# Could The Gimbal Video Show an Atlas V Launch?

Some points here:
The launch of the rocket was at 0104 UTC on the 21st, which means 2004 EST on the 20th.
The inclination of the launch was 16.7 degrees. You can calculate the launch azimuth by using the latitude for Cape Canaveral (28.4 degrees) and the desired inclination. A rough calculation put this at an azimuth of about 95 degrees, which is towards the ESE away from the Jax op areas.
While it is an interesting theory, one would need the actual time of the video, and the plane would have to have been facing towards the South or Southeast for the rocket theory to be plausible.

Some points here:
The launch of the rocket was at 0104 UTC on the 21st, which means 2004 EST on the 20th.
The inclination of the launch was 16.7 degrees. You can calculate the launch azimuth by using the latitude for Cape Canaveral (28.4 degrees) and the desired inclination. A rough calculation put this at an azimuth of about 95 degrees, which is towards the ESE away from the Jax op areas.
While it is an interesting theory, one would need the actual time of the video, and the plane would have to have been facing towards the South or Southeast for the rocket theory to be plausible.
Very true, but given we don't know when the video was taken, nor (if I'm not mistaken) the direction the F-18 was facing, the Atlas launch is consistent with the data we actually have.

According to the DOD:
The three videos (one from 2004 and two from 2015) show incursions into our military
training ranges
by unidentified aerial phenomena. These videos are copies of official Navy
footage taken by naval personnel conducting training missions in controlled military
airspace.
Source, see page 5 of the pdf file.
I don't think the ATLAS V came into their military training range.

According to the DOD:

Source, see page 5 of the pdf file.
I don't think the ATLAS V came into their military training range.
Surely not, but the ATLAS V has a much stronger IR signature than, say, a military jetplane, and it would have been conceivably bright enough to be visible and produce glare into the sensitive F-18 FLIR even if hundreds of miles away. This could be checked easily if we knew the exact time the Gimbal video was taken (*), but until we can do that it remains the best bet on the table, as I see it.

(*) Another, much more complicated way I can think of is to calculate somehow if the IR output of an Atlas V would be enough to induce glare into the F-18 Flir. Not only difficult but there's no way I can see to even gather the basic data about the Flir sensitivity (which is surely classified information) and the Atlas V IR signature (probably not classified, but hard to get anyway). Nah, what we need is the time and date the video was taken (and possibly the airplane heading). In the meantime, I'm tempted to place a small bet...

The DOD states these objects were in their military training range. This is impossible to conclude from the ATFLIR images alone. Since the SA monitor also showed the objects I assume there were more sensors on them than just IR, otherwise the DOD would not make such a firm statement.

The DOD states these objects were in their military training range. This is impossible to conclude from the ATFLIR images alone. Since the SA monitor also showed the objects I assume there were more sensors on them than just IR, otherwise the DOD would not make such a firm statement.
I really don't remember what DoD says about the location of the object, but I have no reasons to doubt what you say. So I agree with you, your observation makes the Atlas V hypothesis less probable that it was before your observation is taken into account. I could agree that we can expect the Dod statement with 75% probability if the Gimbal was inside the test range (then it was no Atlas) and 25% otherwise, because there are many reasons why DoD says what it says (ie.: they may simply be wrong). It's 3:1 odds against the Atlas (75% divided by 25%).

But the Atlas V hypothesis starts with very high probabilities: we're 1) 100% sure it was launched in a timeframe consistent with what we know about the video, 2) I think we can be pretty confident the F-18 FLIR would have seen a glare if pointed to the Atlas V engines even if the distance was large, and 3) we can be pretty confident that IR glare in the FLIR appears as the Gimbal object does. I'd give an 80% probability to assumption 2) being true, but let's say from 80 down to 30%. For assumption 3) I'd say 100%, but let's say 100-50% instead. So with my preferred numbers the probability of Atlas V being the Gimbal is 80%, with the lowest numbers I can possible conceive with the data I have the probability is 15%. The next hypothesis on the list is... uh.. what is it? this is already a difficult question... another F-18 or another nearby jetplane? This looks the best candidate to me and it has long be discussed and apparently refuted (not that I remember the details), do you agree on a probability of 1% that the Gimbal was another nearby jetplane?

So we find (if you agree with my numbers, if you don't, plug in yours and tell me what you find) that the Atlas hypothesis is from 80 times to 15 times more probable than the next leading candidate (another nearby jetplane). Your 3:1 odds observation about the DoD statement would change it to about 26.7 - 5 times more probable (80/3 and 15/3). Any other hypothesis I can think of (expecially from secret Earth technology 'upwards') has probabilities much lower than 'another jetplane', and so with the data I have I still bet on the Atlas. At five to one for sure, but I could rise the stakes up to 26:1 (and still expect a 0.67 gain!).

Last edited:
Mauro loves his probabilities.

I'll go 100% that more information would be nice - and maybe 50% that Mick West is right this moment busying himself investigating this new angle.

