Space Time logo Space Time

Methods and accuracy

How Space Time works out what it shows: the models behind the positions and event times, the frames and time scales, how far the results are from NASA’s JPL ephemerides when measured, and where a view stops being a measurement.

The accuracy figures on this page are not estimates. They are read from the output of the app’s own test suite, which compares Space Time with JPL and fails if the published numbers drift from what it measures. The comparison data was last fetched from JPL on 11 November 2026. This page describes version 1.0.895.

What is computed, and how

WhatHow it is worked outWhere the inputs come from
Sun, Moon and planetsAnalytic series from the open-source astronomy-engine library, with light time, aberration and the observer’s parallax applied.The library’s own models. Checked here against JPL.
Rise and set, Moon phases, eclipses, conjunctions, oppositionsSearches over the same model for the instant a condition is met. Rise and set include standard atmospheric refraction.Computed. Local eclipse circumstances use your saved location.
Asteroids, comets and dwarf planets in the 3D viewA fixed ellipse from published orbital elements. The pull of the planets is left out, so the error grows with time from the date the orbit was fitted.JPL Small-Body Database.
Asteroid and comet tables in the Ephemeris GeneratorNot computed by Space Time. The table is requested from JPL Horizons for your location or observatory code and shown with JPL’s orbit solution number and its 3-sigma position uncertainty.NASA JPL Horizons, fetched when you ask.
StarsCatalog positions. Proper motion is applied where a prediction depends on it, such as an occultation.The HYG compilation (Hipparcos, Yale Bright Star, Gliese) and Tycho-2, with Gaia DR3 parallaxes for the deeper catalog.
Lunar occultationsThe Moon’s place from your location against the star’s place for that date. The Moon is a smooth sphere of radius 1,737.4 km.Star places from the HYG compilation. Checked here against JPL.
Exoplanet transitsA linear clock: a published mid-transit time plus a whole number of periods. Uncertainty is carried forward from the published errors on both.NASA Exoplanet Archive. The paper behind each clock is named on its row.
Variable-star minima and maximaA linear clock from a catalog epoch and period. The catalog gives no uncertainties, so the timing figure shown is a lower bound.AAVSO International Variable Star Index (VSX).
SatellitesThe standard SGP4 propagator on published orbital elements.CelesTrak element sets.
Nebulae, clusters and galaxiesCatalog positions and distances, placed at true scale.The catalog named on each object’s card.

Space Time plans and records observations. It does not measure positions or brightness from your images.

Coordinate frames

  • The 3D view is one scene measured in meters at true scale, with axes fixed to the ecliptic of J2000.
  • Catalog positions are ICRS, which for these purposes is the J2000 equator. A table you load must be in ICRS or J2000. A file that declares another frame is refused, because drawing it as ICRS would put every row in the wrong place.
  • The Sky view, altitude and azimuth use the apparent place for the date, seen from your latitude, longitude and height.
  • Ephemeris tables give both the astrometric J2000 position and the apparent position of date, and label which is which.
  • Far from the present, the sky is turned with the Vondrák long-term precession model.
  • Position angles on the Moon’s edge are measured from celestial north through east, on the true equator of date.

Time scales

  • UTC is the clock everything starts from.
  • Terrestrial Time, which the models run on, is UTC plus 32.184 seconds plus leap seconds from the IERS table. That table is valid until 28 June 2027. Before 1972 the Espenak and Meeus model of Delta T is used.
  • After the table ends nobody knows the offset yet. Space Time holds the last value and moves to the Espenak and Meeus model over 20 years. Event times past the table’s end carry several seconds of uncertainty that no model can remove.
  • Earth’s rotation uses UTC as if it were UT1. The two differ by less than 0.9 seconds.
  • TDB and BJD_TDB are computed for exoplanet transit times, with the light-travel time to the solar-system barycenter for the star in question.
  • On screen, times are shown in the time zone of your observing location, and observing panels name the zone. Exported files use UTC.

Measured accuracy

Positions of the Sun, Moon and planets were compared with NASA/JPL Horizons observer tables, DE441 for 400 cases, seen from London, Tokyo, Nairobi, Santiago and Sydney, on dates from 2000 to 2040.

QuantityCasesTypicalWorstUnitWorst case
Apparent position of date, seen from the observer 400 2.52 20.2 arcseconds Neptune, Santiago, 2010-07-15 19:30 UTC
Astrometric J2000 position, seen from the observer 400 2.87 23 arcseconds Moon, London, 2024-04-08 18:00 UTC
Altitude and azimuth, no refraction 400 3.58 19.8 arcseconds Neptune, London, 2010-07-15 19:30 UTC
Distance from the observer 400 50.9 129 parts per million Moon, Santiago, 2032-01-20 02:00 UTC
TDB minus UTC 4 15.9 21.4 microseconds 2023-02-25 00:00 UTC
Light time to the solar-system barycenter 9 0.867 6.31 milliseconds HD 209458, JD 2460000.50 TDB

Typical is the median. One arcsecond is 1/3600 of a degree, and the Moon is about 1,800 arcseconds wide. For the Moon, the astrometric row compares two slightly different definitions of the position, and that difference accounts for about 20 arcseconds of its entry.

