Loading the measurements…
Cluj-Napoca, Romania · 61 measured winters
Cluj under the lid: the winter nights when it is warmer on the mountain than in the city
Cluj-Napoca is a large city in a valley in Transylvania, in north-western Romania. Weather balloons launched from the city measured the air above it on 5,007 winter nights (November to February) between 1957 and 2012. On 1,726 of them, 34.5%, the air at about 1,500 metres above sea level, some 1,100 m above the city, was warmer than at ground level: the warm air above holds the cold air in the valley like a lid, under which smoke and fog are slow to clear. The picture shows the night with the largest gap in a complete sounding: 17 January 1982.
Choose a nightor scroll the story17 January 1982, 2 a.m.
What the picture shows#
The terrain is real, seen from the north: the valley of the Someșul Mic river with the city in front, Feleac Hill behind it, the Muntele Mare massif at the back. East is on the left. Heights are exaggerated against the width: the picture spans 46 km from east to west, so slopes look much steeper than they are.
The colour of the air shows the temperatures the balloon measured at a few heights; between them the colour is drawn straight from one value to the next. The line on the left shows the same thing as a graph: each dot is one reading from the balloon, with its temperature beside it.
The balloon rose from one place: from 1964, the weather station on Vânătorului Street, at about 410 m. The picture spreads this single profile across the whole relief; 30 km from the station the air can have a different temperature.
17 November 1981, 2 a.m.
An ordinary night: the higher, the colder#
In the same winter, on 17 November 1981, the balloon found +1.0 °C at ground level and −5.5 °C at 1,448 m. This is air stirred by wind: it cools as you climb, and the mountain is colder than the city.
In summer, June to August, the air is warmer at 1,500 m than at ground level on only 5.0% of nights. In winter the share rises to 34.5%.
A night without a lid is not necessarily a night without an inversion: the air can warm with height only in the first tens or hundreds of metres. Of 1,928 winter nights without a lid in which the sounding also has a level below 850 hPa, on 42.0% the air at that level was warmer than at ground level.
17 January 1982, 2 a.m.
The largest gap: −13.3 °C at ground level, +5.4 °C at 985 metres#
At 2 a.m. the balloon measured −13.3 °C at ground level, close to the station thermometer’s −13.5 °C. At 792 m above sea level, 377 m higher, the balloon measured +3.0 °C. At 985 m, +5.4 °C: 18.7 degrees warmer than at ground level, over a climb of 570 m.
The air stayed above freezing up to about 2,900 m. At the 850 hPa level, the one by which the nights are counted (1,569 m that night), it was again +5.4 °C. The nights before and after had gaps of 17.7 and 17.9 degrees.
The 18.7 degrees of 17 January 1982 are the largest gap among the winter soundings with a complete profile. Another sounding, from 17 January 1975, gives a larger one, 19.5 degrees, with +12.6 °C at 850 hPa in the middle of January; it lacks the heights of the levels in between, and its 850 hPa reading has nothing to be checked against.
The mechanism
Why the cold gathers in the valley#
On clear, still nights the ground loses heat and chills the air above it. Cold air is heavier: it drains down the slopes and pools in the valley like water in a basin. Warmer air stays above and holds it in place like a lid. Meteorologists call it a temperature inversion.
Under the lid the air mixes poorly with the air above, so smoke, exhaust fumes and fog clear more slowly. At the airport, on lid nights in the winters of 2016–2025, visibility at 2 a.m. was under 1 km on 21.2% of nights; on the others, on 3.3%.
27 December 2024 – 3 January 2025
Without the balloon: two stations, almost 950 m apart in height#
The last night sounding from Cluj in the public archive dates from 2012. Since then the lid can be followed with two thermometers: the one at the city’s weather station, where the balloon was launched, at 410 m, and the one at Băișoara station, at 1,356 m on the Muntele Mare massif, 34 km from the centre. The picture also shows the airport, on the valley floor at 316 m, and, as small dots, sensors that residents have put up on balconies and in gardens.
The dotted line joins the stations’ thermometers. Above the city’s hills, as far as Băișoara, nobody measures: there the colour of the air is drawn straight from one value to the next.
Scroll: each step is an hour
A week under the lid#
Of 169 hours, 149 were warmer at Băișoara than at the city station. On 29 December at 3 a.m. the city station had −3.4 °C, the airport −5.0 °C and Băișoara +2.2 °C.
On New Year’s Day at 2 p.m. the city station had −3.1 °C, the airport −2.0 °C and Băișoara +8.5 °C. During that week, visibility at the airport was below 1 km in eight of those hours.
