Methods

• A number whose provenance is not stored is not published

  Every value carries the catalogue it came from; every distance carries the method that produced it; every model fit says it is a model fit. Where an upstream gives a quantity without saying what it is, the honest label is the column it came from, and that is what the record says.

• A missing value stays missing

  Nothing is filled in, averaged, or carried over from a similar object. Sentinels are not measurements: an unmeasured mass published as zero, a distance parked at a placeholder, a proper motion sitting on the ceiling of a four-digit column — all read as empty. So does a published uncertainty of exactly zero, which is the same trap wearing the strongest possible claim.

• A derived value falls with what it was derived from

  If a distance is rejected, the absolute magnitude and the luminosity computed from it go too, and an error bar never outlives the value it qualified. Two catalogues describing one object are never blended into a third reading that neither of them published.

A distance from Earth is the number readers quote most and the number most often quoted without its ladder. Each row here is one rung, the count is how many objects stand on it, and the last column is how many of those publish an uncertainty. The record says which rung it is on in distance_method, in words, not as a code.

The last row is not a gap in the data. A planet, a moon, an asteroid or a comet has no fixed distance from Earth — what it has is an orbit, and the orbit is what the record stores. The rest are the stars and galaxies whose catalogue carries no usable parallax and no redshift this page would convert.

• By definition1

  The Sun. Its distance is the astronomical unit, which since 2012 is a defined number of metres rather than a measured one.

• Trigonometric parallax109 399

  The angle the Earth's orbit subtends at the star, inverted to a distance by the catalogue that measured it — Hipparcos, and Gaia where the two could be matched to one star. It is the only rung that assumes nothing about the object itself, which is why it is the rung everything else is calibrated against. These rows carry no error bar: the upstream publishes the distance without one, and the bar it deserves is not this site's to invent. A parallax distance is also not symmetric in the way a naive inversion suggests — the further the star, the more the inversion favours the near side — and no correction for that is applied here either.

• The system distance an exoplanet inherits10 137

  A planet is as far away as its star, so both carry the distance the exoplanet archive publishes for the system — with the column it came from and the paper the archive names for it written into the method, because it is not one kind of measurement. Most are inferred from Gaia parallaxes rather than being one, and the microlensing systems are a Bayesian estimate from a galactic model with no parallax measured anywhere in them.

• Cluster membership318

  An open cluster is nearer than any single one of its stars can prove: the members share a parallax, and averaging over the ones a membership analysis accepts beats every one of them (Cantat-Gaudin et al. 2020).

• The globular-cluster scale103

  Harris's catalogue (2010 edition), which fixes each cluster's distance from its horizontal branch and its reddening — a standard candle, calibrated, not a geometric measurement.

• The planetary-nebula statistical scale106

  Stanghellini et al. (2008), which reads a distance out of a nebula's angular size and surface brightness against a calibrating sample. It is a scale rather than a measurement, and it is the weakest rung the catalogue publishes: individual objects can be wrong by tens of per cent.

• Redshift, converted here9804

  A galaxy's heliocentric redshift, corrected into the frame of the cosmic microwave background and run through the cosmology below to a luminosity distance. Only above the floor, and only where both frames agree there is a Hubble flow to measure.

• Spectroscopic redshift from a galaxy survey21 517

  The same conversion, over a redshift a survey measured from spectral lines. The redshift is the survey's; the distance is this catalogue's, so that two galaxies in one cluster are not divided by two different expansion rates.

• Photometric redshift from a galaxy survey31 437

  The same conversion again, over a redshift fitted to colours rather than measured from lines. The survey gives such a redshift a standard deviation of about 20% of itself against 2.9% for a spectroscopic one, so these distances are good to a fifth and their error bars say so — a photometric distance and a spectroscopic one on neighbouring pages are not comparable numbers.

• A redshift transcribed from a paper39

  The quasars, whose redshift is the whole of their distance: they are too far for a parallax, and no Cepheid or supernova has been found in any of them. The redshift is entered per object from the literature and the conversion is this catalogue's own, so that a quasar's distance and a galaxy's come out of the same integral rather than out of two papers written under two cosmologies.

• A distance the survey measured directly176

  The handful of galaxies whose distance a survey published from a measurement of its own rather than from a redshift, kept exactly as published. There is no expansion rate inside these, so none is divided out of them.

• A pulsar's own parallax410

  The same geometry as a star's, measured a different way: either from how the arrival times of a pulsar's pulses shift as the Earth crosses its orbit, or from watching the source itself move against distant quasars with radio interferometry. A few hundred pulsars have one, and where they do it is preferred over the model below — it contains no model of the galaxy at all.

• Dispersion measure, through a model of the galaxy3868

  A radio pulse arrives later at low frequencies than at high ones, because the free electrons on the way slow the low ones more. The delay counts those electrons exactly, and that count — the dispersion measure — is as precisely known as anything in this catalogue. Turning it into a distance is the uncertain half: it needs a model of how the galaxy's electrons are laid out, and this catalogue uses YMW16. Along a well-mapped line of sight the answer is good to tens of per cent; through a spiral arm nobody has surveyed it can be out by a factor.

