Lessons from Hurricane Katrina

Geotechnical conditions and design flaws both contributed to the failure of the levees in New Orleans.

The social impacts of Hurricane Katrina have been covered extensively in the media. As of this writing (December 2006), at least four hardcover books on the subject have been published in addition to numerous news-paper articles and television programs. The effects of the storm on human life, emergency response, preparedness, and the future of New Orleans and the neighboring regions have properly been the principal focus of much of this coverage, but some attention has also been paid to the engineering aspects of the disaster—what happened, why it happened the way it did, whether the design and construction of the hurricane protection system were adequate, and what should be done in the future.

The volume of material just on the engineering aspects of the disaster is overwhelming. At least five boards of experts are reviewing the engineering problems, and the “draft final” report of one of them, the Interagency Performance Evaluation Task Force (IPET), runs to more than 6,000 pages. This paper provides a general overview of the results of these studies to date and then describes in more detail some issues of geotechnical engineering and the failure of the levees.¹

Boards of Experts
IPET (Interagency Performance Evaluation Task Force) was originally organized by the U.S. Army Corps of Engineers (USACE) and was then expanded to include participation by other government agencies and independent consultants. Prof. L.E. Link of the University of Maryland, formerly head of the USACE Cold Regions Research and Engineering Laboratory, Hanover, New Hampshire, chairs IPET. Shortly after the hurricane, the American Society of Civil Engineers (ASCE) also established a working group to review engineering issues associated with the hurricane. This group has developed a close working relationship with IPET and reviews IPET’s work. Dr. David E. Daniel, president of the University of Texas at Dallas and a member of the National Academy of Engineering (NAE), chairs this committee.

In response to a request from the Assistant Secretary of the Army for Civil Works, NAE and the National Research Council (NRC) appointed an Engineering Review Committee (ERC) to review IPET’s reports. Dr. G. Wayne Clough, president of the Georgia Institute of Technology and a member of NAE, chairs this committee. The author is one of the 16 committee members.

The National Science Foundation (NSF) funded a study by the PEER Institute at the University of California, Berkeley, on the behavior of the levee systems. Researchers from Berkeley and many other institutions participated in studies led by Berkeley professors Raymond B. Seed and Robert Bea (NAE). In addition, the geological engineering staff of the Department of Civil Engineering at Louisiana State University conducted its own evaluation of the design and performance of the levee systems. Professor Ivor van Heeren led the study.

What Happened
The Setting of New Orleans
We begin with a brief look at the geologic circumstances of the New Orleans area. The Mississippi River flows generally from north to south, but it takes a sharp turn to the east in northern Louisiana, passes through Baton Rouge and New Orleans, and disgorges into the Gulf of Mexico at a considerable distance east of the general north-south trend of the river. Over the last several thousand years, the delta has moved from place to place. It has been in its present location only for about the last 1,000 years. Left to its own devices, the river would revert to a more direct north-south pattern of flow, following the Atchafalaya River and discharging near the current location of Morgan City.

Because of the economic importance of Baton Rouge, Morgan City, and New Orleans, a system of levees has been constructed over the years to keep the river in its present bed, as well as to control its frequent flooding. To complicate matters further, flooding can result, not only from large spring flows in the Mississippi River, but also from storms that blow in from the Gulf of Mexico. Thus the designers and operators of a flood-control system for New Orleans are presented with complicated design conditions.

At New Orleans, the river flows essentially west to east (Figure 1—see PDF version for figures). The city itself lies between the river to the south and Lake Pontchartrain to the north. The oldest parts of the city, including the French Quarter, Garden District, Audubon Park, both Tulane and Loyola Universities, and the older residential sections, lie just north of the river along the high ground that forms the natural embankment of the river. About midway between the river and the lake is a gentle rise, called the Gentilly Ridge, that interrupts the general trend of falling elevation as one moves toward the low-lying lake. The region bounded by the ridge and the high ground near the river is thus a bowl, in which rainwater collects during heavy storms.

Early in the twentieth century, pumping stations were built along the ridge to remove floodwater from the basin. As the city expanded north beyond the ridge, the newly inhabited region north of the ridge was exposed to flooding from the lake. Instead of building additional pumping capacity at the lakeshore, the city connected the existing pumping stations to the lake via drainage canals. The city also expanded to the east, particularly into the areas known as New Orleans East and the Lower Ninth Ward. Almost all of these areas lie below sea level and must be protected by a system of levees.

