Vârful Omu · 2,504 m · weather station

A year on the roof of Romania

On top of the Bucegi mountains, at 2,504 m, the Omu weather station has written down the temperature almost every day since 1961 and the depth of snow since 1980. Scroll: a year goes by.

Scroll to move through the year, 1 August to 31 July. The white that rises is the snow measured that day.

1 Aug

Mean: –
Low / high: –
Wind: –
Rain, snow: –
Snow depth: –

Year (1 Aug – 31 Jul)

Station readings: ECA&D, GHCN-Daily · clouds: ERA5 model · rime: drawn

Vârful Omu · 2,504 m · weather station

What is real in the picture

Every number beside the picture is a daily reading from the station: the mean, lowest and highest temperature, the wind, the rain and melted snow, and the depth of snow on the ground. The snow rises on the photograph to the depth measured that day. In the decade view the numbers are medians or means across that decade's winters.

Measured at the station: temperature, wind speed, precipitation, snow depth.

From a weather model, not from the station: cloud cover, humidity and wind direction (ERA5 reanalysis, a model grid cell with a terrain height of about 2,300 m over Omu). They set only the clouds and which way the wind points in the picture (east is to the right; the camera direction is not known). The cups spin at the measured wind speed; rain or snow falls by the day's mean temperature, snow at 0.5 °C or below.

Drawn by a rule: the rime. It grows on cold, humid days and breaks off on days above 0 °C. Nobody measured rime in this file, so its amount is an illustration.

The photograph of the hut was taken in September 2009 by Amorphisman (CC BY-SA 3.0, Wikimedia Commons). The snow is drawn over it at an approximate scale.

The air warmed; the snow season did not get shorter

The question was whether the snow season on Omu has shortened. The test was written down before it was run: the window, the method and what would count as a yes. The test finds no sign of a shorter season.

Across the 42 snow years that pass the reporting check (1980/81 to 2024/25; three winters set aside because the depth readings had gaps), snow lay on the ground, at least 1 cm deep, on a median of 219 days a year. The change is +2.5 days per decade (95 % interval −2.5 … +7.6; Mann–Kendall p = 0.30). By the rule set beforehand, a shorter season is not supported.

Days a year with snow on the ground, each snow year from 1980/81 (August to July). The dashed line is the Theil–Sen trend. Three winters with too many gaps are left out.

The air itself has warmed. Annual mean temperature on Omu rose 0.29 °C per decade since 1961 (95 % interval 0.21 to 0.38; p < 0.001): about 1.8 °C over the 64 years on the trend line.

Annual mean temperature at Omu, each calendar year 1961–2025.

The warming shows in the annual mean. The first snow comes later in this record, but that test only comes near the line (p = 0.053). The count of days with a high at or below 0 °C does not settle it either: fewer in the winters tested, more over 1961–2025, neither clearly. The snow season is not shorter in the 42 winters tested, and the data cannot say more than that.

A decade at a time

This is the table behind the second half of the picture. The picture draws a day-by-day median of each decade on the same photograph. Medians are across the winters in the decade; the snow columns start in 1980 because the file holds no snow depth before then.

Decade (snow years)	Days with snow cover	First snow day	Last snow day	Deepest snow, cm	Mean temp., °C	Days with high ≤ 0 °C
1960s (1961/62–1969/70)	snow depth not recorded in the file				−2.4	150
1970s (1970/71–1979/80)	snow depth not recorded in the file				−2.6	158
1980s (1980/81–1989/90)	202	13 Sep	20 Jun	154	−2.6	168
1990s (1990/91–1999/00)	221	6 Sep	7 Jul	216	−2.4	172
2000s (2000/01–2009/10)	229	8 Sep	17 Jun	175	−1.9	170
2010s (2010/11–2019/20)	213	24 Sep	18 Jun	147	−1.1	158
2020s (2020/21–2024/25)	218	22 Sep	3 Jul	198	−0.7	163

All the tests, with the one that carries the claim first

Snow-cover days per year, 42 winters (the test written down first): +2.5 days/decade [−2.5 … +7.6], p = 0.30
Last day with snow cover: −1.9 days/decade [−7.0 … +3.3], p = 0.51
First day with snow cover: +4.6 days/decade [+0.0 … +8.6], p = 0.053
Days with a high at or below 0 °C, same period, all 45 winters (no snow reading needed): −2.5 days/decade [−5.0 … +0.3], p = 0.083
Days with a high at or below 0 °C, 1961/62–2024/25: +1.4 days/decade [−0.3 … +3.3], p = 0.13
Annual mean temperature, 1961–2025: +0.29 °C/decade [+0.21 … +0.38], p < 0.001

Intervals are Sen's 95 % intervals. Only the first test carries the claim; the rest are descriptions of the same record, and one of them (the first snow date, p = 0.053) is close to the line without crossing it.

