The PyEphem Tutorial

Version 2007.November.1, for PyEphem version and later.

The PyEphem library provides Python programmers with access to the astronomical routines used by the XEphem interactive astronomical ephemeris application; its author, Elwood Charles Downey, has generously granted permission for PyEphem to be built upon his work.

After installing the module, you can use it in Python with the statement:

>>> import ephem

This tutorial assumes that you are familiar with astronomy, but necessarily all of the issues surrounding astronomical calculation; those who find its discussions tedious will probably just want to read over its examples to quickly familiarize themselves with how PyEphem works.

First Steps

PyEphem will compute the positions of celestial bodies on particular dates, and can determine where in the sky they appear from any location on earth.

When using PyEphem, you will usually create instances of the bodies that interest you, compute their position for various dates and perhaps geographic locations, and finally print or display the result. For example, to determine the location and brightness of Uranus on the night it was discovered we simply create a Uranus object and ask where it was on the 13th of March, 1781:

>>> u = ephem.Uranus()
>>> u.compute('1781/3/13')
>>> print('%s %s %s' % (u.ra, u.dec, u.mag))
5:35:45.28 23:32:54.1 5.6
>>> print(ephem.constellation(u))
('Tau', 'Taurus')

Calling compute() sets many attributes of a body, beyond the right ascension, declination, and magnitude printed here; see the PyEphem Quick Reference for other attributes that compute() sets. You see that measurements are formatted as an astronomer would expect: dates are expressed as year, month, and day, delimited by slashes; right ascension as hours of arc around the celestial equator; and declination as degrees north of the equator. The colons between the components of an angle are a compromise — the more traditional 21°46′15.66′′ is not possible with the symbols on a standard computer keyboard.

The code above created and used only one instance of Uranus, but you can also have several going at once. For example, to determine how close Neptune and Jupiter lay as Galileo famously observed them — he was busy watching the Jovian moons he had discovered two years earlier and, though Neptune had moved overnight, he dismissed it as a background star and left its discovery to wait another two hundred years — we create one instance of each planet and compare their positions:

>>> j = ephem.Jupiter('1612/12/28')
>>> n = ephem.Neptune('1612/12/28')
>>> print("%s %s %s" % (j.ra, j.dec, j.mag))
11:48:20.52 2:41:13.6 -1.96
>>> print("%s %s %s" % (n.ra, n.dec, n.mag))
11:49:15.77 2:37:04.5 7.92
>>> print(ephem.separation(j, n))

Notice that, while in our first example we created our Uranus instance and called its compute() method as two separate steps, we have here taken the shortcut of providing dates as we created each planet; in general, any arguments you provide when creating a planet are used to compute() its initial position. The separation() function computes the angle in degrees between two bodies, as measured by their right ascension and declination. In this case, the separation of 0°14′ was small enough to place both planets in Galileo’s field of view.

You can even create several instances of the same body. Let’s compare how far Mars moves in one day at perihelion versus aphelion, and verify that its speed is greater when closer to the Sun:

>>> def hpos(body): return body.hlon, body.hlat
>>> ma0 = ephem.Mars('1976/05/21')    # ma: mars near aphelion
>>> ma1 = ephem.Mars('1976/05/22')
>>> print(ephem.separation(hpos(ma0), hpos(ma1)))
>>> mp0 = ephem.Mars('1975/06/13')    # mp: mars near perihelion
>>> mp1 = ephem.Mars('1975/06/14')
>>> print(ephem.separation(hpos(mp0), hpos(mp1)))

Here we wanted to measure the motion of Mars around the Sun, but separation() normally compares the right ascension and declination of two bodies — which would measure the motion of Mars across the sky of the moving platform of our earth. So instead of giving separation() the Mars instances themselves, we specifically provided the heliocentric longitude and latitude of each instance, revealing how far Mars moved around the Sun regardless of how this motion appeared from earth.

In general separation() can measure the angle between any pair of spherical coordinates, so long as the elements of each coordinate are spherical longitude (angle around the sphere) followed by spherical latitude (angle above or below its equator). Each pair should be provided as a two-item sequence like a tuple or list. Appropriate coordinate pairs include right ascension and declination; heliocentric longitude and latitude; azimuth and altitude; and even the geographic longitude and latitude of two locations on earth.

