The Anniversary Forecast

For the twentieth anniversary, Climate Central did something to that storm that could not have been done when it happened: it ran an attribution analysis. Using its Ocean Climate Shift Index alongside a tropical-cyclone index built on peer-reviewed 2024 methodology, the group calculated that the Gulf waters Katrina crossed at peak intensity were about 0.9°C (1.6°F) warmer than they would have been without human-caused warming, and that temperatures that high were up to 18 times more likely because of it. The consequence for the storm itself: roughly 5 mph of additional peak sustained wind.

"Already back in 2005, climate change was influencing Atlantic hurricanes and their impacts," said Daniel Gilford, the Climate Central scientist behind the tropical cyclone index.

Notice what that does to the genre. A commemoration is supposed to look backward. This one is a measurement, and a measurement of physical conditions is portable. Point the same instrument at today's Gulf and it returns a larger number, because the water is warmer now than it was then. The retrospective is a projection wearing a black armband.

The Small Number That Isn't Small

Five miles per hour. On a 175 mph storm that is under three percent, and the instinct — a reasonable one — is to file it as marginal. A hurricane that reached 170 mph without our help would have been a catastrophe too.

The instinct is wrong, and the reason is arithmetic rather than rhetoric. Wind damage does not scale with wind speed. The force a gust exerts on a structure rises with the square of velocity, and the loss functions that insurers and engineers actually use climb faster still once components start failing outright and the building begins shedding its own pieces into its neighbours. NOAA's rule of thumb, which Climate Central cites, is that 5 mph on a storm of this magnitude raises damages by 25% or more.

That is the translation step most coverage drops. Attribution science reports its findings in the units of physics — degrees, millibars, miles per hour — and those units are undramatic by design. The world receives the finding through a damage function that amplifies. On a storm whose losses are mostly wind, three percent more velocity arrives as a quarter again as much destroyed. Anyone arguing that a small attributed signal implies small attributed harm has skipped the exponent.

Two things that argument does not license, and they are worth fencing off before the next section knocks them down anyway. The 5 mph was measured at peak intensity over open water, and Katrina weakened considerably before Buras; an open-Gulf figure does not transfer one-for-one onto a landfall ledger. And the rule of thumb is a wind loss function. Katrina's $125 billion was overwhelmingly water — surge arriving through failed floodwalls. Twenty-five percent of wind damage is not twenty-five percent of Katrina, and no honest reading of the attribution gets you a dollar figure for this storm.

Which is the deeper problem with using Katrina to carry the climate argument at all.

What Actually Broke

The wind is not the story of Katrina, and pretending otherwise makes the anniversary less useful rather than more.

New Orleans did not drown because a Category 3 hurricane hit it. It drowned because the hurricane protection system failed, and the forensic finding — reached by the Corps' own Interagency Performance Evaluation Task Force, and independently by teams from the American Society of Civil Engineers and UC Berkeley — was that major floodwall sections gave way below their design loads. Sheet piling that did not reach deep enough. Soils weaker than the design assumed. A system assembled piecemeal across decades by different authorities with nobody holding the whole. Roughly eighty percent of the city flooded, and much of that water arrived through breaches that a correctly built wall would have held against the surge that actually showed up.

This matters for the forecast in a way that runs against intuition. If the catastrophe had been purely meteorological — an unstoppable storm overwhelming adequate defences — then climate attribution would be the entire conversation, and the 5 mph would be the headline. It wasn't. The catastrophe was substantially engineered, which means the largest term in the 2005 disaster was a term under direct human control.

That is the good news and the bad news in one sentence. Good, because controllable terms can be fixed, and this one largely was: the $14.4 billion Hurricane and Storm Damage Risk Reduction System — roughly 200 miles of levees, floodwalls, gates and pumps across three parishes, built between 2005 and 2018 — is the first time New Orleans has had a coherent defence against water rather than an assortment of them. Ed Link, the University of Maryland engineer who led the IPET investigation, describes it as the first complete approach the city has ever had.

Bad, because the fix is not a permanent object.

Lead Time Is the Resource Being Spent

Before the levees, though, there is the clock.

It is tempting to hang this on Katrina, and the temptation should be refused. The mandatory evacuation order came on the morning of 28 August, a day before landfall, and the storm had climbed from Category 3 to Category 5 in roughly twelve hours over open water. But the National Hurricane Center's track forecast for Katrina was among the best in its history; the threat to New Orleans was on the record days out. What failed was not the warning. It was the execution, and the fact that roughly a hundred thousand residents had no car and were offered no ride. Call that the third controllable term, sitting alongside the floodwalls — and note that it, too, is a budget rather than a mystery.

