John McKune took this sunset sequence near Waw an Namus in the Libyan Desert on 28th March 2006, the evening before the total solar eclipse there. He saw the setting sun first develop a "tin hat" shape and then shrink to a thin bar on the horizon. Green and blue tips developed which broadened and raced together to form the final "flash".

"As the Sun began to set I would take a quick shot to check the histogram against my exposure settings.  I used a fixed shutter speed of 1/2000s and was changing the aperture manually to try and get an image that was neither over-exposed nor too dim to register proper colors.  The air was so clean and dry that it was difficult to look directly at the Sun even when it was only a sliver above the horizon. I have never seen the sky quite like this before.  As the sun was going down, you could not look at it at all naked-eye; even to the very last moment it was too bright.  I was using my 100-400 lens at 400 with a 1.4x extender.  I also had my shutter release attached and was bursting at 3 frames per second as the upper limb began to flatten out.  The Sun's disk was very clean and steady as it went down; no atmospheric turbulence was present at all.  Just when it was about over and I had decided nothing was going to happen, two brilliant emerald green spots formed at the tips.  These grew into bars which raced towards each other to form a single dazzling spot just before disappearing below the horizon." 

     Other crucial conditions for this particular flash: - "The entire area was extremely flat and dry.  There were broad undulations of no more than a meter in height to the horizon." To the west, 6 km distant there was a shallow depression perhaps 1-2m lower - “That evening was very calm." - John’s camera was less than 2m above the desert surface.

The first clue to what happened is the “tin hat” shape of the sun 22s before the flash. The thin “brim” extensions along the apparent horizon signal that a temperature inversion was present and was sufficiently strong that ‘ducting’ took place. Ducting is when rays are trapped between layers of the inversion and can travel a considerable distance before eventually escaping.

Andrew Young says - "Well, I'm delighted to see this picture, as it shows stuff I've seen in my simulations for years but never caught in reality -- probably because our humidity [in California] is usually too high, and the air is too dirty when there's offshore flow.  What's made this possible is the REALLY CLEAN air out on the middle of nowhere.

What you're seeing is the Sun setting on top of a duct below eye level; ..the top of the duct is no more than a few cm above the ground, and only a meter at most below the camera. And that, in turn, means that the apparent horizon is only a few hundred meters away, which makes these details -- normally invisible -- blatantly obvious.

The line is of course a highly compressed image of the upper part of the Sun. 

Evidently, a radiative inversion had developed in the last minutes of the sunset, because of the extreme dryness of the desert.  Because of
the lack of wind, this inversion is very sharp, so the gradient is enough to make a duct -- but an extremely thin one.  There must have
been a little air movement, so that a few cm of air next to the ground was mixed, allowing just enough room for rays to pass between the
bottom of this very shallow duct and the ground itself.  A very delicate balancing act is required to allow all this to happen below eye level!"