How Often Extremes Return
How often a once in 50 years hottest day, cold night or wettest day of 1951 to 1980 returns in the climate of 2025, at up to 9,645 weather stations. Each station's yearly extremes since 1950 are fitted with an extreme value distribution that moves with global temperature; medians over grid cells with intervals, by continent and country.
Data as of . Last revised (2026-11-17).
Key Facts
- At the typical grid cell a hottest day of the year that came once in 50 years in the climate of 1951 to 1980 comes about once in 9.4 years in the climate of 2025 (95 percent interval between 8.5 and 11 years; 5,418 stations in 472 grid cells).
- At the typical grid cell a warmest night of the year that came once in 50 years in the climate of 1951 to 1980 comes about once in 6.2 years in the climate of 2025 (95 percent interval between 5.5 and 7.5 years; 5,129 stations in 438 grid cells).
- At the typical grid cell a coldest night of the year that came once in 50 years in the climate of 1951 to 1980 comes less than once in 1,000 years in the climate of 2025 (95 percent interval between 810 and more than 1,000 years; 5,217 stations in 447 grid cells).
- At the typical grid cell a coldest day of the year that came once in 50 years in the climate of 1951 to 1980 comes about once in 817 years in the climate of 2025 (95 percent interval between 611 and more than 1,000 years; 5,535 stations in 467 grid cells).
- At the typical grid cell a wettest day of the year that came once in 50 years in the climate of 1951 to 1980 comes about once in 33 years in the climate of 2025 (95 percent interval between 29 and 35 years; 9,645 stations in 415 grid cells).
- The hottest day of the year expected once in 50 years is 1.3 C (2.3 F) warmer in the climate of 2025 than in that of 1951 to 1980 at the typical grid cell (95 percent interval +1.15 to +1.40 C).
- In the fits of 1950 to 2025 the change of the hottest day of the year is clear at 1,975 of 5,418 stations (36 percent) with the false discovery rate held at 5 percent: at 1,637 the old value comes more often and at 338 less often.
- "The climate of 2025" means the mean global temperature of the 4 years to 2025, which was 1.10 C (1.98 F) above the level of 1951 to 1980.
Then and Now at a Glance
the climate of 2025 against 1951 to 1980Take the value that a station's hottest day of the year reached about once in 50 years in the climate of 1951 to 1980. How often does a year reach that same value in the climate of 2025? This page answers that for five yearly extremes at every station with a long enough record. "The climate of 2025" means the level of global temperature in the 4 years to 2025, which was 1.10 C (1.98 F) above its level of 1951 to 1980.
The table gives the typical answer: the median over 5 degree grid cells, weighted by area, with a 95 percent interval. Once in 50 years would mean no change. The shift is how far the value reached once in 50 years has itself moved.
| Yearly extreme | Stations (grid cells) | Once in 50 years in 1951-1980, in 2025 once in | 95% interval, years | Comes more often at | Shift of the once in 50 years value | 95% interval of the shift |
|---|---|---|---|---|---|---|
| Hottest day of the year | 5,418 (472) | 9.4 years | 8.5 to 11 | 84% of stations | +1.3 C (+2.3 F) | +1.15 to +1.40 C |
| Warmest night of the year | 5,129 (438) | 6.2 years | 5.5 to 7.5 | 90% of stations | +1.5 C (+2.7 F) | +1.35 to +1.55 C |
| Coldest night of the year | 5,217 (447) | over 1,000 years | 810 to over 1,000 | 9% of stations | +2.9 C (+5.2 F) | +2.70 to +3.20 C |
| Coldest day of the year | 5,535 (467) | 817 years | 611 to over 1,000 | 9% of stations | +2.1 C (+3.8 F) | +1.90 to +2.40 C |
| Wettest day of the year | 9,645 (415) | 33 years | 29 to 35 | 63% of stations | +6.6 mm (+0.26 in), +7.2% | +5.5 to +8.1 mm |
The Hottest Day of the Year
5,418 stationsAt the typical grid cell a hottest day of the year that came once in 50 years in the climate of 1951 to 1980 comes about once in 9.4 years in the climate of 2025 (95 percent interval between 8.5 and 11 years; 5,418 stations in 472 grid cells). That is 5.3 times as often. The old value comes more often at 84 percent of the stations, weighted by area.
