jplephem 1.1

Earth and Moon

The raw ephemerides provide one position for the Earth-Moon barycenter, and another for the position of the Moon relative to the geocenter. The JPL expects you to combine these values yourself if you want the Solar System location of the Earth or Moon, which gives you the chance to be more efficient by asking the ephemeris for each position only once:

barycenter = eph.position('earthmoon', j)
moonvector = eph.position('moon', j)

earth = barycenter - moonvector * eph.earth_share
moon = barycenter + moonvector * eph.moon_share

High-Precision Dates

Since all modern Julian dates are numbers larger than 2.4 million, a standard 64-bit Python or NumPy float necessarily leaves only a limited number of bits available for the fractional part. An unpublished paper (“AA Technical Note 2011-02”) by the United States Naval Observatory’s Astronomical Applications Department suggests that the precision possible with a 64-bit floating point Julian date is around 20.1µs.

If you need to supply times and receive back planetary positions with greater precision than 20.1µs, then you have two options.

First, you can supply times using the special float96 NumPy type, which is also aliased to the name longfloat. If you provide either a float96 scalar or a float96 array as your tdb parameter to any jplephem routine, you should get back a high-precision result.

Second, you can split your date or dates into two pieces, and supply them as a pair of arguments two tdb and tdb2; one popular approach for how to split your date is to use the tdb float for the integer Julian date, and tdb2 for the fraction that specifies the time of day. Nearly all jplephem routines accept this optional tdb2 argument if you wish to provide it, thanks to the work of Marten van Kerkwijk!

Waiting To Compute Velocity

When a high-level astronomy library computes the distance between an observer and a solar system body, it typically measures the light travel delay between the observer and the body, and then uses a loop to take the position several steps backwards in time until it has determined where the planet was back when the light left its surface (or cloud deck) that is reaching the eye or sensor of the observer right now.

To make such a loop less computationally expensive — a loop that only needs to compute the planet position repeatedly, and can wait to compute the velocity until the loop’s conclusion — jplephem provides a way to split the position_and_velocity() call into two pieces. This lets you examine the position before deciding whether to also proceed with the expense of computing the velocity.

The key is the special compute_bundle() method, which returns a tuple containing the coefficients and intermediate results that are needed by both the position and the velocity computations. There is nothing wasted in calling compute_bundle() whether you are going to ask for the position, the velocity, or both as your next computing step!

So your loop can look something like this:

while True:
    bundle = eph.compute_bundle('mars', tdb)
    position = eph.position_from_bundle(bundle)

    # ...determine whether you are happy...

    if you_are_happy:
        break

    # ...otherwise, adjust `tdb` and then let
    # control return back to the top of the loop

# Now we can re-use the values in `bundle`, for free!

velocity = eph.velocity_from_bundle(bundle)

This is especially important when the number of dates in tdb is large, since vector operations over thousands or millions of dates are going to take a noticeable amount of time, and every mass operation that can be avoided will help move your program toward completion.

Changelog

2013 July 10 — Version 1.1

• Deprecates the old compute() method in favor of separate position() and position_and_velocity() methods.

• Supports computing position and velocity in two separate phases by saving a “bundle” of coefficients returned by compute_bundle().

• From Marten van Kerkwijk: a second tdb2 time argument, for users who want to build higher precision dates out of two 64-bit floats.

2013 January 18 — Version 1.0

• Initial release

References

The Jet Propulsion Laboratory’s “Solar System Dynamics” page introduces the various options for doing solar system position computations: http://ssd.jpl.nasa.gov/?ephemerides

The plain ASCII format element sets from which the jplephem Python ephemeris packages are built, along with documentation, can be found at: ftp://ssd.jpl.nasa.gov/pub/eph/planets/ascii/

Equivalent FORTRAN code for using the ephemerides be found at the same FTP site: ftp://ssd.jpl.nasa.gov/pub/eph/planets/fortran/