# Modified Julian Date (MJD)'s epoch is 0:00 on November 17, 1858 (JD − 2400000.5) has fewer digits than JD. But what's so special about November 1858?

A week ago in the Observatory chat room @HDE226868 wrote:

Happy MJD 60000, everyone!

Thank you!

According to Wikipedia's Julian day; variants the epoch for Reduced JD is 12:00, November 16, 1858 (making it JD − 2400000) and for Modified JD (MJD) it's 0:00 November 17, 1858 (making it JD − 2400000.5)

Theres a note in that table that says that it was introduced by the Smithsonian Astrophysical Observatory in 1957

I can understand the convenience of reducing the number of digits, but what's so special about November 1858?

But what's so special about November 1858?

Nothing, other than that was when JD 2400000 occurred. Look at all of the zeros! The next time we'll see that many zeros will be noon on August 31, 2132.

It is not possible to achieve microsecond level accuracy if one expresses time as a double precision Julian Date. (It will be possible to maintain millisecond level accuracy using a Julian Date for another 100000 years.) Until recently (August 3, 2001), microsecond level precision was possible using a Modified Julian Date.

An alternative is to use the IAU SOFA convention, in which one specifies time as a Julian date using a pair of doubles that conceptually add to the time in question. Typically, the first of the pair is some value that can be represented exactly (e.g., 2460007.5) and the second is an offset from that base time. The recommended practice is to use the Julian date of the J2000 epoch (2451545.0, or noon January 1, 2000) as the first of the two doubles. With this approach, astronomers can easily achieve microsecond level precision until September 18, 2142.

• SOFA actually recommends using 2451545.0 (Jan 1 2000) for the first of the pair. Because most of the algorithms are based on elapsed time since then. Commented Mar 4, 2023 at 13:22
• @GregMiller That's because the SOFA code is chockfull of code such as (date1-DJ00)+date2 where DJ00 is 2451545.0. This means nanosecond accuracy was lost on 22 February 2000, and microsecond level accuracy will be lost on 18 September 2142. I updated my answer. Commented Mar 4, 2023 at 15:21
• ... which makes me wonder why they chose double float numbers in the first place. A regular 64 bit integer would be sufficient for 585 years of nanosecond precision, not just two months. Commented Mar 5, 2023 at 10:46
• @asdfex The problem with integers as represented on computers is they have minimum and maximum values. Some scientists want to look more than 585 years into the past or more than 585 years into the future. Besides, most astronomers do not need nanosecond level accuracy in ephemerides or Earth rotation models. Commented Mar 9, 2023 at 9:21

Nothing special occurred on that day, except that it was 2400000 days after January 1, 4713 BC proleptic Julian calendar. That latter date is significant as it year 1 in the solar, lunar and indiction cycles simultaneously

As the number of days in a year is not a round number, subtracting a large round number from the Julian day results in an epoch that fairly arbitrary. The choice to modify the Julian day number by 2400000 is to allow very simple conversion between mjd and jd.

So the only thing special about this day is that it has an Julian day nubmer of 2400000.

• I also liked your answer to the out-of-focus satellite question and was going to note that it seems that once upon a time I did know of that spiky CCD artifact space.stackexchange.com/a/35058/12102 I think future readers will benefit from seeing your answer if you decide to undelete it.
– uhoh
Commented Mar 6, 2023 at 17:36