1 January 2025, 2 p.m.
Where the picture misleads: the mountain station by day#
By day a mountain station usually reads higher than the open air at the same height; the usual explanation is that the sun warms its slope. On 453 winter days in 1974–1992 when both the stations and the balloon measured at 2 p.m., the two stations indicated a lid on 36.2% of days; the balloon found one on 20.8%.
The +8.5 °C at Băișoara is real for anyone who goes up there. On those 453 days, though, at the same height, the balloon found the air above the city on average 1.7 degrees colder than the station read, and the picture spreads the station’s colour across its whole width. That is why lid nights are counted at 2 a.m.
29 December 2024, 3 a.m.
On lid nights the hills of Cluj have a small thermal advantage#
Zorilor, Europa and Bună Ziua are districts on the slopes south of the centre. Sensor.Community, a volunteer sensor network, has 19 outdoor sensors in Cluj, between 304 and 497 m above sea level. Those in the three hill districts stand at 465–497 m; those in the Mărăști, Gheorgheni and Bulgaria districts, on the valley floor. On lid nights the gap between hill and valley widens by 0.6 degrees in the hills’ favour; by 1.1 degrees if lid nights are chosen by the airport thermometer and not the station’s.
The comparison rests on few sensors: two on the hill and three in the valley, over 185 lid nights. With each of them left out in turn the result comes out between 0.4 and 0.8 degrees, and the hours of one night are not independent measurements. It is an order of magnitude; it says nothing about any particular street. The hills the city has spread onto are only 150 m above the valley, and the warmest air usually sits much higher.
One sensor in the east of the city reads 2.1 degrees above the airport at night even in wind, with large swings from hour to hour. It probably sits near a source of heat. It is left out of the comparison.
22 January 1979, 2 a.m.
How high the warm air sits#
The night in the picture, 22 January 1979, is a typical lid night: −4.5 °C at ground level, and the warmest level, +0.4 °C, at 1,149 m.
On winter nights with a lid, the warmest level the balloon reported usually lay between about 800 and 1,500 m, centred near 1,100 m. Only on 7.9% of nights was it at 700 m or lower, the height of Feleacu, the village on the ridge south of the city.
These heights are approximate: below 1,600 m the soundings usually have only three levels, and the true maximum may lie between them.
1957–2025
61 winters, night by night#
6,188 nights with measurements, from 1957 to 2025. Each square is a winter night, from 1 November to the end of February, and each row is a winter. Orange means it was warmer above than below; blue, colder; the dim squares are nights without a measurement.
On the nights measured by the balloon, December has a lid on 40.0% of nights, January on 38.8%, November on 33.5%, February on 25.0%.
The winters with a dot on the left are measured with the two stations, the others with the balloon. Seven winters are missing: for them the archive has fewer than 60 nights from either the balloon or the stations.
Two instruments, the same nights
The test: stations and balloon agree on 89.6% of nights#
The rows with a dot come from a different instrument. How well two stations stand in for the balloon can be checked on the nights when both the stations and the balloon measured: 898 winter nights between 1974 and 2009.
They gave the same answer on 805 of them. On 46 nights only the stations showed a lid; on 47, only the balloon. Answering “no lid” every time would have been right on 64.1% of nights. Where the gap measured by the balloon was at least 2 degrees either way, the answers matched on 97.4% of nights.
The nine winters from 2016–17 to 2024–25 had a lid on 29.1% of nights on average (the mean of winters, not of nights). The balloon’s 41 winters with at least 80 measured nights average 33.9%, from 15% (1961–62) to 50% (1989–90); eleven of them were below the recent average.
Choose a winter night#
The night you were born, if it was in winter, a Christmas from your childhood, the winter you remember: type the date or tap a square in the grid of winters. The picture shows the air at 2 a.m. that night.
No night chosen. The picture shows 17 January 1982.
Tap a square to choose its night.
Where the measurements stop#
Pollution. A lid keeps in the valley what is emitted in the valley, but thermometers do not say how clean the air is. Sensor.Community’s particle sensors read too high in fog, which is exactly when the lid is on. Official measurements are at calitateaer.ro, the national air quality network.
The temperature of a particular street. The sensors are few and nobody knows where each is mounted: the error of an estimate by address would be larger than the differences between districts.
