Grid independence studies applied to a field-scale computational fluid dynamic (CFD) model using the detached eddy simulation (DES) technique along a reach of the Colorado River in Marble Canyon

Abstract

Grid independence studies have emerged as essential methodological frameworks for comprehending the impact of domain resolution on simulating anisotropic turbulence at the river-reach scale using large eddy simulation models. This study proposes a methodology to assess the loss of information in turbulent flow patterns when coarsening the computational domain, examined in a 1-km transect of the Colorado River along Marble Canyon. Seven computational domain resolutions are explored to analyse the sensitivity of turbulent flow to spatial resolution changes, utilizing the turbulent kinetic energy (TKE) spectrum technique and spatiotemporal analysis of eddy structures via statistical metrics such as root mean square error (RMSE), Kullback-Leibler (KL) divergence, Nash-Sutcliffe model efficiency coefficient (NSE), wavelet power spectrum and grid convergence index (GCI). Based on physical principles and statistics, these metrics quantify information loss and assess domain resolutions. A computational fluid dynamic (CFD) model is developed by employing the detached eddy simulation (DES) technique, with boundary condition (BC) integrating the rough wall extension of the Spallart-Allmaras model in cells near the bed. Evaluation of domain resolutions aims to identify grid cell sizes capturing flow behaviour and hydraulic characteristics, including primary and secondary flows, return currents, shear layers and primary and secondary eddies. The study observes an increase in data representation of the TKE spectrum with finer spatial domain resolution. Additionally, surface analysis, conducted via RMSE, KL and NSE metrics, identifies specific areas within the flow field showing high sensitivity to refining the grid cell sizes.