The dark band is the same count off by a second in either direction. Nobody counts a storm to better than that, which is why a single figure is the wrong output and why this page shows you the range it is really telling you.
Every temperature a person could be standing in, end to end. The line is not flat, but it is close enough to flat that the input you are most likely to get wrong is the one on the stopwatch.
“Divide by 3 for kilometres, by 5 for miles” is stated as a matched pair. It is not one. At 15 °C sound covers a kilometre in 2.94 seconds and a mile in 4.73. So dividing by 3 is out by about 2.0%, and dividing by 5 is out by about 5.4% — more than twice as far, in the same direction.
Both underestimate, which is the right direction to be wrong in: the storm is always at least as far away as the rule says, never nearer. But if you want the mile version to be as good as the kilometre one, the divisor is not 5 — it is 4.7, which nobody is going to do in their head during a thunderstorm. That is the real reason the kilometre form is the one to trust: 3 happens to be a much better rounding of 2.94 than 5 is of 4.73.
Sound is slower when it is cold, and the tool above lets you set the temperature because everybody expects it to. Here is what it buys: across the entire range from −10 °C to 40 °C, the seconds-per-kilometre figure moves it by 9.1%. One second of error on a nine-second count moves it by 11.1%.
So the input the page cannot see is worth more than the input it asks you for — and one second is a generous estimate of how well anyone counts while watching a storm. That is why the answer above is given as a band. A count is worth about half a kilometre either way, and no amount of decimal places changes that.
Light, for completeness, contributes nothing at all: it crosses ten kilometres in 33 microseconds, about one part in a million of the sound’s travel time. Every version of this rule that mentions light arriving “instantly” is right.
Nobody needs to know a storm is 7.2 km away rather than 6.8. The count answers one question: are you already in range? National weather services give the same answer in the same form — the 30/30 rule. If the flash-to-bang is under thirty seconds, go inside. Wait thirty minutes after the last thunder before coming back out.
Thirty seconds is 10.2 km, and it is not an arbitrary round number: strikes are recorded reaching 16 km from the parent storm, well outside the rain, which is where the phrase “a bolt from the blue” comes from.
And the asymmetry that matters most: thunder stops arriving before lightning stops reaching you. Beyond roughly fifteen kilometres, thunder is usually refracted upwards by warmer air aloft and never gets to you at all. So a storm can be visibly close, completely silent, and still in range. “I cannot hear anything” is not the test.
This page is arithmetic, not safety advice. The 30/30 guidance above is quoted from the people whose job it is; nothing here derives it, and nothing here should be used to talk yourself into staying outside.
By three, for kilometres, and trust it more. Sound covers a kilometre in 2.94 seconds at 15 °C, so “divide by 3” is out by about 2%. A mile takes 4.73 seconds, so “divide by 5” is out by 5.4% — the honest mile divisor is 4.7, which nobody will do in a storm. Both err in the safe direction: the storm is never nearer than the rule says.
Less than the counting does. Across the whole range from −10 to 40 °C the seconds-per-kilometre figure moves by 9.1%. Being one second out on a nine-second count moves the answer by 11.1% — and one second is a generous estimate of how well anyone counts while watching a storm. That is why the result is given as a band.
The 30/30 rule, as the national weather services state it: if the flash-to-bang is under thirty seconds — about 10 km — go in, and wait thirty minutes after the last thunder before coming back out. This page quotes that guidance; it does not derive it, and nothing here should be used to argue for staying outside.
No. Beyond roughly fifteen kilometres, thunder is usually refracted upwards by warmer air aloft and never reaches you, while strikes are recorded landing 16 km from the parent storm — well outside the rain. A storm can be visibly near, completely silent, and still in range. Not hearing anything is not the test.
How far away that one was is a curiosity. Standing outside watching a storm, what you want to know is whether it is coming your way and how long you have. A single count cannot answer that; a sequence of them can. Time each strike as you normally would and log it, and the distance against the clock gives you the storm's closing speed directly.
Consecutive strikes are not the same object. They are different discharges in different parts of the same storm, which may be tens of kilometres across. Time a far strike and then a near one and the arithmetic will tell you the storm is closing at seventy kilometres an hour; time them the other way round and it is leaving at the same speed. Neither is true. The scatter between strikes is usually larger than the storm's actual movement over the same couple of minutes, which is why the trend here is fitted through every point you have logged and why it says how confident it is.
For scale: storms typically travel at 20 to 50 km/h, so a genuine approach moves the nearest edge by a few kilometres in the time it takes to log three or four strikes. If the fitted speed comes out at two hundred, you have measured the storm's width, not its motion.
The tracking is interesting; the rule is not. If you can hear thunder at all, you are within range of being struck — thunder carries about 10 km and so do bolts from the side of a storm. The published guidance is 30/30: if the gap is under 30 seconds, go inside; stay inside until 30 minutes after the last thunder. Thirty seconds is about 10 km, which is exactly the distance a strike can reach out of a storm that has not arrived yet — the so-called bolt from the blue, which is the one that catches people who are watching the rain rather than the sky.
Four strikes, logged over three minutes at 15 °C: counts of 24 s, 19 s, 15 s and 11 s — about 8.2, 6.5, 5.1 and 3.8 km.
Fitted against the clock that is roughly 88 km/h of closing speed, which is far too fast for a storm and tells you something useful anyway: you are watching a squall line whose leading edge is sweeping across you, or the later strikes simply happened to land on the near side. Either way the trend is unambiguous and the last count is under thirty seconds.
So the answer is not the speed. It is that the third strike already put you inside the 30/30 rule and the tracking only confirmed it. If the numbers are going down and the count is under thirty, the arithmetic has stopped being the point.
Sound takes 2.94 s to cover a kilometre and 4.73 s to cover a mile. Rounding 2.94 to 3 costs you 2%; rounding 4.73 to 5 costs you 5.4%. The rule presents the two as a pair, and they are not one. Trust the kilometre form.
Thunder is refracted upwards by warmer air aloft and usually stops arriving past about 15 km. Lightning is recorded striking 16 km from the parent storm. So there is a band where you cannot hear a thing and are still in range, and it is the band people walk back out into.
Counting a storm is a two-part reaction time (the flash and the bang), and one second is a fair estimate of the error. That is half a kilometre. Any output with more precision than that is decoration.
The rounding goes the same way in both units: the true distance is always a little further than the rule says. If you are going to be wrong about how close a thunderstorm is, that is the direction to be wrong in.
From a freezing night to a heatwave the answer moves 9.1%. One second of miscounting moves it 11.1%. The slider is there because people look for it, and the honest use of it is to see how little it does.
In a storm producing several strikes a minute you will hear thunder from a flash you did not time, and no arithmetic can tell. The count is trustworthy for an isolated strike and guesswork in a busy sky.
This page has changed in 138 archived releases, but each of those was a site-wide pass, so none is attributable to this tool on its own and none is listed here. That is not a claim that the tool never changed — a release that reworked many pages at once may well have altered this one too. The changelog has them.