The Trindade Island photographs, 16 Jan 1958

Martin Shough

A Preliminary Study of Cloud Displacements
(Feb 2004. Last update 25 Feb 2004)

(index) Part One (Go to Part 2>)

The two photographs considered here are the prints usually believed to be the first and last in the sequence, labelled P1 and P4 by Kentaro Mori. (A total of six photographs were reportedly taken, but two - between P3 & P4 - failed to show the object. Properly therefore these images are P1 & P6, but the two 'empty' frames are not available for study.)

Fig.1 Scans of P1(left) & P4 rendered at very high contrast (courtesy Kentaro Mori)

Each of the four prints contains a large amount of subtle cloud detail. Superficial comparison of P1, P2 & P3 reveals no obvious inconsistencies. However, comparison of P1 & P4 at high contrast reveals a marked difference in the apparent structure of the clouds. This comparison is especially sensitive because of the large area of overlap between the two images. Although it is true that cloud changes between the first and last photographs could reasonably be expected to be greater than those between any other pair, there is a limit to how long the elapsed interval can be whilst remaining consistent with the testimony. The question therefore arises whether P4 has the same date-time group as the sequence P1-P3. If not then fraud is obviously suspected.


Mori tentatively concludes:

The cloud differences seem incompatible with the story of the photos as commonly given. At the very least, the time lapse from the first to the last shot is on the order of minutes, not seconds, and very strong winds account for the changes. At most, the first and the last shot were taken on different days and/or, more probable, the difference in the sky appearance is due to the photographic tricks used to create the UFO.

The present analysis is based on the assumption that, whatever the nature of the 'UFO' image, the prints are unretouched images of genuine cloudscapes (see box below).

Intricate shadow detail on the island shows no perceptible difference in the solar angle between P1 & P4, so the time of day appears to be the same. Therefore, if the difference in cloud structure is incompatible with changes that might occur within a short interval of time then the pictures must have been taken at the same solar time on two different days.

On the other hand, if P4 contains features which can be positively identified as developments of features in P1 then the unchanging solar angle would indicate that such development must have taken place in a short interval of time, consistent with the testimony.

Given the nature of cloud patterns in general, and the fact that P4 in particular shows internal evidence of strong winds at altitude, an exact identification of common features in P1 and P4 is perhaps not realistic. But one might expect to find, at least, a suggestive similarity. In this case the balance of probability would be affected by whether the change indicated is within the probable range of behaviour of the tropical atmosphere over a S. Atlantic island in January.

At present no upper-air data are available for the date and place of the photographs. Therefore the purpose of this exercise is limited to a test of the hypothesis that the two photographs are inconsistent with reasonable and possible meteorological conditions and with the testimony.


The cloud structure in P4 might differ as a result of some special treatment applied to the negative and/or as a result of, say, airbrushing of an intermediate print used to create a photomontage.

A premeditated double-exposure has been widely considered as a possible explanation. In this scenario, a series of UFO images was prepared by Barauna in his darkroom before the trip, so that the images photographed against a blank matte would appear in the middle of genuine cloudscapes later. However this does not help us to understand why P4 would not be a genuine cloudscape taken from the deck of the Almirante Saldanha.

It is known that, according to Barauna, at least some of the six negatives were treated (probably by 'reducing' with potassium ferricyanide, a commonplace darkroom procedure) to compensate for slight overexposure. But such a treatment produces a difference in overall density values which would not alter the basic cloud structure (though it might of course bring out fugitive detail) and does not of itself constitute a "photographic trick" in the sense required.

Another possibility is that the P4 sky was 'created' in the course of producing a later darkroom photomontage using images taken during the trip. In this case the opportunity for inconsistency arises. But if this was done, and if the P4 sky is very inconsistent as a result, one wonders why Barauna departed from the method consistently used to create P1, P2 and P3. The general light conditions, the landscape shadow detail, the solar angle, the sea conditions and the foreground detail (not shown here) are all consistent with a series of four photos taken at the same location at the same time. Why should Barauna deliberately doctor the sky of P4 - in what would have to be a complete and well-executed repainting, not just a bit of haphazard retouching - so as to make it meteorologically inconsistent?

The hypothesis that he used a photo taken at the same solar time on (say) the previous day in different weather would seem much more plausible. This would apply either to premeditated double-exposure or to a later photomontage. This is why the images are treated here as authentic cloudscapes and the problem addressed is one of possible meteorological inconsistency.

Even if the skies are perfectly meteorologically consistent, of course, this would not mean that the 'UFO' images themselves are genuine. However there are serious problems with both double-exposure and photomontage explanations based on quantitative physical-optical and photometric arguments that are beyond the scope of this analysis. If P4 is seriously anomalous and indicates a hoax it still does not mean that we know how the hoax was accomplished.

Cloud types

The division into cloud types is not necessarily very clear, but in very broad terms both photographs show a high altitude striated layer resembling cirrus or cirrostratus with a band of lower altitude cloud resembling cumulus humilis above the horizon in the middle distance.

Cirrus develops characteristically at heights between 20,000'-40,000', probably tending towards the top of the range in tropical areas where the freezing level will be higher. The freezing point decreases with pressure and cirrus forms at -40C or less. Generally the base height tends to increase with reducing latitude, roughly parallelling the mean tropopause height. At latitudes around 70 to 80 cirrus strata form at mean heights around 25,000'or less; near the equator they tend to form at 35,000'- 40,000'. (Although the photographer himself described the cloud as cirrus it is possible that the pictures show the effect of more moderate winds on a dissolving altostratus/altocumulus layer below 20,000' as mentioned later.)

