Records at Risk
The chance that each long-record station's all-time record falls in 2026, for 10,324 stations, tested on 2006 to 2025 against chance alone. Hottest day, warmest night, coldest night, coldest day and wettest day, with watch lists only where the method beat chance.
Data as of . Last revised (2026-11-17).
Key Facts
- Tested on 2006 to 2025, each year from earlier data only, the method that estimates the chance of a new all-time record was better than chance for the warmest night, the coldest night and the coldest day and no better than chance for the hottest day and the wettest day.
- For 2026 the method expects about 173 of the 5,129 warmest night records to fall; chance alone in a climate that did not change would give 61.3. From 2006 to 2025 its error was 2.2 percent smaller than that of chance alone (98 percent interval +0.8 to +3.4%).
- For 2026 the method expects about 19.9 of the 5,217 coldest night records to fall; chance alone in a climate that did not change would give 62.4. From 2006 to 2025 its error was 2.3 percent smaller than that of chance alone (98 percent interval +1.2 to +4.4%).
- For 2026 the method expects about 36.3 of the 5,535 coldest day records to fall; chance alone in a climate that did not change would give 66.6. From 2006 to 2025 its error was 0.9 percent smaller than that of chance alone (98 percent interval +0.2 to +1.9%).
- When the method gave a chance of 20% or more from 2006 to 2025, the record fell in 15.0% of 454 station-years (mean chance given 27.2%).
- By 17 November 2026, 623 all-time records of these kinds had been beaten in 2026 at the stations followed, 0 of them at a station on a watch list.
The Test Comes First
2006 to 2025, 432,811 station-yearsA chance that nobody has checked is worth little, so this page starts with the check. For every year from 2006 to 2025 the method was run on the data before that year only, and the chance it gave for each station was set against what happened: the record was beaten in that year, or it was not. The same was done for chance alone, the baseline: in a climate that does not change, every year on record and the coming one are equally likely to hold the most extreme value, so the chance of a new record is 1 in the number of years plus one.
The score is the Brier score, the mean squared difference between the chance given and the outcome (1 when the record fell, 0 when it stood). The skill is how much smaller the method's score is than the baseline's, in percent. Zero means no better than chance.
| Record | Stations tested | Station-years tested | Records that fell | Expected by the method | Expected by chance alone | Skill against chance (98% interval) | Verdict |
|---|---|---|---|---|---|---|---|
| Hottest day | 6,020 | 76,515 | 1,705 | 1,354 | 979 | +2.3% (-1.0 to +5.4%) | No better than chance |
| Warmest night | 5,739 | 71,479 | 1,634 | 1,419 | 907 | +2.2% (+0.8 to +3.4%) | Better than chance |
| Coldest night | 5,798 | 72,601 | 254 | 237 | 921 | +2.3% (+1.2 to +4.4%) | Better than chance |
| Coldest day | 6,072 | 77,667 | 320 | 447 | 996 | +0.9% (+0.2 to +1.9%) | Better than chance |
| Wettest day | 10,939 | 134,549 | 2,375 | 1,794 | 1,651 | +0.1% (-0.2 to +0.3%) | No better than chance |
| All kinds shown | 11,523 | 432,811 | 6,288 | 5,251 | 5,454 | +1.4% (+0.3 to +2.5%) | Judged per kind |
Hottest day: no better than chance. The skill was +2.3% and its interval includes zero (Brier score 0.0213 against 0.0218; 98 percent interval -1.0 to +5.4%). It expected 1,354 records to fall from 2006 to 2025, chance alone 979, and 1,705 fell. This page gives no watch list for it.
Warmest night: better than chance. The method's error was 2.2 percent smaller than the error of chance alone (Brier score 0.0219 against 0.0224; 98 percent interval +0.8 to +3.4%). It expected 1,419 records to fall from 2006 to 2025, chance alone 907, and 1,634 fell.
Coldest night: better than chance. The method's error was 2.3 percent smaller than the error of chance alone (Brier score 0.0035 against 0.0036; 98 percent interval +1.2 to +4.4%). It expected 237 records to fall from 2006 to 2025, chance alone 921, and 254 fell.
Coldest day: better than chance. The method's error was 0.9 percent smaller than the error of chance alone (Brier score 0.0041 against 0.0042; 98 percent interval +0.2 to +1.9%). It expected 447 records to fall from 2006 to 2025, chance alone 996, and 320 fell.
Wettest day: no better than chance. The skill was +0.1% and its interval includes zero (Brier score 0.0173 against 0.0174; 98 percent interval -0.2 to +0.3%). It expected 1,794 records to fall from 2006 to 2025, chance alone 1,651, and 2,375 fell. This page gives no watch list for it.
