 27: 715. doi:10.1007/s11269-012-0211-1


Manning’s roughness coefficient (n) has been widely used in the estimation of flood discharges or depths of flow in natural channels. Therefore, the selection of appropriate Manning’s n values is of paramount importance for hydraulic engineers and hydrologists and requires considerable experience, although extensive guidelines are available. Generally, the largest source of error in post-flood estimates (termed indirect measurements) is due to estimates of Manning’s n values, particularly when there has been minimal field verification of flow resistance. This emphasizes the need to improve methods for estimating n values. The objective of this study was to develop a soft computing model in the estimation of the Manning’s n values using 75 discharge measurements on 21 high gradient streams in Colorado, USA. The data are from high gradient (S > 0.002 m/m), cobble- and boulder-bed streams for within bank flows. This study presents Gene-Expression Programming (GEP), an extension of Genetic Programming (GP), as an improved approach to estimate Manning’s roughness coefficient for high gradient streams. This study uses field data and assessed the potential of gene-expression programming (GEP) to estimate Manning’s n values. GEP is a search technique that automatically simplifies genetic programs during an evolutionary processes (or evolves) to obtain the most robust computer program (e.g., simplify mathematical expressions, decision trees, polynomial constructs, and logical expressions). Field measurements collected by Jarrett (J Hydraulic Eng ASCE 110: 1519–1539, 1984) were used to train the GEP network and evolve programs. The developed network and evolved programs were validated by using observations that were not involved in training. GEP and ANN-RBF (artificial neural network-radial basis function) models were found to be substantially more effective (e.g., R2 for testing/validation of GEP and RBF-ANN is 0.745 and 0.65, respectively) than Jarrett’s (J Hydraulic Eng ASCE 110: 1519–1539, 1984) equation (R2 for testing/validation equals 0.58) in predicting the Manning’s n.


Open-channel flow Manning’s roughness coefficient Gene-expression programming Artificial neural networks High gradient streams Mountain Rivers 

Copyright information

© Springer Science+Business Media Dordrecht 2012

Authors and Affiliations

  1. 1.River Engineering and Urban Drainage Research Centre (REDAC)Universiti Sains MalaysiaNibong TebalMalaysia
  2. 2.National Research Program, Paleohydrology and Climate Change, U.S. Geological SurveyLakewoodUSA

Personalised recommendations