Methods

This page describes what each analysis does and the standard methods it builds on. The exact rules, thresholds and weights are part of the product: they were tuned on noisy, irregular real-world photometry, and they are tested on every release (see the test suite).

Track features

For each track, the Analysis workspace measures:

Feature	What it says
Magnitude	Typical brightness of the track
Magnitude at 1,000 km	The same brightness brought to a common range, so tracks at different distances compare
Variation	How much the brightness changes during the track
Noise and spread	The measurement noise, and the true scatter of the light curve, both estimated with robust statistics so that a few bad points do not distort them
Peak	How much the brightest points stand out of the track
Phase and geometry	Phase angle (signed as in the simulator), and the directions of the Sun and of the station seen from the object, see below

Standardized magnitude

Brightness depends strongly on the phase angle. The object’s own phase curve is learnt from its history, and each track is compared with it: the standardized magnitude says whether the object was brighter or fainter than usual for that geometry. This is what makes tracks from different nights, stations and geometries comparable.

Period of a track

The rotation period is searched with a generalised Lomb–Scargle periodogram, a standard tool for unevenly sampled data. A period is only accepted when it is statistically significant, covered by enough cycles, and when the light curve folds cleanly at that period (phase dispersion): irregular signals, which also produce peaks, are rejected this way. Light curves with two unequal maxima per turn are recognised, so the period reported is the full rotation. Outliers — thin clouds, a star in the aperture, a lost frame — are left out before the search.

Each period comes with its uncertainty, and each track with a periodicity score, used even when no period is accepted.

Classes

Every track gets one behaviour class:

Class	Meaning
Constant	The light curve does not change beyond the measurement noise: a stabilised or slowly turning object
Periodic	A rotation period is found and the brightness changes significantly over a turn
Aperiodic	The brightness changes, without a reliable period: tumbling, irregular object or difficult conditions
Glint	The track is much brighter than the object usually is for that geometry, or shows a flash that stands out of its usual behaviour
Too short	Not enough measurements to decide

Rotation trend

Periods are followed over time with robust running statistics, so that isolated wrong periods (half or double) do not move the trend. The workspace reports:

• the current period, when recent tracks are periodic often enough to support it;
• the drift, i.e. the spin-up or spin-down rate, in seconds per day;
• abrupt changes and their dates, when the period before and after a date differ significantly.

Behaviour monitor

There is no universal threshold for “abnormal”. The monitor compares the latest tracks of an object with its own history, on the standardized magnitude and on the periodicity score, with robust statistics. A significant departure on one or both gives a watch level, a stronger one a confirmed level; nearby alerts are grouped into one episode.

Equivalent diameter

The size of the Lambertian sphere that would give the same flux at the same phase angle, for an assumed albedo — a classical way to compare objects of unknown shape:

F / F☉ = (2/3) · ρ · R² · p(α) / d²,    p(α) = (sin α + (π − α) cos α) / π

It is computed on the tracks that suit it (glints excluded), and summarised by its most probable value and a 68 % range. The DISCOS mean cross-section, when known, is shown for comparison.

Geometry

The station comes from the stations file, else it is recovered from the angles of its tracks, as in the simulator. The line of sight comes from azimuth and elevation, else from right ascension and declination. In the frame of the object, each track gives the direction of the station and the facet direction: the normal a flat facet must have to mirror the Sun toward the station. Glints with similar facet directions are grouped: they point at the same physical facet. The season (Sun declination, distance to the equinoxes) is kept for each track.

Symmetry

The phase curves on the evening and morning sides are compared where both are observed, and summarised as a symmetry score from 0 to 1.

Signature and neighbours

Each object is summarised by a short signature — brightness, variation, share of periodic tracks, glints, phase behaviour, symmetry and stability. Objects with close signatures are neighbours; the class and shape of an object without metadata are suggested from those of its nearest neighbours with DISCOS metadata.