The rule is fixed in the code. A kind of record counts as better than chance only when the whole 98 percent interval of its skill is above zero. The interval comes from drawing the 20 test years again and again with replacement (4,000 times, with a fixed seed), each year with all its stations, because the stations of one year share that year's weather. 5 kinds are tested at once, so each must pass at 1 percent on its own; that keeps the chance that any kind passes by luck at or under 5 percent. A kind that does not pass is called no better than chance and gets no watch list.
Did a Stated Chance Come True?
all kinds, 2006 to 2025A chance of 5 percent is right when the record falls in about 5 of every 100 such cases. The table groups all 432,811 station-years of the test by the chance the method gave and shows how often the record fell in each group. A method that is well calibrated has a share that fell close to its mean chance given. When the method gave a chance of 20% or more from 2006 to 2025, the record fell in 15.0% of 454 station-years (mean chance given 27.2%).
| Chance given | Station-years | Mean chance given | Records that fell | Share that fell | 95% interval of the share |
|---|---|---|---|---|---|
| under 1% | 284,087 | 0.26% | 1,911 | 0.67% | 0.64% to 0.70% |
| 1 to 2% | 69,061 | 1.4% | 1,176 | 1.7% | 1.6% to 1.8% |
| 2 to 5% | 59,239 | 3.1% | 1,764 | 3.0% | 2.8% to 3.1% |
| 5 to 10% | 16,248 | 6.7% | 979 | 6.0% | 5.7% to 6.4% |
| 10 to 20% | 3,722 | 12.9% | 390 | 10.5% | 9.5% to 11.5% |
| 20% or more | 454 | 27.2% | 68 | 15.0% | 12.0% to 18.6% |
The same for chance alone. Its chances depend only on the length of the station's record, so they fall in few groups and cannot tell a station where a record is likely from one where it is not.
| Chance by chance alone | Station-years | Mean chance given | Records that fell | Share that fell | 95% interval of the share |
|---|---|---|---|---|---|
| under 1% | 131,230 | 0.86% | 1,105 | 0.84% | 0.79% to 0.89% |
| 1 to 2% | 301,581 | 1.4% | 5,183 | 1.7% | 1.7% to 1.8% |
| Record | Chance given | Station-years | Mean chance given | Records that fell | Share that fell | 95% interval of the share |
|---|---|---|---|---|---|---|
| Hottest day | under 1% | 43,873 | 0.25% | 383 | 0.87% | 0.79% to 0.96% |
| Hottest day | 1 to 2% | 11,631 | 1.4% | 258 | 2.2% | 2.0% to 2.5% |
| Hottest day | 2 to 5% | 13,475 | 3.2% | 475 | 3.5% | 3.2% to 3.9% |
| Hottest day | 5 to 10% | 5,713 | 6.9% | 378 | 6.6% | 6.0% to 7.3% |
| Hottest day | 10 to 20% | 1,684 | 12.8% | 186 | 11.0% | 9.6% to 12.6% |
| Hottest day | 20% or more | 139 | 25.7% | 25 | 18.0% | 12.5% to 25.2% |
| Warmest night | under 1% | 38,138 | 0.23% | 274 | 0.72% | 0.64% to 0.81% |
| Warmest night | 1 to 2% | 10,662 | 1.5% | 234 | 2.2% | 1.9% to 2.5% |
| Warmest night | 2 to 5% | 14,682 | 3.2% | 509 | 3.5% | 3.2% to 3.8% |
| Warmest night | 5 to 10% | 6,068 | 6.9% | 401 | 6.6% | 6.0% to 7.3% |
| Warmest night | 10 to 20% | 1,661 | 13.0% | 174 | 10.5% | 9.1% to 12.0% |
| Warmest night | 20% or more | 268 | 27.7% | 42 | 15.7% | 11.8% to 20.5% |
| Coldest night | under 1% | 66,337 | 0.18% | 192 | 0.29% | 0.25% to 0.33% |
| Coldest night | 1 to 2% | 4,791 | 1.4% | 38 | 0.79% | 0.58% to 1.1% |
| Coldest night | 2 to 5% | 1,292 | 2.8% | 17 | 1.3% | 0.82% to 2.1% |
| Coldest night | 5 to 10% | 139 | 6.9% | 3 | 2.2% | 0.74% to 6.2% |
| Coldest night | 10 to 20% | 37 | 14.0% | 3 | 8.1% | 2.8% to 21.3% |