The test as written before it was run

Window: snow years 1 August to 31 July, 1980/81 to 2024/25. Metric: days with snow depth of 10 mm or more. Method: Mann–Kendall trend test (two-sided) with Theil–Sen slope. Decision: a shorter season is supported only if the slope is negative and p < 0.05.

Reporting check: a winter counts only if depth is reported on at least 95 % of its January–March days; if more than 4 of 45 fail, the test is not run. Three failed.

Written on 13 November 2026, before any snow trend was computed. The author had already seen the number of reported snow days per calendar year, which is close to the metric; the test was not changed after that.

Your day on Omu

Pick a date (your birthday, or any day since 1928) and one of the other Romanian weather stations. The page shows the daily mean temperature on Omu and there. Your date stays on your device and is not sent anywhere.

The towns are the 26 other Romanian weather stations in the ECA&D data. Choose the nearest.

What this does not show

Snow depth comes from NOAA's GHCN-Daily for the same station, not from ECA&D, and only from the winter of 1980/81. A day with no depth reading is counted as no snow: in October and November, days with a reading averaged a high of −2.2 °C and days without +3.8 °C. That is an inference from the file, not a statement by the observers.
Three winters (1980/81, 2001/02 and 2013/14) are left out because snow depth was reported on fewer than 95 % of their January–March days. From May 2025 the file holds explicit zeros that earlier years do not; the rule for counting snow days gives the same reading before and after.
Up to ten days missing in the middle of deep snow (20 cm or more on both sides) are drawn as continuing; the tests count it as no snow, as written beforehand. Counts of days with 1 m or more of snow and the precipitation totals in this file change from decade to decade in ways that look like changes in how or where they were measured, so the page does not compare them. The daily rain-and-snow reading is shown as recorded. The deepest snow of a year (72 to 293 cm) varies without a clear trend.
Wind speed has gaps before 1980 (23 % of days missing in the 1960s, 31 % in the 1970s), so the page shows wind only from 1980. Omu and Ceahlău Toaca (1,897 m) both average about 8 m/s (1991–2020); the 11 other stations with wind data average under 3 m/s. Single days with the highest daily mean wind include values that look too high and are not quoted.
Omu is the coldest of the 24 Romanian stations in the file with enough data (mean −1.7 °C in 1991–2020; the next coldest, Ceahlău Toaca, 1.3 °C on the same days). The lowest daily minimum in the file is −38.0 °C, on 10 February 1929.
The series is the ECA&D blended series, not adjusted for changes of instrument or site. A pre-registered interval (the spread of all pairwise slopes) turned out to measure scatter, not the uncertainty of the trend; Sen's interval replaces it here and the original is kept in the notes. Before 1961 the day's mean is the average of the lowest and highest temperature, and 19 % of the highs and 6 % of the lows are missing in 1928–60.
Omu's warming is not a quirk of one instrument: Ceahlău Toaca, the other mountain station, warmed 0.36 °C per decade, and the difference between the two stations has no trend (−0.01 °C per decade, p = 0.34). Sibiu and Cluj-Napoca warmed faster (0.38 and 0.38), Bucharest-Băneasa more slowly (0.23).
The tested window ends with 2024/25 because it was fixed in advance. The winter of 2025/26 is also in the file and had 239 days with snow cover, inside the range of the other winters (171 to 261); adding it gives +2.9 days per decade, p = 0.19.
Rime, clouds and wind direction are not measurements. The cloud, humidity and wind direction come from ERA5 for a point over Omu in a model cell 2,300 m high (the station is at 2,504 m).

Sources

Temperature, wind and precipitation: ECA&D blended series (file created 11 November 2026), station VARFU OMUL (STAID 226) and 26 other Romanian stations. Klein Tank, A.M.G. and coauthors, 2002, Int. J. Climatol., 22, 1441–1453. Data are used here for non-commercial education under the ECA&D data policy.
Snow depth: Global Historical Climatology Network–Daily, station ROM00015280 (VARFU OMUL), NOAA NCEI. Menne, M.J. and coauthors, 2012, J. Atmos. Oceanic Technol., 29, 897–910.
Cloud cover, humidity, wind direction: ERA5 (Hersbach and coauthors, 2020, Q. J. R. Meteorol. Soc., 146, 1999–2049), through the Open-Meteo archive (CC BY 4.0).
Photograph: Amorphisman, “Cabana Omu, Varful Omu.JPG”, Wikimedia Commons, 24 September 2009, CC BY-SA 3.0.