Author(s): Lorenzo Begnudelli; Brett F. Sanders
Linked Author(s): Brett F. Sanders
Keywords: Finite volume; Shallow water equations; Dams; Disasters; Flood modelling
Abstract: Numerical simulation of the 1928 St. Francis Dam failure in southern California was accomplished using a 2D Godunov-type finite-volume shallow-water algorithm run on an unstructured grid of triangular cells. The model was found to be accurate based on historical accounts, including flood maps and arrival time data, and sensitivity analysis was performed to determine factors that control the predictability of flooded area and flood arrival times. Results show that predictions of flood arrival times are sensitive to both mesh resolution and Manning coefficient (used to scale flow resistance), while predictions of flooded area were found to be relatively insensitive to the Manning coefficient. These results suggest that bed resistance controlled the speed of the St. Francis flood while flooded area was controlled by topography and the volume of released water. The study also revealed two types of previously unreported oscillatory surging in the dam-break flood. The first is due to a standing wave that develops in a tortuous reach of channel downstream of the dam. The wave is excited by reflections off canyon walls and accounts for a 30% fluctuation in discharge. The second is due to a mode-two standing wave in the reservoir. This wave is caused by the reflection of dam-break rarefaction waves off reservoir walls, and accounts for only 2-3% fluctuation in discharge. Both oscillations are therefore shown to be physically based and should not be interpreted as spurious oscillations common to many numerical wave models.
Year: 2007