| Coldest night | 20% or more | 5 | 28.2% | 1 | 20.0% | 3.6% to 62.4% |
| Coldest day | under 1% | 62,527 | 0.27% | 186 | 0.30% | 0.26% to 0.34% |
| Coldest day | 1 to 2% | 10,949 | 1.4% | 70 | 0.64% | 0.51% to 0.81% |
| Coldest day | 2 to 5% | 3,978 | 2.8% | 56 | 1.4% | 1.1% to 1.8% |
| Coldest day | 5 to 10% | 181 | 6.1% | 8 | 4.4% | 2.3% to 8.5% |
| Coldest day | 10 to 20% | 17 | 13.8% | 0 | 0.00% | 0.00% to 18.4% |
| Coldest day | 20% or more | 15 | 34.9% | 0 | 0.00% | 0.00% to 20.4% |
| Wettest day | under 1% | 73,212 | 0.35% | 876 | 1.2% | 1.1% to 1.3% |
| Wettest day | 1 to 2% | 31,028 | 1.4% | 576 | 1.9% | 1.7% to 2.0% |
| Wettest day | 2 to 5% | 25,812 | 3.0% | 707 | 2.7% | 2.5% to 2.9% |
| Wettest day | 5 to 10% | 4,147 | 6.4% | 189 | 4.6% | 4.0% to 5.2% |
| Wettest day | 10 to 20% | 323 | 12.3% | 27 | 8.4% | 5.8% to 11.9% |
| Wettest day | 20% or more | 27 | 24.8% | 0 | 0.00% | 0.00% to 12.5% |
Groups without a case are left out. The 95 percent interval of each share is a Wilson interval that treats every station-year as independent. Stations share weather within a year, so the true uncertainty is wider than the interval shows.
Year by Year
2006 to 2025For each test year, the number of records that fell, the number the method expected (the sum of its chances) and the number chance alone expected. A single year can land far from the expected number, because one hot summer or one cold winter moves many stations at once. The method knows the long rise of global temperature. It does not know the weather of the year.
| Year | Records that fell | Expected by the method | Expected by chance alone |
|---|---|---|---|
| 2006 | 42 | 52.9 | 50.4 |
| 2007 | 78 | 58.0 | 51.7 |
| 2008 | 22 | 60.3 | 52.3 |
| 2009 | 49 | 58.1 | 53.1 |
| 2010 | 181 | 58.4 | 55.6 |
| 2011 | 81 | 62.5 | 56.5 |
| 2012 | 129 | 58.6 | 54.9 |
| 2013 | 58 | 56.5 | 49.4 |
| 2014 | 32 | 58.3 | 51.4 |
| 2015 | 107 | 55.9 | 50.1 |
| 2016 | 40 | 67.0 | 50.7 |
| 2017 | 48 | 79.7 | 50.2 |
| 2018 | 123 | 87.2 | 49.3 |
| 2019 | 190 | 85.8 | 45.9 |
| 2020 | 55 | 82.3 | 45.7 |
| 2021 | 164 | 81.1 | 45.3 |
| 2022 | 162 | 72.3 | 44.6 |
| 2023 | 65 | 66.9 | 42.2 |
| 2024 | 48 | 72.0 | 41.5 |
| 2025 | 31 | 80.2 | 38.9 |
| Year | Records that fell | Expected by the method | Expected by chance alone |
|---|---|---|---|
| 2006 | 110 | 65.4 | 46.7 |
| 2007 | 81 | 68.9 | 47.9 |
| 2008 | 37 | 67.2 | 48.5 |
| 2009 | 37 | 63.8 | 49.7 |
| 2010 | 210 | 61.4 | 51.6 |
| 2011 | 113 | 64.5 | 52.9 |
| 2012 | 64 | 61.4 | 51.3 |
| 2013 | 36 | 57.9 | 45.8 |
| 2014 | 45 | 56.8 | 46.0 |
| 2015 | 87 | 56.0 | 46.5 |
| 2016 | 50 | 67.2 | 46.7 |
| 2017 | 65 | 82.4 | 46.1 |
| 2018 | 151 | 91.2 | 45.7 |
| 2019 | 86 | 80.8 | 41.6 |
| 2020 | 82 | 92.2 | 42.4 |
| 2021 | 77 | 84.0 | 42.0 |
| 2022 | 83 | 74.5 | 40.7 |
| 2023 | 70 | 70.5 | 39.8 |
| 2024 | 59 | 74.3 | 38.9 |
| 2025 | 91 | 78.9 | 36.3 |
| Year | Records that fell | Expected by the method | Expected by chance alone |
|---|---|---|---|
| 2006 | 18 | 14.7 | 47.4 |
| 2007 | 7 | 14.4 | 48.7 |
| 2008 | 9 | 14.1 | 49.5 |
| 2009 | 18 | 13.9 | 50.7 |
| 2010 | 10 | 15.3 | 52.7 |
| 2011 | 33 | 16.2 | 54.0 |
| 2012 | 8 | 15.0 | 52.0 |
| 2013 | 4 | 12.4 | 46.6 |
| 2014 | 4 | 12.3 | 46.7 |
| 2015 | 5 | 12.5 | 47.2 |
| 2016 | 7 | 11.8 | 47.4 |
| 2017 | 5 | 10.4 | 46.8 |
| 2018 | 9 | 10.2 | 46.4 |
| 2019 | 20 | 9.9 | 42.1 |
| 2020 | 2 | 9.9 | 42.7 |
| 2021 | 51 | 8.9 | 42.6 |
| 2022 | 9 | 9.4 | 41.4 |
| 2023 | 11 | 9.5 | 40.4 |