The hottest day of the year expected once in 50 years is 1.3 C (2.3 F) warmer in the climate of 2025 than in that of 1951 to 1980 at the typical grid cell (95 percent interval +1.15 to +1.40 C).
| Station | Latitude | Longitude | In 2025 once in |
|---|---|---|---|
| MXN00025045 | 23.97 | -106.60 | 1.0 years |
| MXN00011002 | 20.02 | -100.70 | 1.0 years |
| MXN00006003 | 18.80 | -103.63 | 1.0 years |
| USC00428771 | 40.54 | -112.32 | 1.0 years |
| GME00125410 | 52.52 | 7.31 | 1.1 years |
| JA000047615 | 36.55 | 139.87 | 1.1 years |
| MXN00001019 | 22.12 | -102.35 | 1.1 years |
| JA000047784 | 34.17 | 131.45 | 1.1 years |
| MXN00008097 | 29.18 | -108.13 | 1.1 years |
| RSM00037001 | 44.90 | 37.30 | 1.1 years |
| AU000005010 | 48.05 | 14.13 | 1.1 years |
| USC00043710 | 36.14 | -117.95 | 1.1 years |
| USC00350501 | 43.95 | -120.22 | 1.1 years |
| MXN00032038 | 24.00 | -103.02 | 1.1 years |
| ASN00022008 | -34.37 | 137.67 | 1.2 years |
| HRE00155411 | 42.56 | 18.27 | 1.2 years |
| MXN00020070 | 16.32 | -96.58 | 1.2 years |
| TH000048300 | 19.30 | 97.83 | 1.2 years |
| NZ000093994 | -29.25 | -177.92 | 1.2 years |
| RP000098836 | 6.90 | 122.07 | 1.2 years |
| SPE00120530 | 41.15 | 1.18 | 1.2 years |
| USC00042319 | 36.46 | -116.87 | 1.2 years |
| ASN00065026 | -33.14 | 148.16 | 1.2 years |
| SPE00120620 | 41.52 | -5.73 | 1.2 years |
| SPE00120629 | 41.11 | -1.41 | 1.2 years |
| JA000047638 | 35.67 | 138.55 | 1.2 years |
| MXN00027037 | 17.85 | -93.87 | 1.2 years |
| MXN00014016 | 20.38 | -103.13 | 1.2 years |
| NG000001036 | 14.22 | 1.45 | 1.2 years |
| USC00422592 | 37.77 | -111.60 | 1.2 years |
| JA000047843 | 32.70 | 128.83 | 1.3 years |
| RSM00034720 | 47.20 | 38.95 | 1.3 years |
| FRE00106184 | 49.38 | 1.18 | 1.3 years |
| HRE00105196 | 42.77 | 16.90 | 1.3 years |
| ASN00040068 | -25.93 | 153.19 | 1.3 years |
| JA000047616 | 36.05 | 136.22 | 1.3 years |
| JA000047632 | 35.40 | 136.77 | 1.3 years |
| PO000008575 | 41.80 | -6.73 | 1.3 years |
| MXN00026102 | 27.18 | -109.18 | 1.3 years |
| NLE00152500 | 52.27 | 6.89 | 1.3 years |
| Return period in 2025 | Percent of stations |
|---|---|
| under 5 years | 22.8% |
| 5 to 10 years | 14.5% |
| 10 to 20 years | 14.3% |
| 20 to 35 years | 10.3% |
| 35 to 50 years | 5.8% |
| 50 to 75 years | 6.8% |
| 75 to 100 years | 4.4% |
| 100 to 200 years | 8.3% |
| 200 to 500 years | 5.2% |
| 500 or more years | 7.7% |
The stations spread widely around the typical value, and much of that spread is the uncertainty of a single fit rather than a real difference between places. At the median station the 90 percent interval of its own return period for the hottest day of the year runs from 4.7 to 84 years. In the fits of 1950 to 2025 the change of the hottest day of the year is clear at 1,975 of 5,418 stations (36 percent) with the false discovery rate held at 5 percent: at 1,637 the old value comes more often and at 338 less often.
