It's Always Hurricane Season Somewhere

Dek: We charted every tropical cyclone since 1842 — 13,547 storms. Because the two hemispheres peak about five months apart, the planet is never storm-free: across 56 years of complete records, not one month comes up empty, and even the quietest, May, keeps about five storms turning.

Abstract

Tropical cyclones are usually pictured as a seasonal, regional hazard — the Atlantic "hurricane season," the Pacific "typhoon season." Seen globally, they are closer to a standing feature of the planet. Using the TerraPulse tc catalog — 13,547 storms from 1842 to the present, drawn from IBTrACS best-track history and live U.S. National Hurricane Center fixes — we map where and when tropical cyclones form across all seven basins. Three facts emerge, all from measured storm tracks and none from any model or forecast. First, the Northern and Southern Hemisphere seasons are offset by about five months (Northern genesis peaks in late August, Southern in late January–early February); the offset is stable to within a tenth of a month whether we use the full record or only the reliable satellite era. Second, the hemispheres' alternation means Earth is essentially never without an active storm: across the 672 months of the satellite era (1970–2025), not one had zero storms active anywhere, and the calmest month of the year — May — still averages about five storms turning at once. Third, the North Indian basin shows the textbook monsoon double-peak (a pre-monsoon May peak and a post-monsoon November peak, with the strong cyclones vanishing entirely in July–August) — but only once storms are filtered to hurricane strength; the raw all-systems count buries it under weak monsoon depressions. We report this as a descriptive climatology and a baseline for a standing global season-phase monitor, and we are explicit about the one thing it must not be read as: a trend. Pre-satellite records undercount storms, especially in the Southern Hemisphere, so counts rise across the record for reasons of observation, not climate — which is precisely why the finding here is about timing, which is robust to that undercount, and not about totals, which are not.

1. Why ask this

Ask most people when hurricane season is and they will name a stretch of late summer and autumn. They are right — for their hemisphere. The Atlantic and the western Pacific run their busiest months from roughly August to October. But the Southern Hemisphere's cyclone basins — off Australia, Madagascar, the South Pacific — run their own calendar, and it is half a year out of step. When the north goes quiet, the south is winding up.

Put the two together and a simple question follows: is there ever a moment when the whole planet is quiet? Or does the hand-off between hemispheres keep at least one storm turning somewhere at all times? This is the kind of question a data platform that already follows every storm can simply answer by counting — no modeling, no forecasting, just the observed record of where and when storms have actually formed.

It is also the natural second entry in TerraPulse's "standing-monitor" series, after the Starlink thermosphere pilot: a measured, open-ended monitor of a planetary rhythm, designed to keep running as new storms accrue.

2. Data

The tc catalog. Every storm is one slot in the TerraPulse tc dex — a per-cyclone record carrying its observed track: position, time, intensity where reported, and the basin it formed in. The catalog folds together two measured sources into one series: IBTrACS best-track history, reaching back to 1842, and live National Hurricane Center observed fixes for present-day Atlantic and East-Pacific storms. It holds 13,547 storms across seven basins — western Pacific (4,163), south Indian (2,777), north Atlantic (2,289), north Indian (1,757), east Pacific (1,360), south Pacific (1,198), and a near-empty south Atlantic (3).

Measured reality only. Everything here comes from observed storm tracks — genesis date, genesis basin, and, where an agency recorded it, peak intensity. No forecast, reanalysis, or model field is used. Genesis month and basin are the two most robustly observed attributes of a storm; peak intensity is missing for 43% of the record overall — and for 75% of North Indian storms specifically (mostly older ones) — which matters only for the intensity-filtered part of the North Indian result, flagged below.

One honest limit, stated up front. The number of storms recorded per year roughly doubles across the record — from a median near 57/year before 1970 to about 108/year in the satellite era — because older storms, especially over the empty stretches of the Southern Ocean, were simply never observed. That rise is an artifact of observation, not a change in the climate. It is the reason this paper is about when storms form, not how many: timing is recoverable from an undercounted record; totals are not.

