Independent New Hampshire Journalism · Northfield, NHMeasured again, frame by frame, the September 22 lights line up with five stars of the Big Dipper. The September 10 light flashes 70 times a minute, which GSR’s first story said it did not do.
Audio edition: a synthetic voice reads the article. The charts, frames and fact check are on this page.
Granite State Report’s first look at a reader’s two videos of blinking lights over Rochester said the light in the September 10 clip did not flash. It does. Measured again, one frame at a time, the light flares for a single frame sixteen times in 13 seconds, once every 0.86 seconds. That is the rhythm of an aircraft’s anticollision strobe. Stars do not keep a clock.
The second look settles more than it reopens. The lights at the start of the September 22 clip are five stars of the Big Dipper, and the match tells which way the phone was pointed, the fact the first story said only the reader could supply. The light the reader called a plane at the end of that clip flashes like one: two quick white bursts, then a dimmer reddish pulse, on a cycle of 1.42 seconds. Nothing GSR measured in either video needs an unexplained cause, and the September 10 finding the first story got wrong now points to an aircraft or a drone, not a star.
This follows GSR’s October 12 story, Two Videos of Blinking Lights Over Rochester, Checked Frame by Frame, which published both clips with the sound removed and checked them against the flight records and airport weather for both nights. That story’s September 10 finding is corrected at its top. The figures here come from a fresh measurement of the same two original files. The checks themselves follow GSR’s guide for reader footage.
The flash the first story missed
The video as GSR first published it on October 12: sound removed and trimmed to protect the reader’s privacy. Video: submitted by a reader who asked not to be identified.
The first story’s line on the September 10 clip was plain: “GSR found no flashing rhythm in either one.” From there it reasoned that a small plane flying with its anticollision lights on would show a flash, so the missing flash counted against the two small planes in the flight records and toward a bright star or planet.
This time GSR measured two things in every one of the clip’s 954 frames: the brightness inside a small box around the light, and the brightness of the whole picture. If the phone’s exposure jumps, everything brightens together. If the light itself flashes, only the box does.
Only the box does. In the first 13 seconds the light flares 16 times. Each flare lasts one frame, a twenty-fourth of a second, and throws a pale halo around the light. In those frames the box runs 2.8 times its usual brightness while the picture as a whole changes by less than 1 percent. The flares land every 20 or 21 frames, 0.86 seconds apart on average, or 70 a minute. They do not line up with the video’s compression, which stores a complete picture every 29 frames, so the encoder did not make them.
0:00 to 0:13.5 of the September 10 video at real speed, then 0:00 to 0:02.5 at quarter speed, cropped to the sky and brightened. The circle follows the light; FLASH marks each one-frame flare. Granite State Report analysis of the reader’s original file, sound removed.
The second light, filmed from 0:32 to 0:39 on a different patch of sky, flashes too. GSR counted five flashes about 1.48 seconds apart, roughly 40 a minute, with the box at 1.9 times its usual brightness and the frame as a whole unchanged.
Both September 10 rates sit inside the federal range for airliner anticollision lights, and a night-flying drone must carry flashing lights of its own. The flash cannot say which of those the reader filmed. It does rule out a star.
The flight records in the first story list two small planes in the air near downtown while the clip ran: a Cessna 172 about 1.2 to 1.6 miles northeast at about 3,450 feet, turning, and a Piper PA-28 about 2.5 to 3.4 miles east at about 6,175 feet, heading southwest, each roughly 18 to 26 degrees above the horizon as seen from downtown. Either would carry a strobe. No star shows in the September 10 frames to fix which way the phone faced, so GSR cannot say which plane, if either, is in the video. Drones do not appear in those records at all.
The light also moves. Lined up on the streetlight below the frame, the first 10 seconds draw a smooth track about 300 pixels long. That fits an aircraft, but it is weaker evidence than the flash, because a reader walking would shift the nearby streetlight too.
September 22: the Big Dipper
The video as GSR first published it on October 12: sound removed and trimmed to protect the reader’s privacy. Video: submitted by a reader who asked not to be identified.
