Why Natomas levees flunked

The corps does not require freeboard above this measurement.

Davis said the traditional approach requires 3 feet of freeboard everywhere, essentially ignoring the unique uncertainties at each location. But risk analysis sets heights based on calculations of risks at each location measured. Because of that, 3 feet of freeboard may be overkill on a river reach with gradual and predictable behavior. But it might be too little on a more "flashy" and unpredictable river.

"Risk analysis seeks to replace the 3 feet of freeboard factor of safety with more complete information," said Greg Kukas, chief of hydraulic design for the corps's Sacramento District. "Which, unfortunately, is more complicated, and not just for the public but for the engineers that are performing these processes. But it's a superior process."

The corps began using risk analysis in other parts of the nation in the mid-1990s. But, citing concern about the methods, Congress banned its use in planning new levees around Sacramento. Congress didn't say how projects to bolster existing levees should be approved, so after Hurricane Katrina, the corps began using risk analysis for the certification of local levees.

Countryman and others note that much depends on the assumptions behind the results. Faulty assumptions could lead to faulty results, and critics fear the many layers of mathematical sampling distance the results from reality.

"It's possible there will be samples that will have physically impossible combinations of things," Davis said. "That is the biggest complaint we hear."

Gerald Galloway, an engineer and retired corps brigadier general, agrees that risk analysis can be faulty, especially when applied in an environment as complex as California's.

Galloway, now an engineering professor at the University of Maryland, led a panel of experts that presented a report to the state Jan. 17 on Central Valley flood risk. The panel included input from Countryman and Ronald Stork, an analyst with Friends of the River, and it advises yet another approach to building up area levees.

The team wants levees prepared for the "standard project flood," a calculation that has more in common with traditional methods than with risk analysis. It is determined by estimating the worst combination of weather and hydrologic events "reasonably characteristic" of the region.

This would require levees tall enough to contain even 200- to 500-year floods. But this method is based on worst-case flows predictable from real weather and watershed conditions.

"That's easier to predict in California than trying to do the risk-based analysis," Galloway said.

About the writer:

• Call The Bee's Matt Weiser, (916) 321-1264.