Mauro loves his probabilities.
To be honest.. the course I struggled most with when in University was 'Theory of probabilistic phenomena'! It still gives me nightmares ~40 years later But trust me, never place a bet on anything if you can't calculate probabilities, not even on Gimbal! And no weird theories are needed, just some multiplications and divisions

Last edited:
never place a bet on anything if you can't calculate probabilities
except you're pretty much guessing all of the inputs to your calculation

except you're pretty much guessing all of the inputs to your calculation
Of course I'm guessing, what else could I, or anyone else, do without data? But I try to do justifiable guesses, and at least I have tried to put some numbers onto the the whole thing, which I think is at least a (big) step forward. I was also pretty clear:

(if you agree with my numbers, if you don't, plug in yours and tell me what you find) [this was meant for Itsme, but it's valid in general]

Put your guesses and your list of hypothesis into the calculations and let me know what you find. For example do you think 30% is too high a probability that the FLIR sensor would really experience glare if pointed to the Atlas V engines from some hundred miles? Or do you think 1% is too high a probability that Gimbal was actually a nearby jetplane? Or maybe you think it's too low? Put in your own numbers and see what happens. Then we can compare our guesses and discuss them.

I would even bet (and without calculations!) that for any reasonable choice of numbers Atlas V will yet be the winner, with the data we have (of course it will lose if one assumes for instance that the Atlas engines only have 0.00001% probability of dazzling the Flir, or that the Gimbal was a nearby jetplane with 60% probability, but those kind of numbers are not 'reasonable guesses' in my opinion, they would need to be supported by hard new data (but this is open to discussion too of course)).

Let's compare yours and mine numbers and we'll see what we can find and on what we can agree, but until only my set of numbers is known no further progress is possible: they stand as they are, unchallenged.

Last edited:
Put your guesses and your list of hypothesis into the calculations and let me know what you find.

My bet is that the video was shot at daylight (thus, no Atlas V). I don't have probabilities to plug into your formulas. What I have is that the clouds that can be seen in the video are at low temperature. Basically, the air at 25000 ft (7700 m) has a temperature of about -30 ºC (see https://www.engineeringtoolbox.com/standard-atmosphere-d_604.html , https://www.ngdc.noaa.gov/stp/space...976/us-standard-atmosphere_st76-1562_noaa.pdf ).

A black body at that temperature has its peak IR emission at a wavelength of 12 um, while the radiation in the band ATFLIR is sensible to (3.7-5 um) represents only 0.2% of the total emission which, BTW, won't be high, as the total radiation goes with the fourth power of the temperature. MWIR devices are designed to observe bodies at higher temperatures: A body at 300ºC has its peak emission at 5 um, and emits ~2100 times more energy in the ATFLIR band.

So, if the clouds are seen by the detector, it is very likely to be because of the reflection of the sun's IR radiation on the clouds.

My bet is that the video was shot at daylight (thus, no Atlas V). I don't have probabilities to plug into your formulas. What I have is that the clouds that can be seen in the video are at low temperature. Basically, the air at 25000 ft (7700 m) has a temperature of about -30 ºC (see https://www.engineeringtoolbox.com/standard-atmosphere-d_604.html , https://www.ngdc.noaa.gov/stp/space...976/us-standard-atmosphere_st76-1562_noaa.pdf ).

A black body at that temperature has its peak IR emission at a wavelength of 12 um, while the radiation in the band ATFLIR is sensible to (3.7-5 um) represents only 0.2% of the total emission which, BTW, won't be high, as the total radiation goes with the fourth power of the temperature. MWIR devices are designed to observe bodies at higher temperatures: A body at 300ºC has its peak emission at 5 um, and emits ~2100 times more energy in the ATFLIR band.

So, if the clouds are seen by the detector, it is very likely to be because of the reflection of the sun's IR radiation on the clouds.
This is a perfectly reasonable idea: what you're trying to do is using known data (what the video clip shows) to make a case for the unknown time/date the video was shot to fall in a certain temporal window (daylight), incompatible with the Atlas hypothesis. The only thing missing is a number assigned to the probability the Gimbal video was shot in daylight: if it's > 50% the Atlas hypothesis loses credibility, if it's lower the Atlas hypothesis credibility increases, very easy.

I have no idea which cloud temperature would allow them to be seen by the FLIR, nor I have any idea of how the temperature of clouds at 7700m changes from day to night or from time to time or how they reflect in the IR band, so I really cannot help you on this matter: for what my personal knowledge is able to do I can only conclude that the probability the video was shot in daylight is 50% (50-50% being the default probability assigned to two different hypothesis when one has no idea on how to discriminate between one or the other), and with a 50% probability nothing changes in my calculations.

Conclusion: you have found a perfectly logical way to attack my assumption #1 (that we're 100% sure the Atlas was launched in a timeframe consistent with what we know about the video). You only need to argue (and be able to defend if challenged) a reasonable number for what you call the 'very likely' probability the video was shot in daylight. Or, in other words, would you accept to bet one to one on the video being shot in daylight? Say 100\$ against 100\$ or whatever? Or maybe do you feel so comfortable that you could accept a five to one bet, 500\$ against 100\$ that the video was cast in daylight? Or maybe 1000\$ against 100\$? Just let me know which odds you would accept for this bet and I'll tell you how this number (this new evidence) changes the probability of Atlas being the Gimbal (vs. it being a nearby jetplane), you don't even need to argue about cloud temperatures and Stefan-Boltzmann's law (*)! But yet again, until only my number is on the table it stands as it is, unchallenged.

(*) of course if you can build a case that the video was shot in daylight given what the video shows and what physical laws say your attack on my assumption #1 would become quite strong, even impossible to defend from. That would indeed be an excellent debunking of the Atlas V hypothesis!

Last edited:
Mauro, I advice you to never go to a casino (at least not as a customer, you'd be very welcome as an employee )
I lost track on how you juggle with odds and probabilities but if you want to calculate the probability of a number of uncorrelated things to all be true, you simply multiply the probabilities.

Which gives 0.25 x 1 x 0.8 x 1 = 0.2 for your best case estimates
And 0.25 x 1 x 0.3 x 0.5 = 0.04 for your worst case estimates.

