On August 29, 2005, Hurricane Katrina’s surge pushed against the levees and floodwalls around New Orleans, and within a day most of the city was under water. The usual shorthand says the storm overtopped the defenses. For several of the breaches that flooded the most homes, the post-storm engineering investigations found something different: the walls along the drainage canals failed while the water was still below their tops. This guide uses the Corps of Engineers’ IPET study and the ASCE review of it to separate overtopping from structural failure. It also covers the design assumptions behind the system and what residual risk means for any agency that plans behind a levee.

The storm and the bowl: what reached New Orleans on August 29, 2005

Hurricane Katrina made landfall in southeast Louisiana near Buras, in Plaquemines Parish, early on the morning of August 29, 2005. The National Hurricane Center’s report on the storm rates it a Category 3 at that landfall, after it had reached Category 5 over the Gulf of Mexico the day before. Later that morning it came ashore again near the Louisiana and Mississippi state line. The highest surge struck the Mississippi coast, which suffered damage on a scale that deserves its own treatment. New Orleans lay west of the track, so its surge came from the east and south, through Lake Borgne, the marshes of St. Bernard Parish and the navigation channels cut through them, while the wind also raised Lake Pontchartrain along the city’s northern edge.

Much of New Orleans lies below sea level, inside basins ringed by levees and floodwalls. Rain that falls inside a basin has to be pumped out over the walls, and water that enters through a gap stays until the gap is closed and the pumps remove it. The basins that flooded worst were the Orleans East Bank, which holds most of the older city along with Lakeview, Gentilly and Mid-City, then New Orleans East, then the basin east of the Industrial Canal that takes in the Lower Ninth Ward and most of St. Bernard Parish. Pumped rainwater from the East Bank leaves the city through three outfall canals at 17th Street, Orleans Avenue and London Avenue, which run north to the lake. The canals are open to the lake, so when Lake Pontchartrain rose, the water in them rose too and pressed against floodwalls standing inside residential neighborhoods.

The figure usually published for the flooded share of the city is about 80 percent, and in some neighborhoods water stood for weeks while the Corps closed breaches and pumping resumed. I had no role in any of the investigations described below. Everything here comes from their published reports, and where those reports disagree I say so rather than choose for the reader.

The investigations: IPET, the ASCE review panel and the independent teams

The Corps of Engineers’ own study was the Interagency Performance Evaluation Task Force, known as IPET. The Chief of Engineers set it up in the fall of 2005 to determine how the hurricane protection system had performed. IPET drew its members from the Corps, other federal agencies, universities and private firms. It released a draft final report in June 2006 and issued its full multi-volume report in stages after that. The scope ran from surge and wave modeling, through the performance of each failed structure, to the pumping system and the consequences of the flooding.

Because the Corps was examining a system it had designed and built, its work also went through outside review. At the Corps’ request, the American Society of Civil Engineers convened an External Review Panel, which reviewed IPET’s analyses as they were produced. In 2007 the panel published its own summary, The New Orleans Hurricane Protection System: What Went Wrong and Why. A committee of the National Research Council also reviewed IPET. Two other teams worked independently of the Corps. One was the Independent Levee Investigation Team, led by Raymond Seed and Robert Bea of the University of California, Berkeley, with National Science Foundation support, which issued its final report in July 2006. The other was Team Louisiana, commissioned by the state.

The teams agreed on the central finding: several major breaches were failures of the structures themselves and cannot be explained as a storm too large for the system. They differed on some mechanisms, on the order of events at particular sites, and on how sharply to assign fault. The independent teams were generally more critical of the Corps than IPET was. When the IPET draft came out in June 2006, Lieutenant General Carl Strock, then Chief of Engineers, publicly acknowledged that the Corps’ design had been flawed. For anyone who needs site-level detail, the reports themselves remain available from the Corps and from ASCE, and they should be read in preference to any summary, including this one.

Overtopping, scour and foundation failure: how the investigators sorted the damage

The investigators sorted the damage at each site into distinct failure modes, and that sorting is the core of this article. In the simplest case, surge rises above the crest of a levee or the top of a floodwall and spills over for as long as the water stays high, while the structure remains standing. The flooding from that kind of overtopping is limited by how far the water rose above the crest and for how long, and the inflow stops once the surge falls.