Warm Nights, Cold Extremes and the Wettest Day
the climate of 2025 against 1951 to 1980The warmest night is the highest daily minimum of the year. At the typical grid cell a warmest night of the year that came once in 50 years in the climate of 1951 to 1980 comes about once in 6.2 years in the climate of 2025 (95 percent interval between 5.5 and 7.5 years; 5,129 stations in 438 grid cells). That is 8.1 times as often. The old value comes more often at 90 percent of the stations, weighted by area. The warmest night of the year expected once in 50 years is 1.5 C (2.7 F) warmer in the climate of 2025 than in that of 1951 to 1980 at the typical grid cell (95 percent interval +1.35 to +1.55 C).
The coldest night is the lowest daily minimum of the year. At the typical grid cell a coldest night of the year that came once in 50 years in the climate of 1951 to 1980 comes less than once in 1,000 years in the climate of 2025 (95 percent interval between 810 and more than 1,000 years; 5,217 stations in 447 grid cells). A wait that long cannot be measured from 76 years of data: in plain words, at the typical grid cell cold of that degree has not come back in the climate of 2025. That is less often than before. The old value comes more often at 9 percent of the stations, weighted by area. The coldest night of the year expected once in 50 years is 2.9 C (5.2 F) warmer in the climate of 2025 than in that of 1951 to 1980 at the typical grid cell (95 percent interval +2.70 to +3.20 C). At 56 percent of the stations the old value now comes less often than once in 500 years.
The coldest day is the lowest daily maximum of the year. At the typical grid cell a coldest day of the year that came once in 50 years in the climate of 1951 to 1980 comes about once in 817 years in the climate of 2025 (95 percent interval between 611 and more than 1,000 years; 5,535 stations in 467 grid cells). That is less often than before. The old value comes more often at 9 percent of the stations, weighted by area. The coldest day of the year expected once in 50 years is 2.1 C (3.8 F) warmer in the climate of 2025 than in that of 1951 to 1980 at the typical grid cell (95 percent interval +1.90 to +2.40 C). At 37 percent of the stations the old value now comes less often than once in 500 years.
The wettest day is the largest daily precipitation of the year, rain and the water of melted snow. At the typical grid cell a wettest day of the year that came once in 50 years in the climate of 1951 to 1980 comes about once in 33 years in the climate of 2025 (95 percent interval between 29 and 35 years; 9,645 stations in 415 grid cells). That is 1.5 times as often. The old value comes more often at 63 percent of the stations, weighted by area. The wettest day of the year expected once in 50 years is 6.6 mm (0.26 in), or 7.2 percent, larger in the climate of 2025 than in that of 1951 to 1980 at the typical grid cell (95 percent interval +5.5 to +8.1 mm). In the fits of 1950 to 2025 the change of the wettest day of the year is clear at 97 of 9,645 stations (1 percent) with the false discovery rate held at 5 percent: at 77 the old value comes more often and at 20 less often. Rain is patchy: the wettest day of a year at one gauge depends on where single storms pass, so a station's own change is much harder to see than for temperature.
| Yearly extreme | under 10 years | 10 to 20 years | 20 to 50 years | 50 to 100 years | 100 to 500 years | 500 or more years |
|---|---|---|---|---|---|---|
| Hottest day of the year | 37% | 14% | 16% | 11% | 13% | 8% |
| Warmest night of the year | 55% | 16% | 15% | 6% | 5% | 4% |
| Coldest night of the year | 1% | 1% | 4% | 9% | 29% | 56% |
| Coldest day of the year | 3% | 5% | 13% | 14% | 28% | 37% |
| Wettest day of the year | 10% | 18% | 35% | 18% | 13% | 5% |
Each row adds up to 100 percent of that kind's stations. With no change the stations would gather around 50 years, with about as many below as above.