Computing With Angles

Sometimes you may want to perform computations with times and angles. Strings like '7:45:45.15' are attractive when printed, but cumbersome to add and multiply; so PyEphem also makes times and angles available as floating-point numbers for more convenient use in mathematical formulae.

All angles returned by PyEphem are actually measured in radians. Let us return to our first example above, and examine the results in more detail:

>>> u = ephem.Uranus('1871/3/13')
>>> print(str(u.dec))
>>> print('%.12f' % float(u.dec))
>>> print('%.11f' % (u.dec + 1))

The rule is that angles become strings when printed or given to str(), but otherwise act like Python floating point numbers. The same thing happens when you set an angle: a string is interpreted as degrees or hours (hours if you are setting right ascension, degrees everywhere else); while a float is interpreted as radians.

>>> print(ephem.degrees('90.0'))
>>> print(ephem.degrees(3.141593))

Note that the format operator % can return either value, depending on whether you use %s or one of the numeric formats:

>>> print("as a string: %s, as a float: %f" % (u.dec, u.dec))
as a string: 22:04:47.4, as a float: 0.385366

As an example computation, we can verify Kepler’s Second Law of planetary motion — that a line drawn from a planet to the sun will sweep out equal areas over equal periods of time. We have already computed two positions for Mars near its aphelion that are one day apart (and defined a helpful hpos() function; see above). We can estimate the actual distance it moved in space that day by multiplying its angular motion in radians by its distance from the Sun:

>>> aph_angle = ephem.separation(hpos(ma0), hpos(ma1))
>>> aph_distance = aph_angle * ma0.sun_distance
>>> print('%.13f' % aph_distance)

So, it moved nearly 0.013 AU in a single day (about 1.9 million kilometers). A line drawn between it and the sun would have, roughly, filled in a triangle whose base is 0.013 AU, whose height is the distance to the Sun, and whose area is therefore:

>>> aph_area = aph_distance * ma0.sun_distance / 2.
>>> print('%.13f' % aph_area)

According to Kepler our results should be the same for any other one-day period for which we compute this; we can try using the two Mars positions from near perihelion:

>>> peri_angle = ephem.separation(hpos(mp0), hpos(mp1))
>>> peri_distance = peri_angle * mp0.sun_distance
>>> peri_area = peri_distance * mp0.sun_distance / 2.
>>> print('%.13f' % peri_area)    # the area, to high precision, is the same!

Despite the fact that Mars moves twenty percent faster at perihelion, the area swept out — to quite high precision — is identical, just as Kepler predicted. Some of the tiny difference between the two numbers we got results from our having approximated sectors of its orbit as triangles; the rest comes from the pertubations of other planets and other small sources of irregularity in its motion.

When you use an angle in mathematical operations, Python will return normal floats that lack the special power of printing themselves as degrees or hours or arc. To turn radian measures back into printable angles, PyEphem supplies both a degrees() and an hours() function. For example:

>>> print('%.13f' % (peri_angle * 2))
>>> print(ephem.degrees(peri_angle * 2))

You may find that your angle arithmetic often returns angles that are less than zero or that exceed twice pi. You can access the norm attribute of an angle to force it into this range:

>>> deg = ephem.degrees
>>> print(deg(deg('270') + deg('180')))
>>> print(deg(deg('270') + deg('180')).norm)

Computing With Dates

PyEphem only processes and returns dates that are in Universal Time (UT), which is simliar to Standard Time in Greenwich, England, on the Earth’s Prime Meridian. If you need to display a PyEphem time in your own timezone, use the localtime() function, which returns a Python datetime object:

>>> d = ephem.Date('1984/12/21 15:00')
>>> ephem.localtime(d)
datetime.datetime(1984, 12, 21, 10, 0)
>>> print(ephem.localtime(d).ctime())
Fri Dec 21 10:00:00 1984

As you can see from this result, I am writing this Tutorial in the Eastern Time zone, which in the winter is five hours earlier than the time in Greenwich.

