Based on the value of 365.242199 days for the mean tropical year, and of 29.530588853 days for the mean synodic month, I worked out a calendar that was based on these values (rather than on astronomical observations of actual new moons) that would, using a standard scheme of alternating between months and years of different lengths, the same way that the Gregorian calendar uses a fixed rule for leap years, remain in step with both the sun and the moon.
There are three basic types of year, although the year with an intercalary month, called the long year, can have that extra month at different positions during the year (thus keeping the other months closer to their appropriate seasonal times, by making the interval between intercalary months either 32 or 33 months long).
The types of years used in the calendar are as follows:
No year is both long and leap; the twelfth month only has 30 days in years with only twelve months.
There are two basic types of "cycle" from which this calendar is built. The most common one is the well-known Metonic cycle, consisting of 19 years, seven of which have an extra intercalary month. After seventeen or eighteen of these cycles in a row, however, there is a sufficient build-up of error due to a difference between 19 tropical years and 235 synodic months, that a second type of cycle, consisting of 11 years, four of which have an intercalary month, is used to restore the balance.
The cycles are constructed to be symmetrical, and for each type of cycle, the months followed by an intercalary month are fixed. This means that the months do not quite coincide as well with their proper seasonal placement as would be the case if some variation were allowed, for example, by changing the positions of the intercalary months in the first few and the last few cycles in a stretch of 17 or 18 normal cycles. However, this calendar is complicated enough!
Here are the types of cycle used in the calendar:
These cycles are chosen so as to be symmetric and smooth. The normal cycle, corresponding to the Metonic cycle, is the most common, but occasional short cycles, one short cycle for every 17 or 18 normal cycles, are required to cope with a slight discrepancy between 19 tropical years (6939.60178 days) and 235 synodic months (6939.6883804 days).
A GROUP is either 35 or 52 years long, and is built from the following items:
The types of groups are:
A LONG GROUP consists of:
A SHORT GROUP, which is either an EARLY SHORT GROUP or a LATE SHORT GROUP, consists of:
Finally, we proceed to the highest-level structure in the calendar. Well, almost. A ROUND is 6,479 years long, consisting of five long groups and two short groups, and one extra day is added to one round in every five to make this calendar just about as accurate as the calculations on which it was based, carried out on my trusty old Texas Instruments SR-56 programmable calculator.
A round is built from the following groups, in order:
Every group of five rounds has one round, the third, as a leap round.
This corresponds very closely to the proper average length of a round, consisting of 6,479 tropical years or 80,134 synodic months, of 2,366,404.207 days.
Incidentally, a group of five rounds with one leap round, and a normal round, both take five days longer than an even number of weeks, thus, it will take a full thirty-five rounds, comprising 226,765 years, for the sun, the moon, and the week all to coincide once again.
Well, it's all very well to describe a complicated calendar. But it is not very useful if it isn't possible to say what year it is by that calendar.
Of course, one could start the calendar at any time one liked. But let us suppose a conventional starting point: let each month attempt to begin approximately on the new moon, and let the year begin at the first new moon on or after the vernal equinox; the conventional "first day of spring" that takes place around March 21st on the conventional calendar.
Thus, the first day of the year on this calendar could, at its extreme earliest, fall on a day when the mean new moon takes place at 00:00:01 AM and the actual vernal equinox takes place at 12:59:59 PM.
Given a tropical year of 365.242199 days, the gain and loss through various types of years and cycles is as follows (in parentheses the gain and loss if we assume the year is composed of ideal lunar months, rather than 29 or 30 day months, is shown):
From examining ephemerides, it appears that a fit may be obtained with a special long group that had started in the year 1495. This places us in a round that started in 1224 B.C.
The following set of tables will allow one to determine the Julian Day Number of the beginning of any month within a round by this system.
Displacements of groups within a round:
Displacements of short cycles and stretches within a group:
Displacements of cycles within a stretch:
Displacements of years within a cycle:
Displacements of months within a year:
Thus: March 20, 2004 (J.D. 2453085) is the first day of the fifth year in a leap short cycle. Thus, that leap short cycle began 1447 days earlier, on J.D. 2451638 (April 3, 2000). This leap short cycle followed a special stretch of 17 long cycles, so it occurred 184447 days after the beginning of a special long group, which started on J.D. 2267191 (March 26, 1495, by the Julian calendar). A special long group occurs 996746 days after the beginning of a round, on J.D. 1270445 (April 16, 1235 B.C.).
The two calendars which use a lunar month, and which have an extra month in some years to keep in line with the seasons, that may be familiar to most people are the Hebrew calendar and the Chinese calendar.