So a 80 to 96 % probability that the ATLAS hypothesis is wrong.
That's my gut feeling, too.

This is a perfectly reasonable idea: what you're trying to do is using known data (what the video clip shows) to make a case for the unknown time/date the video was shot to fall in a certain temporal window (daylight), incompatible with the Atlas hypothesis. The only thing missing is a number assigned to the probability the Gimbal video was shot in daylight: if it's > 50% the Atlas hypothesis loses credibility, if it's lower the Atlas hypothesis credibility increases, very easy.

I have no idea which cloud temperature would allow them to be seen by the FLIR, nor I have any idea of how the temperature of clouds at 7700m changes from day to night or from time to time or how they reflect in the IR band, so I really cannot help you on this matter: for what my personal knowledge is able to do I can only conclude that the probability the video was shot in daylight is 50% (50-50% being the default probability assigned to two different hypothesis when one has no idea on how to discriminate between one or the other), and with a 50% probability nothing changes in my calculations.

Conclusion: you have found a perfectly logical way to attack my assumption #1 (that we're 100% sure the Atlas was launched in a timeframe consistent with what we know about the video). You only need to argue (and be able to defend if challenged) a reasonable number for what you call the 'very likely' probability the video was shot in daylight. Or, in other words, would you accept to bet one to one on the video being shot in daylight? Say 100\$ against 100\$ or whatever? Or maybe do you feel so comfortable that you could accept a five to one bet, 500\$ against 100\$ that the video was cast in daylight? Or maybe 1000\$ against 100\$? Just let me know which odds you would accept for this bet and I'll tell you how this number (this new evidence) changes the probability of Atlas being the Gimbal (vs. it being a nearby jetplane), you don't even need to argue about cloud temperatures and Stefan-Boltzmann's law (*)! But yet again, until only my number is on the table it stands as it is, unchallenged.

(*) of course if you can build a case that the video was shot in daylight given what the video shows and what physical laws say your attack on my assumption #1 would become quite strong, even impossible to defend from. That would indeed be an excellent debunking of the Atlas V hypothesis!
"Very likely" as just a subjective statement that shows my confidence in my argumentation. Putting a random high number to that doesn't make it more objective. I could just say "make it as a high as needed to overcome your number"

Mauro, I advice you to never go to a casino (at least not as a customer, you'd be very welcome as an employee )
I lost track on how you juggle with odds and probabilities but if you want to calculate the probability of a number of uncorrelated things to all be true, you simply multiply the probabilities.

Which gives 0.25 x 1 x 0.8 x 1 = 0.2 for your best case estimates
And 0.25 x 1 x 0.3 x 0.5 = 0.04 for your worst case estimates.

So a 80 to 96 % probability that the ATLAS hypothesis is wrong.
That's my gut feeling, too.
Sorry Itsme, but it's you who should stay far from casinos.

1 x 0.8 x 1 = 0.8 is right (it's my 80% a priori estimation)
1 x 0.3 x 0.5 = 0.15 is right (it's my 15%, worst-case a priori estimation)

But when we add in the new evidence (we expect the actual DoD report 75% of the times if the Gimbal was inside the test range, and 25% of the times if the Gimbal was in effect outside the test range) you cannot multiply everything by 0.25. This can be clearly seen if we assume for a moment equal probabilities for the DoD report, 50%-50%: in this case it's evidence neither for Atlas nor for non-Atlas and all the a priori probabilities should not change, while by multiplying everything by 0.5, as you do, all the probabilities halve! And one could go on adding in new irrelevant evidence at 50%-50% probability and quickly bring any hypothesis down to zero probability (including the competing claim!). That's obviously wrong.

What you must do is to transform the 75%-25% probability into odds (75%-25% = 3:1 against Atlas) and then multiply everything by the odds. Try now with 50%-50% probability (no evidence for either side): the odds are 1:1 and multypling by 1 does not change the a priori estimates, as it should be.

If you want a matematical demonstration I can write it down (maybe not in this thread) but it takes a little time.

"Very likely" as just a subjective statement that shows my confidence in my argumentation. Putting a random high number to that doesn't make it more objective. I could just say "make it as a high as needed to overcome your number"

Yes, but why don't you at least try to say a number instead of a qualitative judgment such as "very likely"? Of course you could "make it as a high as needed to overcome my number", but exaggerating numbers so they fit your theory has the nasty side effect of exposing your numbers to criticism, and you'll have to go far beyond some reference to cloud temperatures and appeals to Stefan-Boltzmann if you want to claim, for instance, "I'm 95% sure the Gimbal video was cast in daylight".

That's why I have kept my numbers 'reasonable', and in fact look, at the moment noone has yet criticized any of my numbers.

Just to summarize what happened until now (for my own reference if anything else!):
• @Mendel in post #49 casts doubts on my numbers. That's a perfectly valid thing to do of course, I even asked for it many times in my posts, but he does not propose anything different (or anything at all), so my numbers remain unchallenged.
• @jplaza in post #51 introduces a new evidence (the clouds in the FLIR video vs. the time of day) which is perfectly valid, but he does not propose any probability for his evidence except saying it's "very likely". It's not possible to compare a number with a sentence though (is "big" greater than 1 meter?) and nothing changes.
• @Itsme in post #53 criticizes my mathematics and says I made mistakes. But as it turn out, he makes a mistake instead.
• @jplaza in post #54 defends his use of "very likely". That's good of course, but it does not make "very likely" more comparable to a number that it was before. He also makes an interesting observation about how one could juggle the numbers to fit his pet theory in spite of all the evidence on the planet: in post #56 I explain him why it may not always be a good idea.
Final count as of this post: it's from 26.67 (my preferred numbers) to 5 times (my worst-case numbers) more probable that Gimbal was the Atlas, rather that it was the second most probable thing (a nearby jetplane).

