A genetic programming-based model for drag coefficient of emergent vegetation in open channel flows

Abstract

The estimation of drag exerted by vegetation is of great interest because of its importance in assessing the impact of vegetation on the hydrodynamic processes in aquatic environments. In the current research, genetic programming (GP), a machine learning (ML) technique based on natural selection, was adopted to search for a robust relationship between the bulk drag coefficient (Cd) for arrays of rigid circular cylinders representing emergent vegetation with blockage ratio (ψ), vegetation density (λ) and pore Reynolds number (Rep) based on published data. We utilize a data set covering a wide range of each parameter involved to cover all possible dependencies. A new predictor, which shares the same form with the Ergun-derived formula, was obtained without any pre-specified forms before searching. The dependence of the two parameters in Ergun equation on vegetation characteristics was also estimated by GP. This new Cd predictor for emergent vegetation with a relatively concise form exhibits a considerable improvement in terms of prediction ability relative to existing predictors.