Mark Bernstein provides a link to this article in Civil Engineering called The Creeping Storm. A bit of it goes like this
In the 1980s Joseph Suhayda, then a coastal oceanographer in the civil engineering department at Louisiana State University (LSU), began to seek an answer to this question by simulating storms with a modified version of a hurricane model used by the Federal Emergency Management Agency (FEMA). Suhayda first began modeling the storms to help parishes in southeastern Louisiana determine appropriate flood elevations for FEMA’s National Flood Insurance Program. As his modeling capabilities improved, he began to more closely investigate the level of protection provided by the levees encircling New Orleans.
Suhayda’s model contains a geographic information system overlay that divides a fairly large boundary, from Alabama to Texas, into 0.6 mi (1 km) grids containing information about ground elevations, land masses, and waterways. The FEMA hurricane model does not draw on the same processing power as AdCirc and in general produces more liberal projections of flooding from storm surges. But by solving numerical equations representing a storm’s pressure, wind forces, and forward velocity, Suhayda was able to use the model to predict the storm surge associated with an actual hurricane dozens of hours before it hit land. By subtracting the elevations on a topographical map of coastal Louisiana from those surge values, he was able to approximate the flood risk of a given storm.
In the 1990s, Suhayda began modeling category 4 and 5 storms hitting New Orleans from a variety of directions. His results were frightening enough that he shared them with emergency preparedness officials throughout Louisiana. If such a severe storm were to hit the city from the southwest, for instance, Suhayda’s data indicate that the water level of Lake Pontchartrain would rise by as much as 12 ft (3.7 m). As the storm’s counterclockwise winds battered the levees on the northern shore of the city, the water would easily top the embankments and fill the streets to a depth of 25 ft (7.6 m) or more.
Suhayda’s model is not the only one that describes such a catastrophe. A model called SLOSH (Sea, Lake, and Overland Surges from Hurricanes), which is used by the National Weather Service and local agencies concerned with emergency preparedness, portrays an equally grim outcome should a storm of category 5 hit New Orleans. The SLOSH model does not contain nearly as many computational nodes as does AdCirc, it does not use a finite-element grid to increase the resolution of the nodes on shore, and its boundary is much smaller. Even so, its results are disheartening.
“Suppose it’s wrong,†says Combe, the Corps modeler. “Suppose twenty-five feet is only fifteen feet. Fifteen feet still floods the whole city up to the height of the levees.â€
Experts say a flood of this magnitude would probably shut down the city’s power plants and water and sewage treatment plants and might even take out its drainage system. The workhorse pumps would be clogged with debris, and the levees would suddenly be working to keep water in the city. Survivors of the storm—humans and animals alike—would be sharing space on the crests of levees until the Corps could dynamite holes in the structures to drain the area. In such a scenario, the American Red Cross estimates that between 25,000 and 100,000 people would die.
Some announcer on TV tonight wonders if it’s too dangerous for the help to go in. Pathetic. So much talk about security and looting. People are dying in the streets, in shelters. Water’s everywhere but there is no water. Drop tons of meals-to-eat on the islands of waiting people. Time for “overwhelming aide.” Overwhelming aide. We have the force side down.