Asteroids, dwarf planets and comets with bundled orbits

These follow a fixed ellipse. Within 2.57 years of the date the orbit was fitted, the largest error found was 9.97 arcminutes for an asteroid or dwarf planet and 29.3 arcminutes for a comet, whose orbit Jupiter and its own jets change more. That is good enough to find a bright asteroid in a chart and not good enough to point a narrow-field camera at a faint one. For that, use the Ephemeris Generator or a finder chart, which ask JPL directly.

BodyOrbit fromWorst error (arcminutes)Date of worst case
4 Vesta JPL Small-Body Database 3.78 2029-01-01
2 Pallas JPL Small-Body Database 1.5 2029-01-01
10 Hygiea JPL Small-Body Database 5.42 2029-01-01
101955 Bennu JPL Horizons 0.415 2027-06-01
433 Eros JPL Small-Body Database 0.212 2027-06-01
2060 Chiron JPL Small-Body Database 0.715 2029-01-01
3200 Phaethon JPL Small-Body Database 0.52 2029-01-01
Ceres JPL Horizons 9.97 2029-01-01
Eris JPL Horizons 0.124 2029-01-01
Makemake JPL Horizons 0.21 2029-01-01
Haumea JPL Horizons 0.257 2029-01-01
Sedna JPL Horizons 0.158 2029-01-01
Quaoar JPL Horizons 0.185 2029-01-01
Orcus JPL Horizons 0.162 2029-01-01
Gonggong JPL Horizons 0.036 2029-01-01
67P/Churyumov-Gerasimenko JPL Horizons 4.3 2029-01-01
9P/Tempel 1 JPL Horizons 29.3 2029-01-01

Lunar occultation times

Predicted disappearance and reappearance times were compared with the instant JPL’s Moon puts the star on a smooth limb of the same radius, for 14 contacts at 7 sites. The largest difference was 3.7 seconds and the average 1.1 seconds. Real mountains and valleys on the Moon’s edge are not modeled and add a few seconds more.

Not covered by these checks

  • Comets other than the two in the table above.
  • The times of Jupiter’s moon events, and occultations of planets.
  • Satellite positions, which drift by about a kilometer across the track within two weeks of the element date.
  • Any date outside 2000 to 2040. The same models are used there and have not been compared here.

What is not a measurement

Every source record in the app carries one of five labels, shown on an object’s Evidence tab and on the evidence pages:

  • Observed: an image or a direct measurement.
  • Catalog: a value taken from a published catalog.
  • Computed: calculated from observed or catalog values with a stated method.
  • Reconstructed: filled in between measurements, such as a spacecraft path between known points.
  • Illustrative: drawn to help you see something, and not to be read as data.

Some things are deliberately not to scale or not real, and say so where they appear:

  • Black-hole markers at Sagittarius A* and M87* are far larger than the real horizons, and are labeled with the true size.
  • Exoplanet globes are schematic. Their surfaces are hypothetical, not imagery, and their sizes are enlarged.
  • Predictions are labeled as predictions. A transit time, a variable-star maximum or an occultation is never presented as something observed.
  • A table you load is drawn as given. Rows without a usable distance are shown on the sky only and are never given a distance.

Known limits

  • The Moon’s position from the analytic series trails JPL’s along its orbit by roughly a second or two of lunar motion in this era. Occultation predictions correct for the average of that. What remains is about one arcsecond.
  • Near the horizon the real sky differs from any calculation, because refraction depends on the weather.
  • A time zone worked out from coordinates can be the neighboring zone within a few kilometers of a border. A location chosen from the city search carries its exact zone.
  • About four in ten published exoplanet epochs do not say which time scale they use. Those are read as UTC and marked, and the transit could come about 69 seconds earlier if the authors meant TDB.
  • Light pollution is an estimate from a world sky-brightness atlas, not a measurement at your site.

Citing Space Time

Space Time changes most weeks, so cite the version you used and the date you used it. The version number is shown in the app and on this page.

Space Time, version 1.0.895. Point Dynamics. https://space.pointdynamics.com/. Accessed [date].

For a specific result, also cite the source named on the object’s card or in the exported file: for example JPL Horizons with its orbit solution number, Gaia DR3, or the paper behind an exoplanet’s transit clock. Exported tables carry that information in their header.

Found something wrong? Use the feedback button on any object’s card in the app. Changes to the app are listed on the What’s new page.

Open Space Time