A suggestion for improving the accuracy of LMT of Asia/Tokyo

Fumiyuki Shimizu 清水文行 fumifumi at abacustech.jp
Mon Jan 24 07:49:54 UTC 2011


Thank you for the message.

>>>>> Regarding Re: A suggestion for improving the accuracy of LMT of Asia/Tokyo;
>>>>> Alan Barrett <apb at cequrux.com> adds:

>> With regards to the current LMT of Asia/Tokyo, there are two problems.
>> 1. It is based on 'Tokyo Datum' using Bessel ellipsoid of which
>> equatorial radius, and flattening are less accurate than GRS80.
>> 2. Location of origin point used by 'Rika Nenpyou (until 2009)' was
>> not commonly used for geodetic reference system in Japan.

 > That line in the rules ...

>> # Zone NAME            GMTOFF  RULES   FORMAT  [UNTIL]
>> Zone   Asia/Tokyo      9:18:59 -       LMT     1887 Dec 31 15:00u

 > ... is talking about the local mean time as it was before 31 Dec 1887;
 > it's not the "current" LMT.

I see.  This leads me to believe that there is one more issue of LMT
of Asia/Tokyo in the current(2010o) tz database:

3. The value(9:18:59) of LMT of Asia/Tokyo, which is for time before
1888-01-01 00:00, is based on the location of an instrument of the
Tokyo Astronomical Observatory, which was established in 1888[*12].
[*12] http://www.nao.ac.jp/E/about/history.html

 > The oldest reference in your message seemed to be for the year
 > 1892.  Do you have information about the official time in Tokyo
 > before or during the year 1887?

No, I don't, but it seems correct that the known official time in
Tokyo has started at 1888-01-01 00:00, and lines of Asia/Tokyo seems
correct except for the value 9:18:59.

In the tz database, I believe LMTs are not the official time, but are
mostly the solar time based on the longitude of the typical place,
measured with the current technology, such as GPS, to substitute for
the culmination time of the Sun in the equation of time, one of the
reasons for which may be to fill the absence of the official time. Is
my assumption correct? It is with this assumption that I wrote in my
previous mail that the priority of pursuing accuracy of LMT might be

Please let me revise my explanation about LMT and the definition of
"Tokyo" based on the information you've given me.

The value(9:18:59) of LMT of Asia/Tokyo in tz database(2010o) is based
on 'Rika Nenpyo 1996' as described in the file. 'Rika Nenpyo 1996' set
the site of the center point of the Repsold transit instrument of the
former Tokyo Astronomical Observatory as 'Tokyo,' and displayed the
location in Tokyo Datum. Tokyo Datum is not suitable for a solar time
*even in the past*.

In other words, the value(9:18:59) is neither the official time nor
accurate as the solar time because it is expressed in Tokyo Datum, the
obsolete geodetic reference system in Japan(#1). Most LMTs in the tz
database may be based on locations expressed in WGS84 and its variant,
but the value(9:18:59) of Asia/Tokyo is in Tokyo Datum, which is
9:18:58 in WGS84 as described in the previous mail(#1).

On the other hand, the Tokyo Astronomical Observatory was established
in 1888. If this raises a question 'What is Tokyo?' about
current(2010o) LMT of Asia/Tokyo which is for time before 1888-01-01
00:00, the following information might help:

Edo was renamed to Tokyo on September 3rd, 1868, by "Shousho"(Emperor
expressed himself on official business). [*13]
[*13] http://ja.wikisource.org/wiki/%E6%B1%9F%E6%88%B8%E3%83%B2%E7%A8%B1%E3%82%B7%E3%83%86%E6%9D%B1%E4%BA%AC%E3%83%88%E7%88%B2%E3%82%B9%E3%83%8E%E8%A9%94%E6%9B%B8
(in old Japanese)

It seems that the symbol of Tokyo in 1868 was Edo-jo castle[*14].
To fix problem #3 by using the location of Edo-jo castle,
E 139 45' 15.72" [*15]([*16]),
  => 9h 19m 01s
[*14] http://en.wikipedia.org/wiki/Edo_Castle
[*15] http://ja.wikipedia.org/wiki/%E6%B1%9F%E6%88%B8%E5%9F%8E (in
[*16] http://toolserver.org/~geohack/en/35_41_17.97_N_139_45_15.72_E_scale:20000?pagename=%E6%B1%9F%E6%88%B8%E5%9F%8E

--- tzdata2010o/asia    2010-11-01 22:18:22.000000000 +0900
+++ tzdata/asia 2011-01-22 22:07:54.000000000 +0900
@@ -1280,7 +1280,7 @@
 # ordinances took effect on Jan 1.
 # Zone NAME            GMTOFF  RULES   FORMAT  [UNTIL]
-Zone   Asia/Tokyo      9:18:59 -       LMT     1887 Dec 31 15:00u
+Zone   Asia/Tokyo      9:19:01 -       LMT     1887 Dec 31 15:00u
                        9:00    -       JST     1896
                        9:00    -       CJT     1938
                        9:00    Japan   J%sT
As for the time before September 3rd, 1868, it is uncertain as to how
it should be expressed, because:
1868 Edo was renamed to 'Tokyo.' (no official time)
1457 Edo castle was built. (no official time)

I hope this information helps.

'AT' fields in lines starting with 'Rule Japan' seems to be wrong(it
was 0:00, not 2:00) and I have ordered the book written by 'Shanks &
Pottenger,' which is referenced by those values to search the root
cause of the difference. I might post a patch in the future.

Thank you and best regards,
Fumiyuki Shimizu
Abacus Technologies, Inc.

More information about the tz mailing list