UFORoadmap · version 0.127.3
Future developments.
One rule decides the order: reproduce everything that could have been visible in that sky at that moment — by data where a record exists, by calculation where none does, and never by invention. What follows is a plan, not a promise; the order moves when a real case makes it move.
What the tool already reconstructs, and how each part of it is computed and checked, is on the context page; every source behind it is listed under sources and choices.
Next#
Radar propagation anomalies blocked on data#
Temperature inversions and ducting — the explanation behind a whole family of radar/visual cases. The criterion is a refractivity gradient below −157 units per kilometre, which needs the temperature and humidity profile with height. Checked: Open-Meteo's archive of the reanalysis does not serve pressure levels. Its historical forecast API does, but only from 2022, which is no use to the older cases this is about. The real source is IGRA, the radiosonde archive, with the University of Wyoming soundings and Météo-France's public data as other entry points, and that is what this is waiting on.
Weather balloons considered#
A documented network of launches from the 1940s on — and the Roswell explanation. Datable, placeable, and a shape nobody expects.
Air traffic and contrails before 2022 blocked on data#
From 2022 on, the aircraft around the observer are drawn from adsb.lol's open history, with the trail their engines leave where the air allows one (see the sources page). Before, there is no track at all: Flightradar24's archive is paid and recent. What is left for older cases: published airways and approach procedures (the French eAIP), the calendar of temporary military areas, and the airfields near the observer. For a recent case outside the archive, a KML track exported from a flight tracker is enough to place the aircraft. As for the air aloft, the radiosonde archive the radar anomalies wait on is the only record known for those years.
On the object's side#
A trajectory from directions planned#
Directions are what observers give best. Fitting a great circle through them, for one observer or several, gives the plane of the track and its spread: the COBEPS did it with a stereonet for the Braine-le-Comte case (2015) and found a mean track of N 7.7°. The same fit applies to a track drawn on a registered photograph, which then comes out as an absolute direction to compare with a meteor, a satellite or an aircraft. On a night photograph, the stars themselves can do the registration (astrometry).
Textures on shapes planned#
Blended with the colour rather than replacing it, keyframed and serialized like every other appearance field. Observers describe surfaces — ribbed, matte, mirror-like — and today the format cannot hold that.
Observed entities design open#
Three of the four shipped case files involve one. The open question is whether an entity is a 2D shape (account, like the object) or a 3D decor object. The model says “what reached the observer's eye”, which argues for the shape.
The sound the explanation would have made planned#
An aircraft, a helicopter or a balloon placed by an interpretation would emit a reference sound (for an aircraft, around 100 dB at 20 m), which the tool would carry to the observer to say what they should have heard, to be compared with what they say they heard. The account keeps its relative volume: nobody states decibels. The sound lags behind the image (15 to 22 s for an aircraft at 5,000 m, which has flown more than 6 km meanwhile) and may come from somewhere other than what is seen; it is higher-pitched, hence absorbed faster, ahead of an approaching aircraft; a supersonic aircraft is audible only inside its Mach cone. Played back binaurally on headphones, with the delay, the Doppler shift and the air's filtering, it would become what the observer would have heard from where they stood, and a check would say whether the stated silence fits the proposed explanation (Fedoce, 2014).
Locating a source from several recordings considered#
When several observers filmed, the difference in arrival time of the same sound between their recordings locates the source: acoustic imaging has done it for a long time, and a recording already brings several observers together. It takes enough recordings, clocks that can be synchronised, and a computation that accounts for terrain and wind (Fedoce, 2014).
On the environment's side#
Observing from an aircraft planned#
As already supported from a building or a car. Half of aviation account is given from a cockpit, and a window frame that hides fifteen degrees of sky is part of the account.
Downloaded 3D models, on demand planned#
Real dashboards, real windscreen pillars — to see exactly what angular width a pillar masked. And the inverse case, which matters more: real aircraft seen from the ground and taken for something else, at the angle and with the reflections they actually give.
A video as the scene planned#
A photograph of the place is already there. A video is the same registration with a picture that moves: a texture locked to the timeline, the observer's own footage with the reconstruction over it, frame by frame. And a horizon traced on a picture that fits would hide what passed below it — metres coming back from a photograph the way they come back from a decor crossing.
Reflections of interior lights on glazing planned#
A room lamp, a dashboard light, on a window or a windscreen. The last missing piece of “being inside” — and, on its own, the explanation of a good number of accounts.