The trend. In the soundings the share of lid nights rises: 26.3% up to the winter of 1963–64, while the balloons were launched from the valley, 33.1% in the winters of 1964–1989, 39.4% in the winters of 1990–2012. Over the same years, though, the launch site and the type of sonde changed, and the winters from 2016–17 to 2024–25, measured with the stations, had a lid on 29.0% of nights. From these data one cannot say whether the air changed or only the instruments.
The valley floor and the station hill. The station hill is almost 100 m above the airport, but the mean temperature difference is small and goes both ways: at 2 a.m. in the winters of 2016–2025 the airport and the city station read the same on average (a difference of 0.04 degrees). On lid nights the airport was on average 0.2 degrees colder: at least 2 degrees colder on 10.0% of them and at least 2 degrees warmer on 3.5%.
Faulty soundings. Of 3,696 winter soundings that can be compared with the station’s own ground thermometer, 22 give a ground temperature at least 5 degrees away from it and were left out. One of them, from 19 December 1995, gives −18.5 °C at ground level when the station thermometer read −5.1 °C; it would have been the largest gap in the whole archive. Soundings before 1973 have nothing to be compared with.
How many lid nights each winter had#
Each winter has its own page and picture, to send to someone who lived through it.
| Winter | Nights measured | With a lid | Share | Instrument |
|---|---|---|---|---|
| 2024–25 | 115 | 43 | 37% | stations |
| 2023–24 | 121 | 29 | 24% | stations |
| 2022–23 | 119 | 35 | 29% | stations |
| 2021–22 | 120 | 29 | 24% | stations |
| 2020–21 | 117 | 31 | 26% | stations |
| 2019–20 | 121 | 41 | 34% | stations |
| 2018–19 | 120 | 29 | 24% | stations |
| 2017–18 | 120 | 24 | 20% | stations |
| 2016–17 | 119 | 50 | 42% | stations |
| 2015–16 | 69 | 18 | 26% | stations |
| 2011–12 | 121 | 55 | 45% | balloon |
| 2010–11 | 120 | 57 | 48% | balloon |
| 2009–10 | 120 | 35 | 29% | balloon |
| 2008–09 | 119 | 40 | 34% | balloon |
| 2007–08 | 119 | 52 | 44% | balloon |
| 2006–07 | 120 | 40 | 33% | balloon |
| 2005–06 | 116 | 44 | 38% | balloon |
| 2004–05 | 117 | 41 | 35% | balloon |
| 2003–04 | 98 | 36 | 37% | balloon |
| 2001–02 | 93 | 39 | 42% | stations |
| 2000–01 | 60 | 32 | 53% | balloon |
| 1999–2000 | 104 | 39 | 38% | balloon |
| 1998–99 | 89 | 35 | 39% | balloon |
| 1997–98 | 68 | 31 | 46% | stations |
| 1996–97 | 90 | 44 | 49% | balloon |
| 1995–96 | 79 | 39 | 49% | balloon |
| 1994–95 | 90 | 29 | 32% | balloon |
| 1993–94 | 71 | 29 | 41% | balloon |
| 1992–93 | 69 | 33 | 48% | balloon |
| 1991–92 | 66 | 21 | 32% | balloon |
| 1990–91 | 63 | 20 | 32% | stations |
| 1989–90 | 90 | 45 | 50% | balloon |
| 1988–89 | 106 | 29 | 27% | balloon |
| 1987–88 | 112 | 28 | 25% | balloon |
| 1986–87 | 104 | 46 | 44% | balloon |
| 1985–86 | 110 | 26 | 24% | balloon |
| 1984–85 | 117 | 53 | 45% | balloon |
| 1983–84 | 119 | 49 | 41% | balloon |
| 1982–83 | 116 | 38 | 33% | balloon |
| 1981–82 | 114 | 43 | 38% | balloon |
| 1980–81 | 112 | 32 | 29% | balloon |
| 1979–80 | 117 | 34 | 29% | balloon |
| 1978–79 | 117 | 47 | 40% | balloon |
| 1977–78 | 109 | 36 | 33% | balloon |
| 1976–77 | 112 | 36 | 32% | balloon |
| 1975–76 | 108 | 42 | 39% | balloon |
| 1974–75 | 108 | 29 | 27% | balloon |
| 1973–74 | 105 | 37 | 35% | balloon |
| 1972–73 | 94 | 31 | 33% | balloon |
| 1971–72 | 85 | 23 | 27% | balloon |
| 1970–71 | 97 | 31 | 32% | balloon |
| 1969–70 | 85 | 19 | 22% | balloon |
| 1968–69 | 63 | 29 | 46% | balloon |
| 1966–67 | 62 | 17 | 27% | balloon |
| 1963–64 | 101 | 37 | 37% | balloon |
| 1962–63 | 98 | 26 | 27% | balloon |
| 1961–62 | 100 | 15 | 15% | balloon |
| 1960–61 | 97 | 18 | 19% | balloon |
| 1959–60 | 112 | 28 | 25% | balloon |
| 1958–59 | 101 | 32 | 32% | balloon |
| 1957–58 | 84 | 26 | 31% | balloon |
How the count was made#
Night is the 00 UTC observation, which is 2 a.m. in Romania in winter. Winter runs from 1 November to the end of February.
