A small, soft glow with a bright point at its edge, hanging nearly still in a dark sky, is easy to mistake for something unexplained. It is also a common sight. Often called a "jellyfish" plume, this is a later stage or “upper stage” of a rocket launch.

Enigma #277075

The image above shows an upper stage still firing, roughly 6-9 minutes after lift off and more than 150 km above the ground. The bright point is the engine's hot exhaust, the light that shapes and backlights the cloud around it. The pale halo is exhaust gas, vapor and soot expanding rapidly into near-vacuum, where nothing holds it together. It blooms outward like a flower or a nebula. There is little or no trail. The stage is far away and moving mostly along the line of sight, so it reads as a compact glow rather than a streak.

Timing is what makes it visible. A rocket is easy to miss in full daylight, while the plume loses its distinctive appearance in full darkness. The best window is twilight, when the ground has gone dark but the sun is still shining on the sky overhead. The stage is high enough to catch that sunlight, so its thin exhaust cloud lights up against a dark sky like a lantern. That is why reports cluster around sunrise and sunset, and why most come from witnesses hundreds of kilometers from the pad, and why many come from witnesses hundreds of kilometers from the pad, along or beyond the rocket’s downrange path.

Enigma’s Sighting Data

Identifying these sightings required more than a simple database search. The reports are spread across a dozen shape tags, including light, fireball, circle, sphere, oval, cigar, changing and unknown, so no single field could isolate them. Searching those fields alone would either miss reports or return thousands of unrelated sightings.

Instead, we searched witness descriptions for clues associated with upper-stage burns, including halos, smoke rings, expanding clouds, and objects that split or fade. We then reviewed the reports manually.

Using this approach, we identified 68 upper-stage sightings with video. The footage gives us the strongest basis for comparing shape, timing and location against known launches, so we treat these 68 as the floor, not the full count.

Rocket launches are usually easiest to see near the launch site and along the flight path. In the U.S., that means Florida's Space Coast, where launches from Cape Canaveral climb northeast over the Atlantic and can be seen along the East Coast, and California's Central Coast, where launches from Vandenberg head south over the Pacific and are visible across Southern California and into Arizona and Nevada. Most witnesses are therefore within a few hundred kilometers of one of these launch sites.

Our sightings follow that pattern, but with one important difference: a share of reports come from much farther away. Florida and California together account for about 40% of the set, and these sightings are the closest matches, often occurring within minutes of liftoff and directly under the flight path. Texas, however, accounts for roughly a third on its own, even though none of those sightings match a launch from Texas. Those sightings line up with launches 1,300 to 2,300 km away, with witnesses reporting the glow 1.5 to 2 hours after liftoff, roughly the time it would take the rocket to complete an orbit. Oklahoma, Arkansas and Mississippi show a similar pattern. The distribution suggests that once an upper stage reaches sufficient altitude, its visible burn is no longer confined to the area around the launch site. A single launch can produce sightings across a much wider region, sometimes thousands of kilometers away.

The 68 sightings with video are only the reports we can investigate most directly. The database contains many more reports describing similar imagery that we cannot confirm. Some witnesses began filming after the burn had faded; others captured footage that was too faint or brief to distinguish a plume from an indistinct light or smudge. Many reports have no media at all and exist only as witness descriptions, so we exclude them from the confirmed count.

Based on the descriptions alone, we believe the total number of potential upper-stage reports is closer to three times the number we were able to confirm.

Matching Sightings to Real Rocket Launches

To validate the 68 sightings, we looked beyond our own data and checked each one against independent records of real rocket launches. We used the date and city recorded for each sighting in the Enigma database, converting the date to the sighting’s local time zone. We then compared each sighting against public launch records and news coverage for recent launches not yet included.

A sighting counts as a match only when a launch occurred within one day of the sighting. We then assessed the geographic relationship between the sighting and the launch:

Strong match: The sighting occurred in the same state as the launch site or within the launch’s flight path.

Weak match: A launch occurred within the one-day window.

Of the 68 sightings, 66 matched a real launch within one day. Thirty-five were strong geographic matches, while thirty-one were considered weaker matches. Two sightings had no corresponding launch within the one-day window.

Just because a sighting has a source doesn't make it any less strange to the person who saw it. What these matches change is the question.

The mystery, in these cases, is not necessarily what was in the sky, but how unfamiliar something familiar can look from the ground. An upper stage can be hundreds or even thousands of kilometers from its launch site when its exhaust becomes visible from the ground as a glowing, expanding cloud. To an observer with no reason to connect the two, there is no rocket, no obvious trajectory and often no visible sign of a launch at all. There is only a strange light hanging in the sky. Our matches show that many of these seemingly inexplicable sightings have a traceable origin.

Below, we’ve collected some of the most compelling media from these sightings, showing just how strange an ordinary rocket launch can look from the ground.