The science behind Log On Globe

The Moon's position comes from a truncated version of the ELP2000/Meeus lunar theory, summing periodic terms of its orbit to fix its exact direction and distance from Earth at any moment. The same calculation detects solar and lunar eclipses, by measuring the alignment between the Sun and Moon and the size of Earth's shadow. In the wide Earth views the Moon is drawn closer and larger than reality so it stays visible; in the lunar-mission views it becomes a second, fully to-scale Moon at its true distance and true radius, close enough for a spacecraft to actually fly to and orbit. There's more on this in "Scale and sizes" below.

The 3D globe's orientation is set once when the app loads, roughly facing your own region. What actually moves is the light: the Sun's direction is recalculated every frame from the current (real or simulated) time and used to light the globe, producing the same visual effect as a rotating Earth without the cost of animating it. The optional timezone overlay works the same way, drawn from real timezone boundaries you can switch on in the settings.

The seasons fall straight out of the Sun calculation. As the subsolar point drifts north and south across the year, an equinox is the moment it crosses the equator and a solstice the moment it reaches its furthest point (the Tropic of Cancer or Capricorn). Log On Globe finds each of these instants from the same solar-declination model that draws the day/night line, then stages them as short scenes.

Both kinds come from the same Sun-and-Moon geometry. In a solar eclipse the Moon passes in front of the Sun, and the app renders it on the Sun's own disc: a total blackout ringed by the corona, an annular "ring of fire", or a partial bite, depending on the alignment and the Moon's distance. In a lunar eclipse the Moon passes through Earth's shadow and turns red, and the result varies a lot from one eclipse to the next for two independent reasons. Depth (the umbral magnitude) is how centrally the Moon crosses the shadow: a near-central pass swallows the whole disc in deep red for over an hour, a grazing pass keeps one edge bright and lasts only minutes. Colour is graded on the Danjon scale from 0 to 4, from a bright coppery orange down to a dim, almost invisible dark brown (the darkest eclipses tend to follow major volcanic eruptions). Each catalogued eclipse is rendered from its own real umbral magnitude and Danjon value, so a deep dark eclipse and a shallow bright one genuinely look different.

An aurora scene follows the real chain of events. A solar flare erupts on the Sun, throws a coronal mass ejection (a CME) across space, and the app animates that cloud of particles crossing to Earth. When it arrives it drives a stylised light-curtain effect over the poles, brighter and reaching to lower latitudes for stronger storms. The storm strength is the real Kp / G-scale level of the catalogued event, which is also where the named-storm titles come from.

The background sky is two layers. A faint scatter of tens of thousands of dots is purely decorative, there just to give the sky depth. On top of it sit the real stars: the main stars of all 88 constellations, placed from their true positions in the sky (their right ascension and declination), sized by their real brightness and tinted by their real colour, from the blue-white of the hottest stars to the orange-red of the coolest. They use the same sky-rotation maths as the Sun and Moon, so the whole sky turns correctly with the time of day and the time slider, and familiar shapes (Orion's belt, the Big Dipper, the Southern Cross) land where they should. Hover a star to reveal the lines of its constellation and read its name, type, distance in light-years and magnitude. The soft band arching across the sky is a real diffuse map of the Milky Way from NASA, and a few deep-sky objects (distant galaxies and nebulae) appear as real photographs at their true size on the sky. Star distances are best-known catalogue values, and we don't correct for the tiny drift of positions over the years, which is invisible at this scale.

The ISS is tracked from its real, current orbital elements (a "TLE", or Two-Line Element set), fetched regularly from CelesTrak and propagated forward in time with the SGP4 model, the same standard satellite trackers use worldwide. That gives the station's real position in orbit at any moment, not a simulated approximation. Its size on screen is exaggerated on purpose (see "Scale and sizes" below), since a station the length of a football field would otherwise be a speck next to the Earth.

Historical SpaceX Crew Dragon missions reuse the ISS's real orbital track, using the station's actual position at the time of each mission. The Dragon's short docking and undocking manoeuvres (a few hundred metres, imperceptible at globe scale) aren't physically simulated: they're drawn as a smooth, realistic sweep toward or away from the station. The lunar-mission ships (Apollo's command and lunar modules, Artemis's Orion, Starship) and their launch vehicles (Saturn V, SLS, Falcon 9, Super Heavy) are real 3D models flown along authored trajectories. How those paths are built is the next question.

It depends on the mission. Earth-orbit objects use real data: the ISS and Hubble are propagated from recent TLEs via SGP4, and the James Webb telescope follows a real JPL Horizons ephemeris. The Moon's position, everywhere, comes from the real Meeus lunar theory. The cislunar journeys, though, are a different story. The trip out to the Moon, lunar orbit, a free-return flyby or a near-rectilinear halo orbit, and the return and re-entry are not full orbital mechanics: there's no propagator for a translunar trajectory here. They're authored, realistic curves (smooth parametric and Catmull-Rom arcs, eased at every junction) tuned to match the real timeline and the real published figures (Apollo 11's phase proportions, Artemis II's 6,513 km flyby altitude, and so on). They arc toward the Moon's actual computed position at arrival, and a few real events (like the roughly 58° heading change at trans-lunar injection) are kept deliberately sharp while the rest is smoothed for watchability.

Both are tracked like the other satellites, but on very different orbits. Hubble sits in low Earth orbit and is propagated from its live TLE with SGP4, just like the ISS. The James Webb Space Telescope orbits a point far beyond the Moon (the Sun–Earth L2, about 1.5 million km out), so a TLE doesn't apply; it follows a real JPL Horizons ephemeris instead, and is placed at that true distance. Both models are drawn a little larger than scale to stay visible.

At Log On Globe's scale, one screen unit is one Earth radius, so anything built to true size next to the whole planet would be a single pixel or less. Wherever that would happen, the object is enlarged just enough to read, and only in the wide views where it's needed. The Sun keeps its real distance but its disc is oversized and held at a constant apparent size (with a small extra boost for the ISS and Dragon sunrise framings). The Moon is about five to six times larger than life in the wide Earth views, but snaps to its true radius and true distance in the lunar-mission views, where it's close enough to fill the frame honestly. The ISS and Hubble are drawn about three times their real size. Every crewed spacecraft uses the same rule of thumb, sized to read against black space; the launch stacks (Saturn V, SLS, Falcon 9, Super Heavy) are shrunk further so the whole rocket fits, and capsules start smaller at liftoff and grow back to full size once they've separated. None of this is ever presented as true scale.

It's a mix. The Sun and Moon come from real but simplified astronomical models, typically accurate to a fraction of a degree. The ISS, Hubble and Webb positions are only as accurate as their underlying orbital data, refreshed regularly from public sources. The cislunar mission paths are realistic illustrations rather than integrated orbital mechanics. In short: these models are built for a smooth, believable real-time visualisation, not for navigation or scientific use, and wherever we simplify or enlarge, we've tried to say so.