When the leaves peak in the High Country
Peak color does not arrive here on a date. It arrives on a staircase, and the steps are made of elevation. We already track 18 places, from Beech Mountain at 5,436 feet down to the Wilkes County valleys near 1,001, with a daily temperature record for each one. So we built a model that turns that spread into 18 separate predictions, and we publish them free.
September 28–October 8 to October 27–November 6
Beech Mountain peaks first, North Wilkesboro last, fall 2026
The season, one band at a time
The exposed ground above 5,000 feet turns while the valleys are still green, and the front works downhill from there over the next 40 days. If you are chasing color, this is the whole trip planner. Pick the band, not the weekend.
Every track above runs September 28–November 6, the full predicted season across all 40 days of it. The bar is where that band sits inside it.
Every town we track
Earliest peak first. Each town links to its own page, where the same window sits beside that town's actual forecast with the arithmetic that produced it.
| Town | Elevation | Predicted peak | Against the ridges |
|---|---|---|---|
| Beech Mountain | 5,436 ft | September 28–October 8 | About 3 days ahead of the 5,000 foot ridges |
| Seven Devils | 3,970 ft | October 8–18 | About 7 days behind the 5,000 foot ridges |
| Banner Elk | 3,701 ft | October 9–19 | About 8 days behind the 5,000 foot ridges |
| Blowing Rock | 3,573 ft | October 10–20 | About 9 days behind the 5,000 foot ridges |
| Boone | 3,333 ft | October 12–22 | About 11 days behind the 5,000 foot ridges |
| Foscoe | 3,041 ft | October 14–24 | About 13 days behind the 5,000 foot ridges |
| Deep Gap | 3,002 ft | October 14–24 | About 13 days behind the 5,000 foot ridges |
| Todd | 2,982 ft | October 14–24 | About 13 days behind the 5,000 foot ridges |
| West Jefferson | 2,976 ft | October 14–24 | About 13 days behind the 5,000 foot ridges |
| Jefferson | 2,920 ft | October 15–25 | About 14 days behind the 5,000 foot ridges |
| Burnsville | 2,825 ft | October 15–25 | About 14 days behind the 5,000 foot ridges |
| Vilas | 2,746 ft | October 16–26 | About 15 days behind the 5,000 foot ridges |
| Valle Crucis | 2,677 ft | October 16–26 | About 15 days behind the 5,000 foot ridges |
| Sugar Grove | 2,677 ft | October 16–26 | About 15 days behind the 5,000 foot ridges |
| Spruce Pine | 2,559 ft | October 17–27 | About 16 days behind the 5,000 foot ridges |
| Bakersville | 2,471 ft | October 17–27 | About 16 days behind the 5,000 foot ridges |
| Wilkesboro | 1,024 ft | October 27–November 6 | About 26 days behind the 5,000 foot ridges |
| North Wilkesboro | 1,001 ft | October 27–November 6 | About 26 days behind the 5,000 foot ridges |
Model leaf-v0-draft | Generated July 26, 2026
How we get a date
Three ingredients, in order of how much they move the answer. All of it is arithmetic you can check by hand, which is the point.
1. Daylength sets the clock
What triggers senescence, the shutdown of chlorophyll that lets the yellows and reds show, is mostly how much daylight a tree is still getting. Daylength on a given calendar date is identical year to year, so the base timing is fixed. We encode it as one anchor: peak color at 5,000 feet lands in the first week of October, and we pin it to October 6, 2026.
2. Elevation sets the gradient
Higher ground cools sooner and turns sooner. The documented pattern here is a color front descending roughly 1,000 to 1,500 feet a week. We use 6.5 days per 1,000 feet of drop, which also fits the observed gap between Grandfather Mountain in early October and Boone in the middle of it. This is the dominant term, and the one you can check against any two rows of the table above.
3. Temperature nudges it
A warm early autumn delays the shutdown. A cold snap hurries it along. We compare each town's September temperatures against that town's own 6-year normal, and shift the peak 1.5 days per degree, hard-clamped to 7 days in either direction. The clamp is deliberate. Daylight and elevation own the prediction. Temperature only nudges.
That third term is switched off right now. September 2026 has not accrued enough days to read, so every window above is elevation and climatology only, and these dates will move once the temperature term comes on. We would rather show you the gap than paper over it.
The window itself is the predicted center plus or minus 5 days. Peak is a stretch, not an instant.
What this model cannot do
Run against 2024 and 2025, every elevation band landed inside its predicted window, and the high and middle bands landed within a day or two of center. Read that for exactly what it is: confirmation that the model reproduces the documented elevation gradient. It is not confirmation that the model can tell a warm year from a cool one, day for day, because the published reports describe the season in the same language every year and never give a dated calendar peak. That test needs graded seasons, and this is the first one.
- The gradient is a straight line. Real color fronts bend with slope and aspect. A north-facing cove turns before a south-facing ridge at the same elevation, and none of that is in here.
- Species are ignored. Maples, poplars, oaks, and birches peak on different schedules, and a town's dominant species shifts its real peak.
- The temperature coefficient is a first guess. September mean temperature stands in for the real drivers, which are cool nights, sunny days, first frost, and drought stress. That number gets corrected by grading, not by argument.
- Ground truth is a judgment call. No station reports a peak color number. Grading means reading photo galleries and written reports by eye, which is softer than scoring a temperature against an archive, and we are not going to pretend otherwise.
Then we check ourselves
Every forecast on this site is a claim, and a peak color window is a claim about October. The scoring rules were written and tested in July, before a single observation existed, so nothing gets graded on a curve after the fact: full credit within three days of the observed peak, then six points off per day, on the same Right and Wrong bands the weather scores use. Starting the week of September 21 we read the published color reports every Monday and record what they say.
Nothing is scored yet. The leaves have not turned. When they do, the results land here whether or not they flatter the model.
What we grade against
- Grandfather Mountain's dated fall color gallery. The high-elevation leading indicator, above 5,000 feet, where color turns first.
- WataugaOnline's fall color report. The Boone-elevation check, closest to where most readers live.
These are human-judged reports, not data feeds, which is why we read them rather than scrape them. A source that publishes a prediction rather than an observation is not on this list, however good the prediction is. Grading a forecast against another forecast is not grading.
Why we bothered
The data behind a leaf forecast is elevation and temperature. Both are public, and we were already collecting the second one every morning for 18 places. So the only thing standing between you and a per-town peak color forecast was somebody writing the arithmetic down and agreeing to be graded on it. That is the whole site, really. The same goes for how we score the weather.