Hurricane Katrina made levee protection and evacuation continuity a public-infrastructure accountability test

Summary

• Katrina was a large surge-producing hurricane, not merely a category label. It reached Category 5 strength over the Gulf and made its principal Louisiana landfall near Buras on August 29, 2005, as a Category 3 hurricane. Its size, track, earlier intensity, waves and coastal geometry generated severe water levels over a wide area. Wind category at landfall neither measures surge nor establishes whether a particular structure should have survived.

• Overtopping and structural failure are different mechanisms. Water crossed many low or heavily loaded reaches and eroded their landward sides. At important drainage-canal locations, including the 17th Street and London Avenue canals, I-wall failures began before water reached the wall tops. The engineering question at those sites concerns wall-soil interaction, foundation strength, sheet-pile penetration and design assumptions, not erosion after overtopping. Other breaches had other sequences.

• The protection system did not operate as one coherent system. Federal projects, local levee districts, drainage canals, pump stations, navigation works, incomplete reaches, different design elevations and vulnerable transitions interacted. The Corps controlled major design and construction decisions; nonfederal sponsors held specified operation and maintenance duties; Congress controlled authorization and appropriations; and multiple local and state bodies controlled adjacent assets. Distributed authority does not make responsibility disappear, but it prevents honest analysis from assigning every outcome to one institution.

• Warning existed, while evacuation capability remained unequal. Forecasts and public warnings conveyed extraordinary danger, and many residents left. Yet plans did not provide a dependable end-to-end route for everyone without private transport, for medically fragile people, or for residents of hospitals and nursing homes. A mandatory order is not evacuation continuity unless transport, pickup, destination, records, medication, staffing and return or onward movement are controlled.

• Response failure and engineering failure are separate evidence tracks. Floodwater and infrastructure damage created an extreme operating environment. Congressional, executive-branch, inspector-general and audit reports nevertheless found avoidable weaknesses in situational awareness, command, communications, logistics, mass care and coordination. Those response findings do not prove a floodwall mechanism, and an engineering report does not decide whether a relief mission was timely.

• The human totals require dated definitions. NOAA's January 2023 revision reports 1,392 fatalities associated with Katrina: 520 direct, 565 indirect and 307 of indeterminate cause. NOAA's current estimate is $125 billion in U.S. damage in 2005 dollars, with a 90% confidence interval of $97.4 billion to $145.5 billion. Neither total can be allocated wholesale to the Corps, a breach, an evacuation decision or the federal response.

• Engineering findings did not automatically become compensation. Litigation produced detailed factual records, but statutory flood-control immunity and the Federal Tort Claims Act discretionary-function exception shaped which claims could proceed. In the controlling 2012 appellate opinion discussed here, the Fifth Circuit ultimately reversed plaintiff judgments and held the government insulated from liability on the claims before it. That holding is not a declaration that every design, construction or maintenance choice was sound.

• Rebuilding changed the risk architecture but did not abolish flood risk. The post-Katrina system moved major surge defenses outward, added barriers and closure structures, strengthened levees and floodwalls, and adopted risk-based design practices. A “100-year” level means a modeled 1% annual exceedance probability for specified surge and wave conditions, not a century-long warranty, a hurricane-category guarantee or proof that rainfall, subsidence, sea-level change and operations cannot defeat continuity.

The event must be described as a coupled hazard, not one broken wall

Katrina's physical record begins offshore. The National Hurricane Center's updated Tropical Cyclone Report identifies a peak intensity of 150 knots over the Gulf on August 28 and a Louisiana landfall near Buras at 1110 UTC on August 29 with maximum sustained winds of 110 knots. A second Gulf Coast landfall near the Louisiana-Mississippi border followed at 1445 UTC with 105-knot winds. Those are best-track estimates at specific times and places. They should not be converted into a claim that every neighborhood experienced Category 3 winds or that “Category 3 protection” describes surge resistance.

The report explains why the water hazard was exceptional even though the storm weakened before landfall. Katrina was physically large, its hurricane-force winds covered a broad swath, and waves and northward-propagating swell had developed while it was stronger. High-water evidence was incomplete because gauges failed and many coastal reference structures were destroyed. Subject to that limitation, NOAA reported storm surge of roughly 24 to 28 feet along the most severely affected Mississippi coast, 15 to 19 feet in eastern New Orleans, St.

Bernard and Plaquemines Parishes, and 10 to 14 feet in western New Orleans along Lake Pontchartrain's southern shore.

