寒露, Cold Dew, is not "early October". In 2026 it is 8 October at 14:29 — a specific minute, on a specific clock, and the clock is the one at 120° east. A solar term is a moment rather than a day, which is why the date it lands on changes from year to year and why your almanac and somebody else's can print different days for the same term.
Here is the whole of 2026. These are our own figures, computed for our solar terms tool; all twelve of the major terms match the Hong Kong Observatory's published almanac to the printed minute.
| Term | 2026 date | Time | Term | 2026 date | Time |
|---|---|---|---|---|---|
| 小寒 Minor Cold | 5 Jan | 16:23 | 小暑 Minor Heat | 7 Jul | 09:57 |
| 大寒 Major Cold | 20 Jan | 09:45 | 大暑 Major Heat | 23 Jul | 03:13 |
| 立春 Start of Spring | 4 Feb | 04:02 | 立秋 Start of Autumn | 7 Aug | 19:43 |
| 雨水 Rain Water | 18 Feb | 23:52 | 處暑 End of Heat | 23 Aug | 10:19 |
| 驚蟄 Awakening of Insects | 5 Mar | 21:59 | 白露 White Dew | 7 Sep | 22:41 |
| 春分 Spring Equinox | 20 Mar | 22:46 | 秋分 Autumn Equinox | 23 Sep | 08:05 |
| 清明 Pure Brightness | 5 Apr | 02:40 | 寒露 Cold Dew | 8 Oct | 14:29 |
| 穀雨 Grain Rain | 20 Apr | 09:39 | 霜降 Frost Descent | 23 Oct | 17:38 |
| 立夏 Start of Summer | 5 May | 19:49 | 立冬 Start of Winter | 7 Nov | 17:52 |
| 小滿 Grain Buds | 21 May | 08:37 | 小雪 Minor Snow | 22 Nov | 15:23 |
| 芒種 Grain in Ear | 5 Jun | 23:48 | 大雪 Major Snow | 7 Dec | 10:52 |
| 夏至 Summer Solstice | 21 Jun | 16:25 | 冬至 Winter Solstice | 22 Dec | 04:50 |
All times are at UTC+8. Look at 雨水 and 芒種 in that table: 23:52 and 23:48. In Japan and Korea, an hour ahead, both of those have already tipped over into the next day. Two of 2026's twenty-four terms fall on a different calendar date depending on which side of the East China Sea you read the almanac.
What is being measured is an angle, not a date
The Sun's apparent position, seen from Earth over a year, traces a circle called the ecliptic. Measure around that circle from the spring equinox and you have a number called the Sun's longitude, which runs from 0° to 360° once a year. A solar term is the instant the Sun's longitude reaches an exact multiple of 15°. The Hong Kong Observatory states it in one sentence: the terms divide the ecliptic into 24 equal segments, 15 degrees apart, and at the vernal equinox the Sun's longitude is 0°.
Twenty-four exact angles give twenty-four exact instants. As Taiwan's Central Weather Administration puts it, these are 24 specific moments, not 24 specific days. The calendar date is a convention laid on top of the astronomical instant; the astronomy is fixed, but the convention varies.
The terms alternate between twelve major ones, 中氣, at 0°, 30°, 60° and so on, and twelve minor ones, 節氣, at the odd multiples. That distinction is not decoration: it is the mechanism that decides the leap month in the lunisolar calendar. Each lunar month is supposed to contain one major term, and a lunar month that contains none becomes the leap month of the month before it. Over nineteen tropical years there are 228 major terms and 235 lunar months — we counted the 228 in our own figures and it comes out exactly — so seven months over nineteen years have to be leap months.
The reason a month can miss a major term is that the gaps are not equal. The Earth moves faster along its orbit near perihelion and slower near aphelion, so equal angles take unequal time. In our figures the average gap between neighbouring terms is 15.219 days, ranging from 14.714 to 15.736; between major terms it averages 30.437 days and ranges from 29.441 to 31.457. A lunar month lasts about 29.53 days, so it can fit entirely inside one of the long gaps.
That unevenness was not always there. Before 1645 the year was cut into 24 equal intervals of time — 平氣, mean terms. From 1645 the terms have been equal arcs of the Sun's apparent longitude — 定氣, true terms. Pre-1645 dates in old records differ from a modern recomputation not because the astronomers were careless, but because they were computing a different thing.
No solar term has a fixed date, and the safe dates have moved
We computed all 24 terms for every year from 1900 to 2100 — 4,824 instants — and counted the calendar dates each term can land on.
- None of the 24 falls on a single fixed date.
- 17 of the 24 take any of three dates. Seven take four — 驚蟄, 春分, 立夏, 芒種, 立秋, 白露 and 小寒.
- The solstices and equinoxes are not the stable ones, which is the assumption most people arrive with. 春分 takes 19 to 22 March; 夏至 takes 20 to 22 June.
The more useful result is that the range drifts, so a great many published date ranges are quietly out of date. Every English page that gives Lichun as "4–5 February" is repeating a twentieth-century fact. Across 1900–2100, Lichun last fell on 5 February in 1980 and first fell on 3 February in 2017. In our century it is 3 or 4 February and never the 5th.
| Term | Dates it took, 1900–1999 | Dates it takes, 2000–2099 |
|---|---|---|
| 立春 Start of Spring | 4 or 5 Feb | 3 or 4 Feb |
| 清明 Pure Brightness | 4, 5 or 6 Apr | 4 or 5 Apr |
| 白露 White Dew | 7, 8 or 9 Sep | 6, 7 or 8 Sep |
| 冬至 Winter Solstice | 21, 22 or 23 Dec | 21 or 22 Dec |
The last 6 April Qingming was 1943; the last 23 December solstice was 1951; and the winter solstice did not touch 21 December once before 1988. Our guide to the Chinese festivals set by the sun rather than the moon goes into what that does to Qingming and the solstice as holidays.