Southern Louisiana has been built up out of sediments transported from the interior of the continent down the Mississippi River. Tens of thousands of feet of these soft sediments overlie crystalline bedrock. There is a general pattern of subsidence, complicated by additional settlement whenever a load, such as a levee, is placed on the sediments. Thus the city itself and individual structures settle by different amounts, so the tops of levees may actually be lower than the water levels they are supposed to defend against. Furthermore, because of the general subsidence, it is difficult to establish reliable benchmarks against which to measure the locations of levees.

Katrina’s Track
Hurricane Katrina formed in the Atlantic Ocean in August 2005, moved west across southern Florida and into the Gulf of Mexico, then turned north. On August 29, the storm struck the mainland just east of New Orleans, very close to the border between Louisiana and Mississippi. When the center of the storm was in the Gulf south of New Orleans, its intensity reached Category 5 on the Saffir-Simpson scale. By the time the storm reached the mainland, its intensity had dropped to Category 3, so it is not strictly correct to say that New Orleans was struck by a Category 5 hurricane.

The Saffir-Simpson scale is based primarily on one-minute average wind velocity. It does not take into account rainfall, the storm’s speed and direction, how long the storm remains over an area, the location of the center of the storm relative to the facilities of interest, or the geometry of the region affected by the storm. Although the scale is useful as a shorthand categorization of the severity of a storm, it does not provide a complete description and is certainly not adequate by itself for engineering design.
__________________________________

The tops of levees may be
lower than the water levels
they are supposed to
defend against.
__________________________________

The Flooding
Three phenomena contributed to the flooding in New Orleans. First, the hurricane brought a large amount of rain. Precisely how much of the flooding was due simply to rainfall is still the subject of controversy, but it appears that something like one-third to one-half of the water that wound up in some of the bowls in New Orleans arrived directly from the heavens. Second, some of the levees were overtopped. That is, the level of water in the Gulf, Lake Pontchartrain, and other waterways around the city simply rose higher than the height of the levees, and the excess flowed over them. In many cases the overflow eroded the levee materials, destroying those sections of levee in the process. Third, some sections of levees or canals failed before the water level reached an overtopping level, and the horizontal component of the flow through the breaches added to the damage caused by the flooding. Spectacular examples include the failures of the 17th Street Canal, the London Avenue Canal, and the Inter-Harbor Navigation Canal. These phenomena combined to create flooding of 8 to 15 feet in the Lakeview section just south of Lake Pontchartrain, 10 to 13 feet in New Orleans East, and 12 to 15 feet in the Lower Ninth Ward.

Some levees were overtopped either because they were built to withstand a smaller storm than the one that actually occurred, they had settled below their design elevations, the consequences of the design storm had not been adequately estimated, or a combination of these effects. The levees that failed even before they were overtopped could not withstand even the water levels for which they were designed. All of this suggests major deficiencies in the procedures for the design and construction of the levee system around New Orleans.

Emergency Reaction
It has become commonplace to say that emergency response and reconstruction efforts have been dysfunctional at all levels of government—city, state, and federal. Douglas Brinkley, in his book The Great Deluge (2006), observes that some agencies performed well. For example, the local Society for the Prevention of Cruelty to Animals had a policy that, upon learning of the approach of a Category 3 or greater storm, it would remove the strays under its protection to safety in other cities. The SPCA did precisely that and weathered the storm with no loss of the animals under its care.
__________________________________

Organizations empowered
to protect humans
performed poorly, but not
for want of warning.
__________________________________

Unfortunately, the organizations empowered to protect human beings did not do as well, but not for want of warning. There is a large pre-Katrina literature on the exposure of New Orleans to flooding both from hurricanes and from the Mississippi River. Indeed, Brinkley points out that, immediately after its foundation in 1718, the city was struck by a hurricane, and the original settlers came close to abandoning the site. From June 23 to 27, 2002, the New Orleans Times-Picayune ran a series of articles entitled “Washing Away” that described the damage that could be expected if a major hurricane struck the city. In July 2004, barely more than a year before Katrina, the Federal Emergency Management Administration (FEMA) conducted a multi-agency exercise in emergency planning for the effects of a hypothetical Hurricane Pam. The litera-ture on hazards to New Orleans is written in clear, non-technical, forceful prose, and no thoughtful person should have been surprised by the severity of the storm’s effects or the need for emergency planning.

Although the Hurricane Pam exercise led to the development of some emergency plans, they seem to have been abandoned or modified at the last minute as Katrina approached. The scenes of destroyed houses, drowned people, evacuees huddled in the Superdome amidst crime and filth, and so on are too familiar to require elaboration here. These conditions led to loud and bitter recriminations among government officials and private relief agencies at all levels. As this article is being written, the news still tells of thousands of unused FEMA trailers, uncertainties about reconstruction in New Orleans, questions about the way contracts for reconstruction were let, and continuing chaos in New Orleans’ attempts to recover from the disaster.