The lead-time problem is real regardless, and it is getting worse for the storms that come after this one.

Andra Garner's 2023 analysis in Scientific Reports, working through the National Hurricane Center's HURDAT2 record, found that mean maximum intensification rates for Atlantic storms are up to 28.7% higher in 2001–2020 than in 1971–1990. The probability that a given storm's peak intensification rate reaches 20 knots or more rose from 42.3% to 56.7%. Most starkly: thirteen storms went from tropical storm to major hurricane inside 36 hours during the earlier period; forty-nine did so in the later one.

Rapid intensification is not primarily a damage problem. It is a warning problem. Every hour a storm spends surprising forecasters is an hour subtracted from the interval in which buses can be staged, hospitals emptied, nursing homes cleared, and a million people moved along three highways. Lead time is the resource the response system actually runs on, and warmer water spends it.

The same warming that bought Katrina 5 mph is, in aggregate, buying storms the ability to arrive with less notice than the plans assume. That term appears in no damage estimate at all — and the shortfall lands hardest on the people who need the most notice, which is the same hundred thousand.

Protection Is Not a Thing You Own

The HSDRRS is rated for a "100-year" storm — a surge with a one percent chance of occurring in any given year. It is easy to hear that as a property of the structure, like tensile strength. It isn't. It is a ratio: how high the levee stands over how high the water gets. Both numbers move.

Sections of the system are sinking by nearly two inches a year, faster than the Corps projected during construction. Tulane researchers have found that even the massive concrete works, including the Lake Borgne surge barrier, are subsiding. Meanwhile the Gulf rises. The Corps warned as early as 2019 that the combination could drop the system below its hundred-year standard by 2023. The current official position, per Corps spokesperson Ricky Boyett, is that the system holds that standard through 2057 — with the qualifier carrying the entire load: if adequately maintained.

It is not being adequately maintained. Federal allocations to the Corps have been cut and levee inspection funding for New Orleans eliminated; in the anniversary reporting of August 2025, the Corps said outright that it lacked the money to inspect the system in that fiscal year or the one following. The historian Andy Horowitz makes the blunter observation that storms exceeding the system's stated design capacity have already made landfall in Louisiana within its lifetime.

So the defence decays passively. Do nothing and the rating falls — not because anything breaks, but because the ground drops and the sea climbs. Protection at a fixed standard is not a structure you finish; it is an activity you continue. And the activity is always the first thing cut when budgets tighten, because the structure is still visibly there. A levee that has stopped being maintained looks exactly like a levee.

The Forecast

Line the terms up as they stood in 2005 and as they stand now. The ocean is warmer, so the wind bonus is larger. The intensification statistics have shifted, so the warning interval is shorter. The city is better defended than it was by a wide margin, and that margin erodes every year it isn't paid for.

The honest caveats belong here rather than at the top, because they don't change the direction. Attribution for a single storm two decades gone is modelled, not measured: the Climate Shift Index compares this world against a counterfactual one, and counterfactuals carry error bars. The 18x figure is an upper bound, not a point estimate. The 5 mph is a central value with spread around it. Reasonable people can argue the magnitudes, and attribution methods have improved enough in the past decade that today's numbers will themselves be revised. None of that argument runs the other way. There is no plausible version of the analysis in which the Gulf is cooler than it was, or storms intensify more slowly than they did, or the ground beneath the levees stops sinking.

But laying the terms out flat like that implies they weigh the same, and everything above says they don't. The largest term in 2005 was engineered. The fix worked. Its rating now decays for want of an appropriation, and the inspections that would tell anyone how far it has decayed are themselves unfunded. The variable that most determines whether New Orleans floods again is not 0.9°C. It is a line item.

Which is worth sitting with, because the framing has beneficiaries. When the story is a warming ocean, the responsible party is everyone, which is nobody, and the appropriate response is grief. When the story is an unfunded inspection regime, the responsible party has a name and a committee assignment. The first version is more comfortable for every person in the room, including the ones writing it down — and attribution science, accurate and careful and genuinely improving, is what makes the comfortable version sayable with a straight face. The finding is true. The use it gets put to is a choice.

And the calendar does the rest. Commemoration runs on a twenty-year cycle. Levee inspection runs on an annual one. The round-number anniversary reliably produces a week of coverage about warmer water and a distribution that has moved; the appropriation that actually sets the outcome comes up for a vote in the years with no round number in them, and there are nineteen of those for every one of these. The instrument gets pointed at the Gulf on schedule. Nobody points it at the budget.

That is the sense in which a memorial is a forecast — not only because the physics is portable, though it is, but because the ritual and the maintenance run on clocks that do not match, and the ritual is the one everybody shows up for.