The second clip opens on a moonlit tree line with faint lights above it. The first story showed those lights keeping their spacing while the phone shook, which is how stars behave, but did not name them. Stacking 166 frames from 0:04 to 0:11, lined up on the brightest light, brings out five.
GSR tested those five against the sky over Rochester at 8:28 p.m. that night, using the Yale Bright Star Catalog and PyEphem, free astronomy software, for every star brighter than magnitude 4.5 above the horizon, plus the planets. One group fits: Dubhe, Merak, Phecda and Megrez, the four stars of the Dipper’s bowl, and Alioth, the first star of its handle. With one scale and one rotation applied to the picture, every light lands within 2 to 5 pixels of where the catalog puts its star. The brightness order matches as well, Dubhe brightest and Megrez faintest. No other pairing came close; the runner-up placed only two of the other three lights.
0:04 to 0:11 of the September 22 video at real speed, cropped and brightened. The circles follow the five lights GSR matched to Dubhe, Merak, Phecda, Megrez and Alioth; the lines join them in the Dipper’s shape. Granite State Report analysis of the reader’s original file, sound removed.
That match does what the first story said only the reader could do: it fixes the camera. At 0:04 the phone was aimed north-northwest, at a compass bearing of about 337 degrees, with the middle of the frame about 17.5 degrees above the horizon and the phone held within 6 degrees of upright. Each pixel covered about 0.028 degrees of sky.
The same scale accounts for the two lights the first story tracked from 0:30 to 0:52, which stayed about 539 pixels apart. At 0.028 degrees a pixel, that is 15.2 degrees, the distance between Dubhe and Alioth. GSR’s reading is that the reader was still filming the Dipper. The spacing and the brightness fit, though no third star in those frames pins it down.
On camera the reader says the lights blink and fade completely out. Over the seven seconds GSR measured, Dubhe’s brightness swung between about a third of its average and about one and a half times it, and GSR’s tracker found it in 94 percent of frames. It never flashed on a schedule. The Dipper sat 17 to 28 degrees above the horizon, low enough to shine through a long slant of air, and twinkling is strongest near the horizon, where, as Astronomy magazine put it, “the brightest twinkling stars can flash with color.” The first story also found moonlit haze and high cloud that night, and thin cloud drifting across a star dims it.
Seen from Rochester, the Dipper never sets. Dubhe sits about 61.75 degrees north of the celestial equator, so from 43.3 degrees north it circles the North Star and stays at least about 15 degrees above the northern horizon, by GSR’s calculation. The reader describes a month of sightings GSR has not seen, and nothing here says those were all the Dipper. But a bright star that sits in the same corner of the sky every clear night is the first thing to rule out.
The plane at the end of the clip
Near the end of the September 22 clip the reader turns to a different light and says it is probably an actual plane. The first story agreed and measured its bright pulses 0.62 to 0.79 seconds apart. The closer look splits those pulses into two kinds.
Every 1.42 seconds the light fires a white double flash, two bursts about an eighth of a second apart. Midway between them comes a dimmer pulse whose red channel runs about 1.6 times its green; in the white bursts the two are about even. White, then reddish, then white again: that alternation is where the first story’s spacing came from, so its numbers stand. The double flash repeats 42 times a minute, inside the federal range.
2:09.5 to the end of the September 22 video at real speed, then 2:10.4 to 2:13.6 at quarter speed, cropped and brightened. WHITE marks the double flash and RED the dimmer reddish pulse between. Granite State Report analysis of the reader’s original file, sound removed.
Which airplane, GSR still cannot say. The star match fixes where the phone pointed at 0:04, not two minutes later. The first story’s flight list for that clip includes a United Boeing 777 that crossed the western and north-northwestern sky at 33,000 feet, but tying it to this light would take the camera’s direction at 2:09, and nothing in those frames supplies it.
What is still open
The light the reader saw over Route 108 toward Somersworth, the one that “looked like it was spraying something,” was never filmed, and the tip names no date or hour, so GSR still has nothing to check it against. In the September 22 clip, the stretch from 0:12 to 0:30, where the first story found dotted trails it could not match, and most of the stretch from 0:52 to 2:08 remain unmeasured.