3. Method

For each storm we take its genesis (first observed fix) date and its basin, and assign it to a hemisphere. To measure the seasonal peak we use circular statistics — treating the day of the year as an angle on a clock face, so that December and January sit next to each other rather than at opposite ends of a line. The circular mean of all genesis dates in a hemisphere gives that hemisphere's average peak; the angular gap between the two hemispheres' means is the phase offset.

We test one explicit null hypothesis: that storms form uniformly through the year, with no season at all. The Rayleigh test for circular uniformity rejects it overwhelmingly in both hemispheres (p ≈ 0), confirming what the eye sees — a strong annual cycle. The concentration of genesis dates around the mean (the circular resultant length R, an effect size running 0 for no season to 1 for a single day) is high: R = 0.62 in the North, 0.71 in the South.

For the "is the planet ever quiet" question we count, for every calendar month of every year, how many storms were active (their track spanning that month), and ask whether any month came up empty. We restrict this strict test to the satellite era (1970 onward), where the record is complete enough to trust a zero.

For the North Indian double-peak we stratify genesis by peak intensity — all systems, named strength (≥34 kt), and hurricane strength (≥64 kt) — because the classic bimodal pattern is a property of strong cyclones, not of the weak monsoon depressions that pad the raw count.

4. Results

4.1 The two hemispheres run about five months out of phase

Northern Hemisphere storms cluster in late August (circular-mean genesis on Aug 29, month 7.9); Southern Hemisphere storms cluster in late January to early February (circular mean on Feb 5, month 1.2 — January and February are the two busiest Southern months). The gap between them is 5.3 months — close to, but not exactly, the half-year that a naive "opposite seasons" picture would predict.

What makes this more than a textbook restatement is its stability. Re-computing the offset on only the reliable satellite era barely moves it: 5.26 months on the full 1842-present record, 5.31 from 1970, 5.32 from 1980. The undercount of old storms does not distort the timing at all — exactly as it shouldn't, and a useful demonstration that the phase result is trustworthy even though the raw counts are not.

!The two hemispheres run about five months out of phase

The share of each hemisphere's storms by month. The Northern (cyan) and Southern (coral) curves are almost perfect mirror images, offset by about five months — as one hemisphere's season fades the other's builds.

4.2 It's always hurricane season somewhere

Because the hemispheres alternate, the global storm count never falls to zero. Across the 672 months of the satellite era (1970–2025), not a single one had zero storms active anywhere on Earth. The global cycle has a clear minimum in May (when the Northern season hasn't started and the Southern has wound down) and a maximum in September, with the quiet season running at a fifth to a quarter of the peak — but the floor never touches the ground. Even in May, the quietest month of the year, an average of about five tropical cyclones are turning somewhere at any time. And because incomplete observation can only remove storms, never invent them, an undercounted record is if anything biased toward more empty months; finding none across 56 years is a conservative result — the real planet is even less likely to fall quiet.

!It's always hurricane season somewhere

Average number of storms active in each month across the satellite era. September is the global peak; May is the global lull — but even the lull sits at roughly five active storms, and no month in 56 years has come up empty.

4.3 The Bay of Bengal's monsoon switches off the strong cyclones

The North Indian basin is famous for a double-peaked season: cyclones before the summer monsoon (peaking in May) and after it (peaking in November), with a deep gap during the monsoon itself. In our data that pattern is invisible in the raw count — all systems together show a broad hump cresting in October — because the monsoon months are filled with weak monsoon depressions that the catalog also tracks.

Filter to storms of named strength or above, and the textbook shape snaps into focus: a May peak, a July–August collapse, and a November peak. At hurricane strength the collapse is total — zero hurricane-strength cyclones form in the Bay of Bengal in July or August across the whole record. The monsoon that soaks the subcontinent also shears apart any storm trying to organize over it.

!The monsoon switches off the strong cyclones

North Indian genesis at three intensity thresholds, each scaled to its own peak. The strong-cyclone curves (cyan, coral) plunge to near-zero during the monsoon shut-off; the all-systems curve (gray) stays high, padded by weak monsoon depressions.

4.4 The western Pacific never really stops

One basin resists the whole framing: the western Pacific, which produces storms in every month of the year. Its quietest month, February, still records 42 storms across the record — a lull, not a break. It is the single biggest reason the global count never reaches zero.

