Roof with Snow

H. Daniel Peck identifies the subject as the house across the street at 2000 Fifth Avenue, then also called Evelyn Avenue. O’Keeffe looked across at it from her upstairs room in the Shirley house at 500 Twentieth Street, whose large dormer faced east. Peck reads the county deed index as showing a vacant lot in spring 1916 and a house there, moved onto the site, by the end of that year. The Texas Historical Commission gives the same address and a view from her second-floor window. Peck, pp. 100 and 105, note 4; house identification, PDF p. 56.

The year is disputed. Access O’Keeffe heads its record 1916, the catalogue raisonné date, while its notes say Abiquiú Notebooks dates 1917. Peck dates the watercolor to early 1917. His basis is a label in O’Keeffe’s handwriting assigning this date in the Amarillo Museum of Art’s archives, fixed to the backing of her frame. He adds that the notebooks give 1917, with a separate entry that gives 1915. Neither the label nor the notebooks were inspected here, and the weather does not choose between the years. Peck, p. 109, note 23.

His proposal for the day follows on the next page, in the two sentences quoted below. He hedges it himself in the paragraph after them. The exact day, he writes, can never be known: the winter brought several significant snowfalls, and any of them after mid-February—when the artist moved into the Shirley house—could have been the occasion for this work. His letter for 11 March is cited to the Stieglitz/O’Keeffe archive at Yale. The study tests the two weather claims he makes: the snowfall of 11 March, and snow after the move. Peck, p. 106, and p. 109, note 22.

The window therefore runs on his terms: every day from 18 February to 30 April 1917, 72 days. Von Lintel’s timeline dates the move to 5 February instead, citing O’Keeffe’s letter of 4 February in Greenough’s edition of the correspondence. The window does not choose between the two dates. Timeline p. 20 and p. 54, note 139.

Verdict: Contradicts, on two conditions. The first is that the letter is dated correctly; it was not read here. The second is that Amarillo stands for Canyon. The station’s archive coordinates are about 28 km from the house, and a local shower in Canyon cannot be fully excluded. But an all-day snowfall, with no precipitation at the station and a 73 °F afternoon, is not supported by this record.

The observation hours are not recorded, so a reporting date may not match a calendar day exactly. That is why 10 and 12 March are shown as well: all three dates are dry. These are the historical city observations, marked as CRB data by ACIS. The Weather Service places its downtown office at Taylor and Seventh Streets until the move to the airport in April 1941; the 1917 instrument position has not been surveyed here. Station history; ThreadEx station history; address geocode; format and units.

Verdict on the broader claim: Partially agrees. Snow did fall after the move, so a snowfall in that period could have been the occasion. The record cannot show which snow, if any, she painted. After 11 March it shows no measurable snowfall through 30 April, only traces.

One snowfall lies just outside the window. Amarillo recorded 2.2 inches on 14 February, 4 days before Peck’s move date. Under the 5 February date in Von Lintel’s timeline, it would also fall after the move. February daily data.

The NWS daily-record table lists the same 3 March amount. That is a cross-check within the federal weather record, not an independent second observation. NWS table.

How many days fit the painting? The record supplies a count of measurable snowfall dates at Amarillo. It supplies no defensible count of snow-covered days on this roof, or of possible painting days. A reported zero for new snow does not mean no snow lay on roofs. No melt model or persistence duration is calculated.

Two published readings of O’Keeffe’s letters therefore place different weather on Sunday 11 March: Peck’s all-day snowfall and the timeline’s fine day in the canyon. The station record fits the fine day. The timeline notes that the archive dates letters according to their postmark and Greenough according to their time of composition. A dating offset between letters is one possible explanation. The original letters were not read here, so the study does not decide which letter or date is off. Timeline p. 49, note 10.

These are short passages from modern transcriptions of the artist’s own words, not independent weather observations, and neither passage names this watercolor. No contemporary Canyon or Amarillo newspaper snow report was verified in this bounded search. Archive access failed; that is a gap in this study, not evidence that no report survives.

Peck offered 11 March as a speculation and said so. The test here bears on that speculation only, not on his identification of the house, his reading of the two studies, or his 1917 dating from the artist’s label. The museum’s description concerns colored shapes and a cropped view; no sky color or snow depth is recovered from them. The next useful evidence is the 11 March letter itself, read against its postmark and Greenough’s date, and any Canyon report of snow that winter.