Do fires rise on full-moon nights?
No difference
These are counts of fires the fire service attended, ranging from small to major blazes.
London (since 2018)
- Average (whole period)
- 48.1 per day
- Full moon days (±24h)
- 47.3 per day
- New moon days (±24h)
- 49.0 per day
- Verdict for full moon days
- No difference
- Data
- 2018–2023, 105,323 fires
San Francisco (since 2003)
- Average (whole period)
- 11.3 per day
- Full moon days (±24h)
- 11.3 per day
- New moon days (±24h)
- 11.6 per day
- Verdict for full moon days
- No difference
- Data
- 2003–2025, 95,212 fires
Tokyo (Tokyo Fire Department area)
- Average (whole period)
- 12.9 per day
- Full moon days (±24h)
- 13.1 per day
- New moon days (±24h)
- 13.2 per day
- Verdict for full moon days
- No difference
- Data
- 1989–1990 and 2019–2024, 37,730 fires
Tokyo fires per day, by year
| Year | All days | Full moon | New moon |
|---|---|---|---|
| 2024 | 12.3 | 10.7 | 12.6 |
| 2023 | 11.9 | 12.7 | 11.9 |
| 2022 | 10.8 | 10.5 | 10.3 |
| 2021 | 10.8 | 10.9 | 10.8 |
| 2020 | 10.1 | 10.9 | 10.3 |
| 2019 | 11.2 | 11.3 | 11.9 |
| 1990 | 18.3 | 20.4 | 18.3 |
| 1989 | 17.9 | 17.4 | 19.8 |
1991–2018 are missing because no daily table for those years is published as open data. Fires per day fell about 38% over 30 years. Full-moon days ran above the yearly average in 5 years and below it in 3; with only about 25 full-moon days a year, those swings are chance, and all years together match normal.
Tokyo fires per day, by month
| Month | All days | Normal = 100 |
|---|---|---|
| January | 15.7 | 121 |
| February | 15.2 | 120 |
| March | 14.7 | 112 |
| April | 13.3 | 103 |
| May | 12.2 | 95 |
| June | 10.9 | 85 |
| July | 11.2 | 86 |
| August | 10.9 | 85 |
| September | 10.3 | 80 |
| October | 11.0 | 86 |
| November | 13.3 | 104 |
| December | 16.2 | 125 |
Normal = 100 is each year's daily average. Fires run high in the dry winter and low from summer into autumn.
Tokyo's ten busiest fire days and the moon
| Date | Fires | Phase |
|---|---|---|
| 1990-11-12Monday | 613.3× normal | Waning crescentMoon age 24.5 days |
| 2019-04-13Saturday | 363.2× normal | First quarterMoon age 7.8 days |
| 2023-12-28Thursday | 353.0× normal | Full moonMoon age 15.1 days |
| 1990-01-08Monday | 492.7× normal | Waxing gibbousMoon age 11.0 days |
| 2023-02-28Tuesday | 312.6× normal | First quarterMoon age 7.8 days |
| 2022-03-15Tuesday | 282.6× normal | Waxing gibbousMoon age 12.4 days |
| 2021-01-21Thursday | 272.5× normal | First quarterMoon age 7.9 days |
| 2019-01-29Tuesday | 272.4× normal | Last quarterMoon age 23.1 days |
| 2021-12-27Monday | 262.4× normal | Last quarterMoon age 22.8 days |
| 2020-12-21Monday | 232.3× normal | First quarterMoon age 6.4 days |
Multiples are against that year's daily average. Of the ten days, 0 fell within a day of a full moon and 0 within a day of a new moon.
Source: Tokyo Fire Department statistical yearbook (Tokyo Open Data Catalog), processed by MoonAlibi
Act II: What clearly moved the fire count
Here is the difference this data does show clearly.
×3.2
London's busiest fire day on record was July 18, 2022 — 167 fires, about 3.2x a normal day. What drove fires up was drought, heat, and what happened that day.
The weekday barely moves it: even the busiest day (Sunday) runs only about 1.11x the quietest (Wednesday). What clearly moved the count was the weather and the day's conditions.
In this data at least, dry air and the day's events moved fires far more than anything else. We're still looking for the moon's part.
The folklore that "fires rise on full-moon nights"
Linking the full moon to fire is folklore well known on fire and rescue front lines around the world. The sense that "full-moon nights mean more callouts" belongs to the same family as the crime and ER folklore: the idea that the moon unsettles people and breeds carelessness and trouble. There is also the simple notion that moonlit nights draw more outdoor activity around open flame. So do fires actually rise on full-moon days? This page checks with fire-service data from three cities: Tokyo, London and San Francisco.
How the verdict is computed
- The data is daily counts of fires only from three cities' fire-service open data: Tokyo (Tokyo Fire Department area, 1989–1990 and 2019 on), San Francisco (since 2003, U.S. fire service, NFIRS code 1xx) and London (since 2018, London Fire Brigade, Incident Group = Fire). Non-fire calls (false alarms, rescues) are excluded
- Tokyo comes from the "fires by month and day" table in the Tokyo Fire Department's statistical yearbook. Each yearbook covers one year, and the daily table is published as open data (CSV) only for 1989, 1990 and 2019 onward, so those eight years are used. Each month's daily figures are checked against the table's monthly total before import
- Fire counts have weekday and seasonal patterns (winter heating, summer dryness), so the expected value is the average of the "same weekday × same month," and we compare actual ÷ expected as an index (the same adjustment as the accidents and crime topics)
- Using the lunar phase at each city's local noon, we classify full-moon days (within ±24h of the instant) and new-moon days, and compare each group's mean index against normal (1.00)
See the methodology for the verdict thresholds.
Reading the data carefully
- These are fires the service attended, ranging from small to major blazes; not every fire is recorded
- The page tests one thing: whether recorded fire counts are visibly higher on calendar full-moon days. The front line's "full-moon nights are busy" may show up in ways a daily count doesn't capture
What we can't see yet
This result doesn't mean "the moon has nothing to do with it." Here is what the current method can't see:
- Time of day: the folklore is about the full-moon night. Our data is daily, so a rise confined to a few night hours wouldn't show
- Type of fire: if only certain fires rose — outdoor fires or arson, closest to the folklore's logic — they could be hidden among all the others
- Calls that aren't fires: ambulance and rescue calls, a big part of a "busy night," aren't in these counts
What we want to check next
- Use time-stamped data to compare night hours only
- Split fires by type (structure, outdoor, arson)
- Run the same test on ambulance call counts
Perhaps the fire service's "busy full-moon nights" hold something we haven't yet turned into numbers. We'd like to find it in the data one day.
Sources
- Tokyo Fire Department statistical yearbook (Tokyo Open Data Catalog) (table 22, fires by month and day; CC BY 4.0). Source: Tokyo Fire Department, processed by MoonAlibi
- DataSF — Fire Incidents (San Francisco Fire Department, NFIRS fires only)
- London Datastore — London Fire Brigade Incident Records (UK Open Government Licence, Incident Group = Fire only)
- Moon phase instants computed in-house from Jean Meeus, "Astronomical Algorithms" (UTC)
Last updated: October 7, 2026 00:48 UTC (rebuilt daily)