Last edited:
Making up numbers is less helpful, it fakes precisions where no precision is possible.

The rocket launch is an interesting theory but we have no way of verifying it. There are many aircraft in the skies at any time, but far less rockets. So it's not as likely but trying to put a number it given the date is a pointless endeavour.

There are many aircraft in the skies at any time, but far less rockets. So it's not as likely but trying to put a number it given the date is a pointless endeavour.

It's a pointless endeavour because it leaves out a firm piece of evidence we instead do have: we are 100% sure there was an Atlas V launch near the date the Gimbal video was taken (no matter how many rockets vs. aircrafts there are on average in the sky at any given time). I'm anyway happy that you used a 'probabilistic' reasoning

Last edited:
Ok, let's break it down:
I really don't remember what DoD says about the location of the object, but I have no reasons to doubt what you say. So I agree with you, your observation makes the Atlas V hypothesis less probable that it was before your observation is taken into account. I could agree that we can expect the Dod statement with 75% probability if the Gimbal was inside the test range (then it was no Atlas) and 25% otherwise, because there are many reasons why DoD says what it says (ie.: they may simply be wrong). It's 3:1 odds against the Atlas (75% divided by 25%).
Accuracy on position of the UAP
Why 75%? Where does that number come from? Is it what you expect, based on what?

AFAIK, the DOD has not made public any analysis at all on their 144 UAP events that served as basis for the UAPTF report (which was only a summary of reports they have so far). I personally (subjectively) think that when they refer to "incursions in test ranges" is only a description of the apparent event. But to objectively know if the objects were or not actually where they apparently were, an analysis should be done all the cases and determine if that's right or not. Only then you can calculate a probability on the accuracy of that statement.

I simply don't know what value to assign, so for me this value is unknown, could be anything between 0 and 100%.

But just for fun, my experience tells me that in UFO reports basic estimations like distances, angles, speeds, and so on are usually wrong. But distances measured by devices are often right, unless there is a malfunction. So I am going to say 50%. (I know, this goes in favor of the Atlas V hipothesis.)

But the Atlas V hypothesis starts with very high probabilities: we're 1) 100% sure it was launched in a timeframe consistent with what we know about the video,
Timeframe coincidence
The launch was on Jan 20th, at 20:04 ET. (Jan 21st, 01:04 UTC)
The Navy confirmed the gimbal video was taken on Jan 21st (see post #22).
So, this fact would make the probability of a coincidence 0%.

But, on post #23, @Edward Current suggested the exercise date could be in UTC and someone forgot to put it back to ET, so maybe the dates do actually have a window of coincidence.

So, what probability do we assign to the Navy answering a question in UTC rather than in local time? Do we have any precedents to obtain a reality-based probability? For me this is again unknown, but if we want to play to this game, I am going to say that is unusual, so let's make it 10%.

Anway, let's assume there is in fact a window of coincidence. How long is that window?
Should we take a 1-day window coincidence? Then we have a probability of 100%, as you say, that the plane was on the air the same day the launch happened.

But, how long does a fighter stay in the air during this exercises? mmm... I'm going to take a 1-hour window. As the day has 24 hours, the probability of the fighter flying in an specific 1-hour time window is 1/24 ~ 4.1%. So that number in combination with the 10% of Navy reporting a date in UTC rather than local, is my timeframe probability.

2) I think we can be pretty confident the F-18 FLIR would have seen a glare if pointed to the Atlas V engines even if the distance was large, and 3) we can be pretty confident that IR glare in the FLIR appears as the Gimbal object does. I'd give an 80% probability to assumption 2) being true, but let's say from 80 down to 30%. For assumption 3) I'd say 100%, but let's say 100-50% instead.
IR flare
I would ask for justification of those numbers, but anyway, I can go with them.

So with my preferred numbers the probability of Atlas V being the Gimbal is 80%, with the lowest numbers I can possible conceive with the data I have the probability is 15%. The next hypothesis on the list is... uh.. what is it? this is already a difficult question... another F-18 or another nearby jetplane? This looks the best candidate to me and it has long be discussed and apparently refuted (not that I remember the details), do you agree on a probability of 1% that the Gimbal was another nearby jetplane?
I would say you should do the same exercise of breaking down all the probabilities for the jet solution that lead to that 1%, instead of assigning it directly.

I don't know what my made-up numbers finally result compared to yours. In fact, I don't think it matters, I won't even try to defend these numbers.

These numbers are simply reflecting our own biases. It's like the Drake equation, where you can plug in any numbers you like and get radically different results. These calculations only makes sense if you can objectively determine those probabilities. Otherwise it is numerology

It would me more productive to obtain such basic data like the date and time. Has anyone been able to contact the pilot/WSO who recorded the video? Is it possible to find (via FOIA, I guess) the deck logs of the ships where the planes took off from, and check the date and time the exercise took place?

Or to find go/no-go proofs to discard hypothesis (was the video taken in daylight? Was the plane heading towards the Atlas V?...)

Last edited:
Ok, let's break it down:
I am really, really happy that you posted all that @jplaza, that's the way to go!

Now, let's break it down again.

Accuracy on position of the UAP
Why 75%? Where does that number come from? Is it what you expect, based on what?

