Turns above the North Aegean

What the mission report says

The report is a seven-page mission report headed “Misrep 9329374”, from the 33rd Special Operations Squadron (“Originator (Unit or Squadron): 33 SOS”). USCENTCOM declassified it on 22 January 2026. The aircraft took off from LGLR, Larissa in Greece, at 23:39Z on 26 October 2023, “PROCEEDED TO FRAGGED TASKING”, arrived on station at 03:42Z, returned to base at 10:11Z and landed at OJMS, in Jordan, at 13:09Z. The timeline explains the change of base: “[…] TRANSITED AND LANDING AT OJMS TO REPLACE THE LIGHTNING LINE THAT LANDED AT LGLR YESTERDAY.” The aircraft’s type, tail number and call sign are blacked out, and so are the operation and the unit it supported. It carried a targeting pod listed as “AN/DAS-4”, with “Additional Avionics: AH/GMESH” and “Data Link: LINK 16”; its primary sensor was full-motion video.

The sighting takes one line of the narrative, “AT 0035Z, […] OBSERVED 1X POSS UAP (SEE UAP 1).”, three hours before the aircraft reached its station, and fills the UAP section on pages 6 and 7. Its account in full:

“AT 0035Z, […] WAS EN ROUTE TO THEIR TARGET WHEN THEY SPOTTED A UAP FLYING JUST ABOVE THE SURFACE OF THE OCEAN WATER. THE UAP TOOK MULTIPLE 90 DEGREE TURNS AT AN ESTIMATED 80 MPH. AT 0038Z, […] LOST THE UAP FROM THEIR FEED.”

The description field reads “SEEMINGLY CIRCULAR, TOO SMALL TO MAKE OUT DETAILS”. Initial contact is logged to the second, “270035:12ZOCT23”. The fields around it give “UAP Maneuverability Observations: Sharp 90 degree turns” and “UAP Response to Observer Actions: NONE”, the object as “Solid”, with propulsion “UNK”, “NO” on intelligent control, “NONE” on signatures and “UNK” on advanced capabilities, and the observers’ assessment as “Benign”. The aircraft’s heading is given as “SW”. Its height and speed are left blank, and so are the object’s height, the number of objects, any third-party observers and whether anything interrogated the object. The speed has its own field, “Kinetic Velocity: 80 MPH”, marked “Estimated”.

Where it was

The report gives three grid references in the military grid (MGRS) for the event: the aircraft’s “Friendly Aircraft Location: 35SKD59[…]53[…]”, the “First Coordinate: 35S KD 95[…] 53[…]” and the “Last Coordinate: 35SKD90[…]55[…]”. Each has two boxes marked “1.4a”. We measured the boxes against the digits beside them: each is as wide as three digits and its own margins. Two digits therefore survive in each half of the reference, the tens of kilometres and the kilometres, and every position is known to a square of 1 km.

By our conversion, taking the middle of each square, the object was first seen at about 39.31°N 24.63°E and last seen at about 39.32°N 24.57°E, 5.4 km to the west-north-west. The aircraft was at about 39.30°N 24.21°E, 36 km due west of the first position. Both positions of the object lie in open water between Agios Efstratios and Skyros, 35.6 and 38.6 km from the nearest land. Both are marked “Estimated”, each with a radius of “5” and no unit. The map pin stands on the first.

The point where the crew was to begin its tasking, “36S YC 40[…] 57[…]”, lies in Lebanon by our conversion; the crew saw the object on the way there. The publisher’s title, “Greece, October 2023”, follows the aircraft’s base.

What the recording shows

The recording is DOW-UAP-PR34, in the release 01 video bundle as DOD_111689011.mp4 and on the publisher’s video page. It runs 2 min 57 s, 1920 × 1080 pixels at 30 frames a second. Every overlay field that would give time, position, range or sensor settings is blacked out; what remains is the crosshair, four corner marks around it and a small “N” marker that circles the crosshair. No frame shows the horizon, land or cloud: the camera looks down at the sea throughout. A dark round patch in the middle of the picture, vertical streaks and a few specks stay in place in every frame; they belong to the sensor. Single dark grey frames interrupt the picture here and there, and every one to four seconds after 2:25. By our viewing of the file:

• 0:04. A small white point enters at the lower left.
• 0:04–0:47. It crosses the frame again and again, changing direction sharply (next section).
• 0:15.25–0:15.75. Two points are in view: one stays put, the other leaves to the right, and only that one is seen afterwards.
• 0:36.2–0:55.5. A tracking box stands on the crosshair, not on the point.
• 0:58.6–2:20.87. A small reticle holds the point.
• 2:02.1–2:20.5. A brighter, grainier window in the middle of the picture: the contrast filter.
• After 2:21. We did not find the point again.