| 2024 | 12 | 8.5 | 39.5 |
| 2025 | 12 | 7.5 | 36.7 |
| Year | Records that fell | Expected by the method | Expected by chance alone |
|---|---|---|---|
| 2006 | 14 | 27.5 | 51.2 |
| 2007 | 10 | 26.1 | 52.5 |
| 2008 | 9 | 25.4 | 53.5 |
| 2009 | 4 | 25.4 | 54.1 |
| 2010 | 34 | 26.2 | 56.5 |
| 2011 | 47 | 25.8 | 57.6 |
| 2012 | 5 | 25.3 | 55.9 |
| 2013 | 7 | 21.8 | 50.4 |
| 2014 | 15 | 21.8 | 52.3 |
| 2015 | 7 | 23.1 | 50.8 |
| 2016 | 14 | 22.4 | 51.6 |
| 2017 | 9 | 21.1 | 51 |
| 2018 | 1 | 20.8 | 50.2 |
| 2019 | 20 | 21.0 | 46.6 |
| 2020 | 5 | 20.6 | 46.3 |
| 2021 | 62 | 17.4 | 45.6 |
| 2022 | 14 | 18.4 | 45 |
| 2023 | 5 | 19.9 | 42.7 |
| 2024 | 35 | 17.9 | 41.9 |
| 2025 | 3 | 19.2 | 39.8 |
| Year | Records that fell | Expected by the method | Expected by chance alone |
|---|---|---|---|
| 2006 | 104 | 100 | 91.8 |
| 2007 | 144 | 99.3 | 92.9 |
| 2008 | 118 | 97.4 | 93.1 |
| 2009 | 89 | 97.0 | 93.3 |
| 2010 | 190 | 92.4 | 91.9 |
| 2011 | 192 | 91.5 | 87.3 |
| 2012 | 98 | 89.3 | 86.2 |
| 2013 | 129 | 83.4 | 82.2 |
| 2014 | 115 | 82.0 | 81.5 |
| 2015 | 94 | 82.5 | 81.6 |
| 2016 | 133 | 88.3 | 82.1 |
| 2017 | 130 | 94.6 | 80.6 |
| 2018 | 111 | 101 | 82.0 |
| 2019 | 103 | 94.6 | 78.9 |
| 2020 | 93 | 93.5 | 80.1 |
| 2021 | 132 | 90.2 | 79.3 |
| 2022 | 129 | 84.8 | 76.7 |
| 2023 | 85 | 81.0 | 74.2 |
| 2024 | 113 | 79.6 | 71.8 |
| 2025 | 73 | 72.0 | 63.4 |
2026: Records Expected to Fall
fits through 2025For 2026 the method expects about 173 of the 5,129 warmest night records to fall; chance alone in a climate that did not change would give 61.3. By 17 November 2026, 219 of the 3,322 stations with 2026 data had beaten their warmest night record, 0 of them on the watch list.
For 2026 the method expects about 19.9 of the 5,217 coldest night records to fall; chance alone in a climate that did not change would give 62.4. By 17 November 2026, 2 of the 3,398 stations with 2026 data had beaten their coldest night record, 0 of them on the watch list.
For 2026 the method expects about 36.3 of the 5,535 coldest day records to fall; chance alone in a climate that did not change would give 66.6. By 17 November 2026, 8 of the 3,623 stations with 2026 data had beaten their coldest day record, 0 of them on the watch list.
For the hottest day and the wettest day the method did not beat chance on past years, so the table gives only what chance alone expects and what has fallen.
| Record | Stations | Expected to fall in 2026 by the method | Expected by chance alone | Stations with a chance below the baseline | Stations with 2026 data | Records fallen (data to 17 November 2026) | Of those on the watch list |
|---|---|---|---|---|---|---|---|
| Hottest day | 5,418 | not given: no better than chance | 65.2 | n/a | 3,543 | 230 | n/a |
| Warmest night | 5,129 | 173 | 61.3 | 2,496 | 3,322 | 219 | 0 |
| Coldest night | 5,217 | 19.9 | 62.4 | 4,972 | 3,398 | 2 | 0 |
| Coldest day | 5,535 | 36.3 | 66.6 | 4,652 | 3,623 | 8 | 0 |
| Wettest day | 9,645 | not given: no better than chance | 112 | n/a | 6,309 | 164 | n/a |
By 17 November 2026, 623 records of these kinds had been beaten in 2026 at the stations followed. A record counts as fallen when the 2026 value so far is beyond it; a tie does not count, as in the test. The year is not over, and data from many countries arrive weeks to months late, so these counts are a floor and will rise. The chances were fixed before 2026 began and are not updated as the year goes on.