AFAIK, the DOD has not made public any analysis at all on their 144 UAP events that served as basis for the UAPTF report (which was only a summary of reports they have so far). I personally (subjectively) think that when they refer to "incursions in test ranges" is only a description of the apparent event. But to objectively know if the objects were or not actually where they apparently were, an analysis should be done all the cases and determine if that's right or not. Only then you can calculate a probability on the accuracy of that statement.
My 75% was just a reasonable guess (leaning a little against Atlas). Reasonable because I think noone will deny DoD is wrong (or misleading, or even devious) ~25% of the times.

I simply don't know what value to assign, so for me this value is unknown, could be anything between 0 and 100%.
Excellent! There's anyway a basic rule: if I really don't know, the probability is 50% (that is to say: it does not move the balance of evidence in either direction).

But just for fun, my experience tells me that in UFO reports basic estimations like distances, angles, speeds, and so on are usually wrong. But distances measured by devices are often right, unless there is a malfunction. So I am going to say 50%. (I know, this goes in favor of the Atlas V hipothesis.)
And of course I'd eagerly accept your 50%, but let me stick to my worst-case 75%.

Timeframe coincidence
The launch was on Jan 20th, at 20:04 ET. (Jan 21st, 01:04 UTC)
The Navy confirmed the gimbal video was taken on Jan 21st (see post #22).
So, this fact would make the probability of a coincidence 0%.

But, on post #23, @Edward Current suggested the exercise date could be in UTC and someone forgot to put it back to ET, so maybe the dates do actually have a window of coincidence.

So, what probability do we assign to the Navy answering a question in UTC rather than in local time? Do we have any precedents to obtain a reality-based probability? For me this is again unknown, but if we want to play to this game, I am going to say that is unusual, so let's make it 10%.
You are very reason, this should be factored in and I missed it. An error in the ET/UTC reporting is very much possible, but it does not mean it's also probable. I agree with your 10% figure and I'll use it later to revise my calculations.

Anway, let's assume there is in fact a window of coincidence. How long is that window?
Should we take a 1-day window coincidence? Then we have a probability of 100%, as you say, that the plane was on the air the same day the launch happened.

But, how long does a fighter stay in the air during this exercises? mmm... I'm going to take a 1-hour window. As the day has 24 hours, the probability of the fighter flying in an specific 1-hour time window is 1/24 ~ 4.1%. So that number in combination with the 10% of Navy reporting a date in UTC rather than local, is my timeframe probability.
This is true, but irrelevant, because it can be applied to any possible hypothesis (your 1/24 probability applies equally to the Atlas as to jetplanes as to alien crafts or anything else), so it doesn't change anything.

IR flare
I would ask for justification of those numbers, but anyway, I can go with them.
Very good, less typewriting

I would say you should do the same exercise of breaking down all the probabilities for the jet solution that lead to that 1%, instead of assigning it directly.

I don't know what my made-up numbers finally result compared to yours. In fact, I don't think it matters, I won't even try to defend these numbers.

These numbers are simply reflecting our own biases. It's like the Drake equation, where you can plug in any numbers you like and get radically different results. These calculations only makes sense if you can objectively determine those probabilities. Otherwise it is numerology

It would me more productive to obtain such basic data like the date and time. Has anyone been able to contact the pilot/WSO who recorded the video? Is it possible to find (via FOIA, I guess) the deck logs of the ships where the planes took off from, and check the date and time the exercise took place?

I very much agree with you: if we could obtain more data (expecially the exact date/time) we would make a real progress. But until this happens we're compelled to work with the only data we actually have.

And now, the revised calculation, factoring in what I overlooked and @jplaza rightly made me notice: 10% probability of a reporting error (or of a misleading report) which is needed to make the timeframe of the Atlas launch compatible with the Gimbal video [by the way, this looks to me a point which could yield some clues if investigated, it's surely ways easier than finding the technical specs of the FLIR for instance].

The odds of Gimbal being the Atlas, as compared to the odds of Gimbal being a nearby jetplane, now range from 2.67 times (my preferred numbers + jplaza's 10% observation) to 0.5 times (my worst-case numbers + jplaza's observation). Rise them up to 8 and 1.5 times if you agree with @jplaza that DoD's statements regarding UAPs have a 50% probability of being wrong (or misleading, or even devious) instead of the 25% I used.

I'm still betting on the Atlas, maybe no more than one to one now, though

Last edited:
Well, my point was not about re-doing the numbers, but to show that anyone can put any number and find a justification based on his/her own bias.

Anyway, it's data what its needed.

Well, my point was not about re-doing the numbers, but to show that anyone can put any number and find a justification based on his/her own bias.
That's a very much possible thing one can do, but also the most wrong way of using Bayesian inference I can think of (btw I already responded to the same argument in post #56).

Are you implying I'm behaving that way? I strive not to, but of course I can be fooled by myself as anyone else. That's why I'm glad to tell everybody what my numbers and my reasoning are, so they can all be checked. You found a mistake in my line of reasoning (thanks, by the way), I updated the probabilities accordingly and my confidence in Gimbal being the Atlas went down, do you think this is consistent with 'finding a justification based on my bias'? I could not care less if Gimbal was or was not the Atlas, but I care a lot when I do mistakes in my reasoning.

Are you implying I'm behaving that way? I strive not to, but of course I can be fooled by myself as anyone else.
Not really what I meant, sorry if I gave you that impression. I believe we all have our biases that unconsciously leak into reasonings, even if we try not to.