The second mode starts with overtopping and ends in a breach. Water pouring over a floodwall lands on the soil behind the wall and scours a trench there, and a wall that depends on that soil for support can then lean and fall. Water running over an earthen levee erodes the crest and the landward slope, and a levee built of sand or shell fill can wash away entirely. Once the structure is gone, water keeps flowing in after the surge drops, until the levels inside and outside the basin equalize or crews close the gap.

The third mode is a failure before the water reaches the top. A floodwall or levee can lose stability through its foundation, either by sliding along a weak soil layer or because water pressure works through sandy soil beneath it, while the water against it is at or below the height it was designed to hold. IPET concluded that the most damaging breaches into the Orleans East Bank were of this kind. Its analysis attributed most of the flooding in the city’s main basins to water that came through breaches rather than to water spilling over intact structures.

The distinction matters for planning because the first two modes are exceedances of design capacity. A planner can anticipate those by asking what happens when a storm exceeds the design storm. The third mode is different, because the structure failed to deliver the capacity it was rated for, and a plan that assumes a wall holds until water reaches its top will set its evacuation and warning triggers too late for a wall that fails lower down. Breaches also change the volume of water, since an open breach can run for days and fill a basin toward the level of the lake outside. That is why the deepest and longest-lasting flooding was behind the failed walls.

The 17th Street and London Avenue canals: walls that failed below their tops

The 17th Street Canal breach opened on the east side of the canal, into Lakeview, on the morning of August 29. IPET and the independent teams found that the canal water had not reached the top of the floodwall when the wall failed. The walls along this canal were I-walls. An I-wall is a line of steel sheet piling driven into the crest of an earthen levee and capped with a concrete wall, and it stays upright because the soil around the buried sheet piling holds it.

As the canal rose, water pressure pushed the wall toward the neighborhood and opened a gap between the wall and the levee soil on the canal side. Water filled that gap, so the pressure now acted on the wall all the way down the buried sheet piling instead of only on the part standing above the levee. IPET found that this gap formation, which the original design analyses had not accounted for, combined with a weak clay layer in the foundation beneath the levee. The wall and a section of levee slid toward the neighborhood along a failure surface that passed beneath the tips of the sheet piles.

The London Avenue Canal breached in two places, one near Robert E. Lee Boulevard and one near Mirabeau Avenue, and overtopping did not cause either breach. The foundation soils at these sites included sand. IPET and the other teams found that water pressure from the high canal travelled through the sand beneath the levee and pushed up on the soil on the landward side, and at least one site lost foundation material by erosion through the soil beneath the levee. Gap formation behind the wall contributed here as well, and the teams weighed the mechanisms at each site somewhat differently. Sand from beneath the levees was left spread across streets in the neighborhoods behind the breaches.

The third outfall canal, at Orleans Avenue, did not breach. Water from the 17th Street and London Avenue breaches flowed into Lakeview, Gentilly, Mid-City and the surrounding East Bank neighborhoods. It kept flowing until the basin had filled toward the level of the lake, and the breaches had to be closed before the water could be pumped out.

The Industrial Canal and the St. Bernard levees: where overtopping set off the breaches

The Inner Harbor Navigation Canal, known locally as the Industrial Canal, connects the Mississippi River with Lake Pontchartrain. East of the city it joins the Gulf Intracoastal Waterway and the Mississippi River Gulf Outlet, a deep-draft channel usually called the MRGO. Surge from Lake Borgne moved up those channels toward the city. The investigators disagreed about how much the MRGO itself raised the surge that reached New Orleans, and the question was argued again in federal court. Anyone who needs that point should read the competing analyses in the reports themselves rather than rely on a summary.

Water rose above the floodwalls along the Industrial Canal, and on the east side the walls breached into the Lower Ninth Ward. IPET tied these breaches mainly to overtopping and to the scour it produced behind the walls. The teams did not all agree on the timing and sequence at every Industrial Canal breach. A loose barge ended up inside the neighborhood, and its role in the breach was disputed in later litigation. A rumor that spread in the Lower Ninth Ward, and that every published investigation contradicts, held that the levee was deliberately blown up to protect wealthier parts of the city. The rumor drew on real history, because in 1927 authorities did dynamite the Mississippi River levee at Caernarvon, below New Orleans, and flooded parts of St. Bernard and Plaquemines Parishes to relieve the river at the city. The 2005 investigations found the Industrial Canal breaches consistent with overtopping, scour and structural failure, and none of them reported evidence of explosives.