• A literature value, entered by hand25

  The objects no bulk catalogue carries — the black holes among them — whose distance is transcribed from the paper that measured it.

• Something else1

  A method this page has no paragraph for yet. The row itself still says which one it is.

Turning a redshift into a distance takes three choices, and all three are made here rather than inherited from whichever survey the redshift came from.

• The expansion rate: H₀ = 70 km/s/Mpc

  The round number between the two measurements that do not agree — 67.4 ± 0.5 from the microwave background and 73.0 ± 1.0 from the distance ladder, which are five standard deviations apart. Choosing either would take a side in a live argument on every galaxy page. The ±4% it costs is not left implied by the digits: it is in the error bar on every distance derived this way.

• Flat ΛCDM with Ωm = 0.3153

  The full luminosity distance, not cz/H₀, which is only its first term and runs 13% short at the furthest galaxy this catalogue holds.

• The rest frame of the microwave background

  A heliocentric redshift carries the Sun’s own 370 km/s through the local universe, which at these distances is the same size as the signal. Redshifts are corrected into the background frame through the measured dipole before anything is computed from them.

• A floor at z = 0.0015, tested in both frames

  That is 450 km/s of recession, against the 200–400 km/s of infall toward Virgo and the Great Attractor that nobody removes from an individual galaxy. Below it a “distance” maps the local velocity field rather than the expansion. The test runs in both frames, because a redshift that 370 km/s of correction can push across the floor is one that 300 km/s of the galaxy’s own motion can push back — and where the two frames disagree about whether there is a Hubble flow to measure, no distance is published in either.

One consequence is worth stating plainly: the expansion-rate systematic is the same sign on every distance derived from a redshift. It cancels when you compare two of them and it does not cancel when you compare one against a distance measured any other way. That is also why a survey’s own published distances are recomputed here rather than republished — two galaxies in one cluster divided by two different expansion rates differ by several per cent for no astronomical reason, and no error bar removes a relative offset.

Almost no star has a measured radius or mass. What 59 890 of them carry here is a model fit to a low-resolution spectrum, published by the survey that fitted it, and the record says so in stellar_parameters_method rather than presenting it as a measurement.

A fit the fitter’s own numbers distrust is not published at all. Where the astrometric solution does not fit a single star — the standard threshold for an unresolved companion or a bad fit — the whole set goes, mass, luminosity and age with it, because all of them are derived through that solution’s parallax and the companion that broke it sits inside the aperture the spectrum was taken through.

Two populations are kept and told on themselves instead, because they are wrong in a stated direction rather than unusable: 11 917 fits above 7,500 K, where temperature and interstellar reddening trade off against each other, and 7174 sources the survey’s own classifier calls binaries and its parameter pipeline fitted as one star. Those two sets overlap: 17 874 stars carry a caveat at all, and it is in the same string as the method, so a reader taking the value takes the warning with it.

A further 10 258 stars have no such fit and are described instead by spectroscopic determinations collected from the papers that published them. Those are measurements rather than models, but of a different kind again: each is one group’s analysis of one spectrum, and where a star has several the value here is their median.

A further 3853 stars are exoplanet hosts whose parameters are the exoplanet archive’s composite values — assembled per system from the literature, so the numbers on one page may come from more than one paper.

• Coordinates are J2000 equatorial, as the catalogues published them

  ICRS, epoch J2000.0 in every row. Solar-system bodies are the exception everywhere they appear: their coordinates are apparent and of date, computed for the instant you asked.

• The sky map draws stars where they are now, not where they were

  The map mixes a catalogue of bright stars with a survey whose positions are for epoch 2016.0, so the catalogue’s stars are carried forward to that epoch with their own proper motions before being drawn — otherwise the fastest of them land arcminutes from the survey source underneath. The readout names the frame and the epoch it is quoting. The stored coordinates do not move: this is a rendering epoch, and the record and the API still answer J2000.

• Transit epochs are quoted as published

  An exoplanet transit epoch is published on a dynamical time scale and is used here as though it were civil time. The difference is under about nine minutes — smaller than the transit predictions on this site are useful to, and stated rather than silently absorbed.

The sky calendar is the one part of this site that is computed rather than catalogued. Every date on it except the flybys is found at the moment the page is asked for, by solving for the instant a defining quantity holds, and stated in Universal Time. Nothing on it was typed in from an almanac, which is why no year on it runs out.

• Phases, seasons and oppositions are the instant a longitude is reached

  A full moon is the moment the Moon’s geocentric ecliptic longitude stands 180° from the Sun’s; a quarter, 90° or 270°. An equinox or solstice is the Sun’s apparent longitude reaching a multiple of 90°. An opposition is the moment the Earth and the planet share a heliocentric longitude, which is the same event seen from the Sun; a greatest elongation is the maximum of the angle between an inner planet and the Sun. Each is solved to the second from an analytic ephemeris whose positions are good to about an arcminute, so the times are good to a minute or so and agree with the almanacs to within their own rounding.

