A Community Devoted to the Preservation and Practice of Celestial Navigation and Other Methods of Traditional Position-Finding
From: Frank Reed
Date: 2018 Sep 11, 19:59 -0700
As we've discussed a number of times over the years, the definition of the Greenwich date and GMT itself changed at the beginning of 1925. From the beginning of that year forward, astronomers (and celestial navigators who depended on their data) adopted the civil definition of the day by which the date and the name of the day roll over at midnight. As I write this, in a little over an hour here in the eastern US, it will be midnight, and Tuesday the 11th will become Wednesday the 12th. Before 1925, astronomers used an alternate definition of the day which was convenient for the subject of astronomy: the day rolled over at noon, and thus an entire night's astronomical observations would fall on the same date. An astronomer could observe Jupiter at 10pm, 11pm, 1am, 2am, and all of those observations would be listed on the same calendar date. Of course we still have some lingering date usage on this standard, especially when talking about the weather. We routinely talk about Tuesday night's weather for the weather that starts after dark on Tuesday evening and continues until dawn on Wednesday. It's all called "Tuesday night" when discussing the weather. But technically, ever since 1925, anything after midnight should really be referred to as Wednesday morning... But let's not be picky, right?
Considering the change in the date standard occurred over ninety years ago, you might expect there would be no lasting issues resulting from it. But it turns out there's a little astronomy news today which it seems depends on this change in the definition of the day. The lingering problem today is that we still maintain dates by the "Julian date" running day count, which became common in astronomy in the 18th and 19th centuries (though its zero date is set back before 4700 BC). And this day count, even now, starts at noon in order to maintain compatibility with older observational data. Thus at midnight tonight, the Julian date is 2458373.5. That 0.5 added on the end is the extra half a day that has elapsed since noon. Unfortunately, bad habits developed in the early digital era. Maintaining all those digits in a Julian date caused some technical problems on some computer systems, so it became customary to decapitate or otherwise shorten the date in various ways, for example by subtracting 2400000 from all Julian dates, making today's so-called "Reduced JD" equal to 58373.5 (there have been other variants on this scheme). And on many early computer systems, it was more convenient to count off dates as integers. So drop that 0.5, and we'll fix it in post!
The story in this week's news has been written up in an interesting article on the Sky & Telecope website which characterizes the problem as a "typo" (in a way, it is, but I would call this a data curation problem instead). A binary star in a tight orbit was projected to merge into a single star beginning possibly as early as 2022 based on an old observation that had been mislabeled using the wrong Julian Date. They had forgotten the 0.5 day offset. That slight nudging of the data was enough to bias the entire result and suggest an imminent merger of the two stars. Take away that 0.5 day offset on that single data point, and the possibility of a merger in the immediate future seems to evaporate...