Two facts from the reader would close most of that: where in Rochester the September 10 clip was shot, and which way the phone faced. With those, the strobe can be laid against the flight tracks for that minute, the same way the stars were laid against the catalog.
On camera, the reader says of the lights: “It’s not a plane, it’s not a drone, and it’s not a star.” In the stretches GSR measured, the steady lights were stars of the Big Dipper and the flashing ones flashed the way aircraft lights must. GSR’s own first look got the flashing wrong on September 10, and the correction sits at the top of that story. The fix came from the same test every reader’s clip gets: the original file, every frame, and a clock.
— Granite State Report
Your Turn
Saw lights over Rochester on September 10 or September 22, 2026? A second view from another spot, with the time and the direction you faced, is worth more than a sharper picture. Send the original file through Something You Can’t Explain, or write to granitestatereport@gmail.com. The airlines, the military, Skyhaven Airport and the Pease Development Authority were not asked to comment for this story. Any of them may write to the same address, and any reply will be added here with its date.
Fact check
| # | Claim | Status | Source |
|---|---|---|---|
| 1 | GSR’s October 12 story said of the September 10 clip that “GSR found no flashing rhythm in either one” and reasoned from the missing flash toward a bright star or planet; it measured the September 22 plane’s pulses 0.62 to 0.79 seconds apart and the 0:30 to 0:52 pair at a median 539 pixels apart. | VERIFIED | Granite State Report, “Two Videos of Blinking Lights Over Rochester, Checked Frame by Frame,” post content as stored October 27, 2026. |
| 2 | Both files: iPhone 16 Pro Max, recorded September 10, 2026, 8:44:32 p.m. EDT (39.75 s, 954 frames) and September 22, 2026, 8:28:39 p.m. EDT (135.5 s); no location tags. | VERIFIED | Embedded QuickTime metadata of the reader’s original files, read by GSR. |
| 3 | September 10, first light: 16 one-frame flares between 0:00.29 and 0:13.13, spaced 20 or 21 frames (0.83 to 0.88 s), mean 0.856 s, about 70 a minute. | VERIFIED | GSR measurement of every frame of the original file. |
| 4 | In flare frames the 81 × 81 pixel box around the light averages 2.8 times its brightness in other frames; whole-frame brightness changes by less than 1 percent (ratio 0.993). | VERIFIED | GSR measurement of the original file. |
| 5 | The video stores a full picture (keyframe) every 29 frames, which does not match the 20-to-21-frame flare spacing. | VERIFIED | Frame table of the original file read with ffprobe. |
| 6 | September 10, second light: five flashes between 0:32.71 and 0:38.63, mean spacing 1.48 s, about 40 a minute; box 1.9 times usual brightness; whole frame unchanged (ratio 1.000). | VERIFIED | GSR measurement of the original file. |
| 7 | With the first 10 seconds lined up on the streetlight, the first light moves about 300 pixels; a moving observer would also shift the streetlight. | VERIFIED | GSR measurement (phase correlation on the lower frame, 240 frames). |
| 8 | Transport-category anticollision lights: 40 to 100 cycles a minute; aviation red or aviation white. | VERIFIED | 14 CFR 25.1401(c) and (d), eCFR. |
| 9 | A drone flying at night must have lighted anticollision lighting visible for at least 3 statute miles with a flash rate sufficient to avoid a collision. | VERIFIED | 14 CFR 107.29(a)(2), Legal Information Institute text of the eCFR. |
| 10 | During the September 10 clip a Cessna 172 (about 3,450 ft, 1.2 to 1.6 mi northeast, 20 to 26 degrees up) and a Piper PA-28 (about 6,175 ft, 2.5 to 3.4 mi east, 18 to 23 degrees up) were near downtown. | ATTRIBUTED | GSR, October 12 story, from the adsb.lol archive. |