The Hebrew calendar, used for keeping track of Jewish religious festivals, is calculated according to rules which are made more complicated by the need to adjust some months to prevent certain religious holidays from falling on certain days of the week.
This calendar is based on the 19-year Metonic cycle. Leap years occur in the 3rd, 6th, 8th, 11th, 14th, 17th, and 19th years of the cycle, and the added month is always known as ve-Adar; Adar is the 12th month of the year as the months were numbered in the Bible, but today the Jewish New Year is celebrated on Rosh Hashanah, the first of Tishri, the 7th month in the original numbering, and thus Adar is the 6th month in the current numbering. The month corresponding to Nisan in the Babylonian calendar, Nisannu, was the first month of the year in that calendar as well.
The Metonic cycle in use for the Jewish calendar when this page was first written was as follows:
1 October 2, 1997 5758 8* September 16, 2004 5765 15 September 29, 2011 5772 2 September 21, 1998 5759 9 October 4, 2005 5766 16 September 17, 2012 5773 3* September 11, 1999 5760 10 September 23, 2006 5767 17* September 5, 2013 5774 4 September 30, 2000 5761 11* September 13, 2007 5768 18 September 25, 2014 5775 5 September 18, 2001 5762 12 September 30, 2008 5769 19* September 14, 2015 5776 6* September 7, 2002 5763 13 September 19, 2009 5770 7 September 27, 2003 5764 14* September 9, 2010 5771
but time has marched on, and the new Metonic cycle currently in effect for the Jewish calendar is:
1 October 3, 2016 5777 8* September 16, 2023 5784 15 September 28, 2030 5791 2 September 21, 2017 5778 9 October 3, 2024 5785 16 September 18, 2031 5792 3* September 10, 2018 5779 10 September 23, 2025 5786 17* September 6, 2032 5793 4 September 30, 2019 5780 11* September 12, 2026 5787 18 September 24, 2033 5794 5 September 19, 2020 5781 12 October 2, 2027 5788 19* September 14, 2034 5795 6* September 7, 2021 5782 13 September 21, 2028 5789 7 September 26, 2022 5783 14* September 10, 2029 5790
The names of the months in the ancient Babylonian calendar, and the corresponding month names in the religious calendar of Judaism, as well as those in the Chinese calendar and the ancient calendar of Athens, are listed below for reference:
Babylonian Jewish Athenian Chinese
------------------------------------------------------------------------------------------------
Spring
Aries 1 Nisannu 7 Nisan 10 Munichion 4 Er Yue Pink/Milk/Growing
Taurus 2 Aru 8 Iyyar 11 Thargelion 5 San Yue Flower Moon (Hare)
Gemini 3 Simanu 9 Sivan 12 Skirrophorion 6 Si Yue Strawberry/Mead
Summer
Cancer 4 Du'uzu 10 Tammuz 1 Hecatombaeon 7 Wu Yue Buck/Grain
Leo 5 Abu 11 Ab 2 Metageitonion 8 Liu Yue Sturgeon/Fruit/Corn
Virgo 6 Ululu 12 Elul 3 Boedromion 9 Ch'i Yue Harvest Moon (Wine)
Fall
Libra 7 Tasritu 1 Tishri 4 Maimakterion 10 Pa Yue Hunter's Moon
Scorpio 8 Arahsamna 2 Marheshvan 5 Pyanepsion 11 Chiu Yue Beaver/Oak
Saggitarius 9 Kislimu 3 Kislev 6 Poseideon 12 Shi Yue Cold/Old/Winter
Winter
Capricorn 10 Tebetu 4 Tevet 7 Gamelion 1 Shi Yi Yue Wolf Moon
Aquarius 11 Sabatu 5 Shevat 8 Anthesterion 2 Shi Er Yue Snow/Lenten/Ice
Pisces 12 Addaru 6 Adar 9 Elaphebolion 3 Cheng Yue Worm/Egg/Storm
The Harvest Moon is the Full Moon that comes closest to the Autumnal Equinox, and thus the New Moon that follows it would be the first one on or after the Autumnal Equinox, which is why it is placed at the end of Summer rather than at the beginning of Fall, so that it might be consistent with the months listed here.
The Chinese calendar is based on direct observation, and the intercalary month can occur in any part of the year, as the months are based on the signs of the Zodiac.