East of the Industrial Canal, long reaches of earthen levee along the MRGO and the waterway, protecting St. Bernard Parish and New Orleans East, were overtopped for hours. IPET found that many of these levees had been built with erodible fill. Long stretches were badly eroded or washed away, and nearly all of St. Bernard Parish flooded. In this case overtopping was the trigger, but the scale of the flooding depended on whether the levee could survive being overtopped. A levee armored against erosion lets water over for a few hours and remains in place, whereas one built of sand and shell can become a breach that stays open after the surge recedes. IPET and the ASCE panel both identified this lack of resilience, meaning the system’s inability to take overtopping without collapsing, as a design gap.

Design assumptions: the 1965 authorization, the design storm and the I-wall

Most of the system around New Orleans came from the Lake Pontchartrain and Vicinity Hurricane Protection Project, which Congress authorized in the Flood Control Act of 1965 after Hurricane Betsy flooded parts of the city that year. Its design storm was the standard project hurricane, a hypothetical storm defined with the meteorological data of the time. The common claim that the system was built for a Category 3 hurricane is a later simplification. The design storm was defined by parameters such as central pressure and wind speed and did not correspond to a Saffir-Simpson category, and the investigators noted that the criteria had not been updated as understanding of hurricanes improved. The original plan included barriers at the passes into Lake Pontchartrain. That plan stalled in the 1970s amid environmental litigation, and the Corps turned to raising levees and floodwalls along the lakefront and the canals instead. In 2005, forty years after authorization, the project was still incomplete.

The I-wall was an efficient way to raise protection on a narrow canal levee in a built-up neighborhood without buying more land. Its stability depends almost entirely on the soil it is driven into, so a gap behind the wall or a weak layer beneath it can defeat it at water levels well below its top. Both IPET and the ASCE panel judged the safety margins in the canal wall designs too thin for structures whose failure would flood a city. Readers who want the specific design values should take them from the ASCE summary and the IPET volumes on the floodwalls.

IPET also found that many structures stood lower than their authorized elevations. Part of the shortfall came from subsidence of the ground and the levees, and part came from building some sections to a vertical reference datum that was out of date, which left them shorter relative to real water levels than the drawings implied. Responsibility was divided as well. The Corps designed and built the structures, local levee districts operated and maintained them, and the Sewerage and Water Board ran the pumps and the outfall canals, while separately funded segments were finished at different times. IPET described the result as a system in name only. The ASCE panel called for a single entity accountable for the performance of the whole system. In 2006 Louisiana consolidated the levee districts on each side of the river into two regional flood protection authorities.

The death toll: two counting methods and the gap between them

No single death toll for Katrina is settled, and the figures in circulation differ mainly because they count different things. The most careful study of Louisiana deaths I know of is by Joan Brunkard, Gonza Namulanda and Raoult Ratard, published in Disaster Medicine and Public Health Preparedness in 2008. They reviewed the deaths reported in Louisiana and counted 971 that they attributed to the storm. They counted another 15 among Louisiana residents who had evacuated to other states, for a total of 986.

Other counts are higher. Louisiana’s health department published running totals of deceased victims in 2005 and 2006 that exceeded the study’s figure, and the study’s authors applied their own criteria for linking a death to the storm. The all-state figure most widely repeated, about 1,800 and usually given as 1,833, is associated with the National Hurricane Center’s tropical cyclone report on Katrina by Richard Knabb, Jamie Rhome and Daniel Brown, which was updated after its first release. That number covers every affected state, Mississippi included, so setting it beside the Louisiana study compares a multi-state count with a one-state count that uses different rules. Anyone who needs the current all-state number should check the NHC’s current version of the report, because secondary sources often repeat older figures without a date.

The published counts also do not separate deaths caused by the breaches from deaths caused by wind, surge outside the levee system, or interrupted medical care. The Louisiana study found drowning to be a leading cause and older residents heavily represented among the dead. Some people reported missing were never accounted for. Anyone who wants to say how many people the levee failures killed should state that no published count isolates that figure, and should then give the Louisiana and all-state figures with their sources and counting rules.