Water then encountered a regional assemblage rather than a single ring of identical walls. Earthen levees, concrete I-walls and T-walls, gates, navigation channels, drainage canals, pump stations, road and rail crossings, and natural landscape features all affected the route and duration of flooding. Some defenses were overtopped because the combined surge and waves exceeded their effective elevation. Some landward slopes were scoured after water crossed the crest. Some floodwalls moved before overtopping. Some transitions or lower sections admitted water.

Pumps could remove rainfall and seepage only while power, access, discharge paths and structural conditions remained usable.

The distinction matters because a crest exceeded by water is not the same failure as a wall losing lateral stability below its top. Overtopping can still reveal design, elevation, armoring or system-level shortcomings; it is not synonymous with “the structure performed as designed.” Conversely, observing a breach after the storm does not prove that the initial breach was caused by overtopping. The timeline of water level, wall movement, foundation response and erosion has to be reconstructed for each location.

The human consequences were produced by this physical chain together with exposure, warning, mobility, housing, health, rescue and time. NOAA's 2023 revision classifies 520 fatalities as direct, 565 as indirect and 307 as indeterminate, for 1,392 in total. Direct deaths include physical effects of the storm; indirect deaths can involve conditions such as cardiovascular causes or disruption surrounding the disaster; the indeterminate category exists because available evidence could not resolve the classification.

Even NOAA's statement that most Louisiana deaths were presumably caused by surge-induced flooding and its aftermath does not identify which water route, structure, agency decision or delay caused each death.

Damage requires the same care. NOAA's $125 billion estimate is in 2005 dollars and carries a stated 90% confidence interval. It covers U.S. hurricane damage across affected states and mechanisms, not only New Orleans structures or floodwater from a particular breach. Wind, surge, waves, rainfall, prolonged inundation, fire, debris, infrastructure interruption and business disruption affected different places. An inflation-adjusted comparison, an insured-loss total, a federal outlay and a reconstruction appropriation are different measurements and should not be substituted for one another.

The official engineering record separates failure mechanisms

The Corps established the Interagency Performance Evaluation Task Force after the disaster. The archived IPET Volume I executive summary and overview synthesized work on the storm, structural performance, pumping, consequences and risk. Its most important system finding was not that every component failed in the same way. It was that the protection was a system in name but did not perform as a system because it was incomplete, had inconsistent levels of protection and lacked redundancy. Elevation differences, weak transitions, varied materials and exposure meant that the performance of the whole could be limited by a local weakness.

At the 17th Street Canal and London Avenue Canal, the water did not first pour over the tops of the I-walls and scour out the foundations. The critical failures began at water levels below the wall crests. As canal water rose, lateral pressure pushed the wall and sheet pile away from the retained soil. A gap could develop on the canal side, allowing water pressure to act deeper than assumed and increasing load. Weak foundation strata could then shear or translate.

Site-specific stratigraphy, wall geometry and design analysis determined how the mechanism developed; the concise description should not imply that the two canals were geotechnically identical.

The Corps' later revised final London Avenue Canal memorandum is useful precisely because it preserves that specificity. It notes that failures along the 17th Street and London Avenue canals occurred before overtopping and reports IPET's site-focused causes. It also carries a caution that its analysis preceded a later engineer technical letter for evaluating I-walls. That is an important documentary boundary: a post-disaster repair memorandum can record the adopted engineering basis and subsequent design response, but it is not permission to erase later standards or uncertainty.

Elsewhere, overtopping was central. Along parts of the Inner Harbor Navigation Canal and exposed eastern and southern perimeter, surge and waves crossed levees or walls, attacked the unarmored landward side, created scour and contributed to breaches. At some transition points, differences in crest elevation or structural type concentrated risk. In other reaches, soil erosion, internal erosion, foundation instability or construction condition played roles. Saying “overtopping caused the flooding” is therefore too broad, while saying “design failure caused all flooding” is equally unsound.

Subsidence and elevation control formed another layer. A structure designed to a nominal elevation can lose effective freeboard as soft soils consolidate, organic soils subside, sea level changes, datum methods are corrected or construction falls below intended grade. Inspection that records visible condition without a current elevation survey may miss a major change in hydraulic performance. Yet subsidence should not become a universal explanation: it affected reaches differently and does not replace forensic evidence about the drainage-canal foundation failures.