• A meteor shower is dated by where the Earth is, against J2000

  A shower peaks when the Earth reaches a point in its orbit, which the observing community fixes as a solar longitude — the values on the calendar are the International Meteor Organization’s working list, quoted against the mean equinox of J2000.0, as are the activity spans and the rates. The instant published is when the Sun’s apparent longitude, measured against that same J2000 equinox, reaches the listed value. Measured against the equinox of the day instead, the same number would land nine hours early in the 2020s and drift by twenty minutes a year. The zenithal hourly rate is a normalisation — one observer, a sky showing sixth-magnitude stars, the radiant overhead — and is printed as that rather than as a forecast; the Moon’s illumination on the night is computed beside it because it is the one thing that decides whether the rate means anything.

• An eclipse is published as its geometry

  For a solar eclipse: the instant of greatest eclipse, when the axis of the Moon’s shadow passes closest to the centre of the Earth; the kind at that point; and, for a total or annular one, the latitude and longitude the central shadow falls on then and the fraction of the disc covered there. A partial-only eclipse has no such point and is published without one. For a lunar eclipse: the peak, the kind, the fraction of the Moon inside the umbra, and how long the deepest phase lasts. Whether any of it reaches a particular place is a separate calculation the calendar makes only when a place is named.

• A close pairing is a geocentric minimum

  The angle between two bodies is sampled every six hours across the window and each local minimum refined to under a minute. The separation is measured from the centre of the Earth. For two planets that is what every observer sees; for the Moon, which is near enough that an observer’s own position shifts it by up to a degree against the stars, the calendar says so beside the number. Pairings closer than 15° to the Sun are not listed at all.

• Flybys are read, not computed, and say which window they cover

  A near-Earth asteroid’s orbit is not in any ephemeris this site can run, so its close approaches are taken from JPL’s close-approach data on a stated date, for passes inside twenty lunar distances by bodies brighter than absolute magnitude 22 and for anything at all passing inside the Moon’s orbit. The calendar prints the date the list was cut and the span it covers, because a list of flybys is only as complete as its date — a body discovered afterwards is simply not on it.

• Local hours are for the place you name, and for nothing else

  Given a place, each event is answered with the hours it is both more than 20° up and the Sun more than 12° down, solved from topocentric positions of date. The default calendar names no place and therefore claims nothing local; a row is hidden only when the thing never rises there or the eclipse never reaches there, and low is reported as low rather than removed.

• Which radius

  A radius says which radius it is. Planets and moons carry the volumetric mean radius — for an oblate giant the equatorial one is thousands of kilometres larger — and small bodies carry an effective radius, the radius of the sphere with the same projected area, which is not a claim that the body is round. Where the upstream does not say, neither does the record.

• Which magnitude, in which band

  Magnitudes are apparent unless labelled otherwise and always carry the passband they were measured in. An absolute magnitude means absolute visual for a star and the small-body H for an asteroid or comet; those are different quantities and are labelled as such rather than sharing a column heading.

• Units, once, everywhere

  Masses in kilograms, lengths and orbits in kilometres, distances from Earth in light-years, angles in degrees. Conversions happen at ingest, so nothing on a page is a unit a reader has to guess at, and an uncertainty is converted by the same factor as the value it qualifies.

• Error bars are two numbers

  Uncertainties are stored under the field they qualify as [minus, plus], both positive, because the archives publish them separately and a fit is not obliged to be symmetric. A source that publishes one σ gets two equal halves. They are the upstream’s own, except where a distance was derived here — then the bar is what this arithmetic propagated, and the section above says which terms are in it.

• Star colour is computed, and it excludes reddening

  The colour of a star on the map and on its own page is Planck’s law at the star’s temperature through the standard colour matching functions, taken from the spectral type where there is one and from the colour index otherwise. Interstellar dust reddens the light that arrives, and no correction for it is applied — so the colours are what the star is, not what a telescope would record. Luminosity class and brightness are deliberately not modelled into the hue.

Every merged row is run past a set of physical plausibility rules before it can be published: Kepler’s third law against the orbit, the distance modulus against the two magnitudes and the distance, bulk density against mass and radius, albedo and inclination within their ranges, an eccentricity against the sign of the semi-major axis, a mass against the hydrogen-burning limit, a luminosity against what a star can emit.

Where this catalogue caused a violation, it is fixed at the source and the rule stays behind as a guard matching nothing. Where an upstream publishes two numbers that cannot both be right — most of the 95 flagged rows are a planet whose mass and radius imply a density greater than a white dwarf’s — the row keeps the published values and gains a sentence saying what is wrong with them, on the page and in data_quality. Quote it if you quote the number.

What each rule matched when it was last reviewed is recorded, and a rebuild in which one of them matches more rows than that does not ship. The number going up is the alarm, not the inconvenience. Where two catalogues describe one object differently beyond the tolerance for that quantity, the disagreement is written down and shipped inside the release rather than resolved in silence.

Licences and retrieval dates are also listed on the about page; what is here is the treatment. The three groups are three different jobs: a catalogue creates objects, an enrichment adds fields to objects it did not create, and a survey is stored whole, drawn on the map and given no pages of its own.