PyEphem actually represents dates as the number of days since noon on 1899 December 31. While you will probably not find the absolute value of this number very interesting, the fact that it is counted in days means you can move one day forward or backward by adding or subtracting one. The rules described above for angles hold for floats as well: you can create them with ephem.Date(), but after doing arithmetic on them you must pass them back through ephem.Date() to turn them back into dates:

>>> d = ephem.Date('1950/2/28')
>>> print(d + 1)
>>> print(ephem.Date(d + 1))
1950/3/1 00:00:00

The ephem module provides three constants hour, minute, and second, which can be added or subtracted from dates to increment or decrement them by the desired amount.

You can specify dates in several formats; not only can the strings that specify them use either floating point days or provide hours, minutes, and seconds, but you can also provide the components of the date in a tuple. The following assignments are all equivalent:

>>> d = ephem.Date(34530.34375)
>>> d = ephem.Date('1994/7/16.84375')
>>> d = ephem.Date('1994/7/16 20:15')
>>> d = ephem.Date((1994, 7, 16.84375))
>>> d = ephem.Date((1994, 7, 16, 20, 15, 0))

And to complement the fact that you can specify dates as a tuple, two methods are provided for extracting the date as a tuple: triple() returns a year, month, and floating point day, while tuple() provides everything down to floating point seconds. After any of the above calls, the date can be examined as:

>>> print('as a float: %f\nas a string: "%s"' % (d, d))
as a float: 34530.343750
as a string: "1994/7/16 20:15:00"
>>> print(d.triple())
(1994, 7, 16.84375)
>>> print(d.tuple())
(1994, 7, 16, 20, 15, 0.0)

Any PyEphem function argument that requires an angle or date will accept any of the representations shown above; so you could, for instance, give a three-element tuple directly to compute() for the date, rather than having to pass the tuple through the Date() function before using it (though the latter approach would also work).

Computations for Particular Observers

The examples so far have determined the position of bodies against the background of stars, and their location in the solar system. But to observe a body we need to know more — whether it is visible from our latitude, when it rises and sets, and the height it achieves above our horizon. In return for this more detailed information, PyEphem quite reasonably demands to know our position on the earth’s surface; we can provide this through an object called an Observer:

>>> gatech = ephem.Observer()
>>> gatech.lon, = '-84.39733', '33.775867'

When the Observer is provided to compute() instead of a simple date and epoch, PyEphem has enough information to determine where in the sky the body appears. Fill in the date and epoch fields of the Observer with the values you would otherwise provide to compute(); the epoch defaults to the year 2000 if you do not set it yourself. As an example, we can examine the 1984 eclipse of the sun from Atlanta:

>>> = '1984/5/30 16:22:56'   # 12:22:56 EDT
>>> sun, moon = ephem.Sun(), ephem.Moon()
>>> sun.compute(gatech)
>>> moon.compute(gatech)
>>> print("%s %s" % (sun.alt,
70:08:39.2 122:11:26.4
>>> print("%s %s" % (moon.alt,
70:08:39.5 122:11:26.0

For those unfamiliar with azimuth and altitude: they describe position in the sky by measuring angle around the horizon, then angle above the horizon. To locate the Sun and Moon in this instance, you would begin by facing north and then turn right 122°, bringing you almost around to the southeast (which lies 125° around the sky from north); and by looking 70° above that point on the horizon — fairly high, given that 90° is directly overhead — you would find the Sun and Moon.

Eclipses are classified as partial when the Moon merely takes a bite out of the Sun; annular when the Moon passes inside the disc of the sun to leave only a brilliant ring (Latin annulus) visible; and total when the moon is large enough to cover the Sun completely. To classify this eclipse we must compare the size of the Sun and Moon to the distance between them. Since each argument to separation() can be an arbitrary measure of spherical longitude and latitude, we can provide azimuth and altitude:

>>> print(ephem.separation((, sun.alt), (, moon.alt)))
>>> print("%.8f %.8f %.11f" % (sun.size, moon.size, sun.size - moon.size))
1892.91210938 1891.85778809 1.05432128906

The Sun’s diameter is larger by 1.05′′, so placing the Moon at its center would leave an annulus of width 1.05′′ / 2 = 0.52′′ visible around the Moon’s edge. But, in fact, the center of the Moon lies 0.48 arc seconds towards one edge of the sun — not enough to move its edge outside the sun and make a partial eclipse, but enough to make a quite lopsided annular eclipse, whose annulus is 0.52′′ + 0.48 = 1.00′′ wide on one side and a scant 0.52′′ - 0.48 = 0.04′′ on the other.