Risk reduction after 2005: the rebuilt system and the risk it leaves

After 2005 the Corps rebuilt and expanded the defenses as the Hurricane and Storm Damage Risk Reduction System. The change of name matters, because the old language of protection had encouraged residents and officials to treat the area behind the walls as safe. The new system is designed against a storm with a 1 percent chance of occurring in any year, often called the 100-year storm, which is the standard tied to the National Flood Insurance Program. The work included a surge barrier east of the city across the approach to the Industrial Canal, a large closure and pumping complex on the West Bank, and permanent pump stations at the lake ends of the three outfall canals so that surge no longer rises inside them. Many I-walls were replaced with T-walls, which have a broader base supported by piles, and levees were armored to survive overtopping.

A 1 percent annual standard still leaves real risk, and the arithmetic is simple. The chance of at least one exceedance over a 30-year mortgage is about 26 percent if each year is treated as independent. The ASCE panel and others have questioned whether a 1 percent standard is enough for a major city, and the Netherlands protects its most populated areas to standards set for far rarer events. The Corps has said that settlement of the levees and rising sea levels will require raising parts of the system over time to keep it at the 1 percent standard, so the rating describes the system as built and maintained and is not permanent.

Levees also change the character of the flood that eventually comes. Behind a levee, floods happen less often but arrive faster and deeper when they do, often with little warning, and they can fill a basin that development has turned into dense neighborhoods. Under the flood insurance program, areas behind accredited levees are commonly mapped outside the high-risk zone, so lenders do not require flood insurance there. FEMA’s own guidance tells residents of those areas that the risk is reduced and still present. For an emergency manager, the lesson from 2005 is to plan for a breach at a water level below the design height, with a warning trigger and an evacuation decision that do not depend on the structure performing exactly as rated.

What to do at your agency

• Have your emergency manager search the Corps of Engineers’ National Levee Database for every levee in or upstream of your jurisdiction, list each one’s sponsor and its most recent inspection rating, and file the list in the flood annex by the end of the month.
• Have your emergency manager ask each levee sponsor or district, in writing, for its current flood fight or emergency plan and any breach inundation mapping, and note in the annex which documents were received and which do not exist.
• Have the planner responsible for the flood annex write one paragraph stating the warning and evacuation trigger for a breach at a water level below the levee’s design height, naming who makes the call and on what gauge reading.
• Have your public information officer replace the word protected with reduced risk in the agency’s prewritten flood messages for levee areas, add a sentence recommending flood insurance, and file the revised templates with the joint information center set.
• Have your radio system administrator test one talkgroup or phone path to the levee district or dam operator during their night shift this month, and log who answered, how long it took, and whether the contact list was current.
• Have the coordinator of your hazard mitigation plan check whether the plan’s flood profile treats levee-protected areas as low risk, and add a breach scenario item to the agenda of the next scheduled planning committee meeting.

Takeaways

• Katrina’s surge reached New Orleans on August 29, 2005, and the flooding, usually reported as covering about 80 percent of the city, stood in some neighborhoods for weeks.
• IPET, the ASCE External Review Panel, the Berkeley-led independent team and Team Louisiana agreed that several major breaches were structural failures and could not be explained as a storm simply too large for the system.
• The 17th Street and London Avenue canal walls failed before water reached their tops, through gap formation behind the I-walls, a weak clay foundation layer, and water pressure moving through sand beneath the levees.
• At the Industrial Canal and along the levees protecting St. Bernard Parish and New Orleans East, overtopping set off the breaches, and erodible fill turned overtopping into prolonged open breaches.
• The system rested on a 1965 design storm whose criteria were never updated, structures built lower than authorized, thin margins on the canal walls, and responsibility split among several agencies.
• Death counts depend on method, with 986 Louisiana deaths in the 2008 Brunkard, Namulanda and Ratard study and about 1,800 across all states in the figure associated with the National Hurricane Center report, and no published count isolates deaths from the breaches alone.
• The rebuilt system reduces risk to a 1 percent annual standard, which still means roughly a one in four chance of exceedance over 30 years, so plans behind any levee need a breach scenario and a trigger that does not assume the structure performs as rated.