The short and long tail of Hurricane Names

Part twoThe long tail

§Why do so many retired names start with I?

Because in the Atlantic, the alphabet doubles as a calendar. Each season starts again at A, and storms take letters in order, so a letter's place on the list is roughly a date. The NHC's averages put the eighth named storm at 9 September and the ninth at 16 September, and the season's peak falls on 10 September. Since 1979 the I storm has formed in September, the month of the peak, in 23 of the 36 seasons that reached I (HURDAT2).

We can check this with real numbers. The NHC publishes its storm records as a data file called HURDAT2, so for the 47 seasons from 1979, when men's names arrived, through 2025, we counted how often each letter was used and matched it against the retired list. Greek names are left out. The date column is the median day a storm with that letter first reached tropical-storm strength.

Read it as a multiplication. The number of names a letter loses is roughly how often the letter gets used, times how likely a storm at that point in the season is to be catastrophic. The first five letters are used almost every year, 234 times out of a possible 235, but they mostly name early storms, so only one in 16 is retired.

The back of the alphabet has the opposite problem. M names are retired at the same rate as I names, about one in three (Mitch, Michelle, Matthew, Maria, Michael, Milton and Melissa are among them), but the season only reaches M about half the time. I sits at the sweet spot: used in 36 of 47 seasons, and landing near the peak.

Now follow single slots across the six lists. The I slot of the list used in 2004 held Ivan, retired that year; its replacement Igor was retired in 2010; its replacement Ian was retired in 2022; the fourth occupant, Idris, is due in 2028 (NHC). Another slot went Irene (retired 2011), Irma (retired 2017), Idalia.

In all, five of the six I slots have lost at least two names since 1979: Isabel then Ida, Iris then Ingrid, and Isidore then Ike as well. The sixth belongs to Isaac, used in six seasons since 1988 without being retired (HURDAT2). Timing sets the odds, but geography and luck decide each case.

The clock also runs at different speeds. In 2020 the I storm, Isaias, reached tropical-storm strength on 30 July, about seven weeks ahead of the average, and by mid-September the season had reached V. A busy season slides the alphabet forward, so the letters that meet the peak shift toward the back. With samples this small, a couple of storms could reshuffle the M row; what holds up is the multiplication.

§What happens when the alphabet runs out?

Until 2020 the answer was the Greek alphabet; now it's a backup list of ordinary names. The Greek letters were needed twice. In 2005, Wilma used up the regular list in mid-October, and six Greek names followed, the last of which, Zeta, formed on 30 December (HURDAT2).

In 2020 the list was gone much sooner. Vicky, the 20th name, formed on 14 September, the earliest 20th storm on record by almost a month, and the Greek alphabet came out a few days later (WMO). The season finished with 30 named storms, nine of them Greek, from Alpha, which ended its short life over Portugal, to Iota.

The WMO later listed what went wrong. The novelty of Greek names drew attention away from the actual hazards; the names caused confusion when translated into the region's languages; and Zeta, Eta and Theta sounded alike, came in quick succession and were sometimes active at the same time (WMO).

Then came a retirement paradox. Back in 2006 the committee had decided a Greek letter couldn't really be retired: a destructive Greek-named storm would be recorded on the retired list, but the letter would stay available (WMO). In November 2020, Eta and Iota hit the same stretch of Nicaraguan coast less than two weeks apart, and together they killed at least 272 people and caused more than $9 billion in losses.

In March 2021 the committee retired Eta and Iota anyway and dropped the Greek alphabet for good (WMO). In its place is a supplemental list of 21 ordinary names, from Adria and Braylen to Will (WMO), and those names can be retired and replaced like any others (WMO).

We can replay 2020 under the new rules. Thirty storms minus 21 regular names leaves nine for the backup list: Adria, Braylen, Caridad, Deshawn, Emery, Foster, Gemma, Heath, Isla. Hurricane Iota would have been Hurricane Isla.