For the nineteen years from 2002 to 2020, the date of Chinese New Year was or will be:
2002 February 12 2006 January 29 2010 February 14 2014 January 31 2018 February 16 2003 February 1 2007 February 18 2011 February 3 2015 February 19 2019 February 5 2004 January 22 2008 February 7 2012 January 23 2016 February 8 2020 January 25 2005 February 9 2009 January 26 2013 February 10 2017 January 28
For the 19 years from 2021 to 2039, the sequence is expected to be almost the same, except that 2024 is one day later (February 10), 2026 amd 2027 are one day earlier (February 17, February 6), 2029 is one day earlier (February 13), 2032 is one day later (February 11), and the final three years of the cycle, 2037, 2038, and 2039 are all one day earlier (February 15, February 4, January 24).
Since the cycle of lunar phases is about 29 and a half days, and there are an odd number of lunations in a Metonic cycle, there may be fewer differences in the Metonic cycle after.
In the ancient Babylonian calendar, an extra month was added in the 3rd, 6th, 8th, 11th, 14th, 17th, and 19th years of a 19-year cycle, but while a second Addaru (Adar) was added most of the time, in the 17th year of the cycle, a second Ululu (Elul) was added instead. This somewhat evened out the spacing of intercalary months, keeping the year more closely in harmony with the seasons.
Thus, the time between intercalary months was 3, 3, 2, 3, 3, 2 1/2 and 2 1/2 years. This also made the cycle symmetric; so changing to a second Ululu in the 6th year of the cycle, or replacing the second Addaru in the 8th year of the cycle by a second Ululu in the 9th year would have broken the symmetry again.
Had it been desired to retain symmetry, and stay even closer to the seasons, of course, they could have gone with a second Arahsamna (Marheshvan) in the 6th year of the cycle, and a second Du'uzu (Tammuz) in the 9th year of the cycle, for a spacing between intercalary months of 3, 2 2/3, 2 2/3, 2 2/3, 3, 2 1/2, and 2 1/2 years, but Darius no doubt thought that would be too complicated. (Actually, of course, one could doubt that the idea even crossed his mind.)
In the normal cycle shown above, the extra months are added in the 2nd, 5th, 7th, 10th, 13th, 15th, and 18th years of the cycle, and so the first year of the Babylonian cycle corresponds to the eighth year of the normal cycle, and the first year of the normal cycle corresponds to the thirteenth year of the Babylonian cycle.
Thus, this table shows the possible leap months in a Metonic cycle in these schemes, using the month names of the modern Jewish religious calendar as being somewhat more familiar than those of the ancient Babylonian calendar:
Hebrew Babylonian Modified Normal
3rd year ve-Adar ve-Adar ve-Adar ve-Adar
6th year ve-Adar ve-Adar ve-Marheshvan ve-Kislev
8th year ve-Adar ve-Adar ve-Tammuz ve-Ab
11th year ve-Adar ve-Adar ve-Adar ve-Iyar
14th year ve-Adar ve-Adar ve-Adar ve-Tevet
17th year ve-Adar ve-Elul ve-Elul ve-Tishri
19th year ve-Adar ve-Adar ve-Adar ve-Sivan
The normal cycle and the short cycle, considered in terms of their own starting and ending times as given above, would work like this:
Normal Cycle (19 years) Short Cycle (11 years) 2nd year ve-Tevet 2nd year ve-Shevat 5th year ve-Tishri 5th year ve-Marheshvan 7th year ve-Sivan 7th year ve-Tammuz 10th year ve-Adar 10th year ve-Nisan 13th year ve-Kislev 15th year ve-Ab 18th year vi-Iyyar
In general, the lunisolar calendars are based on the Metonic cycle of 19 lunar years - 12 normal years of 12 months and 7 leap years of 13 months:
19 * 365.242199 = 6939.60178 235 * 29.530588853 = 6939.68838
The ancient Egyptians, who used a vague year of 365 days without any leap years, used a similar cycle to keep in step with that year, of 25 lunar years - 16 normal years of 12 months and 9 leap years of 13 months:
25 * 365 = 9125 309 * 29.530588853 = 9124.95196
Note that here the discrepancy is .05 of a day instead of .08 of a day, so this is a more accurate cycle for its purpose, even if fundamentally less useful.
This calendar was described, using the Carlsberg IX papyrus as a source, in The Calendars of Ancient Egypt by Richard A. Parker.
