In Singapore's air, sound covers a kilometre in 2.86 to 2.87 seconds at each monthly 24-hour mean the Meteorological Service Singapore (MSS) publishes for 1991–2020, by our calculation. Dividing a flash-to-thunder count by three therefore places a strike 4–5% closer than it is, and about 5% closer on a storm afternoon. Most of that gap is the rule's own rounding, which reads about 3% short even in 20 °C air. Below about 20 seconds, one second of miscounting moves the answer more than the rule does.
MSS warns that "lightning strokes from a distant thunderstorm cloud more than 10 km away could still affect that location", a strike it calls a "bolt from the blue". A count turns a delay into a distance and nothing more.
Textbook sound speed falls short in Singapore's air every month
The textbook 343 m/s is the speed of sound in dry air at 20 °C. We computed Singapore's figure with a port of the UK National Physical Laboratory's calculator, which follows Cramer (1993), using MSS's 1991–2020 climate station means.
| Air (MSS 1991–2020 means) | Speed (m/s) | Seconds per km | 343 m/s reads short by |
|---|---|---|---|
| January daily minimum, 24.3 °C, with the month's mean daily maximum humidity of 94.7% | 347.4 | 2.88 | 1.3% |
| May 24-hour mean, the warmest month, 28.6 °C, 82.3% | 350.1 | 2.86 | 2.0% |
| April daily maximum, 32.4 °C, 62.3% (our extrapolation) | 352.2 | 2.84 | 2.6% |
Every afternoon figure here is our extrapolation above NPL's 30 °C limit. Warmth does most of the work: taking relative humidity, the share of the moisture air can hold, from zero to 100% adds 0.50–0.66% to the speed between 25 and 30 °C, while warming dry air from 25 to 33 °C adds 1.33% (above 30 °C, our extrapolation).
Both mnemonics undershoot the distance, the mile rule further still
The Hong Kong Observatory's education page teaches counting the seconds "and dividing the number by 3" for kilometres, which treats sound as 333.3 m/s. The US National Weather Service (NWS) thunder page says to "divide by 5" for miles, which treats it as 321.9 m/s.
In dry 20 °C air, by our calculation, dividing by three already reads 2.9% short and by five 6.3% short; the first matches air at about 3 °C. Singapore's air adds the rest. The three-second rule reads about 4–5% short at MSS's monthly means, and 5–5.4% at afternoon highs; MSS says inter-monsoon thunderstorms "occur in the afternoon and early evening". The mile rule reads 7.4–8.6% short in the same air. At a 30-second count the true distance is 10.42–10.57 km against the rule's 10.00.
A slipped count strays further than the rule's bias below about twenty seconds
One second of counting error moves the answer by 347–352 m at any count, while a rule's shortfall grows with the count. The two meet at the counts below, on our arithmetic. One second is a yardstick; we found no measurement of human counting error.
| Conversion | A one-second slip equals its shortfall at |
|---|---|
| Divide by three (km) | 18.7–24.7 s |
| Divide by five (miles) | 11.6–13.6 s |
| Fixed 343 m/s | 38.3–78.6 s |
Above those counts the rule's bias is the larger error. The 1998 Lightning Safety Group named another: the method "has severe limitations in part due to the difficulty of associating the proper thunder to the corresponding flash".
Until 8 October 2026 our lightning distance calculator multiplied every count by 343 m/s while its page taught both mnemonics. It headlined miles, captioned counts of 30 seconds or less "30-30 rule" as if that were current, and called any count past 60 seconds "Within range", even 99,999 seconds at 34,299.7 km. The revision made that day computes the speed by the NPL method from the reader's temperature and humidity, defaulting to 27.8 °C, our average of MSS's monthly means, and MSS's 82% mean annual humidity. It puts kilometres first, flags anything above 30 °C as extrapolated and gives no distance past 46.0 seconds at its defaults.
Long counts overshoot the usual reach of thunder
"Normally, you can hear thunder about 10 miles from a lightning strike," says the NWS science overview, about 16 km. Its sports page adds that traffic, wind and precipitation can shorten that. Sound covers 16 km of Singapore air in 45.7–46.3 seconds, by our arithmetic. A much longer count lies beyond the range NWS gives; on our reading it more likely pairs one flash with another's thunder.
NWS adds that later thunder comes from "the portions of the channel farther and farther away", so the first rumble comes from the closest part. MSS's alert rings of 6 and 8 km correspond to about 17 and 23 seconds of delay in Singapore's mean air, on our arithmetic.
The flash-to-bang trigger of 1998 misfired in the public's memory
The Lightning Safety Group met in Phoenix in January 1998, and its recommendations appeared in the Bulletin of the American Meteorological Society in October 1999. The first 30 was a count: precautions began when flash and bang were "less than 30 seconds" apart. The second was 30 minutes of waiting "after the last flash or thunder". The paper notes that NOAA in 1994 had pointed to "the first sound of thunder".
William Roeder, listed among that meeting's participants and collaborators, recorded what followed in a January 2008 conference paper. People, he wrote, were remembering the rule as "go to a safe place at this time" rather than "be in a safe place", and "Some meteorologists even taught it that way." After a 2006–07 consensus in the NWS Lightning Safety Awareness Week working group, "the lightning safety community now recommends seeking a safe place when you hear thunder", under the slogan "When Thunder Roars, Go Indoors!" The half-hour wait stayed. No single date marks the change.
An NWS office in Shreveport still teaches the old count. In Singapore, the ANZA Nippers lightning policy, uploaded in July 2025, says "we adhere to the 30/30 rule" beside a stop when MSS's alert service shows lightning "within an 8km radius".
Singapore's lightning alerts sidestep the stopwatch altogether
No seconds count appears on MSS's weather systems page, lightning poster or alert service, or on MetMalaysia's English weather-phenomena page. The first counts "about 167 thunderstorm days and 176 lightning days a year". A lightning day is one with "at least one lightning occurrence" detected at Changi, so it is no island-wide tally. The inter-monsoon months, late March to May and October to November, "account for more than 50% of the lightning strokes in a year".
MSS's Lightning Information Service maps strokes "within the last 30 minutes" inside the 6 or 8 km rings users set. MetMalaysia's page says lightning "can occur as far as 10 miles away from the rain area", about 16 km. Where skin is irritated or hair stands up, it says to "squat down on the ground. Place hands on knees with head in between knees". Roeder's paper describes the US community as "no longer teaching last minute outdoor lightning risk reduction to general audiences".
The Lightning Information Service describes "a network of four lightning sensors located islandwide". NEA's Lightning Observation dataset on data.gov.sg, in metadata dated 19 March 2026, describes "a network of five lightning sensors". We found nothing that reconciles the two.
View author profile → · Editorial policy