The decision history also matters. The Corps' Institute for Water Resources describes its Hurricane Protection Decision Chronology as an effort to document planning, economic, policy, legislative, institutional and financial decisions influencing Greater New Orleans protection. Long project timelines reflected authorization, local cost shares, environmental litigation, changing technical approaches, appropriations, land and rights-of-way, and staged construction on compressible soils.

That complexity explains how the system developed; it does not itself excuse a deficient safety assumption or prove that a delayed segment caused a particular loss.

Protection authority was fragmented but traceable

The U.S. Army Corps of Engineers is the institutional subject because it held central design, construction, technical-standard and federal inspection roles for major authorized projects. It was not the operator of every pump, owner of every levee, evacuation authority, weather forecaster, shelter manager or incident commander. Accountability therefore requires a control map rather than a single name.

Congress authorized the Lake Pontchartrain and Vicinity project in 1965, and federal participation unfolded through later appropriations and project decisions. GAO's early history of the Lake Pontchartrain and Vicinity Hurricane Protection Project describes the project's purpose, evolving barrier and high-level alternatives, construction history and funding. It was published in September 2005, before the mature forensic findings. It is reliable for authorization and funding chronology, not a final statement of why a particular wall failed.

GAO's later statutory and regulatory framework testimony explains the divided roles. For covered levees, the Corps designed and constructed while federal and nonfederal interests shared costs under the governing arrangements; after completion and turnover, nonfederal sponsors had operation and maintenance responsibilities, with Corps inspections and federal rehabilitation authorities under specified conditions. Drainage and pumping involved separate local bodies. An inspection regime could identify maintenance deficiencies, but local maintenance could not correct an unsafe federal design assumption without authority, engineering and money.

Conversely, federal design responsibility did not make the Corps responsible for every local operating decision.

This distribution created interface risk. One organization could measure a wall; another could mow and inspect an embankment; another could operate a pump; another could close a gate; another could issue an evacuation order. If their data, trigger levels and command relationships did not connect, every component owner could report completion while the regional safety case remained incomplete. The most dangerous gaps often sit at transitions: federal to local, levee to floodwall, drainage to surge protection, construction to maintenance, warning to transport, and evacuation to shelter.

Funding is part of the record but not a one-variable causal explanation. A project can be delayed by constrained annual appropriations, yet a specific failure can arise from a design method that more money under the same method would not have corrected. A local sponsor can have maintenance arrears without those arrears causing a below-crest foundation failure. A rigorous review asks which control was required at that location and time, who had the authority to exercise it, what evidence they possessed and whether a feasible correction existed.

The same discipline applies to the Mississippi River-Gulf Outlet. MRGO was a federal navigation channel, while adjacent protection works belonged to flood-control projects with different legal and engineering histories. Channel widening, wetland loss, surge propagation, foreshore protection and levee performance became contested in engineering studies and litigation. “MRGO caused Katrina” is indefensible; Katrina was the hurricane. The narrower questions concern whether operation and maintenance changed surge or wave conditions, whether those changes contributed to specified breaches or depths, and what duties and immunities applied.

Evacuation continuity begins where a warning ends

Katrina was forecast, and the Gulf Coast received urgent warnings. Louisiana and New Orleans used phased and then mandatory evacuation measures. Large numbers of people left before landfall. Those facts refute the idea that there was no warning or no evacuation. They do not prove that the system provided everyone a realistic means of reaching safety.

The House Select Committee's A Failure of Initiative examined preparation and response primarily from one week before through two weeks after landfall. It found that plans at all levels were not executed adequately, that the catastrophic nature of the event was not converted into sufficiently proactive action, and that the response suffered from failures of initiative. It also documented successful individual actions. Its broad institutional judgment should not be rewritten as a finding that every official failed or that one missed decision produced the entire emergency.

The Senate Committee's Hurricane Katrina: A Nation Still Unprepared separately traced preparedness, command and operational problems. The two congressional investigations overlap but are not interchangeable transcripts. Each selected evidence and framed recommendations through its committee mandate. Together they support a conclusion that government planning had contemplated a catastrophic New Orleans hurricane yet had not built a fully executable, integrated capability for the scale that arrived.

An evacuation order is a public instruction. Evacuation continuity is a service chain. The resident must receive and understand the warning, decide in time, reach a pickup point or vehicle, obtain accessible transport, travel on a viable route, arrive at a known destination, remain sheltered with food, water, sanitation and security, preserve medication and medical records, communicate with family, and move onward or return. Failure at any link can strand someone who formally received the order.