His reconstruction of the 25-year cycle proceeds as follows:
L L L L L L L L L
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
--------------------------------------------------------------------------
Thoth 30 30 1 20 9 28 18 7 26 15 4 24 13 2 21 10 30 19 8 27 16 6 25 14 3 22 12
Paophi 30 29 1 20 9 28 18 7 26 15 4 24 13 2 21 10 30 19 8 27 16 6 25 14 3 22 12
Hathor 30 30 * 19 8 27 17 6 25 14 3 23 12 1 20 9 29 18 7 26 15 5 24 13 2 21 11
Choiak 30 29 30 19 8 27 17 6 25 14 3 23 12 1 20 9 29 18 7 26 15 5 24 13 2 21 11
Tybi 30 30 29 18 7 26 16 5 24 13 2 22 11 * 19 8 28 17 6 25 14 4 23 12 1 20 10
Mechir 30 30 29 18 7 26 16 5 24 13 2 22 11 30 19 8 28 17 6 25 14 4 23 12 1 20 10
Phamenoth 30 29 29 18 7 26 16 5 24 13 2 22 11 30 19 8 28 17 6 25 14 4 23 12 * 20 10
Pharmouthi 30 29 28 17 6 25 15 4 23 12 1 21 10 29 18 7 27 16 5 24 13 3 22 11 30 19 9
Pachon 30 30 27 16 5 24 14 3 22 11 * 20 9 28 17 6 26 15 4 23 12 2 21 10 29 18 8
Payni 30 30 27 16 5 24 14 3 22 11 30 20 9 28 17 6 26 15 4 23 12 2 21 10 29 18 8
Epiphi 30 29 27 16 5 24 14 3 22 11 30 20 9 28 17 6 26 15 4 23 12 2 21 10 29 18 8
Mesore 30 (29) 26 15 4 23 13 2 21 10 29 19 8 27 16 5 25 14 3 22 11 1 20 9 28 17 7
Epagomenes 5 29 4 2 5 3 1
In the first three columns, I give the name of each month of the Egyptian vague year, and then the number of days in that month, and then the normal number of days of a lunar month beginning in that month.
In the case of a lunar month beginning in Mesore, it will be 30 days long if it ends in the Epagomenes of the same year, but only 29 days long if it ends in Thoth of the next year.
A leap year can originate from two causes: a 29-day lunar month can begin on the first day of the month in the vague year, causing another month to begin on the thirtieth day of the same month, or a month can begin in the Epagomenes.
The chart of the 25 years in the cycle shows the dates in each month of the vague year where a lunar month begins. An asterisk indicates where lunar months begin on both the 1st and the 30th days of a given month of the vague year.
As can be seen from the chart, when lunar months begin on the first day of a month of the vague year, their length is governed by the normal length applicable to the preceding month; this delays the occurrence of the first days of two lunar months within a single month of the vague year by one month. This is similar to the case of Mesore; when the lunar month beginning in Mesore begins early enough so that it will cause a leap year by having the next month begin in the Epagomenes, it is lengthened to 30 days. This rule can avoid ambiguous cases in the calendar.
In the 23rd year of the cycle, however, that rule is not followed, because following it would delay the dual month from Phamenoth to Pharmouthi, and both of these months have 29 days as the normal length of a lunar month beginning within them, and so, presumably, the usual benefit from the delay is not gained.
Our Gregorian calendar usually has one leap year every four years, and sometimes omits the leap year entirely rather than having some leap years at five year intervals, thus introducing odd numbers in the calendar. When one is adding a whole month, rather than just a day, to the calendar, simplicity may have to take a back seat to minimizing the maximum deviation of the calendar from the seasons.
However, the idea is tempting to use four of these Egyptian cycles in a century, and then add one extra leap year occasionally to keep the calendar in step with the tropical year instead of the vague year.
A conventionalized luni-solar calendar of this type might run as follows:
01 04 06 09 12 15 17 20 23 26 29 31 34 37 40 42 45 48 51 54 56 59 62 65 67 70 73 76 79 81 84 87 90 92 95 98
would be the digits ending each leap year, with the year ending in 99 occasionally being a leap year as well.
Four 25-year cycles adding up to 36499.80782 days, this is short of 100 tropical years by 24.4120777 days. This means that one extra leap month in a century would be the usual case, so the century should perhaps instead have leap years like this:
01 04 06 09 12 15 17 20 23 26 29 31 34 37 40 42 45 48 51 54 56 59 62 65 67 70 73 76 79 81 84 87 90 92 95 97 99
with the year ending in 97 not being a leap year in one in every six centuries, or, as 5.7693708 would be the more exact figure, perhaps not being a leap year four times in every 23 centuries.
A structure with cycles of this type, to reduce the discrepancy between the calendar and its astronomical ideal, by delaying leap days instead of omitting them, has been used in one real calendar.
A proposed algorithmic version of the Persian calendar alternates between cycles 29 and 33 years in length, which begin with four non-leap years followed by a leap year, and then continue with groups of three non-leap years followed by a leap year.
