[LEAPSECS] Predicting the next leap second

[Demetrios Matsakis]

 I think this plot will answer Bob's questions - or at least this round of them:) 

It is the Length of Day (negative of the slope) going back to 1992.
I think Tom has  a good point, that you get negative leap seconds only when the true slope is negative for long enough. 

I am not and have never been party to the actual IERS deliberations - I don't even know how much is public.  But I fully believe that the final decision is based on everything Bob, Tom, and I have pointed out.   

Note that in the attachment all the negative values in the true LOD, that occurred around 2005 (dipping below 3 milliseconds in the shifted red curve of the plot), weren't so serious because (like the so-called post-convention bounces in American politics), they had  a predictable oscillatory contribution.   But the current negative LOD values, if they continue, might be part of a different trend (like the 1988 post-convention jump of G.H.Busch, who went on to solidly win the election).
P.S. If you un-shift the red true-LOD curve in the plot, it falls on top of the other one and makes a big mess.   Truth can be so annoying sometimes.

    On Monday, August 24, 2020, 12:49:51 AM EDT, Seaman, Robert Lewis - (rseaman) <rseaman at arizona.edu> wrote:  
 
 
Hi Demetrios,
 
  
 
Indeed an interesting plot. But if the issue is whether the horizontal trend will continue, or perhaps resume its usual negative slope, or even turn positive, well then, it’s good there is a half second leeway on top of the anticipated -0.4 second UT1-UTC threshold that triggered a decision last time.
 
  
 
It would be fascinating to finally see a negative leap second (by all means, everybody please discuss the implications), but it’s hard to interpret a roughly zero slope within 0.2 s offset of nominal as a crisis. What has been the maximum (negative or positive) slope over the past five decades? What has been the maximum annual deviation over that time from the extrapolated trend? How often has the second derivative been positive versus negative? (For either the red or blue lines.)
 
  
 
Rob Seaman
 
Lunar and Planetary Laboratory
 
University of Arizona
 
--
 
  
 
From: Demetrios Matsakis via LEAPSECS
 
Date: Sunday, August 23, 2020 at 8:32 PM
 
  
 
Here is  a plot that might interest some of you.  The blue curve  is  UT1-UTC, and you can see when the last leap second was inserted.  The goal is to be sure it never goes below -0.9, and the IERS obviously took no chances back then.   But remember that if they are making a decision 5 months in advance, the time of interest is 11 months later.   (The decision made last July was to forestall being too negative on the next June 30.)
 
  
 
But UT1-UTC is distorted by seasonal and lunar terms (the largest one being a 19 year cycle, with peak to peak about 0.3 seconds).  You can see a seasonal variation in the blue curve.  The red curve shows what happens if you take them out (the formulas are in the IERS Standards).   To predict what the IERS will decide about a possible June 30 leap second next year, I recommend you take your attention from the blue to the red, decide how to extrapolate the red curve to December 2021, and then map it back to the blue.   Then you can place your bet.  But will you win?  Don't ask me.
 
  
 
Demetrios, who is actually being paid to work on predicting this kind of thing.
 
  
 



   
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://pairlist6.pair.net/pipermail/leapsecs/attachments/20200824/1c129bec/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: LOD and LOD_ut2s.jpg
Type: image/jpeg
Size: 842250 bytes
Desc: not available
URL: <https://pairlist6.pair.net/pipermail/leapsecs/attachments/20200824/1c129bec/attachment.jpg>