[LEAPSECS] Cheating means more planning, not less

[M. Warner Losh]

In message: <20081229165202.GP2263 at fysh.org>
            Zefram <zefram at fysh.org> writes:

: Richard B. Langley wrote:

: >How accurate

: >are the predictions (especially the long-term ones) really?

: >One would have to compare

: >one of the historical empirical functions with actual UT1 data.

: 

: We discussed this in 2007-01 in a thread titled "UT1 confidence".

: No firm answers were forthcoming regarding present IERS capability.

: 

: PHK noted that the accuracy estimation formula in Bulletin A gives

: unbelievable results if applied to periods of decades.  We don't know

: how far out that formula, or the DUT1 estimation formula, are intended

: to be applied.


Yes.  I think he said that they worked well out about 10 years, but
that we should graph the actual vs historical predictions to make
sure...


: Steve Allen pointed at some interesting papers, of which the most relevant

: was <http://www.ucolick.org/~sla/leapsecs/torino/arias_3.pdf>.  This paper

: looks at (retrospectively) predicting UT1-TAI two and three years ahead,

: and how well those predictions match reality.  These predictions were

: made with a fairly naive algorithm, which in the short term performs

: much more poorly than what IERS does.  The three year predictions were

: all correct to within 1.0 s.


I'd note that Bullitin A data is available, and one could graph the
performance of different time lines vs actual over a period of the
last few years.  I was thinking of doing this data crunching myself,
but time has gotten away from me...

Warner