Evaluation of a 3D unstructured-mesh finite element model for dam-break floods.

This paper aims to comprehensively understand a 3D model for simulating dam-break floods through three test cases. The advantages and characteristics of this model are evaluated from different perspectives. Firstly, vertical inertial is considered in the 3D model. In the existing 1D and 2D flooding models, vertical inertia is usually ignored, resulting in unreliable results in high vertical inertia situations. A circular dam break case was designed and widely used for testing 2D models. Here it has been adopted to test the 3D model. Evident difference has been observed when compared with 2D modelling results. In order to reduce the vertical inertial in this case, we scaled the vertical size by 1:100, while kept the original horizontal size, then amplified the results according to the Froude number scaling rule. The results of the scaled model is very close to 2D modelling results. Therefore, the difference between the 3D and 2D modelling results are mainly caused by the vertical inertia. Thus it can be seen that 3D modelling techniques are needed when vertical inertia is not negligible. Secondly, the use of unstructured mesh makes this model more powerful in operating with complex topography, which is very important in real-world application. Five similar cases with different topography demonstrated that the complexity of topography affects the numerical solution process directly. However, this 3D model has powerful ability in utilizing very complex terrain. Thirdly, sensitivity of the model to the thickness of the artificial thin layer in dry areas ( d 0 ) and mesh resolution ( Δx ) have been analysed with a realistic dike-break flooding case. Smaller d 0 produces more realistic results but costs more computational time. The model is more sensitive to mesh resolution when the terrain is more complex. Therefore, a multi-scale mesh with high-resolution in areas with complex terrain and low-resolution in flat regions is a good choice for this model. Also, trade-offs between modelling precision and efficiency should be considered when choosing a proper value of d 0 and mesh resolution. [ABSTRACT FROM AUTHOR]