Not really what I meant, sorry if I gave you that impression. I believe we all have our biases that unconsciously leak into reasonings, even if we try not to.
No worries! And I agree with you on this point too

I think we can rule out the gimbal video showing a rocket launch because the FLIR looks a constant 2 degrees below the aircraft the entire time.

The January 20 launch was for a MUOS-3 satellite. I could not find much for this satellite, but since it is part of a family of MUOS satellites, I assume the launches are basically the same, because that makes life easier for the engineers. Accordingly, I assume the following link for MUOS-5 applies to MUOS-3 as well. https://spaceflight101.com/atlas-v-muos-5/launch-profile/ Here we see the target orbit was GTO.
From the following video, at booster engine cutoff the rocket has an altitude of 84 miles, and is 277 miles downrange.

For a GTO orbit, the Atlas V user guide provides two different ascent profiles on pages 2-19 and 2-21 that match the video.
https://www.ulalaunch.com/docs/default-source/rockets/atlasvusersguide2010a.pdf?sfvrsn=f84bb59e_2

The sample profile on page 2-21 shows that at 260 seconds the rocket has an altitude of 170 km, at 1040 seconds, it has an altitude somewhere between 166 and 22,178km (side note, ULA PLEASE put altitude readings in your launch broadcasts).

Thus there are some questions that hopefully someone more versed in geometry than I am can answer. If a jet is at 25000 feet, how far away is it from an object that is at 170km, if the target is 2 degrees below the jet. Is this distance reasonable for the FLIR to pick up? Is there a sweet spot that works for the entire 30sih second video clip? I dont think the first 260 seconds of rocket flight will work, as the altitude gain is significant.

Here is a good video by one of the pilots involved in the gimbal video.

Of special note. The gimbal video was the first time they saw a gimbal object (21:00). Also, they were flying out of Jacksonville Florida, 140 miles north of Cape Canaveral. I assume pilots would made be aware of the associated flight restrictions that accompany a rocket launch if they were anywhere close to it. Here is a sample TRF for a launch coming up on the 15th. https://tfr.faa.gov/save_pages/detail_1_6952.html

That's why I have kept my numbers 'reasonable', and in fact look, at the moment noone has yet criticized any of my numbers.

Just to summarize what happened until now (for my own reference if anything else!):
• @Mendel in post #49 casts doubts on my numbers. That's a perfectly valid thing to do of course, I even asked for it many times in my posts, but he does not propose anything different (or anything at all), so my numbers remain unchallenged.
Fallacy: I said your numbers are guesswork.
Your rebuttal is that unless I can come up with valid numbers, yours are valid, and that's not logical. You need to support your own numbers, and you can't. You don't even have probabilities that sum to 100%.

My 75% was just a reasonable guess (leaning a little against Atlas). Reasonable because I think noone will deny DoD is wrong (or misleading, or even devious) ~25% of the times.
Your support is "no-one will deny". I deny it. Now what?

Sorry Itsme, but it's you who should stay far from casinos.

1 x 0.8 x 1 = 0.8 is right (it's my 80% a priori estimation)
1 x 0.3 x 0.5 = 0.15 is right (it's my 15%, worst-case a priori estimation)

But when we add in the new evidence (we expect the actual DoD report 75% of the times if the Gimbal was inside the test range, and 25% of the times if the Gimbal was in effect outside the test range) you cannot multiply everything by 0.25. This can be clearly seen if we assume for a moment equal probabilities for the DoD report, 50%-50%: in this case it's evidence neither for Atlas nor for non-Atlas and all the a priori probabilities should not change, while by multiplying everything by 0.5, as you do, all the probabilities halve! And one could go on adding in new irrelevant evidence at 50%-50% probability and quickly bring any hypothesis down to zero probability (including the competing claim!). That's obviously wrong.

What you must do is to transform the 75%-25% probability into odds (75%-25% = 3:1 against Atlas) and then multiply everything by the odds. Try now with 50%-50% probability (no evidence for either side): the odds are 1:1 and multypling by 1 does not change the a priori estimates, as it should be.

If you want a matematical demonstration I can write it down (maybe not in this thread) but it takes a little time.
I get the feeling I'm never going to convince you, so this is my final attempt:
If the probability of this being ATLAS is 25% according to a DOD statement the total probability of this being ATLAS will never rise above 25%. So, if you take this DOD statement into account you really have to multiply everything by 0.25. The DOD statement is a new piece of information (well, the statement is, not the guestimated percentage) that affects your probabilities in a negative way. You may not like it but this is simply how the math works.

I think we can rule out the gimbal video showing a rocket launch because the FLIR looks a constant 2 degrees below the aircraft the entire time.

Not necessarily: if the aircraft nose is pointing upwards the object can very well be 2° 'below' the airplane, yet at a greater height

Fallacy: I said your numbers are guesswork.
Your rebuttal is that unless I can come up with valid numbers, yours are valid, and that's not logical. You need to support your own numbers, and you can't. You don't even have probabilities that sum to 100%.
This is more a critique of Bayesian inference in general than a critique of my application of it to Gimbal. I'm writing a small primer on Bayesian inference (I'd like to post it in the 'practical debunking' section): that will be a better place to discuss this.

Your support is "no-one will deny". I deny it. Now what?
@jplaza already denied it in post#59, proposing 50% instead: I just made the calculations with both values (at the end of post #60). Tell me what you think the relaiability of DoD statetements about UAPs is and I'll make the same, so at least on can see how much that evidence influences the odds, isnt'it a cool thing to know?