So far, other than numerous reports in the press, no independent public examination has been done of what worked and what went wrong in the recovery and reconstruction effort. However, the engineering aspects of design, construction, and operation of the levee system have been reviewed.

Taskforce Guardian
As an interim measure, USACE set up Taskforce Guardian, whose mission was to restore the hurricane protection system to a satisfactory condition in anticipation of the 2006 hurricane season. The goal was to repair and upgrade the system as much as possible in the time available.

Some additional pumping stations and gates were built at the lakeside entrances to the drainage canals, and erodable materials in the levees were replaced. USACE officers stated explicitly that the system would not stand up to a Category 5 hurricane but should resist the more modest events that could be expected during the hurricane season. The construction of an ultimate robust system would have to wait for the completion of the IPET studies and the development of appropriate design criteria. The task force completed its work very shortly after the due date of June 1, 2006. In the actual event, no hurricanes struck the mainland in 2006.

Engineering Review of the Disaster
Although the five engineering review bodies operated independently, all of them reacted to IPET’s work, the largest and best funded of the study groups. Volumes 2 through 7 of the IPET “draft final” report divide the results into six broad sections: Geodetic Vertical and Water Level Datums; The Hurricane Protection System; The Storm; The Performance of the Levees and Floodwalls; The Performance of Interior Drainage and Pumping; The Consequences; and Engineering and Operational Risk and Reliability Analyses. The report is more than 6,000 pages long so far, and the last section on risk and reliability has not yet appeared. The NAE/NRC team and the ASCE team both organized their comments to conform to the IPET pattern.

In reviewing a study of this sort, it is important to keep in mind that critiques may address the quality of the design and construction for facilities in place at the time of the hurricane or the adequacy of the post-hurricane investigations described in the report. For example, one might find that the wind and water levels used for the design calculations were not adequate but that the hind-casting calculations of the hurricane’s effects described in the report are excellent. The NAE/NRC team and the ASCE team had similar reactions, with some minor differences of detail.

The detailed study of the geodetic levels revealed considerable room for confusion and error. Two different benchmark levels were used in creating the levee system. In some cases, the water levels were expressed against one benchmark and the height of the levees against another. Furthermore, there seemed to be no consistent effort to monitor subsidence of the levee system or its components. The focus was on building the levees to the “authorized” elevations without considering whether the corresponding water elevations were measured from the same base or whether subsequent settlement and subsidence might make the authorized levels irrelevant.

All of the review teams found that the hurricane protection system was a system in name only. Planning and system-wide design were confused by political and short-term economic considerations over a period of several decades. In such an environment, it is almost impossible to plan rationally, and a number of basic questions were not addressed. For example, detailed design was based on the parameters of a “standard project hurricane” (SPH), which changed over the years as the city experienced more hurricanes. Although the annual risk represented by the SPH was never clear, it was invoked in congressional authorizing legislation. Similarly, little effort was made to establish target factors or margins of safety that corresponded to actual risks.

The IPET study included an extensive computerized calculation to recover the history of Hurricane Katrina, which yielded detailed descriptions of the wind velocities and water elevations throughout the storm. These calculations confirmed that the particular track taken by the storm caused it to pass near the eastern system of levees and to become positioned near the eastern entrance to Lake Pontchartrain, which generated surges of water that overtopped the eastern levees and drove water up the drainage canals. This part of IPET’s work has shown that the behavior of hurricanes can be hind-cast or predicted with modern finite-element and finite-difference methods and that they should be used in future to develop design loadings.
__________________________________

Two different benchmarks
were used in creating
the levee system.
__________________________________

Evaluating the performance of the protection systems involved detailed studies of the mechanics of various failures. IPET dealt primarily with the failures of the 17th Street Canal and the London Avenue Canal, but some attention was focused on other failures as well. These back-analyses of levee failures were difficult because water rushing through the breaches eroded much of the evidence, making it nearly impossible to determine the precise conditions at the time of failure. The IPET group used conventional limit-equilibrium analyses with circular failure surfaces, finite-element simulations, and centrifuge simulations to study the levee failures (Figures 2 and 3). Other groups relied primarily on limit-equilibrium analyses.

The soil profile differs from location to location, but the situation at the 17th Street Canal failure is typical. The levees, which run north-south along the sides of the canal, are composed of clayey soils resting on a layer of peat (called marsh material in the IPET report), which in turn lies over lacustrine clay and then a sandy beach deposit. To attain the desired level of protection, “I-walls” (steel-sheet piling or reinforced-concrete panels) extending several feet above the levee embankments were embedded in the levees.