Turns against water that does not move

The table lists the sharp changes of direction we measured by following the point frame by frame through the first 47 seconds. Directions are on the screen; speeds are in pixels of the 1920-pixel frame per second.

Time	Before	After	Change on the screen
0:05.8	left, about 340 px/s	right, about 85 px/s	reversal
0:12.0–0:12.8	right and up	left	a curve of about 140°
0:17.0–0:17.5	right, slowing down	left and up, about 75 px/s	about 140°
0:23.2–0:23.8	left and up, about 80 px/s	right and up, 120–150 px/s	about 120°
0:26.8–0:28.0	right and up, almost stopping	down, up to 220 px/s	about 120°, through a near stop
0:30.0–0:30.7	down, about 95 px/s	left and up, about 190 px/s	about 100°
0:32.2–0:32.8	left and up, about 190 px/s	right and down, 180–250 px/s	reversal
0:35.1–0:35.5	right and down, about 175 px/s	left, 250–290 px/s	about 160°
0:37.9–0:38.8	left, about 270 px/s, stopping	right, slowly	reversal
0:42.1–0:42.4	down, stopping	right and up	about 140°
0:46.0–0:46.9	right, stopping	left and up, about 60 px/s	about 120°

These are angles on the screen, not over the sea. The camera looks down at the water at a slant, and the angle at which it looks is blacked out, so a turn on the water can appear wider or narrower on the screen than it was. The change closest to a right angle on the screen is the one at 0:30.0, about 100°.

A point can swerve on the screen without swerving at all, if the camera swings while it watches. To test that, we used the sea itself. We traced thin strips of the picture through time, thirty frames a second, as space-time slices, or kymograms: a strip of sea that stays put draws lines parallel to the time axis, one that moves draws slanting lines. We left out the part of each strip that the point itself was crossing. By our measurement, from 0:04 to 0:47 the texture of the sea does not move across at all and moves up or down by at most 1 pixel per 0.2 seconds, 2 in two single moments, while the point crosses the frame at 80–400 pixels a second and turns. The method does see movement when there is some: the slices show a drift of 1 pixel per 0.2 seconds at 0:13.4–0:14.4 and 0:35.8–0:38.0.

That the sea stays put in the frame means the sensor held its aim steady on the sea surface; it does not mean the aircraft stood still. Our finding is that the turns are the point’s own movement against the water in the picture, not a swing of the camera. That is all it gives: without a range and a field of view it says nothing of the point’s speed or size.

The method has limits. A single blank frame, as at 0:28.5, breaks the comparison for a moment. At 0:15.25–0:15.75 two points are in view, and we count no turn there. The “N” marker, which turns smoothly clockwise around the crosshair from −17.5° to +53.8° over the recording, shows no jump at any of the turns; on its own it would not settle the question, since in a narrow field of view a fraction of a degree carries the point across the screen.

82 seconds in the reticle

The publisher’s description speaks of a lock. By our viewing, two different symbols appear, and only the second holds the point.

At 0:36.2 a box appears on the crosshair and grows to about 126 × 130 pixels, with a blacked-out label beside it. It stays on the crosshair while the point is elsewhere: until about 0:47.5 the point lies outside it, at 0:37–0:44 some 120–340 pixels to its left, and only then drifts into it. The box is not a lock on the point.

At 0:55.6 the box gives way to four corner brackets, 114 × 116 pixels around the crosshair, which close in on the point: about 50 pixels across at 0:57.1, 34 at 0:58.2. From 0:58.6 a small reticle holds the point, 10–25 pixels left of and 20–35 pixels below the centre of the crosshair, until 2:20.87. The point shows plainly inside it at 1:01–1:10 and grows fainter towards 2:00. The contrast filter comes on at 2:02.1 and goes off at 2:20.5; the point is still visible at 2:04 and not at 2:15. The reticle slides about 15 pixels to the right and is gone at 2:20.90. The lock lasts 82.3 seconds, from 0:58.6 to 2:20.87.

