On the sensitivity of hurricane storm surge simulation to domain size

• Impact of model domain size on storm surge simulation under various hurricane and topographic conditions is investigated.
• A threshold domain size is defined to determine the minimum model domain size.
• Variation of threshold domain size under various hurricane and topographic conditions is investigated.
• A polar graphic chart is presented to indicate the threshold domain size for possible hurricane parameters.

The impact of model domain size on the simulated storm surge under various hurricane and topographic conditions is investigated by using a three-dimensional storm surge model. Idealized experiments are designed to simulate storm surge with different domain sizes under various hurricane intensities, radius of maximum wind (RMW), translation speeds and landfall directions, coastal ocean bottom slopes, and continental shelf extents. The results show that, in general, the simulated storm surge first increases with increasing domain size and then approaches a constant value. A “threshold domain size” is defined accordingly, so that differences among model results with different domain sizes larger than the threshold domain size are negligible. The threshold domain size is shown to be insensitive to hurricane intensities, but increases linearly with increasing hurricane RMWs. It also increases with increasing translation speeds and decreasing bottom slopes, when hurricanes approach land perpendicularly. The threshold domain size can also be affected by other factors such as the direction of the storm track of a landfalling hurricane and the extent of continental shelf. Considering the complex dependence of the threshold domain size on various hurricane parameters, a polar graphic chart is created for estimating the threshold domain size for possible hurricane landfall directions and RMWs. A real case study of hurricane Charley (2004) is conducted with different domain sizes. Model results indicate that the threshold domain size estimated from the idealized experiments is reasonable for practical applications.