5. What this says, and what it doesn't

It says: globally, tropical cyclones are a standing feature, not a seasonal one. The Northern and Southern Hemisphere seasons are locked about five months apart, and the hand-off between them keeps at least one storm — usually several — turning somewhere on Earth at every moment of the year. The quietest the planet gets is May, and even then roughly five storms are active. The seasonal structure within basins is real and sharp (the Rayleigh test annihilates the no-season null), including the intensity-gated monsoon double-peak of the Bay of Bengal.

It does not say anything about whether storms are becoming more frequent, stronger, or shifting their timing over the decades. The record's rising storm count is an artifact of improving observation, not a climate signal, and we have deliberately built the entire result around timing, which survives that undercount, rather than totals, which do not. Nor is the North Indian double-peak a claim about the full 1842 record: it rests on the intensity-filtered subset — 447 of the 1,757 North Indian storms, every one of them from 1972 onward, the only ones with a measured intensity.

This is a baseline — the measured shape of the planet's storm calendar — against which a standing monitor can watch for change, and against which the arc's later reports (the freshwater pulse, the barometric autopsy, the microseism) will be set.

6. Limitations

Observational undercount. Storms per year roughly double from the pre-satellite to the satellite era. This is why we report timing, never totals or trends.
Intensity missingness. Peak intensity is absent for 43% of storms overall and 75% of North Indian storms, concentrated in older records; the intensity-present North Indian storms are entirely post-1971, so the intensity-stratified double-peak is wholly a modern-era statement (447 of 1,757 NI storms). It is reported descriptively — the July–August hurricane-strength count is simply zero — without a formal bimodality test.
Basin assignment. Each storm is counted in its genesis basin; the handful of storms that cross basin boundaries keep a single identity, so there is no double-counting, but a storm's later life may occur elsewhere.
South Atlantic. With only three storms in the entire record, the South Atlantic is a genuine near-absence — noted as a curiosity, not analyzed.
Descriptive, not inferential about change. This is a climatology. It establishes the baseline rhythm; it does not test any hypothesis about that rhythm changing.

7. Outlook: a standing season-phase monitor

The tc catalog updates live as new storms are fixed, so this climatology is the frozen backdrop for a running monitor: at any moment it can answer where on Earth is it hurricane season right now, and place the current global storm count against the 184-year seasonal envelope. As the arc's instrument-fingerprint layer deepens, the same per-storm records carry the river pulses, pressure minima, lightning, and seismic hum each storm leaves behind — but the season-phase backbone is the first, and it is complete today.

8. Conclusion

We mapped 13,547 tropical cyclones over 184 years and found that the planet keeps a storm calendar with no blank pages. The two hemispheres run about five months out of phase, their alternation leaves no month of the satellite era without an active storm, and even the year's quietest moment holds about five. Within that global rhythm the basins keep their own sharp seasons, down to the Bay of Bengal's monsoon shutting off its strongest cyclones for two months every summer. It is, measurably, always hurricane season somewhere.

Data and reproducibility

scripts/extract.py: reads the tc dex via dex_reader.read_slots, produces one clean per-storm row (genesis year/month, basin, hemisphere, peak intensity, lifespan). Writes data/storms.parquet.
scripts/analyze.py: basin × month climatology, hemisphere circular statistics and phase offset, the active-storm "always a season" test, the intensity-stratified North Indian bimodality, and the coverage/era assessment. Writes data/results.json.
scripts/viz.py: the three figures. Writes www/season-calendar.png, www/never-zero.png, www/monsoon-switch.png.
Data products: data/storms.parquet (13,547 storms), data/results.json.

References

Knapp, K. R. et al. (2010). The International Best Track Archive for Climate Stewardship (IBTrACS). Bull. Amer. Meteor. Soc.
U.S. National Hurricane Center, CurrentStorms.json observed-fix feed.
Fisher, N. I. (1993). Statistical Analysis of Circular Data (Rayleigh test, circular mean).
Li, R. C. Y. & Zhou, W. (2013). Seasonal and intraseasonal modulation of North Indian Ocean tropical cyclones.