And that was just a start. In the year after Katrina, two independent investigations and the corps's own 25-million-dollar study painted a detailed picture of flaws in the planning, design, and construction of the levee system. The corps, in its defense, says it was hamstrung by a political process that tied the project to what the local sponsor wanted and, more important, could afford. "Basically, you had political influence on significant engineering decisions," says the corps's project manager for the hurricane protection system, Al Naomi. "We went from fighting surge at the Rigolets and Chef Menteur passes, to fighting surge at the lakefront, to fighting surge in the heart of a major American city. Failure at the Rigolets would have had far less consequences than failure on 17th Street."

TALLER, STRONGER FLOODWALLS now glisten in the breaches, their clean white concrete contrasting starkly with the still ruined neighborhoods behind them, while massive new black floodgates are poised to close the canals at the lakefront. The rebuilt hurricane protection system gives returning New Orleanians some sense of security. But the corps has yet to fix what many see as the weakest link in the system, the 76-mile (122 kilometers) ship channel called the Mississippi River Gulf Outlet—Mr. Go to the locals—which the corps dug east of town in the late 1950s and early 1960s.

On a steamy summer afternoon with squalls in the offing, coastal scientists Paul Kemp of Louisiana State University and John Lopez of the Lake Pontchartrain Basin Foundation set out by boat to inspect the "funnel," formed east of town by the levees lining the MRGO and another channel that converges with it, the Gulf Intracoastal Waterway. Computer models run by Kemp's colleagues at LSU show the funnel raised Katrina's massive surge by more than three feet in the Industrial Canal, overtopping and destroying floodwalls protecting the Lower Ninth Ward. Farther east, the storm surge hammered through more than eight miles (13 kilometers) of the MRGO levees, which in turn wiped out much of St. Bernard Parish.

St. Bernard residents had been clamoring for years for the corps to close the little-used channel they call the "hurricane highway." Touted as a shortcut to the port for ocean freighters, the channel instead destroyed tens of thousands of acres of wetlands. It brought in salt water that killed marsh plants, while the wakes of the few ships eroded the banks of the channel, widening it from 500 feet (150 meters) to almost a half mile (one kilometer) in places. One lesson of Katrina is simple, says Lopez: Close MRGO.

The corps says it now intends to do so. But when or how the channel might be shut down is anyone's guess. Congress has yet to give a green light. "If we don't close MRGO," says Lopez, "it might be time to do what my wife says and move to Kansas."

Though the corps denies that the channel amplified Katrina's surge, everyone agrees that its levees—St. Bernard's primary hurricane defense—failed miserably. The corps insists the structures simply weren't high enough to withstand Katrina's 17 feet (five meters) of surge and six-foot (two meters) waves. But at many of the breaches, the levees were built of weak sand and shell dredged from the canal itself. Kemp believes the shell-sand sections began to collapse as soon as the waves started breaking on them, long before the main surge hit. He also notes that where these levees were fronted by intact wetlands or trees, they survived. Where they ended directly in the water, they failed.

Old ways die hard in the bayou. Even after the dramatic failure of the shell sand in the levees, independent investigators found corps contractors using the same material to rebuild them. Only after the discovery was made public did the corps barge in yellowish clay from Mississippi to cap the levees. And parts of the new structures still have no buffer against erosion.