Vrancea, Romania · weather and earthquakes, 2012–2026
There is no earthquake weather: 1,039 deep Vrancea earthquakes sorted by heat, rain and the phase of the Moon
People say the ground moves after oppressive heat, after heavy rain, or when the weather turns suddenly. On the hottest 20% of days in the last 14 years (in Bucharest, against normal for the date), Vrancea had 19.7 deep earthquakes per 100 days; on all other days, 19.9. The gap is smaller than chance alone produces.
earthquakes per 100 days (Bucharest weather)
Each small square is one day between 1 March 2012 and 30 June 2026, coloured by how warm it was against the normal for its date. Each dot is a day with a deep Vrancea earthquake of magnitude 3 or more; larger dots mark days with two or three. Pick a criterion and the days line up in order; the bars above count earthquakes per 100 days.
Every day, with its earthquakes
A commune can also be picked on the commune map.
The numbers in the chart
| Group of days | Days | Earthquakes | Per 100 days |
|---|---|---|---|
| Coldest | 1,046 | 204 | 19.5 |
| Cool | 1,046 | 197 | 18.8 |
| Ordinary | 1,045 | 231 | 22.1 |
| Warm | 1,046 | 201 | 19.2 |
| Hottest | 1,046 | 206 | 19.7 |
#Where the idea of "earthquake weather" comes from
In the fourth century BC, Aristotle wrote in his Meteorology that the earth shakes because wind enters its depths, and that "most earthquakes and the greatest are accompanied by calm". The explanation was abandoned long ago. The idea that an earthquake can be read in the sky stayed.
On 20 October 2026, after a magnitude 4.8 earthquake, Știri pe Surse in Chișinău wrote that the idea that earthquakes are preceded by an unusual atmosphere, oppressive heat or spells of heavy rain "has been around for a long time". The United States Geological Survey answers briefly that there is no such thing as earthquake weather.
The belief names three things that can be measured: heat, heavy rain and a sudden change in the weather. All three were compared with the 1,039 deep Vrancea earthquakes, day by day, for each of Romania's 3,181 municipalities and communes.
#The weather on the days of the big earthquakes since 1950
On 14 January 1963 at 20:33, Bucharest had just had a day with a high of −7.2 °C. On 1 August 1985 at 17:35, the thermometer had reached 34.2 °C. On both days, Vrancea had a magnitude 5.8 earthquake.
Since E-OBS records begin in 1950, Vrancea has had 24 earthquakes of magnitude 5.5 or more, on 23 days (the two of 30–31 May 1990 fall on the same weather-day). On the axis, each sits by how warm its day was in Bucharest against the days within a week of the same date, in 31 nearby years (1950–1980 for the oldest). Of those days, 14 were in the warmer half and 9 in the colder, as a tossed coin might give.
All 24 earthquakes, with the day's weather
| Date (weather-day) | Magnitude (Mw, ROMPLUS) | High in Bucharest | Warmer than | Rain |
|---|---|---|---|---|
| 16 January 1950 | 5.7 | 2.8 °C | 65% | 0.0 mm |
| 20 June 1950, 03:18 (weather-day: 19 June 1950) | 5.9 | 24.7 °C | 16% | 0.0 mm |
| 3 August 1952 | 5.5 | 31.5 °C | 68% | 0.0 mm |
| 1 May 1955 | 5.8 | 24.2 °C | 74% | 0.0 mm |
| 19 August 1959 | 5.5 | 25.5 °C | 19% | 0.0 mm |
| 26 January 1960 | 5.7 | −1.8 °C | 23% | 0.0 mm |
| 13 October 1960, 04:21 (weather-day: 12 October 1960) | 5.9 | 23.0 °C | 82% | 0.0 mm |
| 14 January 1963 | 5.8 | −7.2 °C | 5% | 1.7 mm |
| 10 January 1965, 04:52 (weather-day: 9 January 1965) | 5.8 | 5.8 °C | 83% | 0.0 mm |
| 2 October 1966 | 5.9 | 26.9 °C | 90% | 0.0 mm |
| 20 August 1973 | 6.0 | 28.8 °C | 55% | 0.0 mm |
| 1 October 1976 | 6.0 | 18.9 °C | 29% | 0.0 mm |
| 4 March 1977 | 7.4 | 13.3 °C | 85% | 0.0 mm |
| 18 July 1981, 03:02 (weather-day: 17 July 1981) | 5.5 | 28.6 °C | 43% | 0.0 mm |
| 25 January 1983 | 5.6 | 3.6 °C | 60% | 0.0 mm |
| 1 August 1985 | 5.8 | 34.2 °C | 92% | 0.0 mm |
| 31 August 1986, 00:28 (weather-day: 30 August 1986) | 7.1 | 32.6 °C | 96% | 0.0 mm |
| 30 May 1990 | 6.9 | 16.5 °C | 4% | 27.9 mm |
| 31 May 1990, 03:17 (weather-day: 30 May 1990) | 6.4 | 16.5 °C | 4% | 27.9 mm |
| 27 October 2004 | 6.0 | 20.7 °C | 86% | 0.0 mm |
| 14 May 2005, 04:53 (weather-day: 13 May 2005) | 5.5 | 22.6 °C | 38% | 0.0 mm |
| 24 September 2016, 02:11 (weather-day: 23 September 2016) | 5.5 | 20.9 °C | 27% | 0.0 mm |
| 28 December 2016, 01:20 (weather-day: 27 December 2016) | 5.6 | 6.5 °C | 67% | 0.0 mm |
| 28 October 2018, 03:38 (weather-day: 27 October 2018) | 5.5 | 22.0 °C | 86% | 0.0 mm |
Families remember three of them. On 4 March 1977, the high in Bucharest had been 13.3 °C, a warm day for March. On 30 August 1986, the day before the 00:28 earthquake, it had been 32.6 °C, warmer than 96% of days within a week of the same date in 1971–2001. On 30 May 1990 it rained 27.9 mm and was cold for May: 16.5 °C.