I get the feeling I'm never going to convince you, so this is my final attempt:
If the probability of this being ATLAS is 25% according to a DOD statement the total probability of this being ATLAS will never rise above 25%. So, if you take this DOD statement into account you really have to multiply everything by 0.25. The DOD statement is a new piece of information (well, the statement is, not the guestimated percentage) that affects your probabilities in a negative way. You may not like it but this is simply how the math works.
Ah I instead get the feeling it will be you to be hard to convince, but oh well. I'll post the demonstration of the formula (hopefully today) in another thread, there will be a better place for this discussion.

In support of Itsme's point:
But when we add in the new evidence (we expect the actual DoD report 75% of the times if the Gimbal was inside the test range, and 25% of the times if the Gimbal was in effect outside the test range) you cannot multiply everything by 0.25. This can be clearly seen if we assume for a moment equal probabilities for the DoD report, 50%-50%: in this case it's evidence neither for Atlas nor for non-Atlas and all the a priori probabilities should not change, while by multiplying everything by 0.5, as you do, all the probabilities halve! And one could go on adding in new irrelevant evidence at 50%-50% probability and quickly bring any
The evidence isn't irrelevant.
I could agree that we can expect the Dod statement with 75% probability if the Gimbal was inside the test range (then it was no Atlas) and 25% otherwise, because there are many reasons why DoD says what it says (ie.: they may simply be wrong). It's 3:1 odds against the Atlas (75% divided by 25%).

Now, hypothetically,
IF Gimbal was inside the test range, THEN I might expect the DoD to say so with 99% certainty (because there are many sensors surveying this area).
IF Gimbal was outside the test range, THEN I might expect the DoD to say it's inside with 25% probability due to spurious sensor readings etc.

What you are using in your computation is the inverse conditional probability, i.e. you are claiming that
IF the DoD reports Gimbal inside the range, THEN it is 25% probable that it was outside.
And that's not supported by the two above conditionals.
I'll illustrate why.

Assume an a priori of 50/50 for Gimbal to be outside the test range.
We can now compute 4 probilities:
• Gimbal inside, report says inside 50%×99%=0.495
• Gimbal inside, report says outside 50%×1%=0.005
• Gimbal outside, report says inside 50%×25%=0.125
• Gimbal outside, report says outside 50%×75%=0.375
With this, the probability that the DoD says it's inside is 0.62.
IF the DoD says it's inside, it's outside 0.125/0.62=20.2% of the time, not 25%.
Obviously this changes if you choose different priors. (And obviously there is no data to support the 99% and 25% conditional probabilities in the first place.)

This is more a critique of Bayesian inference in general than a critique of my application of it to Gimbal.
No, it isn't. Bayesian inference is fine if you don't invent numbers, but rather derive them from actual knowledge.
Tell me what you think the relaiability of DoD statetements about UAPs is and I'll make the same, so at least on can see how much that evidence influences the odds, isnt'it a cool thing to know?
I don't know the odds, that's my whole point. But neither do you, so saying it's not 25% is a pretty good bet.

I think we can rule out the gimbal video showing a rocket launch because the FLIR looks a constant 2 degrees below the aircraft the entire time.
If an aircraft is at an altitude of 15000 feet or higher, normally the horizon will appear at 2⁰ below horizontal or lower, because Earth is a big ball. (via https://www.metabunk.org/curve/ )
Because the rocket travels around the planet, it will get close to the horizon and eventually disappear behind it.

Last edited:
I don't know the odds, that's my whole point. But neither do you, so saying it's not 25% is a pretty good bet.
If something is unknown, then the logical probability to assign is 50%. My 25% was an assumption skewed against the Atlas (at three to one, not exactly great odds for poor Atlas). It will take me a little more time before answering the rest of the remarks in posts #72 and #73.

If an aircraft is at an altude of 15000 feet or higher, normally the horizon will appear at 2⁰ below horizontal or lower, because Earth is a big ball. (via https://www.metabunk.org/curve/ )
Because the rocket travels around the planet, it will get close to the horizon and eventually disappear behind it.
Is the horizon visible in the Gimbal video?

This is the geometry I'm thinking of (sorry for my lousy graphics skills), it does not seem strange to me, but I may be very mistaken; nor I did any checks on the compatibilty of angles and ranges (and the drawing is not at scale!):

Last edited:
Good morning everyone.
This conversation really has a strong informative and didactic character. I think all the considerations are perfectly valid. The identification of gimbal still remains unknown because it's the footage itself lacking in essential informations. However, my hypothesis was fundamentally based on two notable indications.
The first is the launch date. No rockets are fired from the KSC every day. So statistically this has a significant impact.
What I find curious, however, is what the pilots observed on the SA display. That is, a wedge formation of smaller objects following the larger object. A situation even more consistent with what happened when the 5 SRBs were jettisoned.

However, associating that IR signature that appears in the video would require very accurate analysis which must be based on data that we don't have. First of all, the distance. Because a rocket in 30 seconds travels a considerable vertical trajectory that could be limited to a field of view of only one degree only if the object were found at a considerable distance.

Finally, I wanted to give my idea on two doubts raised.

Navy training areas overlap with the USAF / USSF Eastern Range. Activities are regulated via NOTAM. It is therefore difficult to determine if a rocket is making a incursion into one of these areas. Theoretically it always does.