Private-car assumptions placed unequal risk on people without cars, money for fuel or lodging, physical ability to travel, or a support network outside the region. Poverty, disability, age, language, work obligations, caregiving and distrust affected decisions. Some residents underestimated the hazard; some chose to remain; some could not leave. Those categories must not be merged into “refused evacuation.” A plan needs a documented denominator: who may need transport, where they are, what assistance they require, which vehicles and drivers are assigned, which receiving sites have capacity, and when the last safe departure occurs.

The Superdome was used as a shelter of last resort, not designed as a complete long-duration destination for the post-flood population that assembled there. The Convention Center later became another concentration point without the planned continuity expected of an equipped shelter. Roof damage, power loss, sanitation failures, supply uncertainty, communications breakdown and changing population counts compounded conditions. Accounts of suffering were sometimes mixed with rumors; accountability should rely on verified conditions and official timelines rather than repeat every early report as fact.

Hospitals and nursing homes exposed a specialized gap. GAO's review of federal evacuation assistance for health facilities found ambiguity about how federal help connected a facility to the point where the National Disaster Medical System assumed transportation. Facility administrators also faced limited vehicles, receiving capacity, staffing and clinical risks in moving patients. The lesson is not that every patient should always be moved early. Evacuation itself can harm fragile people.

The control is a clinically informed decision with triggers, transport categories, destination acceptance, records, medication, power, oxygen, staff and contingency shelter-in-place capability.

Command, communications and logistics prolonged exposure

The flood disabled ordinary systems on which emergency doctrine depended. Roads were submerged, public-safety facilities were damaged, commercial communications failed, power and fuel were scarce, and local responders were themselves victims. These conditions explain why information and movement were difficult. They do not answer whether federal, state and local systems had built adequate redundancy for a catastrophe whose defining feature was loss of normal infrastructure.

The executive branch's Federal Response to Hurricane Katrina: Lessons Learned identified 17 critical challenges and 125 recommendations. It described shortfalls in national preparedness, integrated military support, communications, logistics and evacuation, search and rescue, public health, human services, mass care and public communication. As an administration review, it records the executive branch's own reconstruction and reform position. Congressional reports and independent audits remain necessary checks; the White House report is not a judicial allocation of fault.

The DHS Inspector General's performance review of FEMA's disaster-management activities found that FEMA faced extraordinary scale while also identifying planning, staffing, logistics and operational weaknesses. FEMA did mobilize resources and support many missions. The accountability issue is whether pre-positioning, commodity tracking, contracting, personnel qualification and field authority were adequate for a catastrophic incident, and whether a request-based assistance model could accelerate when state and local capacity was visibly overwhelmed.

Communications had three separate requirements: operability within an organization, interoperability across organizations, and sufficient capacity under extreme demand. The FCC Katrina Panel's report and recommendations on communications networks documented damage from flooding, power loss and fuel shortages as well as failures of redundant pathways. Radios that survive physically may still be unable to communicate across agencies. A shared frequency is not enough if users lack credentials, programming, batteries, repeaters, procedures or an exercised communications plan.

Loss of communications then contaminated situational awareness. Population estimates at shelters changed; reports of levee breaches, rescue needs and supply status moved slowly or inconsistently; leaders at different levels operated from different pictures. A common operating picture cannot depend on one network or dashboard. It needs field reports that can travel by radio, satellite, data, courier or prearranged offline protocol, with time, source, confidence and geographic reference preserved.

Military support revealed command and integration issues rather than lack of effort. GAO's review of military response planning and exercises found that active-duty, National Guard and reserve forces supplied major capabilities but that plans and exercises had not resolved coordination, damage assessment, communications, search-and-rescue and logistics problems. Constitutional and statutory roles, gubernatorial control of Guard forces, federal command and mission assignments matter.

The repair is not a vague instruction to “send the military sooner”; it is a pre-negotiated command structure, capability inventory, triggers and lawful mission set.

The Coast Guard provides an important positive comparison. GAO's observations on Coast Guard preparation and response describe pre-staged communications, plans that allowed units to operate with limited guidance, adaptive use of radios and messaging, and extensive rescue activity. Success did not mean the Coast Guard solved the whole response. It shows that doctrine built around degraded communications, delegated initiative, trained crews and immediately usable assets can preserve continuity when central direction is impaired.