Anyone who remembers 1977 and 1986 has reason to link earthquakes with heat. Anyone who remembers 1990 has reason to link them with cold and rain. Two or three unforgettable days do not make a rule. A rule only shows when thousands of ordinary days are counted.
#Your day: the weather and the earthquakes on a given date
A birthday, a heatwave, the night you felt an earthquake. A day runs from 6 in the morning to 6 the next morning, so that an earthquake after midnight sits with the weather of the evening before. Daily magnitudes are those of the INFP event catalogue and can differ by a few tenths from those in the table of big earthquakes.
#In all 3,181 municipalities and communes, the same result
Each commune has its own weather, read from the roughly 11 km grid square it falls in. The map colours it by the ratio between earthquakes on its hottest days and on all its other days: brick red would mean more earthquakes in the heat, blue fewer.
The ratio stays between 0.90 and 1.16 everywhere, and no commune rises above what chance can produce. With the days shuffled at random 2,000 times, at least one commune looks "significant" in 46% of the draws, because neighbouring communes share their weather. The real data give none.
#What of the weather reaches 100 kilometres down
Vrancea's earthquake nest lies 90 to 150 kilometres under the mountains, in a slab of tectonic plate sinking into the mantle. To move anything there, the weather would have to get through the rock.
Heat does not get through. A single day's warming falls to a third within the top 17 centimetres of ground and is practically gone below a metre; the difference between summer and winter falls to a third within the top 3.2 metres. The nest is nearly 28,000 times deeper.
Air pressure does get through, because a weather system presses on an area hundreds of kilometres wide. But a strong storm changes the pressure by a few kilopascals at most, about as much as the tide the Moon raises in the rock twice a day. The rupture of an earthquake releases stresses of a few megapascals, a thousand times more. Rock at 90 km is under nearly 2.7 gigapascals, a million times more.
Rain can trigger earthquakes, but close to the surface. At Hochstaufen in the Bavarian Alps, water from heavy rain seeps down through cracks and raises the water pressure in the rock around faults a few kilometres under the mountain. Vrancea's earthquakes happen 20 to 40 times deeper.
#The one result past the threshold: heavy rain at Focșani
Of the 32 comparisons between weather and earthquakes, one comes out under the usual 5% threshold. On the 189 days with over 10 mm of rain at Focșani, Vrancea had 28.0 earthquakes of magnitude 3 or more per 100 days, against 19.6 on other days (ratio 1.43, p = 0.003). Out of 32 comparisons, chance usually gives one or two under 5%; if the tests were independent, a value this small would turn up about once in ten (they are not fully independent, since they share the same earthquakes). On heavy-rain days in Bucharest the ratio is 1.04; for earthquakes of magnitude 3.5 or more, at Focșani itself, 1.19 (between 0.70 and 1.91), nothing out of the ordinary.
Rain cannot reach 90–150 km. One possible explanation lies with the instruments. Taking all recorded earthquakes, aftershocks included, by size, the ratio between heavy-rain days and other days climbs step by step: 0.56 below magnitude 2.5, 0.68 from 2.5 to 2.8, 0.90 from 2.8 to 3 and 0.99 from 3 to 3.2. That fits the idea that the faintest earthquakes are lost more often in the noise rain and wind make in seismographs, but these data do not measure the noise and cannot prove it.