| 11 | The five lights at 0:04 to 0:11 of the September 22 clip match Dubhe, Merak, Phecda, Megrez and Alioth: residuals 2 to 5 pixels after one scale and one rotation, 0.09 degree root-mean-square; brightness order matches catalog magnitudes (1.79 to 3.31). | VERIFIED | GSR computation: Yale Bright Star Catalog, 5th edition (JSON conversion), positions by PyEphem for 43.30 N, 70.98 W at 00:28:46 UTC September 23. |
| 12 | Searching every star brighter than magnitude 4.5 above the horizon plus the planets, the only full match was the Dipper; the runner-up placed two of the three remaining lights. | VERIFIED | GSR computation, same inputs. |
| 13 | From the match: frame center at bearing about 337 degrees, 17.5 degrees up; roll 5.5 degrees; 0.0282 degree per pixel. | VERIFIED | GSR computation (least-squares fit). |
| 14 | 539 pixels at 0.0282 degree per pixel is 15.2 degrees, the computed Dubhe–Alioth separation that evening; that the 0:30 to 0:52 pair is those two stars is GSR’s reading. | ATTRIBUTED | GSR computation (separation VERIFIED); identification is a GSR reading. |
| 15 | Over 0:04 to 0:11, Dubhe’s brightness ranged from 0.30 to 1.51 times its mean (standard deviation 21 percent); tracker detection in 94 percent of frames; no periodic flash. | VERIFIED | GSR measurement of the original file. |
| 16 | At 8:28 p.m. the five Dipper stars were 17.4 to 27.8 degrees up, bearings 322 to 338 degrees. | VERIFIED | GSR computation, PyEphem with the Yale Bright Star Catalog. |
| 17 | Dubhe’s declination is +61° 45′; from 43.30 N it never drops below about 15 degrees. | VERIFIED | Yale Bright Star Catalog, HR 4301; GSR calculation (43.30 − (90 − 61.75)). |
| 18 | “The brightest twinkling stars can flash with color”; twinkling is largest near the horizon. | VERIFIED | Stephen James O’Meara, “Simply Scintillating,” Astronomy, October 24, 2011. |
| 19 | A waxing gibbous moon stood in the south-southeast and Portsmouth reported broken cloud at 25,000 feet on September 22. | ATTRIBUTED | GSR, October 12 story (U.S. Naval Observatory; Iowa Environmental Mesonet METAR archive). |
| 20 | The September 22 plane: white double flashes 0.12 to 0.13 s apart at 2:10.54, 2:11.96, 2:13.37 and 2:14.79, every 1.42 s (about 42 a minute); dimmer pulses midway with red-to-green ratio about 1.6, against about 1.0 in the white bursts. | VERIFIED | GSR measurement of the original file. |
| 21 | The reader’s narration: the lights are not a plane, drone or star (quoted); they blink and fade out (paraphrased); the last light is probably a plane (paraphrased). | ATTRIBUTED | The reader’s narration on the September 22 original file. |
| 22 | A United Boeing 777 (UAL42) crossed the western and north-northwestern sky at 33,000 feet during the September 22 clip. | ATTRIBUTED | GSR, October 12 story, from the adsb.lol archive. |
| 23 | The reader wrote that one light over Route 108 toward Somersworth “looked like it was spraying something”; no date or time given. | ATTRIBUTED | The reader’s written tip, GSR drop box, October 7, 2026. |
Send original files through Something You Can’t Explain. Identities are protected on request.
granitestatereport@gmail.com · (603) 931-9264 · Granite State Report, 43 Sargent Street, Northfield, NH 03276
Editor’s note. The videos, their narration and their embedded camera data are the reader’s, who asked not to be identified; the measurements of them are Granite State Report’s. Star positions are from the Yale Bright Star Catalog computed with PyEphem; flight and weather details are from GSR’s October 12 story, which drew on the adsb.lol archive, the U.S. Naval Observatory and the Iowa Environmental Mesonet. Federal rules are quoted from the Electronic Code of Federal Regulations. Corrections: Granite State Report corrects verified errors promptly and appends a dated note identifying what changed and when; correction notes are never removed.