Logistics was not merely the existence of supplies. Water, food, fuel, medical materiel, buses and generators had to be ordered, staged outside the hazard, assigned, transported over changing routes, received by a functioning local node and distributed to a changing population. A truck dispatched is not a delivery. Performance records should show request time, approval, load, location, estimated arrival, actual handoff and exception reason. Without that chain, leaders can mistake movement in the network for service at the point of need.

Displacement and housing cannot be reduced to one Katrina count

Displacement began with evacuation and expanded when homes became inaccessible or uninhabitable. It continued through hotels, emergency shelters, host families, temporary units, rental assistance and relocation across many states. The relevant population changes by date and definition. Someone away for three nights, a household barred from a flooded neighborhood, and a family still in temporary housing years later are all displaced in different senses.

GAO's disaster-housing review reported that Hurricanes Katrina and Rita together displaced more than one million people and that an estimated 300,000 homes were destroyed or made uninhabitable by Katrina. The report's combined-storm displacement total must not be silently relabeled “Katrina evacuees.” It also reported extensive shelter use across states and found that federal agencies and the Red Cross lacked sufficiently developed plans for catastrophic shelter and temporary housing. The number is useful for scale, not for assigning individual causation.

Housing continuity requires more than a temporary unit. Location determines access to work, school, health care, transit and community. A household placed far from services may be formally housed but functionally stranded. Program transitions can create cliffs when rental support ends or eligibility changes. Data systems need to connect assistance without exposing personal information broadly, reconcile duplicate or changing household records, and preserve an appeal route when automated matching fails.

Displacement also complicates democratic and institutional continuity. Residents need records, benefits, mail, schooling, voting access and information about neighborhoods and rebuilding. Local governments lose staff and revenue while facing extraordinary service demands. Hospitals and employers may not reopen until residents return; residents may not return until services and jobs exist. Recovery is therefore a coordinated network problem, not a sequence in which housing can wait until engineering is complete.

The Corps' role in that network remains specific. Floodwater removal, debris missions assigned by FEMA, navigation restoration and rebuilding of protection affected whether neighborhoods could reopen. FEMA, HUD, state programs, local planning authorities, insurers and private landlords controlled other parts. A resident's prolonged displacement may reflect several interacting conditions. It should not be assigned automatically to a floodwall mechanism merely because the initial home flooding began at a breach.

Courts, immunity, claims and compensation occupy a different track

Engineering findings ask what happened physically and why. Government investigations ask whether planning and operations met public duties. Civil litigation asks whether named plaintiffs prove jurisdiction, duty, breach, causation and damages under applicable law, subject to immunity. Administrative compensation and disaster assistance apply statutory or program rules. These tracks can use overlapping facts while producing different outcomes.

The Fifth Circuit's substituted 2012 opinion in In re Katrina Canal Breaches Litigation illustrates the boundary. Claims concerned MRGO, drainage canals and specific properties. The court distinguished immunity under section 702c of the Flood Control Act from the Federal Tort Claims Act discretionary-function exception, and it treated flood-control and navigation activities differently within its analysis. It ultimately held that application of the discretionary-function exception insulated the government from liability on the claims before it, reversing plaintiff judgments and affirming judgments for the government.

That procedural and substantive result must be stated whole. A district court had made negligence and causation findings for certain MRGO plaintiffs; the appellate court did not turn every underlying engineering concern into a damages award. For the drainage-canal claims, flood-control immunity independently mattered. For other plaintiffs, property-specific causation mattered. The final appellate disposition is not proof that no negligent conduct occurred anywhere, and an engineering criticism is not proof that a claimant can overcome sovereign immunity.

Compensation after Katrina came through multiple channels: federal disaster assistance, National Flood Insurance Program payments, private insurance, Community Development Block Grant recovery programs such as Louisiana's Road Home, appropriations, contracts, settlements and litigation. Each had its own eligible harm, proof, offsets, caps, subrogation rules and appeal procedures. Aggregate federal spending is not the same as damages paid by the Corps, and a grant is not a tort judgment.

Insurance disputes also turned on whether damage resulted from covered wind or excluded flood, policy language, adjustment evidence and state law. A neighborhood could experience both. Program administrators sometimes used modeled damage or categorical rules to move claims faster, while courts required policy- and property-specific proof. Accountability requires publishing the rule and preserving review rather than describing every payment as full compensation.

The deepest remedial problem is evidentiary asymmetry. Residents may have lost deeds, policies, receipts and medical records in the flood while institutions retained engineering, claims and inspection data in separate systems. A fair process should use authoritative government data to reduce the claimant's burden, disclose adverse evidence, document automated matches and provide human reconsideration. That is data sovereignty in practical form: affected people can see and challenge the records used to determine their recovery without requiring uncontrolled release of personal data.