That replay also shows why the old design was fragile. An overflow list is only needed in the wildest seasons, and the wildest seasons are the ones most likely to ruin names, so an overflow that can't retire names fails exactly when it's used. The new backup list hasn't been needed yet: 2021 ran out the regular list precisely, with Wanda forming on 30 October, and no season since has passed 21 (HURDAT2).

§Why are some typhoons named after puddings and constellations?

Because in the western North Pacific, 14 countries and territories each contributed ten names, and many chose words over people. The Japan Meteorological Agency, which assigns them, publishes the whole list with meanings. Japan's ten names are all constellations: Koinu is Canis Minor, "puppy"; Kujira is Cetus, the whale; Tokei is Horologium, the clock.

Hong Kong gave the list Dim-sum and Tsing-ma, the bridge to its airport, and Macao gave Bebinca, a milk pudding. The Philippines chose some menacing words, like Hagupit ("lash") and Lupit ("cruel"), and the US slot draws on Pacific languages: Etau is Palauan for "storm cloud". Names here get retired too; at its 58th session the Typhoon Committee retired eight at once, among them Ragasa and Bualoi, a Thai dessert (WMO).

The bigger difference is how the list is used. The 140 names run as one continuous queue, so if the year's last storm takes one name, the next year's first storm takes the next (JMA). The region averages about 25 named storms a year (JMA, 1991–2020), so the queue comes back around in about 140 ÷ 25 ≈ 5.6 years. That's close to the Atlantic's six, but with a side effect: since nothing resets in January, no slot on the list is chained to the peak of the season, and the I problem can't form.

The North Indian Ocean goes further and never reuses a name. Its 2020 list has 169 names, 13 from each of 13 countries, and the rules say each must be neutral on politics, religion, culture and gender and no longer than eight letters.

The first list, with names from eight countries, came into use in September 2004. It held 64 names, and by April 2020 every one except the last, Amphan, had been used, a pace of about four a year. At that pace, 169 names last roughly 40 years, so a list adopted in 2020 could still be in use in the 2060s.

Some storms carry two names at once. The Philippine weather agency, PAGASA, gives its own local names to storms inside its area of responsibility, drawing on four sets of 25 names that repeat every four years. The 2013 super typhoon that killed about 6,300 people was Haiyan to the world and Yolanda at home.

Others used to change names at a border. In 1996, Hurricane Cesar crossed Nicaragua into the Pacific and, under the rules of the time, carried on as Douglas. In 2000 the Hurricane Committee decided that a storm which stays a tropical cyclone through the crossing keeps its name, and in 2016 Otto became the first Atlantic storm to do so (NHC).

Otto's tracking number still changed, from AL162016 to EP222016, as its forecasters noted the night it crossed (NHC). The name follows the storm; the number follows the ocean. Other regions settled on the same principle: Australia kept the name Fiji's forecasters gave Cyclone Yasi in 2011, and a storm crossing Thailand into the Bay of Bengal keeps its South China Sea name.

§Why can anyone buy a name for a low over Berlin?

Because Berlin names every high and low on its weather map, not just dangerous ones, and since 2002 it has sold the naming rights to pay for student weather observations. The practice began in 1954, when a student at the Free University of Berlin, Karla Wege, suggested naming the pressure systems over central Europe (Wetterpate); the first names, Anka and Albert, appeared on 1 November 1954 (FU Berlin).

For decades lows got women's names and highs got men's. Hardly anyone outside Berlin noticed until February 1990, when the violent storms Vivian and Wiebke made the names national news (Wetterpate). In 1998 a public debate about pairing women with bad weather ended in a compromise: the genders now swap every year, with male highs and female lows in even years and the reverse in odd ones (DWD).

In November 2002 the university and the Berliner Wetterkarte association launched Aktion Wetterpate, "weather godparent". For 2026, naming a high costs €390 and naming a low costs €290 (FU Berlin). Highs cost more because they are steadier and linger on the map longer, according to Germany's national weather service, the DWD.

