Nephosthe weather behind the paintings

Every painted sky happened on a real day.

Give Nephos a painting with a place and a date. It gathers the weather records, newspapers and letters of that time and shows, source by source, what they confirm and what they cannot tell you.

See how it works Browse the studies
Claude Monet, The Break-Up of the Ice, January 1880: broken ice covering the Seine beside a winter bank of trees.
Le Petit Journal, 5 January 1880: the headline LA DÉBACLE and the report beneath it. Le Monde illustré, 10 January 1880: La Débâcle de la Seine. The last lines say the break-up began on 2 January and was complete at Pont Picard at ten o'clock on the 3rd.

Paris-Montsouris, lowest reading

−23.9 °C

10 December 1879, still the station’s record

FindingColdest December
on record

the ice goes out,
2–3 January

On the desk: Monet, January 1880

The museum dates Monet’s Seine breaking up at Vétheuil to January 1880. The month before was the coldest December in the Paris record, and two newspapers dated the break-up to 2 and 3 January.

Open this study

How Nephos reads a painting

Four steps, the same for every study. Constable’s cloud study of 21 September 1822 runs through them here.

  1. Constable's cloud study of 21 September 1822.

    “Sepbr 21. 1822. ½ p’st one o clock.”

    Start from what is written

    A museum record or the artist’s own note gives a place and a date. Without one, there is nothing to check.

  2. Howard's register, September 1822. Howard’s notes
    “17–21. Fine.”     Reconstruction, 13:00 UTC
    ENE 57°, 5.9 m/s
    cloud 61 ± 33 %

    Gather the records of those days

    Weather registers, station thermometers, newspapers, letters, weather maps and modern reconstructions, each kept with its page, place and distance.

  3. The note saysHowardModel
    wind EastNEENE
    warm65 / 49 °F14.0 °C
    ½ past onecannot saycannot say

    Test each claim, one source at a time

    Each phrase of the note is a separate claim. A near value is not a match, and two notes by one hand are not two witnesses.

  4. Agrees Partially agrees Contradicts Unknown

    Give a verdict, and say what cannot be known

    One of four labels per claim, never a score. No claim has yet earned “Agrees”; most verdicts are “Partially agrees” or “Unknown”.

Study · Claude Monet, January 1880

The winter the Seine froze

Claude Monet, The Break-Up of the Ice: broken floes crowd the Seine beside bare, damaged trees.
The Break-Up of the Ice, 1880. Museu Calouste Gulbenkian, inv. 451. Photo: Szilas, Wikimedia Commons, public domain; cropped to the canvas and colour-balanced.

The Gulbenkian places this thaw at Vétheuil in January 1880; the Thyssen notes that the Seine had frozen in the winter of 1879. Le Monde illustré reported that the ice began to break on 2 January and that by 3 January the break-up was complete at Pont Picard. Here, broken floes crowd the river beside the trees.

According to the museum, Monet sketched these elements outdoors and later refined the painting in his studio. Other versions followed, including a canvas made entirely indoors. Five museums now hold paintings from this series, which continued after the ice had gone.

Thirty-three nights of frost, then the thaw

Each red point is one night’s lowest temperature at Paris-Montsouris, 60 km from Vétheuil; the pale line is each day’s highest. The ice broke up four days after the thermometer first stayed above freezing.

Point at a night to read it.

The coldest December in 154 years

Each bar is one December’s average temperature in Paris, 1872 to 2025. December 1879, the month before this painting, sits alone at −6.6 °C. The next coldest, 1890, was −2.6 °C. Decembers now average +5.9 °C.

Point at a year to read it.

The ice goes out, in the papers

Le Petit Journal, 5 January 1880, front page column headed LA DÉBACLE: Dimanche 4 janvier 1880.

Le Petit Journal, 5 January 1880, page 1

“After December’s extreme cold, after the accumulated snow and ice, further misfortunes and disasters had to be expected. Reality, alas, far exceeds the forecasts and fears.”

Translation by Nephos. Source: gallica.bnf.fr / BnF.

Le Monde illustré, 10 January 1880, La Débâcle de la Seine.

Le Monde illustré, 10 January 1880, page 22

“The break-up had begun on 2 January; on the 3rd, at ten o’clock, it was complete at Pont Picard.”

Translation by Nephos. Source: gallica.bnf.fr / BnF.

A label a museum could hang beside it

Claude Monet (1840–1926)

The Break-Up of the Ice, January 1880

Monet painted the break-up of the Seine in January 1880. The month before was the coldest December in the Paris record: an average of −6.6 °C, with −23.9 °C on 10 December, still the station’s lowest reading. It froze for 33 nights in a row. Decembers there now average +5.9 °C. Paris has not reached −10 °C since 1997.

Source: Météo-France daily lowest and highest temperatures, Paris-Montsouris (station 75114001).

What this temperature record cannot tell you

  • This record says nothing about the sky.
  • The thaw bounds the scene, not Monet’s painting days.
  • Montsouris is a city station. Some of the warming reflects Paris’s growing urban heat island, not the whole region.
Open the full study

Study · John Constable, 21 September 1822

Three witnesses to one wind

On a Saturday afternoon in September 1822, Constable dated two sky studies ninety minutes apart and wrote the wind on both. Luke Howard’s register noted it too, and a modern reconstruction estimates it.