By Continent and Country
the climate of 2025 against 1951 to 1980| Region | Hottest day | Warmest night | Coldest night | Coldest day | Wettest day |
|---|---|---|---|---|---|
| World | 9.4 years (5,418) | 6.2 years (5,129) | over 1,000 years (5,217) | 817 years (5,535) | 33 years (9,645) |
| Africa | 4.2 years (20) | 2.8 years (18) | 73 years (18) | over 1,000 years (20) | n/a |
| Asia | 8.4 years (398) | 4.5 years (195) | over 1,000 years (190) | over 1,000 years (410) | 26 years (261) |
| Europe | 8.1 years (1,217) | 5.4 years (1,211) | over 1,000 years (1,216) | over 1,000 years (1,248) | 34 years (2,012) |
| North America | 19 years (3,592) | 8.6 years (3,493) | 739 years (3,582) | 224 years (3,644) | 29 years (4,791) |
| Oceania | 7.7 years (191) | 7.7 years (193) | 275 years (193) | 278 years (213) | 54 years (2,572) |
| South America | n/a | 24 years (19) | 316 years (18) | n/a | n/a |
| Country | Hottest day | Warmest night | Coldest night | Coldest day | Wettest day |
|---|---|---|---|---|---|
| United States | 28 years (3,105) | 8.5 years (3,017) | 663 years (3,118) | 146 years (3,162) | 32 years (4,231) |
| Australia | 6.8 years (180) | 7.8 years (191) | 244 years (186) | 269 years (201) | 54 years (2,565) |
| Russia | 10 years (450) | 6.2 years (452) | over 1,000 years (457) | over 1,000 years (460) | 31 years (462) |
| Germany | 2.3 years (248) | 6.1 years (246) | over 1,000 years (249) | 908 years (252) | 46 years (353) |
| Sweden | 11 years (96) | 3.8 years (99) | over 1,000 years (92) | over 1,000 years (94) | 25 years (330) |
| Mexico | 16 years (231) | 40 years (204) | 294 years (218) | 94 years (229) | 22 years (284) |
| Netherlands | 1.6 years (14) | 1.8 years (14) | 668 years (14) | over 1,000 years (14) | 18 years (268) |
| Canada | 12 years (236) | 5.7 years (253) | over 1,000 years (226) | over 1,000 years (236) | 27 years (246) |
| Norway | 8.5 years (59) | 8.0 years (59) | 590 years (61) | 747 years (59) | 19 years (203) |
| China | 9.8 years (182) | 4.9 years (173) | 933 years (170) | over 1,000 years (173) | 26 years (169) |
| Japan | 4.0 years (108) | n/a | n/a | over 1,000 years (128) | 25 years (83) |
| Finland | 4.3 years (47) | 3.5 years (60) | over 1,000 years (56) | over 1,000 years (59) | 9.7 years (101) |
| United Kingdom | 7.3 years (73) | 4.3 years (71) | 349 years (73) | 385 years (74) | 46 years (69) |
| Spain | 6.5 years (56) | 4.4 years (52) | 995 years (55) | over 1,000 years (54) | 107 years (54) |
| France | 2.5 years (39) | 1.9 years (38) | 319 years (37) | 321 years (39) | 36 years (39) |
| Thailand | 14 years (36) | n/a | n/a | over 1,000 years (37) | n/a |
| Puerto Rico [United States] | 23 years (12) | 4.1 years (11) | 127 years (12) | 106 years (10) | 60 years (25) |
| Estonia | 3.9 years (16) | 2.8 years (16) | over 1,000 years (16) | 670 years (16) | 40 years (24) |
| Romania | 5.2 years (21) | 5.6 years (23) | 456 years (23) | 549 years (21) | 34 years (19) |
| Kazakhstan | 9.5 years (20) | n/a | n/a | 458 years (21) | n/a |
| Argentina | n/a | 24 years (18) | 392 years (17) | n/a | n/a |
| Ukraine | 2.0 years (12) | n/a | n/a | 837 years (14) | n/a |
| Croatia | 2.1 years (11) | 2.8 years (12) | 208 years (13) | 219 years (13) | 32 years (13) |
| Slovenia | 2.3 years (10) | n/a | over 1,000 years (10) | 985 years (10) | 76 years (10) |
Each cell is the typical return period in the climate of 2025 of the value that came once in 50 years in 1951 to 1980, with the number of stations in brackets. A continent is listed with at least 10 grid cells and a country with at least 10 stations; n/a means fewer. The downloads add the intervals and the shifts. Stations are not spread evenly over a country, and a country with few grid cells has a wide interval or none.