The DWD puts the yearly total at about 55 highs and 140 lows. Berlin uses all 26 letters, so lows go through the alphabet more than five times a year and highs about twice. If every slot sold at 2026 prices, that would be 55 × €390 + 140 × €290, or about €62,000 a year, paying for student weather observers at the Berlin-Dahlem station, which keeps a 117-year climate record going.

What you buy is a lottery ticket. Your low might become Kyrill, the destructive windstorm of 2007, or a few days of drizzle; the DWD's example of a lucky sponsor was a high called Petralilli, named on 17 September 2025 and still on the map three weeks later. More than 4,400 people from 17 European countries, plus Brazil, Japan, the US and South Africa, have taken part (FU Berlin). Since 2021, weather services in Germany, Switzerland, Austria, Poland, Czechia, Slovakia and Hungary have used these names for significant lows, so a bought name can end up in official warnings (Wetterpate).

Next door, UK storm names answer a different question. The UK, Irish and Dutch weather services, which have named storms together since 2015 (the Dutch joined in 2019), give a storm a name only when it could cause medium or high impacts, so the trigger is a forecast of consequences rather than a wind speed. The names come from the public: more than 50,000 suggestions went into the 2025/26 list, and the 2026/27 list includes Shaun, after Shaun the Sheep.

The number of named storms swings widely, from a record 12 in 2023/24 to six in 2024/25. And once any European service names a system, the others keep that name, including former hurricanes such as Ophelia, which hit Ireland in 2017 (Met Éireann). So one corner of Europe offers three answers to the question of when weather deserves a name: when the wind crosses a line, when the damage might, and always.

§Do hurricanes with women's names really kill more people?

The famous study said yes; the reanalyses say its data can't show it. In 2014, Kiju Jung and colleagues published a paper in PNAS titled Female hurricanes are deadlier than male hurricanes. They took the 94 hurricanes that hit the US from 1950 to 2012, removed Katrina and Audrey as outliers, and had coders rate each name from very masculine to very feminine.

Their model said that among highly damaging storms, more feminine names went with more deaths: renaming a severe storm from Charley to Eloise would take its predicted toll from about 15 to about 41. Six experiments backed the story. Participants judged a hypothetical Hurricane Alexander riskier than Hurricane Alexandra, and said they'd be more likely to evacuate for Christopher than for Christina.

The critics' first objection was history. From 1953 to 1978 the lists held only women's names, so the old female storms were also the old storms. When Björn and Sören Christensen reran the model on storms after the switch to mixed names, nothing was significant; the original authors had themselves reported only a marginal result for that period. Steve Maley found that four storms, Diane, Camille, Agnes and Sandy, accounted for 732 of the dataset's 1,900 deaths, and that four of the six deadliest storms of the study period came from the era when every name was female.

The second objection was fragility. The Christensens noted the model predicted almost 20,000 deaths for Sandy, which killed 159. Daniel Malter showed the effect vanished once one interaction the authors had left out was added, and Laura Bakkensen and William Larson found it disappeared or reversed once the population in harm's way was taken into account. The original authors replied to each critique and stood by their result.

The cleanest verdict came in 2020, from Uri Simonsohn, Joseph Simmons and Leif Nelson. Instead of arguing over the one right analysis, they listed every analysis of the data they judged reasonable (which storms to drop, how to measure femininity and severity, which model to fit) and ran all 1,728 of them. Only 37 gave a significant result in the predicted direction.

Then came the step that decided it. They shuffled the names randomly among the storms, reran all 1,728 analyses, and repeated that 500 times. In 425 of the 500 shuffled datasets, at least 37 analyses came out significant anyway, a P-value of 0.85: the real names did about as well as random ones.

That's the mechanism the gender study reveals: with enough reasonable choices, some path through the data finds a pattern in noise. What remains open is smaller. People in the experiments did rate a male-named imaginary storm as slightly scarier, but critics have questioned whether answers about imaginary storms tell us what people at risk actually do. The claim that the gap shows up as deaths hasn't held up in independent reanalyses.