At 4,972 of the 5,217 stations the chance for the coldest night is below the baseline. That means the method finds a new record there less likely in 2026 than it would be in a climate that did not change: in the station's own record the yearly coldest night has moved away from the old record as the global temperature yardstick rose. A watch list for such a record names the stations where a new record is least unlikely, and most of its chances are still small.
At 4,652 of the 5,535 stations the chance for the coldest day is below the baseline. That means the method finds a new record there less likely in 2026 than it would be in a climate that did not change: in the station's own record the yearly coldest day has moved away from the old record as the global temperature yardstick rose. A watch list for such a record names the stations where a new record is least unlikely, and most of its chances are still small.
| Country | Stations | Warmest night | Coldest night | Coldest day |
|---|---|---|---|---|
| All countries | 10,324 | 173 against 61.3 (5,129) | 19.9 against 62.4 (5,217) | 36.3 against 66.6 (5,535) |
| United States | 4,430 | 57.4 against 33.6 (3,017) | 7.4 against 34.7 (3,118) | 25.1 against 35.3 (3,162) |
| Australia | 2,579 | 7.1 against 2.8 (191) | 1.8 against 2.7 (186) | 0.8 against 3.0 (201) |
| Russia | 500 | 25.8 against 5.3 (452) | 0.6 against 5.4 (457) | 0.6 against 5.5 (460) |
| Germany | 356 | 12.5 against 3.1 (246) | 0.2 against 3.1 (249) | 0.2 against 3.2 (252) |
| Sweden | 343 | 5.4 against 1.2 (99) | 0.1 against 1.2 (92) | 0.1 against 1.2 (94) |
| Mexico | 328 | 8.3 against 3.2 (204) | 7.9 against 3.5 (218) | 6.9 against 3.6 (229) |
| Canada | 291 | 12.3 against 3.4 (253) | 0.2 against 3.1 (226) | 0.3 against 3.2 (236) |
| Netherlands | 277 | 0.9 against 0.2 (14) | 0.0 against 0.2 (14) | 0.0 against 0.2 (14) |
Each cell gives the records the method expects in 2026, then what chance alone expects, then the number of stations in brackets. The countries are the 8 with the most stations followed, each with at least 30. Only kinds that beat chance in the test are shown.
Watch Lists for 2026
25 stations per kindA watch list names the 25 stations where the method gives the highest chance of a new record in 2026. It exists only for the warmest night, the coldest night and the coldest day, the kinds that beat chance in the test. There is none for the hottest day and the wettest day. The range is the 5th to 95th percentile of a bootstrap of the station's fit. A chance below 1 percent is shown as under 1 percent, because the test cannot tell smaller chances apart. Being on a list is not a forecast: most listed records will still stand at the end of the year.