Post-Katrina reform created new duties and new proof questions

Congress enacted the Post-Katrina Emergency Management Reform Act of 2006 as Title VI of Public Law 109-295. It strengthened FEMA's statutory identity and mission, created or formalized leadership and preparedness functions, addressed regional structures, logistics, disability coordination, communications and catastrophic response, and required a national preparedness system. Enactment demonstrates a legal response to identified weaknesses. It does not prove that every capability was funded, integrated, exercised or sustained.

The physical system changed as well. Temporary repairs first aimed to restore pre-storm protection, but GAO warned that restoration was not a comprehensive future strategy. Its September 2006 report on levee repairs and enhancements found that the Corps had rapidly repaired many damaged miles while relying on temporary measures in places and continuing contracts after the nominal June 1 restoration date. GAO called for an integrated strategy and an evaluative organization. A declared percentage restored to “pre-Katrina level” was therefore a construction milestone, not proof of uniform reliability.

The Greater New Orleans Hurricane and Storm Damage Risk Reduction System later added or strengthened perimeter levees, floodwalls, surge barriers, gates and pump complexes and removed many interior reaches from direct surge exposure. The Corps' HSDRRS program page says the last major project, the Permanent Canal Closures and Pumps, was completed in May 2018 while further armoring continued. It defines the 100-year level as risk reduction against surge with a 1% chance of occurring or being exceeded in a given year and explicitly says the system reduces rather than eliminates risk.

That language is more honest than a category guarantee. Annual exceedance probability depends on storm populations, tracks, sizes, intensities, waves, water levels and model assumptions. A 1% annual probability is not “once every 100 years”; independent annual chances accumulate over time. Certification at that level also does not cover every rainfall event, drainage failure, river flood, future subsidence or scenario beyond the design basis.

The Corps' published HSDRRS hurricane design guidelines show how post-Katrina lessons entered hydraulics, geotechnical work, levees, structures, mechanical and electrical systems, surveys and design details. A guideline is an important control because it converts lessons into repeatable requirements. Compliance still needs project records, independent review, construction quality evidence, as-built surveys, instrumentation and maintenance.

Reform must include natural and operational systems. Wetlands and barrier landscapes can influence waves and surge but are not an absolute shield. Pumps and gates need fuel, power, operators, communications, access and tested modes for conflicting rainfall and surge conditions. Levees need settlement monitoring, vegetation and animal-burrow control, erosion protection, closure exercises and corrected elevations. The strongest wall can be defeated by an unclosed gate or an unknown utility penetration.

Control rights define the accountability evidence

The lesson becomes actionable when every critical control has an owner, trigger and inspectable record.

