NavList:

Dear Paul,

Thanks for your reply in which you state that "it's not possible for Horizons to calculate refracted semidiameter" .

Is it really the case?

(1) - As a complement to the values published by Geoff Hitchcox, Chuck Varney published a first attachment listing JPL Horizon results from which we can extract :

1.1 - Refracted First Contact at 15:36:08.0 with Sun SD + Moon SD = 0.52055° = 31.233' .

Remark 1.1 :

Own result for refracted Augmented SD's at first contact UT is 31.233' (no difference here).

Own result for unrefracted Augmented SD's at first contact UT is 31.252'

1.2 - Refracted Last Contact at 18:51:10.5 with Sun SD + Moon SD = 0.52120° = 31.272'

Remark 1.2 :

Own result for refracted Augmented SD's at last contact UT is 31.282' (difference of 0.01' here).

Own result for unrefracted Augmented SD's at last contact UT is 31.293'

Undoubtly such JPL Horizon results are applicable to our real world 1st and last contact times. Accordingly they have been computed with both Augmented and Refracted Sun & Moon SD's.

(2) - Chuck Varney's second attachment indicates that Sun and Moon SD's are not affected whatsoever by refraction since they are all listed with the very same values every time, i.e. the refracted values.

On the other hand we know for sure - from here and here - that refraction always affects Sun and Moon SD's by at least 0.25" .

(3) - I therefore keep thinking that :

3.1. - The results of refracted 1st and last contact UT's given by JPL Horizons are realistic and reliable values based on actual Augmented and Refracted Sun and Moon SD's (probably fully spherical Moon ???). Computing such quantities is a must in order to publish reliable real world values.

3.2 - The JPL Horizons published results for unrefracted Moon and Sun SD's are exactly the same as their refracted values counterparts and such results labeled as "unrefracted" are not correct. They simply can not be correct when compared to their refracted counterparts published values.

Let's try to find a reasonable explanation to this.

It is quite true that in the real world, unrefracted values have quite limited interest because they can not be compared to observations, at least to the ones made from the Earth surface.

Rather than performing a new set of  unrefracted specific computations - again of quite limited interest and practical comparison value - it looks like the JPL Software Developers used a short cut and simply copy-pasted the refracted values into the unrefracted column instead. Nonetheless such values listed as "unrefracted" are incorrect. Accordingly for the times of unrefracted 1st and last contacts the resulting JPL Horizon published results are incorrect also.

3.3 - Finally closing the loop onto this first post which triggered this lengthy discussion :

With the exception of the differential refraction effect described in part (2) of this post and assuming Lady Moon to be a perfect sphere, Refraction does NOT modify the times of 1st and last contact in a Sun eclipse.

Kermit

PS : If anyone disagrees about the conclusion given in 3.3 here-above ... "  please shoot it down if there is a flaw in it !

Thanks to your wording AdrianF :-)