Stations With the Clearest Changes
false discovery rate 5 percentThese are the stations where a change is most firmly established, not where it is largest. Only stations that pass the test described below are listed, and they are ranked by the cautious end of their 90 percent interval. A change at a single station can also come from a move, a new instrument or its surroundings. The station pages compare each station's yearly mean temperature with its neighbours to find such changes, and a station where one is suspected is left out here; "not tested" marks a station with too few neighbours for the comparison. The comparison uses mean temperatures, so it can miss a change that only touched the extremes.
Hottest day: the old value comes more often
| Station | Country | Years used | Once in 50 years in 1951-1980 | In 2025 once in | 90% interval, years | Once in 50 years in 2025 | Station change check |
|---|---|---|---|---|---|---|---|
| Grants AP | US | 59 from 1954 | 38.1 C (100.6 F) | 1.0 years | 1.0 to 1.0 | 44.7 C (112.5 F) | none found |
| Buffalo Bill Dam | US | 56 from 1950 | 33.9 C (93.0 F) | 1.0 years | 1.0 to 1.1 | 39.9 C (103.8 F) | none found |
| Ixpalino | MX | 51 from 1961 | 40.6 C (105.1 F) | 1.0 years | 1.0 to 1.1 | 43.6 C (110.5 F) | not tested |
| Ciudad Obregon (Dge) | MX | 50 from 1961 | 44.1 C (111.4 F) | 1.0 years | 1.0 to 1.1 | 49.9 C (121.8 F) | none found |
| Cave Junction 1 WNW | US | 53 from 1963 | 41.8 C (107.2 F) | 1.0 years | 1.0 to 1.1 | 49.1 C (120.4 F) | none found |
| Pericos | MX | 62 from 1951 | 34.9 C (94.8 F) | 1.0 years | 1.0 to 1.2 | 39.2 C (102.6 F) | none found |
| Tooele | US | 76 from 1950 | 38.5 C (101.3 F) | 1.0 years | 1.0 to 1.2 | 43.1 C (109.6 F) | none found |
| Mulege | MX | 50 from 1959 | 42.1 C (107.8 F) | 1.0 years | 1.0 to 1.3 | 48.8 C (119.8 F) | not tested |
| Nieves | MX | 53 from 1950 | 37.1 C (98.8 F) | 1.1 years | 1.0 to 1.3 | 45.5 C (113.9 F) | not tested |
| Tottori | JP | 66 from 1951 | 37.3 C (99.1 F) | 1.0 years | 1.0 to 1.3 | 40.9 C (105.6 F) | not tested |
Of the 1,975 stations with a clear change of the hottest day of the year, 1,097 are left out of the list because a comparison of their yearly mean temperature with neighbouring stations points to a station change.
