NavList:

I've found an interesting article on lunar distance observations from 1889 Australia. The author is one E.J. White. He's a land-based lunarian writing to the "Royal Society of Victoria" in Melbourne extolling the virtues of longitude by lunar distances.

 

White notes in his article that this method isn't used at sea any longer, but he believes that it should be required on any vessel with fewer than three chronometers. He asks why the method has fallen into disuse. He notes that complaints about long calculations are not relevant since Thomson's Tables, as well as others, let a navigator clear a lunar in "ten minutes" (I would say twenty minutes but it depends on how you divide up the work). White also believes that the bad reputation of lunars is partially a hangover from the earlier period when the tables in the Nautical Almanac were intrinsically inaccurate due to incomplete models of the Moon's motion. That's no longer the case by 1889. Primarily he blames the continued inaccuracy of lunar observations on bad sextants, saying that "eccentric error" and "division error" (which we know collectively as "arc error") are increasingly common and frequently amount to several minutes of arc in sextants from this period. Contemporary (c.1889) navigational practice does not demand high-quality sextants. He blames it on "excessive competition". White then suggests that the best way to measure this arc error, apart from using collimating devices at an observatory, is to shoot lunars at a known longitude. He has some good practical advice and some good reading recommendations, too.

 

For clearing lunar distances White recommends Chauvenet's method. He is the first person from this period that I have found who actually used Chauvenet's method. He compares it with Borda's method and Thomson's Tables, which he says shouldn't necessarily be trusted for high accuracy. In fact in my opinion, if interpolated, clearing a lunar by Thomson's Tables should yield very nearly the same result as with Chauvenet's method.

 

White then details the lunar distance observations he has made in the past three years: forty-two lunars in 1887, 1888, and 1889. He lists the time and measured distance (only approximately, not enough detail to re-work them today) and also lists the error of the resulting longitude in seconds of time. We can convert this to an approximate error in the measured lunar by dividing by 120. That is, an error of 12 seconds in the longitude is approximately equivalent to an error of 0.1 minutes of arc in the measured lunar distance. Calculating from his listed numbers, I find he has a standard deviation in Greenwich time of about 30 seconds which is equivalent to about 0.25 minutes of arc in the lunar distances. If we take his lunars in sets of four and average them (which I consider the best approach with lunars), the results are generally within 0.1 minutes of arc. I would note that these results are very similar to my own experience.

 

In his conclusion, White writes that he hopes his paper "may contribute toward rescuing this once favorite method of finding the longitude from the neglect into which it has undeservedly fallen". Of course, these are the words of a lunarian, and it's the sort of sentimental hope you find again and again in middle and late 19th century articles on lunar distances. Lunars by this late date were already long over and done with. Right at the beginning of his paper he has stated the obvious alternative: "it is criminal to rely upon [the chronometer method] for long voyages when less than three chronometers are carried". So carry three chronometers then... It may well have been cheaper in 1889 to supply a vessel with three chronometers than to provide a navigator with the skill and high-quality sextant required to work with lunar distances.

 

I found White's article originally while digging around in Google Books. Since it's short, I've also made it available on the web here: www.HistoricalAtlas.com/lunars/ejwhite.

 

-FER
www.HistoricalAtlas.com/lunars