[ntpwg] Further to the timestamping issue

[Rob Seaman]

Replies to leapsecs.

On Jun 19, 2008, at 1:33 PM, Marshall Eubanks wrote:

> Leapseconds are driven by torques at the core mantle boundary, which  
> we cannot see directly
> and basically have information about only through UT1 and LOD  
> measurements. In practice, leap seconds cannot be predicted much  
> more than 6 months to a year in advance because of that.

Modeling of Earth rotation has become exquisitely good.  (Don't  
neglect information from seismology.)  Discussions on LEAPSECS suggest  
that useful predictions are possible several years in advance.  The  
horizon can be lengthened further by loosening the amplitude of DUT1 a  
skosh.  A consensus appears likely around a set of parameters allowing  
for about a decade lookahead.

Note that the related science of helioseismology has progressed to the  
point that sunspots can be "seen" on the far side of the Sun.  Hidden  
engines are studied quite productively throughout the sciences.

> AND, keep in mind that in the "jerk" seen around 1905 would have  
> meant several _negative_ leap seconds in one year, which would I am  
> sure be a fun experience if repeated today.

We don't have leap seconds due to current variations, but due to  
integrating the already accumulated LOD offset of a few milliseconds.   
As the tidal offset grows quadratically, it has become virtually  
impossible for a negative leap second to occur, since such a jerk as  
you describe would have to persist long enough to overcome the  
accumulated temporal moment arm.  The effective LOD zero point is in  
the nineteenth century, not the 1970's.

Rob Seaman
NOAO