Modeling the physics of storm surges

cated on the coast and 44% of the world’s population lives within 150 km of the ocean.1 Unfortunately, coastal regions are often low-lying and thus susceptible to an increase in seasurface elevation. A storm surge is a potentially devastating rise in the sea surface caused by extratropical cyclones or by tropical cyclones such as hurricanes and typhoons. Surges can lead to large loss of human life, destruction of homes and civil infrastructure, and disruption of trade, fisheries, and industry. Since tropical cyclones have lower interior pressures and higher wind speeds than extratropical cyclones, they typically produce significantly higher surges than extratropical cyclones do. Their effects extend across the western Atlantic Ocean, the Gulf of Mexico, and the western Pacific and Indian oceans. Hurricane Katrina, the 2005 storm that struck southern Louisiana and Mississippi in the US, and Cyclone Nargis, which made landfall in Myanmar in 2008, provide ample evidence of the ruinous effects of storm surges. As in most scientific fields, the early foundation of surge prediction relied heavily on observational data and on relationships that were suggested by those data. Surge events, however, are rare, and the associated historical data are limited. As described in the box at right, such limitations are one reason planners did not effectively act on warnings that New Orleans is vulnerable to hurricanes. Furthermore, the forces that drive surges are different from one storm to another, and a storm’s characteristics evolve. For all those reasons, direct deductions about the regional and local impacts are of limited use. The advent of digital computers in the 1960s led to discrete computational solutions of the governing equations, a significant advance over earlier empirical methods. But those early calculations were limited by the scope and scales of the modeled physics, the modest size of the regions over which the computations were performed, and the lack of spatial resolution of critical phenomena influencing the solutions. As a result, they required extensive caseand regionspecific calibration or tuning of the boundary forcing and model parameters. Indeed, those early codes were at the crossroads of physics simulators, which rely on fundamental underlying physics, and tuned interpolators, which rely on observational data. Unfortunately, extreme storms such as Katrina and Nargis often transcend the limits of local calibrations. Moreover, reliance on such calibrations precludes using models to answer important questions related to variations in the system itself or meteorological forcing mechanisms. An important example concerns the role of wetlands in coastal protection, which is difficult to assess if the predictive model is specifically tuned to work for the existing wetland configuration and for a select set of storms. Empirical rules of thumb based