Magnetic declination planned#
A bearing taken with a compass is magnetic, and the difference with true north depends on the place and the year: several degrees in France in the 1950s. The recording should say which north a bearing refers to, and the tool correct it from the geomagnetic model.
Sound on its way to the observer planned#
Between the source and the ear, the air absorbs more the higher the pitch, depending on the temperature and humidity the records already give (50 dB per kilometre at 4,000 Hz, at 20 °C and 30 % humidity); wind bends sound down downwind and up upwind; a temperature inversion can duct it or cut it off from the ground, which the upper-air profiles already archived would tell. The loaded terrain gives the shadow zones behind a ridge or an embankment, and the echoes off a cliff. Above all, it takes the noise the observer was listening through: a car with the fan and the radio on is around 75 dB inside, closed windows remove 15 to over 30 dB, and a sound some ten decibels below that ambient noise is not heard. In an example drawn from a GEIPAN investigation, a driver thus passes under a wide-body aircraft taking off without hearing it (Fedoce, 2014).
On the witness's side#
Reconstructing the sky is half the work; the other half is knowing how much each part of an account can be trusted. Witness perception and memory have been measured for decades, on events whose truth is known: re-entries, fireballs, launches, aircraft. What those studies agree on is meant to become checks in the tool. Every lead, its sources and how firmly it is established.
Recognition before description planned#
The idea this tool comes from: Roger Shepard proposed in 1968 to reconstruct a sighting by having the observer recognise it among images, after a free account and before any drawing, since people recognise far better than they describe. The reconstruction exists; what is missing is the protocol: a free account first, then a fixed set of rendered views with decoys, and the choice recorded.
Each statement with its known error planned#
Directions, counts, layout and time are the reliable part; distances, altitudes and speeds are not. On re-entries seen by hundreds of people, from the Condon report's analysis of Zond IV (1968) to the police reports studied by GEIPAN, stated distances are off by factors of ten or more whatever the truth, while the time is right within minutes. A low light is placed at the horizon, an elevation is overestimated, and a stated size is usually too large, though by how much is still disputed.
Candidates weighted by what they usually are planned#
Of more than a thousand night lights that Allan Hendry identified in 1976-1977, a third were stars and planets, a fifth advertising planes, a fifth aircraft. Each common cause has a signature (a silent plane, a light that “left” when it went out, a “hover” that was an aircraft coming straight on), and each has to be tested before a sighting is called unexplained, including with the date shifted, since a wrong date is a common error.
Poher's matrix of known phenomena, brought up to date planned#
In 1977, as GEPAN was being set up, Claude Poher drew up a memory aid for investigators: 33 known phenomena against 55 features of a sighting, each pair rated from “very possible” to “impossible”, the explanations with the fewest impossibilities coming first. Most of those features a recording already holds or computes (time of day, weather, elevation, apparent size and speed, path, duration), so the matrix, already an application on rr0.org, can run on its own and hand the scene a first list of candidates to test. Its ratings are an expert's judgement of 1977, though, and the misidentifications have changed since: advertising planes have become rare while drones, lasers, sky lanterns and satellite trains have appeared. The work is to add the new ones, check every rating against cases whose explanation is known, and weigh least the features observers estimate worst, distance first.
What the observer could hear planned#
No two ears are alike: high pitches are lost with age, down to missing the front of an approaching aircraft, and musical practice helps to recognise a sound. A sound is located within a few degrees straight ahead, far less well in elevation or behind, and its distance is underestimated. The recording should keep what Gaelle Fedoce recommended to GEIPAN: the observer's age and hearing, their usual noise surroundings, and for the sound heard its direction, loudness, pitch, duration and what it sounded like, a sound that recalls another one saying more than a timbre picked from a list. The leads and their sources.
Independent observers, and the versions of an account design open#
Several observers confirm each other only if they did not talk, were questioned apart and reported before the press did; a group watching one light converges on one story. The dated versions of an account, what the observer had read or watched before each one, and whether a detail first appeared under hypnosis, are all part of what the recording should hold.
How the order gets decided#
By which misperception the item lets someone actually test, and by whether the data or the physics for it exists. Something with no record behind it does not get invented — it either gets computed from first principles, or it stays a hand-placed object that the interface admits is hand-placed.
If you have a case that needs something on this list — or something that is not on it — say so. A real dated, placed, sourced sighting moves an item up this page faster than anything else can.