Dear Roger,

No one seems to have responded to your interesting letter to
the Editor of the BSS Bulletin.

May I join in?


SEASONAL MARKERS

You write about Seasonal Markers on analemmatic sundials.
Seasonal Markers are two special points on the major axis
of the ellipse (of hour marks).  If you stand on one of
these points and look in the direction of a date point
on the date line you will be facing the direction of
sunrise on that date.  With the other you face sunset.

You refer to two articles on Seasonal Markers:

 1. ... by M.K. Head in the June 2007 BSS Bulletin

 2. ... by R.T. Bailey via http://www.walkingshadow.info

Taken together these two articles can be summarised in a
single line.  Unfortunately I couldn't find that line in
either article :-(

What I want to know is where these markers should be placed...


ALL ANYONE NEEDS TO KNOW ABOUT SEASONAL MARKERS

  Position = a . (coslat/cosdec) . sqrt(coslat^2 - sindec^2)

  a = semi-major axis
  coslat = cos(latitude)
  cosdec = cos(declination)
  sindec = sin(declination)

This tells you how far out along the major axis the markers
should be placed, given latitude and declination (season?).


WHAT DO WE LEARN FROM THIS?

A telling sentence in the Head article begins...

   The mathematics becomes cumbersome...

Three cosines and a sine?  Cumbersome?

Simple inspection of this expression explains almost everything
in both articles and lots more besides.  Let's have a look:

 1.  First we can see that the position is dependent on
     the declination which rather spoils the story but...

 2.  Fortunately the declination is confined to the range
     +/- 23.5 degrees which is small enough for sindec^2
     to be ignored, at least when the latitude is small
     and coslat^2 dominates the term in the square root.
     The cosdec denominator is close to one.

 3.  At the equator when latitude=0 the entire expression
     reduces to a, the markers are at the ends of the
     major axis and are don't change with declination.
     Of course the minor axis is zero but that doesn't
     matter here.

 4.  At the equinoxes when declination=0 the expression
     reduces to  a.coslat^2  so there IS a proper placement
     at the equinoxes.  There is no need for the contrivance
     of an angle of 0.01 degrees.

 5.  At higher latitudes, when coslat^2 loses its dominance
     over sindec^2 two things happen:

      i) The markers get much closer to the date line

     ii) There is much greater variability with declination

 6.  When you get into the arctic (or antarctic) you are in
     deep trouble.  The term under the square root can go
     negative and the whole concept falls apart!  [This is
     simply because there are seasons when there is no
     daily sunrise and sunset.]


BEST POSITION

Your message considers the best fixed positions for the markers
acknowledging that most of the time the correct positions will
be displaced slightly from the fixed positions.

You say:

   ... there is a difference of opinion on minimizing error
   from Head, Sonderegger and Bailey.  Head suggest calculating
   the seasonal marker using the solstices.  Unfortunately this
   is an extreme point.  Sonderegger suggests using declination
   20.2ยบ.  This minimizes error assuming a Gaussian normal
   distribution for that error.  But this error is not normal
   but harmonic.  My recommendation is to calculate the seasonal
   marker position for both extreme positions, the solstice and
   the equinox* to define the range and selecting the average
   value in between as the seasonal marker point.

You add:

 * Not exactly the equinox due to zero and infinity problems.
   Use a declination of 0.01.

Quite unnecessary!  There are no zero and infinity problems.
Use a.coslat^2.

Hmmm.  Can this game accommodate a fourth player?

You are quite right to dismiss using the solstices which give
an extreme point, moreover it is imaginary if you are in the
arctic or antarctic!

You are also quite right to dismiss the reference to a normal
distribution.  The error isn't a random variable.  It is
thoroughly deterministic and the term "distribution" is
inappropriate

Your method improves on the others but doesn't allow for the
fact that the declination hovers round the solstices and zips
across the equinoxes.  It needs biasing towards the solstices.
The Head solution is not wholly misplaced.  We can do better...


WHAT DISTRIBUTION DO YOU WANT?

Although the error isn't a random variable there IS a random
variable which one MAY consider.  This is the time of year...

You can say, "Let t be a random variable which is some
instant uniformly distributed in time between one vernal
equinox and the next".

Now any function of a random variable is itself a random
variable with a proper mean and so on.

The expression now turns into a function of time:

  a . (coslat/cos(dec(t)) . sqrt(coslat^2 - sin(dec(t))^2)

THIS is the expression to integrate over a whole year and
whose average should be found.  THAT result is where the
markers should be placed.


ANALOGY WITH GREENWICH MEAN TIME

It is important NOT to plot the position against declination
because the declination doesn't vary uniformly with time.

Something similar is the case of "mean time"...

If you plot the equation of time against time for a whole
year, the average will be zero but...

If you plot the equation of time against declination for a
whole year the average will NOT be zero (though it is VERY
close!).

If you ornament an hour-line on a sundial with the familiar
figure-of-eight analemma and do the job properly, the hour-line
doesn't (quite) separate the analemma into equal areas.

Just a passing thought.

Enough...

Frank


---------------------------------------------------
https://lists.uni-koeln.de/mailman/listinfo/sundial

Reply via email to