At 3.2–3.5 the ratio jumps to 1.77, without a good explanation: the steps were chosen after the Focșani result was known, and a jump on a single step can still be chance. What can be said: no number in these data shows rain setting off earthquakes at Vrancea's depth.
#Every comparison
The ratio of earthquakes per 100 days in each group of days to the rest; 1 means the same. In brackets, the 95% confidence interval. The 5% threshold comes from shuffling days within the same month; the one cell that passes it is marked. Data from 1 March 2012 to 30 June 2026.
| Days | Focșani, M 3+ | Bucharest, M 3+ | Focșani, M 3.5+ | Bucharest, M 3.5+ |
|---|---|---|---|---|
| The hottest 20% of days | 0.95 (0.81–1.11) | 0.99 (0.84–1.15) | 0.94 (0.73–1.20) | 0.98 (0.76–1.25) |
| The day after one of the hottest | 1.05 (0.90–1.22) | 1.11 (0.95–1.29) | 1.03 (0.80–1.31) | 0.91 (0.70–1.17) |
| Hot and dry | 0.97 (0.82–1.13) | 1.00 (0.85–1.17) | 0.95 (0.73–1.22) | 0.97 (0.75–1.25) |
| Summer, above 30 °C | 0.87 (0.72–1.04) | 0.88 (0.74–1.05) | 0.81 (0.59–1.08) | 0.91 (0.68–1.20) |
| The coldest 20% of days | 1.06 (0.91–1.23) | 0.98 (0.83–1.14) | 1.09 (0.85–1.38) | 1.00 (0.78–1.27) |
| Sudden change in temperature (top 10%) | 1.15 (0.94–1.40) | 0.89 (0.71–1.10) | 1.19 (0.86–1.60) | 0.86 (0.60–1.21) |
| Rainy days (1 mm or more) | 1.05 (0.90–1.23) | 0.92 (0.79–1.07) | 1.00 (0.78–1.28) | 0.93 (0.72–1.18) |
| Heavy rain (over 10 mm) | 1.43 (1.07–1.89) | 1.04 (0.78–1.37) | 1.19 (0.70–1.91) | 1.08 (0.68–1.64) |
By month, between 17.6 (June) and 24.7 (December) earthquakes per 100 days; the differences between months are the size chance produces (p = 0.55).
#The full moon and the tide in the rock
The Moon does not only pull on the seas. It pulls on the whole planet, and the rock stretches and squeezes slightly twice a day. The stress it adds at depth is a few kilopascals, as much as a storm, and it reaches the Vrancea nest.
If the Moon pushed earthquakes, they would gather at full and new moon, when tides are largest. The spiral puts each of the 1,059 deep earthquakes of magnitude 3 or more from March 2012 to October 2026 (more than the 1,039, which stop on 30 June, where the weather data end) at the angle of the Moon's phase at its moment: new moon at the top, full moon at the bottom, years from the centre outwards.
In the two days around full moon, 58 earthquakes occurred, and around new moon 84: together 142, against about 143 by chance. The Schuster test on the doubled angle, which looks for gathering at full and new moon at once, gives p = 0.14: nothing. The Moon's phase is only a rough stand-in for the tide; an effect tied to the exact tidal stress on Vrancea's faults is not tested here.
In 2007, Cadicheanu, van Ruymbeke and Zhu looked in Vrancea's earthquakes of 1981–2005 for a link with the 12-hour tides, a related question. Over all years together, they found the tidal phases scattered at random; some one-year windows came out significant, and the authors read them as a sign before larger earthquakes. When many windows are tried, some come out significant by chance alone, and the data here show no such link.
#The compass before an earthquake
The third story says the compass needle, or the magnetic field, grows restless before an earthquake. The Surlari geomagnetic observatory, north of Bucharest, measures the field every minute. On the day before each of the 155 Vrancea earthquakes of magnitude 4 or more from 2012 to 2026, the field's restlessness crossed the threshold of the most restless 5% of days in 4.5% of cases, as often as on any other day.
This measure at this station shows no link. Other researchers (Stănică et al., 2011) use ultra-low-frequency signals at Provița de Sus, with Surlari as the reference station, and report anomalies before some earthquakes; their method is not tested in these data. The same Surlari magnetometer does, on the other hand, record the Bucharest metro passing.
#Play the earthquake forecaster
Two days, with the weather of each in the chosen place. On one, Vrancea had a deep earthquake; on the other, none. Eight pairs, drawn at random from the real data.
Which day had the earthquake?
#Questions
Is there such a thing as earthquake weather?
Not in Vrancea's data. From March 2012 to June 2026, the hottest 20% of days had 19.7 deep earthquakes per 100 days and all other days 19.9 (Bucharest weather; at Focșani the ratio is 0.95). For sudden changes in temperature and the full moon, the data show no clear change in the count either. The one result past the threshold, heavy rain at Focșani, is the size chance gives across 32 comparisons and disappears for earthquakes of 3.5 or more.