The type of 'fair weather' cumulus suggested occurs typically within the first few thousand feet and tends to indicate that the air below the condensation level is convectively unstable, whilst the air above is convectively stable. Instability in this sense refers to upward or downward accelerations of air parcels due to 'runaway' changes in buoyancy relative to the surrounding air. If a parcel is compressed in descent and becomes denser than its medium it will tend to sink further; if it expands in ascent so as to become lighter than its medium it will tend to rise. In these conditions disturbed air parcels will tend to move further and further from equilibrium. The opposite effect occurs when the induced acceleration is in the opposite direction to the initial displacement, and the air is said to be convectively stable, leading to horizontal stratification. The cumulus in P4 appears to be lower (and/or more distant) and more ragged than in P1. If the photographs are of the 'same' sky then this difference might be suggestive of convective instability, triggered possibly by subsidence in the lee of the mountains.

One other significant point is the presence of what appear to be local foreground clouds, probably also associated with the mountain topography. These are much more evident in P1 than in P4. Indeed much of the high cirrus in P1 is obscured by them, but enough can be seen to indicate that: a) In P4 the texture of the cirrus is more striated in appearance than in P1, suggestive of higher winds; b) The striations faintly visible through the foreground cloud in P1 make the same roughly 10 angle with the horizon as those in P4.

(a) Comparison of cirrus features

The images used for this preliminary study are those used by Mori, scanned from first generation prints, rendered at high contrast to bring out cloud detail. The approximate regions selected for closer study are outlined in blue below. This choice is not wholly arbitrary, since large areas of P1 are partially obscured by veils of foreground cloud. Only in the upper central region indicated is the cirrus structure visible in P1 with any clarity.

Fig.2. The approximate regions of cirrus selected for comparison

Each of these regions contains the darkest (clearest) sky area on the print, in approximately the same relation to a distinctive three-pronged or arrow-shaped patch of cloud. The two regions are shown below, rotated to a common horizon angle.

Fig.3 Top: P1(a). Bottom: P2(a). Contrast enhanced. Approx. 27 degree x 4 degree sections from upper central sky area, horizon aligned. P2 slightly softened by two steps (ImageFolio processing).

The pair of images below has been further contrast-enhanced. The horizontal stratification is more distinct in P4(a), as though 'smeared' by winds; nevertheless it is quite easy to interpret P4(a) as a development of the structure visible in P1(a).

Fig.4. P1(a) & P4(a) at extreme contrast

The patches of cloud in the bottom left-hand corner of the P1(a) image are identifiable from the uncropped print in Fig.1 as mountain cloud in the foreground, not associated with the high cirrus layer. In P4(a) the region of cirrus concerned appears completely free of obscuration. The images below show the approximate effect of superimposing the same foreground cloud on P4(a).

Fig.5. A suggestion of how P4(a) might appear with similar cloud intervening in the foreground.

(b) Comparison of cumulus features

Fig.6. The approximate regions of cumulus selected for comparison


Fig.7. Regions of P1(b) & P4(b) rotated to the same horizon angle


Fig.8. P1(b) & P4(b) at increased contrast


Fig.9. P1(b) & P4(b) at extreme contrast

The cloud group in P4(b) is at a lower elevation and/or increased distance and appears to merge into a bright horizon haze, making its undersurface difficult to discern. It is more broken than the group in P1(b) and shows a diagonal skewing suggestive of a vertical wind shear.

Inferences from comparisons

a) Cirrus match: In my opinion it is reasonable to say that this match can be considered good. The patterns compared in Fig.4 above have several features in common although relative proportions of some of these features have changed somewhat. These are identifiable in the approximate outline tracing in Fig.10 below.


As mentioned, the arbitrariness in the selection of regions of the two prints for comparison is much reduced by the fact that only in the selected region of P1 is the cirrus layer visible with any distinctness through the obscuring foreground cloud. The usable region of P1 is forced on us. Although there is a relatively large area of cirrus visible on P4, the probability of finding a region of similar pattern by chance must be influenced by the fact that the number of degrees of freedom is reduced in this way.

The overall angular displacement of P4(a) from P1(a) is in the region of 12 degrees. Together with the more developed lateral streakiness in P4(a), this would indicate strong winds at altitude, possibly rising during the interval between the photos. It remains to be seen if such a displacement is consistent both with the likely meteorology and with the time lapse implied by the testimony.

b) Cumulus match: Clearly there is little or no arbitrariness in the selection of regions of cumulus, inasmuch as almost all of the cumulus in both P1 and P4 falls inside the regions selected. But at first sight the features appear less distinctive. There is a band of broken low cloud in a similar area of both prints, but beyond that the prospects for a detailed match look poor.

Nevertheless, at extreme contrast in Fig.9 the gross structure of both patterns is revealed as broadly similar. This structure is brought out in the approximate outline tracing in Fig.11, a central denser mass (A) flanked by smaller parcels of broken cloud (B & C).


The bulk lateral displacement of the cumulus group from P1(b) to P4(b) appears to be much smaller than that of the cirrus features in P1(a) and P1(b), only a few degrees; but the bulk vertical displacement of both is comparable, and if anything is larger in the case of the cumulus.

Go to Part 2>