Comparison of simplified physically based dam breach models

Abstract

Three simplified physically based earthen embankment breach models, NWS BREACH (Revision 1), HR BREACH (Version 4.1) and DLBreach, are used to calculate the breaching of twelve dams, and the results are compared against the measured data and the predictions by three parametric breach models. It is found that NWS BREACH may have large errors for cohesive embankments, since it uses a noncohesive sediment transport model and does not consider headcut erosion as a typical mode of cohesive dam breach. HR BREACH considers headcut and surface erosion modes and adopts various surface erosion equations for noncohesive and cohesive soils. DLBreach adopts a nonequilibrium total-load sediment transport model and headcut erosion model for noncohesive and cohesive embankment breaching, respectively. All the three physically based models can handle overtopping failure of homogeneous and composite dams, as well as piping failure. HR BREACH and DLBreach consider both one- and two-sided widening, whereas only DLBreach allows subbase erosion. The comparison shows that DLBreach has best overall performance. Sensitivity studies show that sensitivity of these three models to soil erodibility is case dependent, but overall, DLBreach and HR BREACH are more sensitive than NWS BREACH. In addition, it is demonstrated that an adequate physically based breach model can perform better and provide more detailed results than a parametric model.