The Earth's rotation axis tends, like the axis of a gyroscope, to maintain its orientation in inertial space. However, due to external forces acting upon the Earth, this axis nevertheless exhibits a slow, large-scale motion, collectively known as precession and nutation. Nutation is divided into a predictable part from a "nutation theory," a long series of trigonometric terms depending on time, and derived from motions of the Moon and models of the Earth, plus corrections called the Celestial Pole Offset, which are the measured departures (two-dimensional on the celestial sphere) of the celestial rotation pole from the theory. [1] These corrections, of order 0.03 seconds of arc, are of no importance to most users, but by monitoring them, a better theory of nutation may be developed in the future.

When using, instead of an inertial reference frame, a frame attached to the body of the solid Earth (a so-called Earth-centred, Earth-fixed or ECEF reference frame), the rotation axis isn't invariable either. This variation, which is only a few metres measured on the Earth's surface, is called Polar motion.

It consists of two quasi-periodic components and a gradual drift, mostly westward, of the Earth's instantaneous rotational axis or North pole, from a conventionally defined reference axis, the CIO (Conventional International Origin ), being the pole's average location over the year 1900.

The two periodic parts are a more or less circular motion called Chandler wobble with a period of about 435 days, and a yearly circular motion. There is also a slow drift which is less well known. These motions are illustrated on the polar motion map from the International Earth Rotation and Reference Systems Service.

On that figure, which is not terribly well explained, the Earth's angular momentum vector points at the cross-hairs or axis origin, while North pole positions relative to that are illustrated in contrasting ways. The mean pole position from 1900 to over a period from 1900 to 1997 (it says 1996, but it shows 1997!) is the wiggly line. The variations from the mean are shown as more or less loopy motions from 1995 (it says 1994) to 1997 (it says 1996, but it shows 1997!).

One can see that the mean displacement far exceeds in magnitude the wobbles. This can lead to errors in software for Earth observing spacecraft, since analysts may read of a 5 meter circular motion and ignore it, while a 20 meter offset sits there, fouling the accuracy of the calculated latitude and longitude. The latter are, of course, determined based on the International Terrestrial Reference System, which follows the polar motion.