Coldest night: the old value comes less often
| Station | Country | Years used | Once in 50 years in 1951-1980 | In 2025 once in | 90% interval, years | Once in 50 years in 2025 | Station change check |
|---|---|---|---|---|---|---|---|
| Lhasa | CN | 63 from 1955 | -17.3 C (0.9 F) | over 1,000 years | over 1,000 | -11.3 C (11.7 F) | not tested |
| Ust-Moma | RU | 74 from 1950 | -63.1 C (-81.6 F) | over 1,000 years | over 1,000 | -54.4 C (-65.9 F) | not tested |
| Jakutsk | RU | 69 from 1950 | -60.0 C (-76.0 F) | over 1,000 years | over 1,000 | -47.5 C (-53.5 F) | not tested |
| Tengchong | CN | 68 from 1951 | -4.5 C (23.9 F) | over 1,000 years | over 1,000 | -0.4 C (31.3 F) | not tested |
| Elsinore | US | 59 from 1950 | -8.5 C (16.7 F) | over 1,000 years | over 1,000 | -3.0 C (26.6 F) | none found |
| Hay River A | CA | 63 from 1950 | -48.3 C (-54.9 F) | over 1,000 years | over 1,000 | -39.1 C (-38.4 F) | not tested |
| Birni N'Konni | NE | 57 from 1950 | 8.4 C (47.1 F) | over 1,000 years | over 1,000 | 12.0 C (53.6 F) | not tested |
| Aranyaprathet | TH | 62 from 1952 | 8.6 C (47.5 F) | over 1,000 years | over 1,000 | 14.0 C (57.2 F) | none found |
| Litchfield Park | US | 62 from 1950 | -8.7 C (16.3 F) | over 1,000 years | over 1,000 | -3.0 C (26.6 F) | none found |
| Bomnak | RU | 68 from 1950 | -51.9 C (-61.4 F) | over 1,000 years | over 1,000 | -44.2 C (-47.6 F) | not tested |
Of the 3,085 stations with a clear change of the coldest night of the year, 1,695 are left out of the list because a comparison of their yearly mean temperature with neighbouring stations points to a station change.
In the fits of 1950 to 2025 the change of the warmest night of the year is clear at 2,856 of 5,129 stations (56 percent) with the false discovery rate held at 5 percent: at 2,705 the old value comes more often and at 151 less often.
In the fits of 1950 to 2025 the change of the coldest night of the year is clear at 3,085 of 5,217 stations (59 percent) with the false discovery rate held at 5 percent: at 50 the old value comes more often and at 3,035 less often.
In the fits of 1950 to 2025 the change of the coldest day of the year is clear at 1,920 of 5,535 stations (35 percent) with the false discovery rate held at 5 percent: at 84 the old value comes more often and at 1,836 less often.
How the Return Periods Are Computed
For each station and calendar year the page uses five values: the hottest day of the year (the highest daily maximum of the year), the warmest night of the year (the highest daily minimum of the year), the coldest night of the year (the lowest daily minimum of the year), the coldest day of the year (the lowest daily maximum of the year), the wettest day of the year (the largest daily precipitation of the year). A year counts when it has at least 300 days of readings of that kind, from 1950 to 2025.
The yearly values of a station are set against a yardstick of global warming: the mean global surface temperature of each year and the 3 years before it, from the NOAAGlobalTemp series. A straight line is fitted to the yearly values against the yardstick. What is left around the line is described by a generalised extreme value distribution, the usual distribution for the largest value of a year, fitted by the method of L-moments. The page then assumes that this distribution moves along the line as global temperature changes, while its width and shape stay the same. The cold extremes are fitted upside down and the wettest day on its logarithm, so its shift is a percentage.
From the fit come the value reached once in 50 years at the yardstick of 1951 to 1980, and how often that same value is reached at the yardstick of 2025. A return period of 10 years means a chance of 1 in 10 in each year, not an event every tenth year. The fit puts no number on a return period beyond 10,000 years, and this page words everything beyond 1,000 years as "over 1,000 years", because 50 to 76 years of data cannot tell such long waits apart.
