Fall foliage webcams from Acadia to the Blue Ridge

At 17:31 on 2 October the camera above the sugar maples at Underhill, Vermont had most of the hillside in rust and copper, one wedge of green still holding in the middle of the shot, and a dirt track running up between the trees into the fog. The frame carried that stamp in its top left when I pulled it at 01:41 UTC on 2026-10-03. A new one lands about every 30 minutes.

That camera belongs to a research network rather than a tourist board, and it is the one I would open first this month. The lists of foliage webcams I checked today are mostly links grouped by state, which is fine until you click one and get a dead player or a view of a car park. So here are the ones that were writing frames this morning, what each actually points at, and when in the day it is worth opening.

The sugarbush at Underhill, and a red maple at Boone

Proctor Maple Research Center belongs to the University of Vermont and sits at 403 m on the western side of Mount Mansfield. The camera looks north-east across the sugarbush, over the grey metal roof of one of the research buildings, with a power line cutting the shot at head height. It has been photographing those trees since 2008-04-15, and the archive is 351,829 images deep, which makes it older than every other camera here by eight years. Sunrise there on 2026-10-03 is 06:51 and sunset 18:28, per Open-Meteo for 44.525, -72.866, so the useful window is the whole working day, and the fog sits in that valley first thing.

Eight hundred miles south, Appalachian State's highlands camera at Boone is the best single tree I found. Its 18:02:05 frame on 2 October has two maples gone fully scarlet against a wall of hemlock and white pine, the nearer one still green through its bottom third, late sun coming in low from the left and a bank of haze washing out the ridge behind. The site is at 1,016 m, which is why it turns weeks before the valleys around it, and the camera reported its own temperature as 48.0 in the same overlay. Sunset at Boone is 19:07, an hour later than Vermont, so it is the one still in daylight when the northern cameras have gone dark.

Above the canopy at Bartlett, and below it

The top-of-tower camera at Bartlett Experimental Forest in the White Mountains gives you the view no road gives you: the canopy from above, running out to the ridges. At 22:40 UTC on 2 October it was scarlet, orange and gold in patches with dark spruce spires standing through it, two or three house lights on in the valley and the mountains stacked in blue haze behind. Beech, hemlock and striped maple, at 285 m, photographed every half hour since 2016-12-13 for 100,282 images. There is a second camera lower down the same tower looking into the canopy rather than across it, and the pair together are the clearest demonstration I know of that a forest turns from the top down. I wrote about this camera's liveness check in a post about Waterville Valley on 2026-10-01.

Bartlett also answers the question people actually want answered, which is when to come. Researchers draw regions of interest on the frame and reduce each one to a green chromatic coordinate, the green channel's share of the total brightness, and publish the daily series. I pulled the deciduous series and the evergreen one this morning and took, for each of the eight complete years from 2017 to 2024, the date the deciduous value fell halfway from its mid-September level to its mid-November floor. The median is day 273, which is 30 September. The earliest was 2020-09-21 and the latest 2021-10-06, so a fortnight of spread either side of the end of the month. Over the same window the evergreen region falls about half as far, which is the arithmetic behind every photograph of a red hillside with green spikes in it.

A hundred and twenty-seven towers framed the same way

The New York State Mesonet runs 127 standard weather stations, roughly 17 miles apart, with at least one in every one of New York's 62 counties, and each one has a camera on a 30 ft tower. We carry 162 of their cameras, counted from our own catalogue on 2026-10-03 and filed under New York. Four of them sit in the Adirondacks and were all written within a second of each other at 01:39 UTC: Piseco in the southern lakes, Cold Brook west of them, Ticonderoga over on Lake Champlain and Malone up near the Quebec line.

Be clear about what you get. The Piseco frame at night is an infrared grey rectangle of wire stock fence, a numbered snow stake reading 0 to 3, an instrument enclosure at the left and the treeline across the top third. Nobody sited these for the view. What they have instead is sameness: identical tower, identical framing, identical quarter-hourly cadence, which means the Malone frame and the Piseco frame this week can be set against each other and against the same two frames next week, and the colour moving south down the state is a thing you can actually see rather than a thing a map asserts.

Acadia after dark, and the summits at first light

The Jordan Pond House camera in Acadia had written a frame a minute before I asked, stamped 10/2/2026 9:41:00 PM with 62.0 degrees and wind SSW at 6.0 mph burned along the bottom. At that hour it is monochrome infrared: the lawn, the tea house at the right with three windows lit, empty tables and chairs still out on the grass. Come back after 07:00 Eastern and it is one of the best framed views in Maine, looking up the pond at the Bubbles.

Vermont's mountain cameras are a daylight proposition and the file sizes say so plainly. At about 01:34 UTC Jay Peak's mountain camera was a 20,408 byte image, which is lights and cloud, while Stowe's gondola camera came back as 1,218 bytes and Mount Snow's summit as 2,308: an almost empty file is an almost black picture. In daylight they are the fastest moving things on this list, a new frame every five minutes, and they sit high enough to turn a week or two before the valley floor. Our Vermont page listed 189 live cameras today, 62 of them at ski areas.

What we are doing next

A still image on a timer is the right format for watching a slow thing, and the wrong one for watching weather cross a ridge. The job we have queued is to make the per-camera pages show the last few frames rather than only the current one, so that a half-hourly camera reads as a sequence and you can see the change between Tuesday and Friday without keeping a tab open.

The thing I have not solved is the comparison across sites. Proctor and Bartlett and Boone each publish a green chromatic coordinate, but the three cameras have different sensors, different exposure behaviour and different angles to the sun, and I do not think the absolute numbers are comparable between them. The shapes of the curves are. If you know of a correction that makes the levels comparable too, I would like to read it.