In astronomy and navigation, the celestial sphere is an abstract sphere that has an arbitrarily large radius and is concentric to Earth. All objects in the sky can be conceived as being projected upon the inner surface of the celestial sphere, which may be centered on Earth or the observer. If centered on the observer, half of the sphere would resemble a hemispherical screen over the observing location.
The celestial sphere is a practical tool for spherical astronomy, allowing astronomers to specify the apparent positions of objects in the sky if their distances are unknown or irrelevant. In the equatorial coordinate system, the celestial equator divides the celestial sphere into two halves: the northern and southern celestial hemispheres.
Because astronomical objects are at such remote distances, casual observation of the sky offers no information on their actual distances. All celestial objects seem equally far away, as if fixed onto the inside of a sphere with a large but unknown radius,^{[1]} which appears to rotate westward overhead; meanwhile, Earth underfoot seems to remain still. For purposes of spherical astronomy, which is concerned only with the directions to celestial objects, it makes no difference if this is actually the case or if it is Earth that is rotating while the celestial sphere is stationary.
The celestial sphere can be considered to be infinite in radius. This means any point within it, including that occupied by the observer, can be considered the center. It also means that all parallel lines, be they millimetres apart or across the Solar System from each other, will seem to intersect the sphere at a single point, analogous to the vanishing point of graphical perspective.^{[2]} All parallel planes will seem to intersect the sphere in a coincident great circle^{[3]} (a "vanishing circle").
Conversely, observers looking toward the same point on an infiniteradius celestial sphere will be looking along parallel lines, and observers looking toward the same great circle, along parallel planes. On an infiniteradius celestial sphere, all observers see the same things in the same direction.
For some objects, this is oversimplified. Objects which are relatively near to the observer (for instance, the Moon) will seem to change position against the distant celestial sphere if the observer moves far enough, say, from one side of planet Earth to the other. This effect, known as parallax, can be represented as a small offset from a mean position. The celestial sphere can be considered to be centered at the Earth's center, the Sun's center, or any other convenient location, and offsets from positions referred to these centers can be calculated.^{[4]}
In this way, astronomers can predict geocentric or heliocentric positions of objects on the celestial sphere, without the need to calculate the individual geometry of any particular observer, and the utility of the celestial sphere is maintained. Individual observers can work out their own small offsets from the mean positions, if necessary. In many cases in astronomy, the offsets are insignificant.
The celestial sphere can thus be thought of as a kind of astronomical shorthand, and is applied very frequently by astronomers. For instance, the Astronomical Almanac for 2010 lists the apparent geocentric position of the Moon on January 1 at 00:00:00.00 Terrestrial Time, in equatorial coordinates, as right ascension 6^{h} 57^{m} 48.86^{s}, declination +23° 30' 05.5". Implied in this position is that it is as projected onto the celestial sphere; any observer at any location looking in that direction would see the "geocentric Moon" in the same place against the stars. For many rough uses (e.g. calculating an approximate phase of the Moon), this position, as seen from the Earth's center, is adequate.
For applications requiring precision (e.g. calculating the shadow path of an eclipse), the Almanac gives formulae and methods for calculating the topocentric coordinates, that is, as seen from a particular place on the Earth's surface, based on the geocentric position.^{[5]} This greatly abbreviates the amount of detail necessary in such almanacs, as each observer can handle their own specific circumstances.
These concepts are important for understanding celestial coordinate systems, frameworks for measuring the positions of objects in the sky. Certain reference lines and planes on Earth, when projected onto the celestial sphere, form the bases of the reference systems. These include the Earth's equator, axis, and orbit. At their intersections with the celestial sphere, these form the celestial equator, the north and south celestial poles, and the ecliptic, respectively.^{[6]} As the celestial sphere is considered arbitrary or infinite in radius, all observers see the celestial equator, celestial poles, and ecliptic at the same place against the background stars.
From these bases, directions toward objects in the sky can be quantified by constructing celestial coordinate systems. Similar to geographic longitude and latitude, the equatorial coordinate system specifies positions relative to the celestial equator and celestial poles, using right ascension and declination. The ecliptic coordinate system specifies positions relative to the ecliptic (Earth's orbit), using ecliptic longitude and latitude. Besides the equatorial and ecliptic systems, some other celestial coordinate systems, like the galactic coordinate system, are more appropriate for particular purposes.
The ancients assumed the literal truth of stars attached to a celestial sphere, revolving about the Earth in one day, and a fixed Earth.^{[7]} The Eudoxan planetary model, on which the Aristotelian and Ptolemaic models were based, was the first geometric explanation for the "wandering" of the classical planets.^{[8]} The outer most of these "crystal spheres" was thought to carry the fixed stars. Eudoxus used 27 concentric spherical solids to answer Plato's challenge: "By the assumption of what uniform and orderly motions can the apparent motions of the planets be accounted for?"^{[9]}
A celestial sphere can also refer to a physical model of the celestial sphere or celestial globe. Such globes map the constellations on the outside of a sphere, resulting in a mirror image of the constellations as seen from Earth. The oldest surviving example of such an artifact is the globe of the Farnese Atlas sculpture, a 2ndcentury copy of an older (Hellenistic period, ca. 120 BCE) work.
Observers on other worlds would, of course, see objects in that sky under much the same conditions – as if projected onto a dome. Coordinate systems based on the sky of that world could be constructed. These could be based on the equivalent "ecliptic", poles and equator, although the reasons for building a system that way are as much historic as technical.

Wikimedia Commons has media related to Celestial spheres. 