According to the problem of the visibility of the clouds. The analysis is very precise and exhaustive. However, meteorological satellite images show that cloudy IR signatures are visible even at night.
https://commons.erau.edu/cgi/viewcontent.cgi?article=1018&context=db-applied-aviation

According to the problem of the visibility of the clouds. The analysis is very precise and exhaustive. However, meteorological satellite images show that cloudy IR signatures are visible even at night.
https://commons.erau.edu/cgi/viewcontent.cgi?article=1018&context=db-applied-aviation
That is a very specific process that involves detection in two different (and narrow) bands: 3.9 um and 11 um, and then calculating the difference between them.

External Quote:
At night the difference between the 3.9 and 11 micron channels detects emissivity differences rather thanabsolute temperature
It is not a single image from a single band, but a process that involves two images from two different bands, and finding a contrast when the same process is done on radiation coming from the ground.

From https://www.weather.gov/media/crp/GOES_16_Guides_FINALBIS.pdf
External Quote:

ABI Band #7
3.9 microns IR (“Shortwave Window Band”)

Contains daytime solar reflectance component

Primary Uses:
• Low stratus and fog (especially when differenced with the 11.2-micron IR channel taking advantage of emissivity differences)
• Fire/hot spot detection and volcanic ash
So, its primary use as independent channel is the detection of fire and hot spots, since "hot" bodies emit most of its radiation including that band. By itself, it is not enough to detect clouds respect to the ground.

External Quote:

ABI Band #14
11.2 microns IR (“IR Longwave Band”)
• Differenced with the 3.9 micron near IR channel for low stratus and fog detection
"Cold" bodies emit most of their IR radiation in the 8-12 micron wavelenghts, that's why this is the "traditional" IR window.

In the case of ATFLIR it is a IR device operating in a continuous wide band from 3.7 to 5 um. That is, it receives energy from all those wavelenghts, but does not distinguish among them. In any case, it is centered in the "hot bodies" band, and (as the 3.9 um channel above) is also affected by the reflected sunlight.

Can ATFLIR see clouds at night? It is not specifically designed for that. An Earth Observation satellite is. There are several factors that make them different. Number of sensors, bands, optical apertures, specific processes... ATFLIR needs a short time to be able to deliver realtime images, while from a satellite you can have longer exposures to ensure detecting small differences or weak sources.

Comparing different systems with different purpouses is not straightforward.

Last edited:
That is a very specific process that involves detection in two different (and narrow) bands: 3.9 um and 11 um, and then calculating the difference between them.

External Quote:
At night the difference between the 3.9 and 11 micron channels detects emissivity differences rather thanabsolute temperature
It is not a single image from a single band, but a process that involves two images from two different bands, and finding a contrast when the same process is done on radiation coming from the ground.

From https://www.weather.gov/media/crp/GOES_16_Guides_FINALBIS.pdf
External Quote:

ABI Band #7
3.9 microns IR (“Shortwave Window Band”)

Contains daytime solar reflectance component

Primary Uses:
• Low stratus and fog (especially when differenced with the 11.2-micron IR channel taking advantage of emissivity differences)
• Fire/hot spot detection and volcanic ash
So, its primary use as independent channel is the detection of fire and hot spots, since "hot" bodies emit most of its radiation including that band. By itself, it is not enough to detect clouds respect to the ground.

External Quote:

ABI Band #14
11.2 microns IR (“IR Longwave Band”)
• Differenced with the 3.9 micron near IR channel for low stratus and fog detection
"Cold" bodies emit most of their IR radiation in the 8-12 micron wavelenghts, that's why this is the "traditional" IR window.

In the case of ATFLIR it is a IR device operating in a continuous wide band from 3.7 to 5 um. That is, it receives energy from all those wavelenghts, but does not distinguish among them. In any case, it is centered in the "hot bodies" band, and (as the 3.9 um channel above) is also affected by the reflected sunlight.

Can ATFLIR see clouds at night? It is not specifically designed for that. An Earth Observation satellite is. There are several factors that make them different. Number of sensors, bands, optical apertures, specific processes... ATFLIR needs a short time to be able to deliver realtime images, while from a satellite you can have longer exposures to ensure detecting small differences.

Comparing different systems with different purpouses is not straightforward.
You surely have a point here, and in any case it's very well written, thumbs up. I agree with all what you say, but I feel there is one link missing yet. Given the FLIR can surely see clouds in daylight (the whole argument rests on this) the difficult part is to determine if it can't do the same at night. You proposed the FLIR can see the clouds in daylight because they reflect the direct IR radiation from the sun (if I'm not mistaken). This may very well be, but is difficult to demonstrate (for me, at least, ie. I have no clue what the IR reflectance of cloud tops may be, among many others things). More factors could be involved (was it a full moon night?) But I'd say a provisional 2:1 odds in favor of daylight is surely deserved here, which brings us to:

The odds of Gimbal being the Atlas, as compared to the odds of Gimbal being a nearby jetplane, now range from 1.33 times (my preferred numbers + jplaza's 10% observation + daylight observation) to 0.25 times (my worst-case numbers + jplaza's 10% observation + daylight observation). Rise them up to 4 and 0.75 times if you agree with @jplaza (and apparently @Mendel, if I'm not mistaken) that DoD's statements regarding UAPs have a 50% probability of being wrong (or misleading, or even devious) instead of the 25% I used.

Not necessarily: if the aircraft nose is pointing upwards the object can very well be 2° 'below' the airplane, yet at a greater height
The 2° is relative to the horizon, not to the aircraft. If the aircraft tilts up, the number won't change. (That's different from the left–right angle number, which is relative to the fuselage and does change when the aircraft turns.)

Replies
25
Views
6K
Replies
25
Views
8K
Replies
13
Views
12K
Deleted User P
D
Replies
20
Views
14K