Watch List: Warmest Night
5,129 stations, median 67 years in the fit| Station | Country | Record | Set in | Chance in 2026 | Range | By chance alone | 2026 so far |
|---|---|---|---|---|---|---|---|
| Yautepec | Mexico | 21.1 C (70.0 F) | 2013 | 86% | 74 to 96% | 1 in 60 | No data yet |
| Kunming | China | 20.6 C (69.1 F) | 2004 | 64% | 29 to 91% | 1 in 76 | No data yet |
| Torrevieja | Spain | 29.0 C (84.2 F) | 2003 | 61% | 29 to 95% | 1 in 100 | No data yet |
| Mexicali (Dge) | Mexico | 33.3 C (91.9 F) | 2002 | 58% | 31 to 88% | 1 in 68 | No data yet |
| Nevada City | United States | 22.8 C (73.0 F) | 2022 | 55% | 39 to 71% | 1 in 134 | Standing |
| Krasnodar | Russia | 26.7 C (80.1 F) | 2012 | 51% | 22 to 82% | 1 in 80 | No data yet |
| Rinconada | Mexico | 25.6 C (78.1 F) | 1994 | 47% | 22 to 77% | 1 in 62 | No data yet |
| Madrid - Retiro | Spain | 26.2 C (79.2 F) | 2022 | 45% | 31 to 60% | 1 in 107 | Standing |
| San Luis Potosi (Dge) | Mexico | 21.7 C (71.1 F) | 2004 | 42% | 12 to 81% | 1 in 64 | No data yet |
| Irkutsk | Russia | 20.4 C (68.7 F) | 2016 | 41% | 20 to 65% | 1 in 140 | No data yet |
| Willow Grove NAS | United States | 27.2 C (81.0 F) | 2010 | 40% | 11 to 77% | 1 in 68 | No data yet |
| Rena - Haugedalen | Norway | 18.1 C (64.6 F) | 2010 | 40% | 11 to 68% | 1 in 57 | No data yet |
| Graz | Austria | 22.3 C (72.1 F) | 2021 | 40% | 22 to 62% | 1 in 133 | No data yet |
| Charlottetown A | Canada | 21.3 C (70.3 F) | 2009 | 40% | 11 to 71% | 1 in 84 | No data yet |
| Jingdezhen | China | 31.2 C (88.2 F) | 2012 | 40% | 23 to 65% | 1 in 75 | Standing |
| Ciudad Lerdo (Dge) | Mexico | 27.8 C (82.0 F) | 2011 | 39% | 15 to 80% | 1 in 62 | No data yet |
| Kerman | Iran | 27.0 C (80.6 F) | 2008 | 39% | 13 to 69% | 1 in 76 | No data yet |
| Culiacan (Dge) | Mexico | 29.4 C (84.9 F) | 2009 | 38% | 14 to 66% | 1 in 54 | No data yet |
| Penticton A | Canada | 22.5 C (72.5 F) | 1994 | 37% | 13 to 76% | 1 in 73 | No data yet |
| Primorsko-Ahtarsk | Russia | 27.5 C (81.5 F) | 2001 | 37% | 16 to 62% | 1 in 76 | No data yet |
| Germansen Landing | Canada | 15.5 C (59.9 F) | 1997 | 37% | 10 to 68% | 1 in 64 | No data yet |
| La Florencia | Mexico | 28.9 C (84.0 F) | 1995 | 37% | 11 to 74% | 1 in 59 | No data yet |
| Phoenix AP | United States | 36.1 C (97.0 F) | 2023 | 36% | 22 to 54% | 1 in 93 | Standing |
| Reno AP | United States | 25.0 C (77.0 F) | 2023 | 36% | 21 to 58% | 1 in 90 | Standing |
| Simao | China | 22.5 C (72.5 F) | 2012 | 35% | 11 to 70% | 1 in 76 | No data yet |
A list like this one, drawn up for each year from 2006 to 2025 from earlier data only, would have named 500 station-years. The record fell in 72 of them (14.4%), and the chances given added up to 109. Among all stations it fell in 2.3% of station-years.
Watch List: Coldest Night
5,217 stations, median 67 years in the fit| Station | Country | Record | Set in | Chance in 2026 | Range | By chance alone | 2026 so far |
|---|---|---|---|---|---|---|---|
| Nieves | Mexico | -10.0 C (14.0 F) | 2011 | >99% | 98 to 100% | 1 in 68 | No data yet |
| Santa Rosa | Mexico | 2.8 C (37.0 F) | 1986 | 80% | 46 to 100% | 1 in 55 | No data yet |
| Ocozocoautla | Mexico | 1.7 C (35.1 F) | 2002 | 61% | 23 to 88% | 1 in 76 | No data yet |
| San Pedro | Mexico | 0.0 C (32.0 F) | 1996 | 49% | 23 to 75% | 1 in 64 | No data yet |
| Santo Domingo | Mexico | 5.0 C (41.0 F) | 2004 | 44% | 14 to 72% | 1 in 70 | No data yet |
| Calvillo (Dge) | Mexico | -7.8 C (18.0 F) | 2006 | 43% | 13 to 74% | 1 in 55 | No data yet |
| Rinconada | Mexico | -8.3 C (17.1 F) | 1998 | 40% | 16 to 79% | 1 in 62 | No data yet |
| Parkes (Macarthur Street) | Australia | -4.2 C (24.4 F) | 1997 | 31% | 6 to 58% | 1 in 57 | No data yet |
| Altamirano (Smn) | Mexico | -2.2 C (28.0 F) | 1998 | 30% | 10 to 54% | 1 in 73 | No data yet |
| Armeria | Mexico | 7.2 C (45.0 F) | 2011 | 30% | 2 to 64% | 1 in 66 | No data yet |
| Huitzilac | Mexico | -6.1 C (21.0 F) | 1997 | 29% | 12 to 66% | 1 in 53 | No data yet |
| Coyame (Cfe) | Mexico | 7.8 C (46.0 F) | 2006 | 25% | 5 to 50% | 1 in 61 | No data yet |
| East London | South Africa | -1.6 C (29.1 F) | 2016 | 21% | 10 to 44% | 1 in 86 | No data yet |