Which clock the official time is on, and when that mattered here
The mainland calendar has a national standard behind it, GB/T 33661-2017, drafted by the Purple Mountain Observatory, and it reckons solar-term times in East-longitude-120° standard time — UTC+8. So does the Hong Kong Observatory. Everyone in UTC+8 therefore reads the same date today, and that includes Singapore, Kuala Lumpur, Hong Kong, Taipei and Beijing.
The agreement between these tables is better evidence than it first looks. The Hong Kong Observatory notes that its astronomical data comes from the HM Nautical Almanac Office in the United Kingdom and the United States Naval Observatory. So when its figures match the Purple Mountain Observatory's, that is a UK and US derived ephemeris agreeing with an independently computed Chinese one. On the 28 instants we could compare all three ways, Hong Kong and Purple Mountain agree on every single one, and our own figures agree with Hong Kong on 164 of 168 printed minutes — the four exceptions all sit within a second of a rounding boundary. Third-party almanac sites are less reliable: one gives 寒露 2026 as 14:28 where all three authorities give 14:29. And since the official bodies publish to the minute, a page printing seconds is claiming a precision the published record does not carry.
Today's agreement across UTC+8 is recent, though, and Singapore and Malaysia are the reason it is worth a paragraph. Singapore and Malaya kept their clocks at UTC+7:30 until they went forward half an hour at 23:30 on 31 December 1981. Running all 4,824 instants through the historical record of time zones, 55 solar terms fell in that half-hour after Beijing midnight — so for eighty years, on 55 occasions, the term that Beijing called today was yesterday in Singapore and Kuala Lumpur. Since 1 January 1982 that has happened zero times, and it cannot happen again while the offset holds.
It is still live one border away. Vietnam computes its own lunisolar calendar on UTC+7, so its terms sit an hour earlier. Of the 201 winter solstices in our span, ten fall between midnight and 00:59 Beijing time, and on those the solstice is a day earlier in Hanoi — which matters more than it sounds, because the month containing the winter solstice is the eleventh month in both calendars, so a one-day shift there can move the whole year. The 1984 solstice fell at 00:22 Beijing time: 21 December in Hanoi, 22 December in Beijing. That is the documented mechanism behind Vietnamese Tết 1985 arriving a full month before Chinese New Year that year — 21 January against 20 February. The next solstice in that window is 2050.
Further out, the effect is routine. Under today's time-zone offsets, a term falls on a different date 4.2% of the time in Jakarta, Bangkok and Hanoi, 33% in London, 54% in New York and 67% in Los Angeles.
Nine minutes against four hours forty
But do the terms describe the local weather? A Hong Kong government body gives the most direct answer: the Observatory's own wording is that the 24 solar terms reflect, to some extent, the climate over central China in ancient time.
We put a number on the gap. At each term the Sun's declination is a fixed quantity that does not depend on the year, so the daylight length at each term can be computed for any latitude. Using the same horizon the Observatory uses for sunrise and sunset:
| Place | Shortest day | Longest day | Swing across the year |
|---|---|---|---|
| Singapore | 12h03 | 12h12 | 9 minutes |
| Kuala Lumpur | 11h56 | 12h18 | 22 minutes |
| Jakarta | 11h46 | 12h29 | 43 minutes |
| Bangkok | 11h19 | 12h56 | 1h37 |
| Manila | 11h16 | 12h59 | 1h44 |
| Hong Kong | 10h46 | 13h30 | 2h44 |
| Zhengzhou, on the Yellow River | 9h49 | 14h29 | 4h40 |
The entire annual swing the system was built to track is, in Singapore, nine minutes of daylight. Where the terms were named it is four hours and forty. In Jakarta the sign even flips: because Jakarta is south of the equator, the summer solstice is its shortest day, at 11h46. Singapore's own weather service describes a climate of two monsoon seasons with high and uniform temperatures and little variation month to month — there is no four-season structure here for the names to map onto, and three of the names (霜降 frost, 小雪 light snow, 大雪 heavy snow) describe events that do not happen at sea level anywhere in ASEAN.
None of which makes the terms wrong here. The instant of 冬至 is the same instant in Singapore as in Beijing, and the dates in the table above are correct for anyone on UTC+8. The astronomy travels, but the agricultural meaning attached to the names does not. UNESCO's inscription in 2016 recognises exactly that distinction — the element is called "The Twenty-Four Solar Terms, knowledge in China of time and practices developed through observation of the sun's annual motion", and its description places the system's origin in the Yellow River region.
The 2026 table and the 1900–2100 figures are computed by the astronomical routine our own tool ships, on 24 September 2026; all twelve of 2026's major terms match the Hong Kong Observatory's published 2026 almanac sheet to the printed minute, and the 28 instants we could compare against the Purple Mountain Observatory's own published calendar data match it as well. The definition and the ecliptic geometry are the Observatory's; the "moments, not days" wording is Taiwan's Central Weather Administration; the mainland standard is GB/T 33661-2017, drafted by the Purple Mountain Observatory; the climate description is the Meteorological Service Singapore's; the element title and origin are UNESCO's own. The historical clock changes come from the standard time-zone database, and the daylight figures are ours.
View author profile → · Editorial policy