Control point	Principal controller	Evidence required	Accountability lesson
Hazard characterization	NOAA, FEMA, Corps and technical partners within their mandates	Dated track, intensity, surge, wave and rainfall forecasts; model version; uncertainty; update log	Wind category alone did not describe surge loading
Project authorization and funding	Congress, executive budget authorities, Corps and nonfederal sponsors	Authorized purpose, design level, schedule, cost share, appropriations, scope changes and unresolved gaps	A decades-long joint project could remain incomplete and uneven without one visible system safety case
Levee and floodwall design	Corps and responsible design agents	Geotechnical data, wall-soil model, load cases, independent review, assumptions, factors of safety and design changes	Below-crest I-wall failures require foundation and interaction evidence, not an overtopping label
Construction and elevation	Corps, contractors and sponsors within project roles	Material tests, pile records, inspection reports, as-built elevation, datum, nonconformance and acceptance	Nominal design height is not proof of actual or current effective elevation
Operation and maintenance	Levee districts, drainage and sewerage bodies, Corps inspection programs and other owners	Inspection frequency, defects, settlement, repairs, closures, vegetation, burrows, penetrations and escalation	Turnover divides duties but should not divide the evidence chain
Evacuation decision	Governor, mayor, parish leaders and emergency managers within authority	Trigger, forecast basis, order time, zone, route status, public message and exception handling	An order must be early enough for the slowest assisted movement
Assisted transport	State and local transport/emergency agencies with federal support	Registrant denominator, pickup sites, accessible vehicles, drivers, manifests, destinations and missed pickups	Private-car evacuation cannot stand in for a whole-population plan
Medical evacuation	Facilities, health authorities, HHS/NDMS and transport partners	Clinical categories, destination acceptance, transport, staff, medication, records, power and shelter-in-place fallback	Movement itself carries risk and must be managed end to end
Shelter continuity	Local/state authorities, FEMA, Red Cross and facility operators within assigned roles	Capacity, population count, food, water, sanitation, security, medical support, communications and onward plan	A refuge of last resort is not automatically a sustainable shelter
Emergency communications	Public-safety agencies, communications providers, FCC and DHS coordination bodies	Operability, interoperability, capacity, backup power, fuel, alternate paths, credentials and exercises	Redundancy must survive common-cause loss of power, towers and backhaul
Incident command and situational awareness	Local and state command, DHS/FEMA and assigned federal leaders	Named command roles, delegation, common operating picture, source confidence, requests, decisions and handoffs	Multiple coordination titles without clear field authority delayed unity of effort
Rescue and relief logistics	Local responders, Coast Guard, FEMA, Defense, Guard and mission partners	Capability inventory, mission assignment, dispatch, route, delivery, exception and completion	Resources counted nationally are not capability delivered locally
Claims and remedy	FEMA, insurers, HUD/state recovery programs and courts within jurisdiction	Eligibility rule, causation standard, evidence used, offsets, decision time, explanation, review and payment	Assistance, insurance, settlement and tort damages are not interchangeable
Reform verification	Corps, FEMA, sponsors, auditors and elected oversight bodies	Baseline, owner, deadline, funding, design proof, exercise result, residual risk and independent review	Completion of a project or statute is an input; reduced loss and preserved continuity are outcomes

No single controller owns every row. That is the point. A regional safety case should connect the rows so that a forecast updates protection operations and evacuation triggers; transportation status updates shelter planning; structural status updates rescue routes; communications degradation activates alternate reporting; and claims systems can use authoritative exposure data after the event.

Durable protection must be demonstrated in operation

For structural protection, proof begins with current geometry and condition. Every critical reach should have a traceable as-built record, common vertical datum, settlement trend, geotechnical basis, known penetrations and closure inventory. Instrument thresholds should identify abnormal movement or seepage. Inspection findings need owners and deadlines, and unresolved high-risk defects should be visible to the authority responsible for evacuation decisions.

For system performance, exercises must combine surge, rainfall, power loss and communications loss. Gate and pump tests conducted in fair weather are necessary but limited public evidence. Operators should demonstrate manual and remote modes, alternate power, fuel resupply, staffing depth and decision rules when closing a surge gate conflicts with pumping interior water. After-action records should distinguish an exercise artifact from a real control weakness and track correction to retest.

For evacuation, officials should measure the assisted population rather than assume it. Registries are useful but incomplete because people move, lose phones, decline registration or develop needs after the list closes. Plans need multiple intake routes and neighborhood verification. A full-scale exercise should time the slowest pickup-to-destination path, including wheelchair boarding, medical transfer and companion or service-animal needs.

For public warning, message delivery is not comprehension. Warnings should identify the hazard, location, deadline, transport option and trusted source in accessible formats and relevant languages. Officials should test reach among people without broadband, commercial power or English proficiency. Corrections must propagate across channels quickly enough that an outdated message does not continue circulating after routes or shelters change.

For command, every level needs a succession rule and decision log that survives connectivity loss. The record should show who had authority to request federal support, who could redirect resources, when a catastrophic-incident assumption replaced routine request processing, and which unresolved decision constrained the field. Shared data must retain provenance and uncertainty; a precise number without a source can misdirect scarce rescue assets.

For relief, the denominator belongs at the delivery point. Publish not only how many trucks, meals or liters entered the pipeline but how many people were at each site, when supplies arrived, how long stock lasted and why shortfalls occurred. Privacy is not a barrier to aggregate operational accountability. Personally identifiable rescue and medical data should remain access-controlled while de-identified performance evidence is retained for review.

For long-term risk, certification should be renewed against subsidence, sea-level conditions, updated storm science, land-use change and component aging. The public needs plain statements of what the system is designed to reduce, what it does not cover and what evacuation remains necessary. Residual risk is not a footnote. It is the basis for insurance, zoning, building elevation, continuity planning and personal decisions.

What remains uncertain

The precise initial sequence at every damaged reach is not equally documented. Some sites had instruments, high-water marks, soil samples and visible remnants; others were altered by floodwater, emergency work or later reconstruction. Official engineering conclusions are strongest at the intensively investigated canal breaches and broader at some overtopped reaches. New analysis should identify whether it revises a site finding or merely applies a new model.