Constable, Cloud Study, Yale Center for British Art, B1981.25.116.
½ past one. “wind very fresh at East, but warm.” Yale Center for British Art, B1981.25.116. Public domain, CC0.
Constable, Cloud Study, The Courtauld, P.1952.RW.69.
3 o’clock. “brisk wind to east, warm and fresh.” The Courtauld, P.1952.RW.69. CC BY-NC 4.0.
Where the wind came from on 21 September 1822 Constable wrote East (90 degrees). Luke Howard's register, kept at Tottenham or Stratford, gives north-east (45 degrees). The 20th Century Reanalysis gives east-north-east, 57 degrees, for 13:00 UTC, with a one-sigma range of 14 to 86 degrees.
  • Constable, on the sheet: East
  • Luke Howard’s register, kept at Tottenham or Stratford: NE
  • 20th Century Reanalysis, a modern reconstruction for a 1° cell at 13:00 UTC: ENE, 57°. The shaded wedge is its likely range, 14° to 86°.
Partially agrees

Neighbouring directions from three sources that never met. Close, but not the same, and none was taken at Constable’s spot.

September 1822, as Luke Howard printed it

Howard kept a daily weather register near London and published it in The Climate of London (1833). Wind and the first thermometer’s highest and lowest, in °F, transcribed from the page. See the page

Howard’s notes: Ninth Mo. 1–7. Fine. 8. Cloudy: windy. 9, 10. Fine. 11. Cloudy: a little rain in the evening. 12. Fine. 13. Cloudy and fine. 14. Cloudy. 15. Cloudy: windy. 16. Overcast. 17–21. Fine. 22. Cloudy: windy: swallows begin to congregate. 23. Cloudy: rainy. 24. Rainy. 25. Fine. 26. Cloudy: fine. 27. Bleak. 28–30. Fine.

The month, against the long record

The Central England Temperature series, a regional Midlands record that runs back to 1772, puts September 1822 at 12.4 °C: the 45th coolest of 254 Septembers. The 21st itself averaged 11.6 °C.

The 1822 study against the Central England record

Go further into that afternoon

Every study, hung by year

From Turner’s red sky to Homer’s hurricane. Each one is a page you can read, cite and check.

American painters

U.S. Weather Bureau daily map for 14 May 1892, north-eastern states. The painting itself is not reproduced.

1892

Oliver Hazard Perry View of Stonington, May 14, 1892, at 5am

Inscribed on the front in Perry’s hand. The Weather Bureau’s 8 a.m. map puts Block Island, 35 km away, at 44 °F under a clear sky; sunrise at Stonington was 4:29 standard time.

“5am” partially agreesFlorence Griswold Museum, Old Lyme
Winslow Homer, After the Hurricane, Bahamas: a figure among wreckage on a beach under a dark sky.

1899

Winslow Homer After the Hurricane, Bahamas

NOAA’s hurricane database records no tropical storm or hurricane in the Atlantic between 4 November 1898 and 26 June 1899, the winter the museum gives for the painting.

Hurricane unknownArt Institute of Chicago

All studies on one page The 18 object records

Seventeen skies, dated in Constable’s hand

Between 1821 and 1833 Constable wrote the date and hour, and often the weather, on his sky studies. Each note is read against the registers kept on the same days.

The full register Compare any two studies

Where the record pushes back

The cases where the evidence corrected Nephos, or refused to say more. Each keeps the claim, the pushback and the verdict together.

Five discrepancies, kept

  1. Partially agrees

    A machine reading, checked. Howard’s wind for the 21st was first read by text recognition. The page image confirmed NE, which is not Constable’s East, and the verdict was lowered.

  2. Partially agrees

    58 per cent against 61 ± 33. Paint measured on a photograph is not cloud measured in the sky, however close the numbers.

  3. Partially agrees

    Two notes by one hand. Yale and the Courtauld agree with each other because Constable wrote both.

  4. Contradicts

    “31 September.” The Ashmolean’s sheet names a day that does not exist. The museum’s 1 October is plausible and unproved.

  5. Unknown

    Turner and Tambora. A red sky may follow a volcano. “May” is where the evidence stops.

The discrepancy packet

Every reading keeps its history

When the page image changed Howard’s wind from a provisional reading to a checked one, the Yale comparison went from agreement to partial agreement. Both states stay public: the machine record before, the verified record after, and the page image between them.

Corrections and versions

What Nephos will not claim

  • That a weather cell 111 by 68 kilometres speaks for a hilltop.
  • That wide uncertainty rounds to certainty.
  • That a painted cloud is a measurement of the atmosphere.
  • That Constable always painted exactly what was in front of him.
  • That the register of dated studies is complete.
The full list

Earlier readers

John Thornes dated Constable’s cloud studies from historical weather records in the late 1970s and catalogued the skies in John Constable’s Skies (1999); the Yale sheet is his no. 52. Nephos adds a record for every claim that can be cited, corrected and read by machines.

Prior scholarship

Have a painting with a place and a date?

Pick one. Nephos returns the weather records of its days, a record page your object page can link, and a short piece your museum can publish. Corrections are welcome too.

What a museum getsUse it in teaching

Enquiry slip

Painting
title, artist, museum and accession
Place
where it was painted, as closely as known
Date
a day, a month or a range; inscriptions and letters help

What comes back

  • A labelAbout sixty words, every number sourced, ready for a wall or an object page.
  • A record pageEvery claim, witness and verdict, with a suggested citation and a printable summary.
  • A piece to publishA short account of the painting’s day for your blog or newsletter, drafted for your review.
  • Machine-readable dataJSON-LD for each record and, where you publish IIIF, annotations on your own image.
  • Corrections the same dayThe earlier version stays on the record.