No turn happens during the lock, by our measurement. The point stays in the reticle; the sea does not move across and drifts down the screen, mostly by 0–2 pixels per 0.2 seconds and at moments faster, without changing direction. Against the water the point therefore moves one way only, up the screen, and slowly compared with the minute before, in pixels that nothing on the recording converts into metres.

80 mph without a range

The speed appears twice in the report, as “AT AN ESTIMATED 80 MPH” in the account and as “Kinetic Velocity: 80 MPH” with “Kinetic Velocity Accuracy: Estimated”. The report does not say how it was estimated. Nothing measured a range: the field “Observation Interrogation of UAP (yes/no; if yes list sensor(s) and returns)” is blank, and so are the aircraft’s height and speed and the object’s height. A camera without a range gives angles, not speeds. The recording, with every data field blacked out, carries neither range nor field of view, so its pixels per second cannot be turned into miles per hour.

The grid references allow one rough check, by our calculation. The first and last positions are 5.4 km apart. Initial contact was at 00:35:12Z, and the crew lost the object “AT 0038Z”, sometime between 00:38:00 and 00:38:59: 2 min 48 s to 3 min 47 s. That is an average of 53–72 mph, or 43–85 mph allowing for the 1 km squares. It neither confirms nor contradicts the 80 mph. Both positions are “Estimated”; their radius of “5” has no unit, and if it means kilometres or nautical miles it is larger than the distance moved. The report does not say how the positions were obtained. And a straight line between two points understates the length of a path with turns in it.

Two days later, straight towards Crete

On 29 October the same squadron filed DOW-UAP-D35, with its recording DOW-UAP-PR35. The aircraft had taken off from Larissa at 15:04Z on 28 October, worked the same tasking point and was on its way back: “AT 0811Z, […] WAS RTB WHEN THEY SPOTTED A UAP FLYING JUST ABOVE THE SURFACE OF THE OCEAN WATER. THE UAP FLEW STRAIGHT ABOVE THE OCEAN TOWARDS LANDS. AT 0811Z, […] LOST THE UAP FROM THEIR FEED.” (p. 7). The two reports come from the same template, list the same targeting pod and describe the object in the same words: “SEEMINGLY CIRCULAR, TOO SMALL TO MAKE OUT DETAILS”, “Solid”, “Benign”. Side by side, from what both reports say and both recordings show:

	27 October: D33 and PR34	29 October: D35 and PR35
In the mission	“EN ROUTE TO THEIR TARGET”, 00:35:12Z	“WAS RTB”, on the way back, 08:11Z
Movement	“Sharp 90 degree turns”	“NONE”; “FLEW STRAIGHT … TOWARDS LANDS”
Speed	“80 MPH”, estimated	“30 MPH”, estimated
Lost	“AT 0038Z”	“AT 0811Z”, within the same minute
Where, by our conversion	open sea, 35–39 km from land	1–2 km off the north-eastern tip of Crete; the last position on land
Recording	2 min 57 s, 1920 × 1080, 30 frames a second	24 s, 800 × 444, 10 frames a second
Background	sea only	cloud over the sea, then the coast
Reticle	0:58.6–2:20.87	none
Against the background	many sharp turns	a straight line

By our conversion, D35’s first and last positions lie 2.2 km apart, both within the minute 08:11Z; at 30 mph that distance takes about 2 min 47 s. The same caveats apply: both positions “Estimated”, radius “5”, no unit.

PR35 runs 24 seconds. After a change of view at 0:02.2–0:02.6 it shows white cloud over the sea and, near the crosshair, a small white point two or three pixels across. The sensor keeps that point within about 90 pixels of the crosshair from 0:03 to about 0:17, while the clouds, and with them the whole scene, slide some 300 pixels across the screen. On the screen the point’s path kinks at 0:04.7–0:05.1 and at 0:05.8–0:07.3, exactly when the whole scene jumps: at 0:05.8–0:07.3 the clouds move up by about 84 pixels. Measured against the clouds, the point moves in a straight line at an even rate, about 34 pixels a second to the right and 22 down (0:03.0–0:08.7), towards the side from which land comes into view. PR35 thus shows how the camera’s own movement makes turns on the screen; PR34 shows turns while the camera’s aim stays put. We last see the point at about 0:17; the picture darkens at 0:17.5–0:18, and land fills the middle of the frame at 0:19–0:20.

Two things in PR35 go unmentioned in the publisher’s description: a white, lobed shape with small fragments, at the upper right at 0:03, which drifts with the clouds (3–5 pixels a second against them), passes under the top black box at 0:12.3–0:12.9 and leaves the frame at the top at about 0:14; and a second small point that crosses the view at 0:12–0:17.