| Prospect Reservoir | Australia | -0.8 C (30.6 F) | 2010 | 19% | up to 47% | 1 in 55 | No data yet |
| Sahuaripa | Mexico | -8.3 C (17.1 F) | 2011 | 19% | 2 to 52% | 1 in 73 | No data yet |
| Temascales | Mexico | -10.0 C (14.0 F) | 2006 | 18% | up to 50% | 1 in 70 | No data yet |
| La Angostura | Mexico | -1.1 C (30.0 F) | 1965 | 17% | 6 to 39% | 1 in 53 | No data yet |
| Dzaore | Russia | -40.9 C (-41.6 F) | 2005 | 15% | 4 to 34% | 1 in 85 | No data yet |
| Abelardo L. Rodriguez | Mexico | -8.3 C (17.1 F) | 2006 | 14% | 3 to 49% | 1 in 58 | No data yet |
| Meacham | United States | -35.0 C (-31.0 F) | 2004 | 12% | 4 to 23% | 1 in 58 | Standing |
| El Burrero | Mexico | 0.0 C (32.0 F) | 2013 | 12% | 3 to 35% | 1 in 64 | No data yet |
| Sain Alto | Mexico | -13.3 C (8.1 F) | 2006 | 10% | up to 25% | 1 in 58 | No data yet |
| Kimberley | South Africa | -9.9 C (14.2 F) | 2021 | 10% | 2 to 34% | 1 in 79 | No data yet |
| Longerenong | Australia | -5.0 C (23.0 F) | 1997 | 9% | 2 to 18% | 1 in 62 | Standing |
| Horsham Polkemmet Rd | Australia | -5.6 C (21.9 F) | 1976 | 8% | 2 to 23% | 1 in 57 | No data yet |
A list like this one, drawn up for each year from 2006 to 2025 from earlier data only, would have named 500 station-years. The record fell in 12 of them (2.4%), and the chances given added up to 28.5. Among all stations it fell in 0.35% of station-years.
Watch List: Coldest Day
5,535 stations, median 67 years in the fit| Station | Country | Record | Set in | Chance in 2026 | Range | By chance alone | 2026 so far |
|---|---|---|---|---|---|---|---|
| Presa Requena | Mexico | 10.0 C (50.0 F) | 2010 | 73% | 50 to 93% | 1 in 61 | No data yet |
| Rinconada | Mexico | 0.0 C (32.0 F) | 1999 | 65% | 38 to 88% | 1 in 62 | No data yet |
| La Tinaja | Mexico | 7.8 C (46.0 F) | 2012 | 54% | 18 to 90% | 1 in 53 | No data yet |
| Calderones | Mexico | 5.0 C (41.0 F) | 1997 | 51% | 20 to 81% | 1 in 55 | No data yet |
| East London | South Africa | 5.2 C (41.4 F) | 2001 | 49% | 34 to 65% | 1 in 86 | No data yet |
| La Florencia | Mexico | 11.7 C (53.1 F) | 1989 | 30% | 9 to 58% | 1 in 59 | No data yet |
| San Carlos | Mexico | 2.8 C (37.0 F) | 1997 | 27% | 2 to 58% | 1 in 51 | No data yet |
| Frontera Hidalgo | Mexico | 23.9 C (75.0 F) | 2005 | 25% | 4 to 51% | 1 in 66 | No data yet |
| X-Pichil | Mexico | 18.9 C (66.0 F) | 2002 | 21% | 2 to 58% | 1 in 52 | No data yet |
| Huerta Vieja | Mexico | 9.4 C (48.9 F) | 2004 | 21% | 5 to 53% | 1 in 63 | No data yet |
| Presa Potrerillos | Mexico | 4.4 C (39.9 F) | 2010 | 20% | 4 to 49% | 1 in 68 | No data yet |
| Tutonchany | Russia | -57.0 C (-70.6 F) | 1994 | 18% | 1 to 42% | 1 in 62 | No data yet |
| Los Castillos | Mexico | 10.6 C (51.1 F) | 2002 | 14% | 4 to 31% | 1 in 66 | No data yet |
| Santa Maria del Mar | Mexico | 9.4 C (48.9 F) | 2007 | 13% | up to 41% | 1 in 56 | No data yet |
| Healy | United States | -18.9 C (-2.0 F) | 2025 | 11% | 4 to 21% | 1 in 126 | Standing |
| Sanya | China | 14.5 C (58.1 F) | 2025 | 11% | 3 to 20% | 1 in 63 | Standing |
| Las Pocitas | Mexico | 10.0 C (50.0 F) | 2006 | 11% | up to 39% | 1 in 62 | No data yet |
| Cedral | Mexico | 3.9 C (39.0 F) | 1997 | 11% | up to 26% | 1 in 71 | No data yet |
| Jasso | Mexico | 5.6 C (42.1 F) | 2010 | 11% | up to 32% | 1 in 63 | No data yet |
| Irapuato | Mexico | 8.9 C (48.0 F) | 1978 | 10% | 3 to 30% | 1 in 93 | No data yet |
| Atlatlahuacan | Mexico | 15.0 C (59.0 F) | 1946 | 10% | 2 to 26% | 1 in 88 | No data yet |
| Gate | United States | -16.1 C (3.0 F) | 2025 | 10% | 2 to 21% | 1 in 68 | Standing |
| Holly | United States | -18.9 C (-2.0 F) | 1996 | 10% | 4 to 15% | 1 in 123 | Standing |
| California | United States | -18.3 C (-0.9 F) | 2024 | 9% | 2 to 20% | 1 in 73 | Standing |
| Paso de San Antonio | Mexico | 6.7 C (44.1 F) | 1997 | 9% | up to 26% | 1 in 69 | No data yet |
A list like this one, drawn up for each year from 2006 to 2025 from earlier data only, would have named 500 station-years. The record fell in 18 of them (3.6%), and the chances given added up to 30.9. Among all stations it fell in 0.41% of station-years.