The sky positions computed by PyEphem take into account the refraction of the atmosphere, which bends upwards the images of bodies near the horizon. During sunset, for example, the descent of the sun appears to slow because the atmosphere bends its image upwards as it approaches the horizon:

>>> = '1984/5/31 00:00'   # 20:00 EDT
>>> sun.compute(gatech)
>>> for i in range(8):
...     old_az, old_alt =, sun.alt
... += ephem.minute * 5.
...     sun.compute(gatech)
...     sep = ephem.separation((old_az, old_alt), (, sun.alt))
...     print("%s %s %s" % (, sun.alt, sep))
1984/5/31 00:05:00 6:17:36.8 1:08:48.1
1984/5/31 00:10:00 5:21:15.6 1:08:36.3
1984/5/31 00:15:00 4:25:31.6 1:08:20.0
1984/5/31 00:20:00 3:30:34.2 1:07:56.5
1984/5/31 00:25:00 2:36:37.8 1:07:22.7
1984/5/31 00:30:00 1:44:04.6 1:06:32.2
1984/5/31 00:35:00 0:53:28.7 1:05:17.0
1984/5/31 00:40:00 0:05:37.8 1:03:28.3

We see that the Sun’s apparent angular speed indeed decreased as it approached the horizon, from around 1°08′ to barely 1°03′ each five minutes.

Since atmospheric refraction varies with temperature and pressure, you can improve the accuracy of PyEphem by providing these values from a local forecast, or at least from average values for your location and season. By default an Observer uses 15°C and 1010 mB, the values for these parameters at sea level in the standard atmosphere model used in aviation. Setting the pressure to zero directs PyEphem to simply ignore atmospheric refraction.

Once PyEphem knows your location it can also work out when bodies rise, cross your meridian, and set each day. These computations can be fairly involved, since planets continue their journey among the stars even as the rotation of the earth brings them across the sky; PyEphem has to internally re-compute their position several times before it finds the exact circumstances of rising or setting. But this is taken care of automatically, leaving you to simply ask:

>>> print(gatech.next_setting(sun))
1984/5/31 00:42:22
>>> print("%s %s" % (sun.alt,
-0:15:45.8 297:20:43.7

Functions also exist for finding risings, transits, and — just for completeness — the moment of “anti-transit” when the object lies along the meridian directly under your feet. See the section on transit, rising, and setting in the Quick Reference for more details.

Loading Bodies From Catalogues

So far we have dealt with the planets, the Sun, and the Moon — major bodies whose orbits PyEphem already knows in great detail. But for minor bodies, like comets and asteroids, you must aquire and load the orbital parameters yourself.

Understand that because the major planets constantly perturb the other bodies in the solar system, including each other, it requires great effort — years of observation yielding formulae with dozens or hundreds of terms — to predict the position of a body accurately over decades or centuries. For a comet or asteroid, astronomers find it more convenient to describe its orbit as perfect ellipse, parabola, or hyperbola, and then issue new orbital parameters as its orbit changes.

The PyEphem home page provides links to several Asteroid, Comet, and Satellite Catalogs of orbital elements. Once you have obtained elements for a particular body, simply provide them to PyEphem’s readdb() function in ephem database format and the resulting object is ready to use:

>>> yh = ephem.readdb("C/2002 Y1 (Juels-Holvorcem),e,103.7816," +
...    "166.2194,128.8232,242.5695,0.0002609,0.99705756,0.0000," +
...    "04/13.2508/2003,2000,g  6.5,4.0")
>>> yh.compute('2003/4/11')
>>> print(
C/2002 Y1 (Juels-Holvorcem)
>>> print("%s %s" % (yh.ra, yh.dec))
0:22:44.58 26:49:48.1
>>> print("%s %s" % (ephem.constellation(yh), yh.mag))
('And', 'Andromeda') 5.96

(Unfortunately, the library upon which PyEphem is build truncates object names to twenty characters, as you can see.) Each call to readdb() returns an object appropriate for the orbit specified in the database entry; in this case it has returned an EllipticalBody:

>>> print(yh)  
<ephem.EllipticalBody 'C/2002 Y1 (Juels-Holvorcem)' at 0x...>

For objects for which you cannot find an entry in ephem database format, you can always create the appropriate kind of object and then fill in its orbital parameters yourself; see the PyEphem Quick Reference for their names and meanings. By calling the writedb() function of a PyEphem object, you can even get it to generate its own database entry for archiving or distribution.