What the release says about objects low over the water

One report in the release explains a low-flying object over water. DOW-UAP-D55, a one-page summary from 18 November 2016, has a P-8A crew watching “an unidentified low-flying object 55 nm northwest of Latakia from an unknown origin, traveling at approx 500KTS on a southeasterly heading outbound from KCTG, for ~2 minutes”. Its timeline calls it “a possible missile launch” that “appeared to be in sea skim mode”, and the task group’s comment ends: “it is assessed to be standard activity consistent with the assessed activity of the KCTG” (p. 1). The report does not spell out KCTG; the map on the same page marks Russian warships, among them “(RUS) KUZNETSOV CV-063”. That object flew straight at about 500 knots.

Other reports describe objects over water without explaining them. DOW-UAP-D27 has an object “FLYING STRAIGHT JUST OVER THE WATER AT SPEED” (p. 6) and offers a “POSS REFLECTION FROM THE OBJECT IN THE WATER” for the pole beneath it, not for the object (p. 7). DOW-UAP-D101 has “two approximately four-foot diameter cold orbs between 0 to 20 feet above the surface of the water” (p. 2). DOW-UAP-D38 has a crew following an object that “appeared to make erratic moments above the water” and noting that “the sensor operator was continuously manipulating the sensor to maintain eyes on the object. This is apparent by the waves of the water in the background being visible and not being visible” (p. 1): a remark about the picture, not an explanation of the object. The only other explanation the release gives for small objects crossing a sensor’s view, “ALL 4X REMAINING UAP FIT CLOSER TO THE PROFILE OF POSSIBLE BIRDS” (DOW-UAP-D10, p. 6), concerns objects over land.

Nothing in the release explains an object like the one in D33.

What the publisher got wrong

Both recordings’ pages give the date taken as “10.01.2023”, 1 October 2023 in the American order, and both identifiers begin with 231002. The reports put the events on 27 and 29 October.

For 0:07–0:19 of PR34 the description has the point moving back and forth “as the sensor pans to track it”. By our measurement the sea does not move in the frame at that time: the camera does not pan, the point moves.

For 0:20–1:00 it says the point “remains generally centered within the sensor field-of-view”. By our viewing the point reaches the upper part of the frame at 0:27 and is 200–340 pixels left of the centre at 0:37–0:42; it stays near the centre only from about 0:45.

It gives the reticle as “01:00-02:01” and has it drop its lock at “02:22”. The reticle holds the point from 0:58.6 to 2:20.87, through the contrast filter, and the box that appears at 0:36.2 is not on the point.

Why this scores where it does

Witnesses score 2: one crew, trained to watch through its sensors, which the report does not name or count, and no one else. Radar scores 0: nothing interrogated the object. Optics scores 3: a recording of nearly three minutes shows the point for over two of them, lets the camera’s own movement be checked against the sea and holds the point for 82 seconds in a reticle; but every overlay that would give range, field of view or time is blacked out, and the object is a point with no shape. Physical trace scores 0.

Multi-channel scores 1: the report and the recording come from the same crew and the same sensor, so they are one channel, not two; D35 and PR35 describe another event. Documentation scores 3: a full mission report with times to the minute, initial contact to the second, and grid references to 1 km for the aircraft and the object; but the aircraft’s height and speed and the object’s height are blank, and the radius of each position has no unit. Anomaly scores 1: the turns are real in the picture, not the camera’s, but without a scale neither the speed nor the size of what turned is known; the report itself answers “NO” on intelligent control and “UNK” on advanced capabilities, and calls the object “Benign”. Data today scores 2: the report and the recording are public, but what would test the claim, the range, the field of view and the sensor’s telemetry, is blacked out or was never recorded.

That puts the case at 1.4 overall, class D, with the status insufficient data.

What would settle it

The overlays blacked out on the recording. The field of view, the slant range, where the sensor pointed and where the aircraft was, frame by frame, would turn the pixels of the turns into metres and the 80 mph into a measurement.

The grid references without their redactions, and a unit for the radius. They would show where the object was, and whether “First” and “Last” are its positions at all.

The aircraft’s height, left blank in the report. With the pointing angle it would give the range.

The data link. The report lists “Data Link: LINK 16”; the picture on that link for 00:35–00:38Z might show whether anything else held the object.

AARO’s file on the event, if it has one and if it concluded anything.