However, at the end of a period of 2820 years, consisting of groups of one 29 year cycle followed by three 33 year cycles, the final cycle is made 37 years long, instead of replacing a 29 year cycle by a 33 year cycle, and distributing the cycles evenly by following some 29 year cycles by four (instead of three) 33 year cycles. Thus, the principle of gradualism is not followed consistently into this higher level, but instead the suddenness found in our Gregorian calendar appears at that level.
I believe this would work out as follows:
Instead of 21 cycles of 29,33,33,33 followed by one cycle of 29,33,33,37, there would now be a total of 21 cycles in 2,820 years; four cycles of 29,33,33,33,33 and seventeen cycles of 29,33,33,33.
And, so, to follow the pattern of the rest of the calendar, first the pattern
29,33,33,33,33, 29,33,33,33, 29,33,33,33, 29,33,33,33, 29,33,33,33, 29,33,33,33
would occur once, and then the pattern
29,33,33,33,33, 29,33,33,33, 29,33,33,33, 29,33,33,33, 29,33,33,33
would occur three times.
Perhaps a way to make a lunisolar calendar simpler would be to put the variation in an associated solar calendar, because we're used to the kind of interpolation it requires.
That is: 235 lunations take 6939.68838 days, so one could think in terms of a Metonic year of 365.246756866 days, one-nineteenth of this period.
This solar calendar would perhaps work like the Gregorian calendar, but slightly changed: omit leap years for the century years, but then put them back in two-thirds of the centuries.
Then, the real solar calendar would also be similar to the Gregorian calendar, but it would be more accurate, and mesh better with this calendar, by following the rule given by the Orthodox church: omit the leap years for the century years, but then put them back in two centuries out of every nine.
So we have a strictly Metonic cycle with respect to the Metonic solar calendar, and over a span of 900 years, the Metonic calendar will move with respect to the solar calendar by four days.
Nineteen years in the Julian calendar, depending on the year on which they begin, may contain either five leap years or four, to span either 6,940 days or 6,939 days. The Metonic calendar will be the one used to determine the lengths of the lunar months; if a span of nineteen years would have 6,938 days due to an omitted leap year, instead, the start of the next span is delayed one day: since spans of 6,939 days are the less common ones, that should be the very next span thereafter as two of them will not be consecutive - and so it is not necessary to add, with the delay continuing until it can be restored by turning a span of 6,940 days into one of 6,939 days.
But the names of the months, and how they are divided into years, would be determined by the Solar calendar.
The goal of this construction will be to construct a conventional calendar which approximates the Chinese calendar, which, since 1645, has been determined by complicated astronomical calculations. In this respect, it might be noted that in one epoch used for the Chinese calendar, the year that will start on February 10, 2013 is considered to be the year 4711.
As a very rough approximation to the actual time the Sun spends in each sign of the tropical zodiac, the Solar calendar can be constructed as follows:
Taurus, Gemini, Cancer, Leo and Virgo have 31 days;
Libra, Scorpio, Sagittarius, Capricorn, Aquarius, and Pisces have 30 days;
and Aries has 30 days, except in leap years, where it has 31.
To minimize the February 29 birthday problem, the calendar could attempt to keep the same lunar months as 29 days and 30 days each year. However, this runs into the problem that it would further bias which months would be duplicated as intercalary months.
6,940 days amounts to 235 lunations of which 110 are 29 days long, and 125 are 30 days long; the other case is 6,939 days, with 111 lunations of 29 days, and 124 of 30 days.
Splitting them neatly between the 7 intercalary lunations and the other 228 lunations poses a minor dilemma.
Just making the 7 intercalary lunations all 30 days long leaves us with 110 lunations of 29 days and 118 lunations of 30 days, instead of a half-and-half split of 114 and 114, so we have four extra long months to distribute.If, instead, we make all the intercalary lunations only 29 days long, we now have eleven extra long months to distribute.
So having two lengths of intercalary month won't be enough to make all the years of each of the two types otherwise the same as each other if we exclude the alternative of adding the leap day to the seven years to which an extra month has already been added.
The solution is to avoid conventionalizing the calendar in this respect. Alternate 30-day months and 29-day months through the cycle independently of the year.
The Muslim world uses a purely lunar calendar, in which the year always consists of 12 lunar months. This makes it more difficult to relate to the purely solar Gregorian calendar, in some ways, than a luni-solar calendar which attempts to keep in step with the solar year, despite the greater complexity of luni-solar calendars.
Thus, 12 Metonic cycles, each containing 19 solar years, for a total of 228 solar years, would contain 235 lunar years.
Thus, since 1 A. H. began in the year 622, we can have a short table of corresponding years:
A.D. A.H. 622 1 (July 16, Julian) 850 236 1078 471 1306 706 1534 941 1762 1176 1990 1411 (July 24, Gregorian)