An essential ingredient in stability analysis is the description of soil strength. The general consensus is that the design of the levees was not based on conservative estimates of soil strength. The original geotechnical reports showed very wide scatter in the measured values of shear strength. This scatter, rather than indicating variability of the soil, is more likely to reflect errors in drilling, sampling, and testing procedures. More modern investigations using piezocones yielded much less scatter. The original designers not only failed to use conservative soil strengths for the analyses, but also failed to account adequately for the fact that the soil under the toe of an embankment should be weaker than the soil under the crest, where the weight of the embankment compresses it to a greater density.

All of the investigations showed that the levees were either unstable or marginally stable. This is not surprising because all of the analysts knew that the levees had in fact failed. However, different groups arrived at different mechanisms of failure. Centrifuge tests, carried out at Rensselaer Polytechnic Institute and the USACE Engineer Research and Development Center, demonstrated failure along a nearly horizontal plane near the top of the lacustrine clay. In these tests, the I-wall moved downstream as a result of the pressure of rising water in the canal and opened a gap between the wall and the upstream portion of the levee embankment. Water flowing into that gap then applied the full hydrostatic head to the wall. The IPET limit-equilibrium analyses could not recapture failure conditions until they too assumed the existence of the gap and upstream water levels higher than those observed in the field during the storm. The Berkeley PEER group concluded that the failure had not occurred through the clay but had followed a weak seam in the peat. Other groups postulated variations on these failure mechanisms.

In summary, examinations of the levee failures have revealed several plausible failure mechanisms, including a weaker than anticipated clay layer, a soft zone in the marsh material, the effects of a “crack” between the wall and the upstream part of the embankment, differential settlement of the levees and neighboring structures, seepage forces, and even the consequences of toppled trees pulling up their roots.

All of these explanations are at least plausible, but it is hard to identify the actual culprit. Different effects may be dominant at different sections of a levee, and more than one effect may have contributed to several failures. Some sections may have held, not because they were safe but because another section of the levee failed first and relieved the load. In any case, it is clear that the design of many of the levees was, at best, marginal and that the procedures for designing and building levee systems need to be improved. Design criteria, such as safety factors, must be based rationally on the consequences and probabilities of failure and not simply on past practice.

Conclusions for the Future
All five review panels agree in one way or another that the engineering of the levee system was not adequate. The procedures for designing and constructing hurricane protection systems will have to be improved, and the designing organizations must upgrade their engineering capabilities. The levees must be seen not as a system to protect real estate but as a set of dams to protect people.

There must be independent peer reviews of future designs and construction. The Board of Inquiry into the 1928 failure of the Saint Francis Dam concluded that no dam whose failure could cause such a large loss of life should be based on the judgment of one man and that independent review boards should be required for all dams in California. Almost identical language can be found in the report of the federal review panel for the 1977 failure of Teton Dam. Hurricane Katrina demonstrates again that independent peer review of the design, construction, and operation of critical infrastructure systems should be required. Current indications are that USACE is adopting this philosophy.

Emergency response and reconstruction efforts after Katrina were poorly organized, even dysfunctional. Recovery efforts continue to be plagued by confusion, disorganization, and recriminations. Unfortunately, these aspects of Hurricane Katrina have received much less systematic attention than the engineering issues. There is no “Interagency Response and Recovery Evaluation Task Force” corresponding to IPET. Adequate responses to future disasters will require that the failures of emergency procedures and recovery efforts for Hurricane Katrina be subjected to the same kind of independent, public examination as the engineering performance.

References
Brinkley, D. 2006. The Great Deluge. New York: HarperCollins Publishers.
USACE (U.S. Army Corps of Engineers). 2006. Performance Evaluation of the New Orleans and Southeastern Louisiana Hurricane Protection System. Report of the Interagency Performance Evaluation Task Force. Washington, D.C.: USACE.


Footnotes
¹ For more information on IPET, see https://ipet.wes.army.mil/ and http://www.usace.army.mil/; http://www.asce.org. Information about the University of California, Berkeley, PEER Institute is available online at: http://peer.berkeley.edu/PEERCenter/about.html; the Louisiana State University Department of Civil Engineering is available online at: http://www.cee.lsu.edu/eng/ceeweb.nsf/index. For information on the joint NAE/NRC New Orleans Regional Hurricane Protection Projects, see http://www8.nationalacademies.org/cp/projectview.aspx? key=335. The New Orleans Times-Picayune is available online at http://www.timespicayune.com.