Case Study of the Big Bay Dam Failure: Accuracy and Comparison of Breach Predictions

The Big Bay Dam embankment failure occurred on March 12, 2004, releasing $17,500,000{\mathrm{m}}^3$ $(14,200\text{acre}\text{-}\mathrm{ft})$ of water. In all, 104 structures were documented as being damaged or destroyed as a result of this failure. No human lives were lost. This paper documents data gathered and analyses performed on the hydraulics of the failure. High water levels from the failure were marked and measured. A HEC-RAS unsteady flow model was developed. Using observed breach geometry, HEC-RAS provided results that agreed with the measured high water marks from $-0.02\text{to}-0.90\mathrm{m}$ and $0.01\text{to}0.62\mathrm{m}$ with associated modeled flow depths ranging from $9.3\text{to}5.7\mathrm{m}$ (from $30\text{to}19\mathrm{ft}$ ). A peak breach flow of $4,160{\mathrm{m}}^3∕\mathrm{s}$ $(147,000{\mathrm{ft}}^3∕\mathrm{s})$ was predicted at the embankment. Breach peak flow prediction equations were found to substantially underpredict the peak flow indicated by HEC-RAS for this failure. HEC-RAS modeling utilizing predicted breach geometry and formation time also underpredicted the peak flow, but by a lesser amount. The National Resources Conservation Service models WinTR-20 and TR 66 were also assessed. WinTR-20 results compared reasonably well with the high water marks for this failure. TR-66 results did not compare well.