There is one other database format with which PyEphem is familiar: the NORAD Two-Line Element format (TLE) used for earth satellites. Here are some recent elements for the International Space Station.

>>> iss = ephem.readtle("ISS (ZARYA)",
...  "1 25544U 98067A   03097.78853147  .00021906  00000-0  28403-3 0  8652",
...  "2 25544  51.6361  13.7980 0004256  35.6671  59.2566 15.58778559250029")
>>> = '2003/3/23'
>>> iss.compute(gatech)
>>> print("%s %s %s" % (iss.rise_time, iss.transit_time, iss.set_time))
2003/3/23 00:00:50 2003/3/23 00:03:26 2003/3/23 00:06:01

The transit_time for an artificial satellite is actually defined in PyEphem as the moment at which it is at highest altitude, not the moment at which it crosses (transits) the local meridian.

Note that earth satellites are fast movers — in this case rising and setting in less than six minutes! They can therefore have multiple risings and settings each day, and the particular ones you get from rise_time and set_time depend on the particular time of day for which you ask. Repeating the above query eight hours later gives complete different results:

>>> = '2003/3/23 8:00'
>>> iss.compute(gatech)
>>> print("%s %s %s" % (iss.rise_time, iss.transit_time, iss.set_time))
2003/3/23 08:03:40 2003/3/23 08:08:25 2003/3/23 08:13:10

When calling compute() for an earth satellite you should provide an Observer, and not simply a date and epoch, since its location is entirely dependent upon the location from which you are observing. PyEphem provides extra information about earth satellites, beyond the ones available for other objects; again, see the PyEphem Quick Reference for details.

Fixed Objects, Precession, and Epochs

The simplest kind of object to create from a catalog entry are fixed objects, for which a constant right ascension and declination are specified. These include stars, nebulae, global clusters, and galaxies. One example is Polaris, the North Star, which lies at the end of Ursa Minor’s tail:

>>> polaris = ephem.readdb("Polaris,f|M|F7,2:31:48.704,89:15:50.72,2.02,2000")
>>> print(polaris.dec)
Traceback (most recent call last):
RuntimeError: field dec undefined until first compute()

We are able to create the object successfully — why should asking its position raise a runtime error? The reason is that fixed objects, like planets, have an undefined position and magnitude until you call their compute() method to determine their position for a particular date or Observer:

>>> polaris.compute()    # uses the current time by default
>>> print(polaris.a_dec)
>>> print(ephem.degrees(ephem.degrees('90') - polaris.a_dec))

Much better; we see that the North Star lies less than forty-five arc minutes from the pole. But why should we have to call compute() for something fixed — something whose position is considered permanent, and which should not move between one date and another?

The reason is that, while fixed stars and nebulae are indeed nearly motionless over the span of human civilization, the coordinate system by which we designate their positions changes more rapidly. Right ascension and declination are based upon the orientation of the earth’s pole — but it turns out that the pole slowly revolves (around the axis of the ecliptic plane) like the axis of a whirling top, completing each revolution in roughly 25,800 years. This motion is called precession. Because this makes the entire coordinate system shift slightly every year, is not sufficient to state that Polaris lies at 2h31m right ascension and 89:15° declination; you have to say in which year.

That is why the Polaris entry above ends with 2000; this gives the year for which the coordinates are correct, called the epoch of the coordinates. Because the year 2000 is currently a very popular epoch for quoting positions and orbital parameters, compute() uses it by default; but we can provide an epoch= keyword parameter to have the coordinates translated into those for another year:

>>> polaris.compute(epoch='2100')
>>> print(polaris.a_dec)

Thus we see that in another hundred years Polaris will actually lie closer to the pole that it does today. (The '2100' is the same year/month/day format you have seen already, missing both its month and day because we are not bothering to be that specific.) If you enter subsequent years you will find that 2100 is very nearly the closest approach of the pole to Polaris, and that soon afterwards they move apart. For much of the twenty-five thousand year journey the pole makes, there are no stars very near; we may have been lucky to have held the Age of Exploration as the pole was approaching as convenient a star as Polaris.