How the Chance Is Computed
For each station and each kind of record the method takes one value per calendar year: the highest daily maximum temperature (the hottest day), the highest daily minimum temperature (the warmest night), the lowest daily minimum temperature (the coldest night), the lowest daily maximum temperature (the coldest day) and the largest daily precipitation total (the wettest day). The record is the most extreme of those yearly values up to 2025, whatever the year. The fit uses the years from 1950 to 2025 with enough days of data.
The yearly values are regressed on a yardstick of global temperature, the NOAAGlobalTemp annual mean of land and ocean smoothed over several years, and a generalised extreme value distribution is fitted to what the regression leaves. For the wettest day the fit works on the logarithm of the amount. From the fit and the yardstick the method computes the chance that the value of 2026 is beyond the record. Readings sit on a grid, whole degrees Fahrenheit at most stations in the United States and tenths of a degree elsewhere, so the chance is that of passing the record by half a step of the station's own grid: a tie is not a new record. The yardstick of 2026 is its latest published value, so no further warming is assumed. Repeating the fit on resampled years gives the range. The fits are the ones behind the return periods page, and only fits that pass its quality rules are used here.
The baseline needs no fit. In a climate that does not change, a station with 99 years on record has a chance of 1 in 100 that the next year is the most extreme of all: every one of the 100 years is equally likely to hold the record. The baseline counts every year with a value, also the years before 1950 that the fit does not use.
The test repeats the whole method for each year from 2006 to 2025 with the data before that year only. It uses every station and kind whose fit passed the quality rules with the data before that year, whether or not its fit is published today, so a station is not dropped from the test for what happened later. The expected number of records in a year is the sum of the chances of its stations.
What the Numbers Cover and Where They Stop
The method knows two things: the station's own history and the rise of global temperature. It does not know whether 2026 is an El Nino year, how the season has gone so far or what any forecast says. The chance for a station is the same on 1 January and on 30 November. It is a statement about the climate the station is now in, not a weather forecast.
Years are calendar years. A cold season that spans New Year is split over two years, and so is the southern summer. A record here is the most extreme yearly value in this site's copy of GHCN-Daily; a national weather service may hold older or corrected values for the same place.
Temperatures in the United States are recorded in whole degrees Fahrenheit, steps of about 0.6 C. A record there falls only when a value is at least one step beyond it, and ties are common. A tie is not a new record on this page, in the test or in the count for 2026.
Of the 10,324 stations in 77 countries, 4,430 (43%) are in the United States, 2,579 (25%) in Australia and 500 (5%) in Russia. The counts of records expected and fallen mostly describe those networks, not the land area of the world. Neighbouring stations share weather, so the stations are far from independent.
The test years are not independent of the design of the method. The chance for each past year used only earlier data, but the choices behind the method, such as the yardstick and the quality rules, were made in 2026 with all the years in view. That flatters the method to an unknown degree. The 20 test years are also few. The years ahead are the fair test, and this page counts them as they end.
A station move, a new instrument or a change of observing time can shift a station's extremes and its record; the series are used as measured. The page describes how the yearly extremes of these stations move with global temperature and does not claim a cause for any single record.
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. The chances come from fits of each station's yearly extremes against the NOAAGlobalTemp global temperature series. The back test, the scores, the expected counts and the watch lists are computed by this site from GHCN-Daily; they are not an official NOAA product and not a weather forecast.