The contribution of each water pathway to each property and death is also uncertain. Floodwaters mixed after entering protected basins. Timing, depth and velocity differed by neighborhood. A person classified as an indirect fatality may have been affected by evacuation stress, interrupted care or post-storm conditions without evidence tying the death to one agency. The 307 indeterminate classifications in NOAA's 2023 total are a reason to preserve uncertainty, not distribute them proportionally.

Evacuation choice cannot be reconstructed household by household. Surveys, registrations and interviews have coverage and memory limits. “Stayed” does not reveal whether someone had transport, received the order, trusted the destination, cared for another person or believed sheltering in place was safer. Policy should respond to observed barriers without inventing a motive for every resident.

The damage total changes with price year, geography, included losses and later data. NOAA's estimate and confidence interval are appropriate for broad event scale. Insured loss, federal assistance, public infrastructure cost and claimant compensation answer different questions. This article does not add them into a larger total.

Legal disposition is claim-specific. The Fifth Circuit opinion controls the consolidated federal claims it decided and its immunity analysis; it does not adjudicate every state, local, contractor, insurer or levee-district claim. A settlement may reflect litigation risk and limited funds rather than a factual allocation of all harm. A disaster grant can be payable without negligence and denied without proving the applicant suffered no loss.

Finally, future performance is necessarily uncertain. HSDRRS has operated during later storms, but no past event can prove performance under every track, surge, wave, rainfall, subsidence and operational combination. Models estimate annual probability under assumptions that must be maintained and updated. The honest reform claim is reduced risk supported by current evidence, not invulnerability.

The accountability test

Katrina was neither an unknowable surprise nor a simple experiment in which one defective wall produced every loss. It was a forecast large hurricane whose surge and waves met an incomplete, inconsistent and insufficiently redundant protection system; whose flooding disabled ordinary communications and transport; and whose remaining residents required a rescue, shelter and relief capability larger than plans could reliably deliver.

The engineering judgment must stay site-specific. Extensive overtopping and landward erosion occurred in exposed reaches. The 17th Street and London Avenue I-wall failures began below their tops through distinct wall-foundation mechanisms. Subsidence, elevations, transitions, construction condition and maintenance mattered in different places. Combining all of them under “overtopping” conceals design failures; combining all of them under “bad design” conceals hydraulic exceedance and operational interfaces.

The institutional judgment must preserve distributed control without accepting ownerlessness. The Corps held central design and construction authority for major federal works. Congress, nonfederal sponsors, local drainage bodies, state and city emergency leaders, FEMA, DHS, health agencies, the military, communications providers and shelter partners controlled other indispensable decisions. Each should answer for the control it held and for whether it connected its evidence to the wider system.

The legal judgment must remain separate. Official engineering reports can identify unsafe assumptions. Congressional and executive investigations can find preparation and response failures. Courts then apply jurisdiction, immunity, causation and damages rules to defined claims. Assistance and compensation programs apply still other rules. None of these tracks should be presented as if it automatically decided the others.

The post-Katrina standard is therefore an evidence chain: current hazard models to current elevations; design assumptions to independent review; construction records to operating condition; warning to assisted transport; patient origin to receiving bed; shelter population to supplies; field observation to common operating picture; request to delivered resource; property exposure to explainable remedy; and every reform promise to an exercised, independently reviewable outcome.

The public-infrastructure accountability test is not whether officials can say the system is complete. It is whether residents can know the residual risk, whether the weakest transition is visible before the storm, whether people without cars can reach safety, whether hospitals can move or sustain patients, whether responders can operate when normal networks fail, whether courts and programs explain the remedy they can and cannot provide, and whether protection continues to earn confidence as land, water and institutions change.

Source notes

This article gives primary weight to the National Hurricane Center for storm and updated fatality and damage estimates; to IPET and later Corps records for engineering mechanisms and system lessons; to GAO for project roles, evacuation, military, Coast Guard, housing and rebuilding audits; to congressional, executive-branch, DHS Inspector General and FCC records for preparedness and response findings; to the Fifth Circuit for the legal holding; and to enacted law and Corps program materials for reform. Dated estimates retain their scope.

Engineering, response, legal remedy and reform evidence are kept on separate tracks, and no source is used to assign every death, displacement or dollar of damage to one institution or defect.