Today a dim star in Draco named Thuban lies more than twenty degrees from the pole:

>>> thuban = ephem.readdb("Thuban,f|V|A0,14:4:23.3,64:22:33,3.65,2000")
>>> thuban.compute()
>>> print(thuban.a_dec)

But in 2801 BC, as the Egyptians built the pyramids, Thuban served as their pole star, while Polaris lay further from their pole than Thuban lies from ours today:

>>> thuban.compute(epoch='-2800')
>>> print(thuban.a_dec)
>>> polaris.compute(epoch='-2800')
>>> print(polaris.a_dec)

Realize that in these examples I have been lazy by giving compute() an epoch without an actual date, which requests the current position of each star in the coordinates of another epoch. This makes no difference for these fixed objects, since their positions never change; but when dealing with moving objects one must always keep in mind the difference between the date for which you want their position computed, and the epoch in which you want those coordinates expressed. Here are some example compute() calls, beginning with one like the above but for a moving object:

This is probably useless: it computes the current position of halley, but returns coordinates relative to the direction the earth’s axis was pointing in the year 1066. Unless you use a Conquest-era star atlas, this is not useful.
halley.compute('1066', epoch='1066')
This is slightly more promising: it computes the position of halley in 1066 and returns coordinates for the orientation of the earth in that year. This might help you visualize how the object was positioned above contemporary observers, who considered it an ill omen in the imminent conflict between King Harold of England and William the Bastard. But to plot this position against a background of stars, you would first have to recompute each star’s position in 1066 coordinates.
This is what you will probably use most often; you get the position of halley in the year 1066 but expressed in the 2000 coordinates that your star atlas probably uses.

When planning to observe with an equatorial telescope, you may want to use the current date as your epoch, because the rotation of the sky above your telescope is determined by where the pole points today, not where it pointed in 2000 or some other convenient epoch. Computing positions in the epoch of their date is accomplished by simply providing the same argument for both date and epoch:

>>> j = ephem.Jupiter()
>>> j.compute(   # so both date and epoch are now
>>> print("%s %s" % (j.a_ra, j.a_dec))  
8:44:29.49 19:00:10.23
>>> j.compute('2003/3/25', epoch='2003/3/25')
>>> print("%s %s" % (j.a_ra, j.a_dec))
8:43:32.82 19:03:32.5

Be careful when computing distances; comparing two positions in the coordinates of their own epochs will give slightly different results than if the two were based on the same epoch:

>>> j1, j2 = ephem.Jupiter(), ephem.Jupiter()
>>> j1.compute('2003/3/1')
>>> j2.compute('2003/4/1')
>>> print(ephem.separation(
...     (j1.a_ra, j1.a_dec),
...     (j2.a_ra, j2.a_dec)))   # coordinates are both epoch 2000
>>> j1.compute('2003/3/1', '2003/3/1')
>>> j2.compute('2003/4/1', '2003/4/1')
>>> print(ephem.separation(
...     (j1.a_ra, j1.a_dec),
...     (j2.a_ra, j2.a_dec)))   # coordinates are both epoch-of-date

Comparing coordinates of the same epoch, as in the first call above, measures motion against the background of stars; comparing coordinates from different epochs, as in the second call, measures motion against the slowly shifting coordinate system of the earth. Users are most often interested in the first kind of measurement, and stick with a single epoch the whole way through a computation.

It was for the sake of simplicity that all of the examples in this section simply provided dates as arguments to the compute() function. If you are instead using an Observer argument, then you specify the epoch through the observer’s epoch variable, not through the epoch= argument. Observers use epoch 2000 by default.

Finally, make sure you understand that your choice of epoch only affects absolute position — the right ascension and declination returned for objects — not the azimuth and altitude of an object above an observer. This is because the sun will hang in the same position over Atlanta whether the star atlas with which you plot its position has epoch 2000, or 1950, or even 1066 coordinates; the epoch only affects how you name locations in the sky, not how they are positioned with respect to you.