The balloon. NOAA’s IGRA archive holds 38,157 soundings from Cluj-Napoca, from 1951 to 2012. The count starts in 1957, when the night soundings moved to 00 UTC, and uses the soundings that have both the ground level and the 850 hPa level with its height, 1,480 m above sea level on average: 5,007 winter nights, without the 22 rejected soundings (see above). A night has a lid if the temperature at 850 hPa is higher than at ground level. The test does not find every temperature inversion: those that end well below 1,500 m fall outside it, so 34.5% is the frequency of the lid as defined here; low inversions are not counted. The grid shows only winters with at least 60 measured nights; 184 nights from sparser winters are in the overall count but not in the grid.
Ground level in the soundings. Until 29 February 1964 the soundings report the ground at 333 m; from 1 May 1964, at 416 m, then between 410 and 420 m, a sign that the launch site moved from the valley onto the hill. Each time, the comparison is with the ground of that sounding. The type of sonde also changed several times.
The stations. From NOAA’s ISD archive: the Cluj-Napoca weather station (410 m, where the balloon was launched) and Băișoara (1,356 m). A night has a lid if at 00 UTC Băișoara is warmer than the city station. The nine winters from 2016–17 to 2024–25 have 1,072 nights in all, 311 of them with a lid (29.0%); between 24 and 50 per winter. The same rule fills four winters in which the balloon has too few nights: 1990–91, 1997–98, 2001–02 and 2015–16. With the airport thermometer (316 m, from the airport’s hourly reports) in place of the station, the count would be 327 of 1,071 (30.5%). The METAR archive up to and including 2011 largely repeats the city station’s data, so the airport appears here only from 2012. Vlădeasa, at 1,837 m and 60 km to the west, was warmer than the city station on 14.1% of the 1,065 nights for which it has data. The ISD-Lite archive used here has not been updated since August 2025, and the archive that replaces it at NOAA (GHCNh) does not include Băișoara station. The winter of 2025–26 is left out: its observations exist in other archives and first have to be checked and matched to the series used here.
The sensors. Hourly means of outdoor BME280 sensors from the Sensor.Community archive, for the winters of 2020–2025. A sensor is used if, in night hours with wind of at least 10 knots at the airport, its offset from the airport is stable; those with a spread above 1.5 times the network median are rejected. The result compares the same sensors with themselves: the hill-minus-valley difference on lid nights minus the hill-minus-valley difference on other nights, over 1,004 and 2,505 hours. The result is 0.6 degrees; with a second variant of the rule (wind of 8 knots, more sensors), 0.6.
The terrain comes from the Copernicus GLO-30 model. The area drawn is 46 km from east to west and 34 km from north to south, cut into 30 strips; each ridge in the picture is the highest ground in one strip.
Sources#
- NOAA NCEI, Integrated Global Radiosonde Archive (IGRA) v2, station ROM00015120 Cluj-Napoca: all soundings, 1951–2012. Public domain. ncei.noaa.gov
- Iowa Environmental Mesonet, archive of METAR reports for LRCL (Cluj airport): temperature, wind, visibility; used from 2012. mesonet.agron.iastate.edu
- NOAA NCEI, Integrated Surface Database (ISD-Lite): Cluj-Napoca (151200), Băișoara (151630) and Vlădeasa 1800 (151190), hourly observations, 1973–2025. ncei.noaa.gov
- Sensor.Community, archive of BME280 sensors in Cluj-Napoca, 2020–2025. Open Database License (ODbL) 1.0; the derived table on this page may be reused under the same terms. archive.sensor.community
- The relief: produced using Copernicus WorldDEM-30 © DLR e.V. 2010-2014 and © Airbus Defence and Space GmbH 2014-2018 provided under COPERNICUS by the European Union and ESA; all rights reserved. dataspace.copernicus.eu
- American Meteorological Society, Glossary of Meteorology: “inversion” and “cold pool”, for the description of the mechanism. glossary.ametsoc.org