A station's fit is used with at least 50 usable years, at least 15 of them from 1950 to 1985 and at least 15 in the last 30 years, a shape of the distribution inside the usual range (-0.5 to 0.5), and no yearly value outside what the fitted distribution allows. Hottest day: 5,418 of the 13,235 stations with at least 30 usable years pass every rule. Warmest night: 5,129 of the 12,814 stations with at least 30 usable years pass every rule. Coldest night: 5,217 of the 12,814 stations with at least 30 usable years pass every rule. Coldest day: 5,535 of the 13,235 stations with at least 30 usable years pass every rule. Wettest day: 9,645 of the 26,341 stations with at least 30 usable years pass every rule. A station's own interval comes from fitting again to 400 random redraws of its years and taking the middle 90 percent of the answers.
The typical value of a region is a median over 5 degree grid cells: first the median of the stations in a cell, then the median of the cells, each weighted by its area, so that a country with many stations does not decide the result. Return periods are averaged as yearly chances. The 95 percent interval of a region comes from redrawing its grid cells 1,000 times; a region with fewer than 10 cells gets none. It shows the spread between cells and does not include the uncertainty of each station's own fit, and neighbouring cells share their weather, so the real uncertainty is wider than the interval.
With thousands of stations some will show a change by chance alone. A station's change is called clear when a two sided test of its slope against zero, made from its own interval, passes with the false discovery rate held at 5 percent over all stations of that kind (the method of Benjamini and Hochberg): of the stations called clear, about 5 in 100 are expected to be false alarms.
What the Numbers Cover and Where They Stop
The station network is uneven. Of the 5,418 stations with a publishable fit for the hottest day of the year, 3,105 (57 percent) are in the United States, 450 (8 percent) in Russia and 248 (5 percent) in Germany. The median over grid cells limits the weight of dense networks, but large parts of Africa, South America and Asia have few stations with 50 years of data, and the oceans have none: "the world" on this page is the land with long records. The tables give the number of stations behind every value.
The extremes are taken per calendar year. That cuts the southern summer and the northern winter in two, so one hot or cold season can supply the extreme of two years. This adds scatter to the fits of southern hot extremes and northern cold extremes; no bias in one direction is known.
Most stations in the United States report whole degrees Fahrenheit, steps of 0.56 C. A shift smaller than one step cannot be read at a single station, and values at such stations often tie.
The yardstick is the global mean temperature, not the local one. Each station gets its own slope against it, so regions that warm faster or slower are allowed for, but the fit assumes that a station's extremes follow global temperature in a straight line and that the width and shape of their distribution did not change. Anything else that changed a station's readings over the same decades ends up in its slope: a move, a new instrument, a new observing time, a growing city, irrigation. The records are used as measured.
A value that comes once in 50 years is estimated here from 50 to 76 years of data, so the answer for one station is uncertain. At the median station the 90 percent interval of its own return period for the hottest day of the year runs from 4.7 to 84 years. Single stations should be read with their interval; the medians over many grid cells are steadier.
This page describes how yearly extremes at stations moved together with global temperature from 1950 to 2025. It does not show what caused the change at any station, and it is not a forecast.
Download the Data
Source: NOAA National Centers for Environmental Information, Global Historical Climatology Network Daily (GHCN-Daily). Records are computed by World Extreme Weather Monitor and are not official NOAA products. (US Government open data (NOAA NCEI)) Source: NOAA National Centers for Environmental Information, NOAA Global Surface Temperature Dataset (NOAAGlobalTemp), Version 6.1, annual land and ocean time series (Huang, B., X. Yin, M.J. Menne and R.S. Vose, 2026, doi:10.25921/vvaa-wq11). Used as the measure of global warming; the fits to station data are computed by World Extreme Weather Monitor and are not NOAA products. (US Government open data (NOAA NCEI Open Data Policy, citation requested)) Data as of 2026-11-17. Extreme value fits, return periods, shifts, medians over grid cells and intervals computed by this site from GHCN-Daily yearly extremes and the NOAAGlobalTemp annual series; not an official NOAA product.