Are there more earthquakes in summer or in winter?
By month, between 17.6 (June) and 24.7 (December) earthquakes per 100 days; the differences between months are the size chance produces (p = 0.55).
Can a storm trigger an earthquake?
A storm's pressure does reach the depths, but it is a thousand times smaller than the stresses an earthquake releases. For some shallow faults close to failure, researchers have found links with heavy rain, and in Taiwan links between typhoons and slow slip on faults. For Vrancea's nest at 90–150 km, the data show nothing.
Does the full moon bring earthquakes?
Not in Vrancea. The two days around full moon had 58 earthquakes, against about 72 expected by chance; around full and new moon together, 142 against 143.
Why does the weather where I live matter, if the earthquake is in Vrancea?
Only as a sample of the region's weather. Cold fronts, heatwaves and heavy rain cover hundreds of kilometres, so the weather in almost any Romanian town follows the weather over Vrancea. The result is the same everywhere: your hot days did not have more earthquakes.
Can earthquakes be predicted from the weather?
No. If hot days brought 50% more earthquakes, a simple test on the same data would almost certainly have shown it (over 99.9% chance). In Bucharest, the 95% confidence interval runs from 16% fewer to 15% more earthquakes on hot days, and the central estimate is 1% fewer.
On the same subject
#How it was counted
Earthquakes: the INFP catalogue, deep earthquakes (60 km or more) in the Vrancea box (45.2–46.1° N, 25.9–27.0° E), magnitude 3 or more, 1 March 2012 – 30 June 2026. The catalogue effectively starts on 1 March 2012; from 15 to 20 August 2014 it has no event from anywhere, and those days are left out. Of earthquakes within 3 days and 30 km of each other, only the first is kept, so a possible aftershock is not counted twice: 1,039 remain of 1,263. Vrancea's deep nest has few aftershocks, so some of those removed are simply close in time; the results do not change.
Weather: E-OBS v34.0e, daily maximum temperature and rainfall on a 0.1° grid, for the square each commune falls in. "Hot" means hot for its date, against the 1991–2025 mean for that day and against the mean of that year, so that recent warmer years do not pass for hot days. A day runs from 6 to 6 in the morning, Romanian time.
Each ratio has a 95% confidence interval, and significance comes from a test with the days shuffled at random 2,000 times, only within the same month of the same year. If hot days had brought 20% more earthquakes, a simple test on the same data would have caught it 68% of the time; with 50% more, over 99.9% of the time.
By day, surface noise hides some small earthquakes, including some of magnitude 3, and days are longer in summer than in winter. That may explain why June has the fewest recorded earthquakes and December the most. So every comparison is repeated for magnitude 3.5 or more, with the same result.
Moon phases come from the JPL DE422 ephemeris, at the time of each earthquake. For the large earthquakes since 1950, dates, times and moment magnitudes come from the ROMPLUS catalogue. The INFP catalogue files are not redistributed.
The page publishes only values computed from E-OBS (each day's departure from normal, the change from one day to the next, a rain class), not the E-OBS data. "We acknowledge the E-OBS dataset and the data providers in the ECA&D project (https://www.ecad.eu)"; Cornes et al., 2018.
#Sources
- INFP, Romanian Seismic Network: FDSN event service (doi:10.7914/SN/RO)
- INFP, ROMPLUS catalogue (the large earthquakes since 1950)
- E-OBS v34.0e, ECA&D and Copernicus: daily maximum temperature and rainfall, 0.1° grid
- Cornes et al. (2018), "An Ensemble Version of the E-OBS Temperature and Precipitation Data Sets", JGR Atmospheres
- NASA JPL, DE422 ephemeris (Moon phase)
- INTERMAGNET, Surlari geomagnetic observatory (SUA), minute data
- ANCPI municipality and commune boundaries, via geo-spatial.org (26 March 2025 version; Romanian Open Government Licence)
- Aristotle, Meteorology, Book II, ch. 8 (tr. E. W. Webster)
- Știri pe Surse (Chișinău), 20 October 2026: "Myth or truth? Why people say the weather changes before a major earthquake" (in Romanian)
- USGS: "Is there earthquake weather?"
- Hainzl et al. (2006), "Evidence for rainfall-triggered earthquake activity", GRL
- Allmann and Shearer (2009), "Global variations of stress drop for moderate to large earthquakes", JGR
- Cadicheanu, van Ruymbeke and Zhu (2007), "Tidal triggering evidence of intermediate depth earthquakes in the Vrancea zone", NHESS
- Stănică et al. (2011), electromagnetic signals before Vrancea earthquakes, NHESS