Longitude data from the almanac were added up to the end of 2019 and the fit was updated. It was then discovered that the sunpy python package has a routine to calculate the start time for a given carrington rotation. The sunpy times match the Astronomical Almanac values closely enough for our needs, so sunpy is now used to calculate the CRN start and stop times. Our thanks to the maintainers of the package. Note that these times do not match the times when HGLONG changes from 0 to 360 in the SOHO orbit files.

Longitude data from the almanac were added for years past 2001. A plot of residuals using the 2001 fit shows that the fit is not as good for recent dates as it is for the original time period. So a new fit was done with the same fitting formula. The residuals are more stable with the new fit.

Traditionally the start and stop times of Carrington rotations are derived assuming that the solar rotation period as viewed from the Earth is a constant. This assumption is perfectly adequate for most studies. We've had several inquiries asking how we derived these times, so for academic purposes we describe below both the original derivation and the presently-used technique that accounts for the variable length of Carrington rotations.

The following quote (emphasis added) is from from the "Carrington and Bartels Calendars" page at Stanford University:

The rotation of the Sun and the orbital motion of the Earth are in the same direction (both counterclockwise when viewed from north of Earth's orbital plane). As a consequence the apparent (synodic) solar rotation period is a bit longer than the true (sidereal) period. (Hint: it may help to think of the sideREAL period as the 'real' period.) If you think about it you can convince yourself that the sidereal period is equal to the synodic period divided by (1+frac), where frac = 27.2753/365.25 = the fraction of a year the Earth moves in its orbit during one synodic solar rotation.Lord Carrington determined the solar rotation rate by watching low-latitude sunspots in the 1850s. He defined a fixed solar coordinate system that rotates in a sidereal frame exactly once every 25.38 days (Carrington, Observations of the Spots on the Sun, 1863, p 221, 244).

Thesynodicrotation rate varies a little during the year because of the eccentricity of the Earth's orbit; the mean synodic value is about 27.2753days. See the back of an Astronomical Almanac for details.

The Earth is farthest from the Sun in July and closest in January. Objects that are farther from the Sun have a smaller angular velocity (a longer 'year') than do closer objects. Hence the synodic period is a bit smaller in July than in January.

Previously we determined the start and stop times of Carrington rotations by assuming
a constant period of 27.2753 days. The phase was taken from the *Astronomical Almanac*,
which gives a value for Carrington longitude of 349.03 degrees at 0000 UT
on 1 January 1995. One can then derive the Carrington longitude in degrees (call it OLD)
as a function of time:

OLD = 349.03 - (360.* X / 27.2753),

where X is the number of days since 1 January 1995. It is understood that OLD is to be
taken modulo 360. Note that the Carrington longitude decreases as time increases. If one
now compares the values of OLD with the values listed in the Almanac one finds
reasonable agreement, with maximum discrepancies of about 4 hours.

To get a better estimate of the start and stop times, we find the difference between
OLD and the values listed in the *Astronomical Almanac*, and then fit the
difference (ALMANAC - OLD) with a sine-cosine series:

Fit#1 = f + X/g + a*SIN(2*πX/e) + b*SIN(4*πpi;*X/e) + h*SIN(6*πpi;*X/e)

+ c*COS(2*πpi;*X/e) + d*COS(4*πpi;*X/e) + i*COS(6*πpi;*X/e)

In Figure 1 the red data points are the values of (ALMANAC -
OLD), and the blue line reperesents Fit#1. The improved estimate is then

NEW = OLD + Fit#1

The values of (NEW - ALMANAC) are plotted in Figure 2. The maximum
discrepancy is now about 2 minutes. Notice that the data points between Days ~ 350 and
700 appear disjoint from the rest of the data set. This appears to be caused by a small
offset in the Almanac tables for 1996. Each volume of the Almanac presents data for the
stated year, for the last day of the previous year and the first day of the next year.
The last day of the previous year is referred to as "January 0", and the first day of
the subsequent year is called "December 32". This repetition of data allows a comparison
of data from the 1995, 1996, and 1997 almanacs, shown in
Table 1.

It appears that the longitudes listed for 1996 is systematically very slightly larger
than for either 1995 or 1997. The fitting procedure described above can be generalized
to allow for a different offset in 1996. We define:

Fit#2 = Fit#1 + j for the year 1996

Fit#2 = Fit#1 for all other years

NEW2 = OLD + Fit#2

Values of (ALMANAC - NEW2) are shown in Figure 3. The residuals are less than 1 minute nearly all of the time. The function NEW2 is then used to generate the Carrington rotation start and stop times in our plots and listings.