Home Library Glossary Tracking & Space Surveillance Covariance
📡 Tracking & Space Surveillance

Covariance

Also known as: Covariance Matrix, Position Uncertainty, Error Ellipsoid

Quick answer

Covariance is the formal statement of how uncertain a satellite's estimated position and velocity are — a matrix that draws an "error ellipsoid" around the predicted state. It is the raw material of collision-risk maths: without realistic covariance, a probability of collision cannot be computed honestly.

📘 Full definition✓ Reviewed 2026-09-07
Covariance is how orbit determination admits what it does not know. An estimated state vector is never exact — observations are noisy, force models imperfect — and the covariance matrix quantifies that doubt: a 6×6 grid whose diagonal holds the variance of each position and velocity component and whose off-diagonal terms record how the errors correlate. Geometrically it defines an uncertainty ellipsoid around the predicted position. The ellipsoid is rarely a sphere: for a typical LEO object it is stretched dramatically along the direction of travel, because a small error in orbital energy converts into a growing along-track timing error — the satellite is almost exactly on its track, but slightly early or late. Covariance also grows with prediction time, inflating as drag and model errors accumulate. Its central application is conjunction assessment: the probability of collision is computed by overlapping two objects' covariance ellipsoids at the time of closest approach, so the honesty of the answer rests entirely on the realism of the input covariances. An overconfident (too-small) covariance can hide genuine risk; a bloated one drowns operators in false alarms — which is why "covariance realism" is one of the quiet central problems of space traffic management.
Form
6×6 matrix
position + velocity uncertainties
Shape
Ellipsoid, along-track dominant
timing error grows fastest
Grows with
Prediction horizon
drag is the main inflator in LEO
Feeds
Probability of collision
core input to every Pc figure

Understanding Covariance

Why the ellipsoid points along the track

Orbital dynamics converts energy errors into timing errors. Estimate a satellite's semi-major axis a few metres high and you have its period wrong by milliseconds; within hours those milliseconds compound into a position error of kilometres — all of it along the direction of flight. Radial and cross-track errors, by contrast, oscillate rather than accumulate. The result is the signature cigar-shaped ellipsoid, often tens of times longer along-track than across it. Close-approach geometry inherits this shape: a conjunction that looks alarming in raw miss distance may carry most of its separation across the narrow axes of both ellipsoids, or vice versa, which is precisely what the probability-of-collision integral untangles.

Covariance realism — the hard part

A covariance is itself an estimate, and making it truthful is notoriously difficult. Filters produce formal covariances that reflect only the errors they model; unmodelled effects — atmospheric density surprises during a geomagnetic storm, unannounced manoeuvres, sensor biases — leave the formal number too optimistic. Operators therefore validate covariance empirically, comparing predictions against later truth and inflating the matrix until it matches observed error statistics. During strong solar activity, LEO covariances can balloon within hours as drag forecasts degrade, temporarily blinding screening systems at exactly the moment risk rises.

See it live The conjunction feed screens close approaches between tracked objects — every event on it is, underneath, a story about two overlapping uncertainty ellipsoids. Open the conjunction feed →
📖 Learn More

Frequently Asked Questions

A surface of equal probability: the satellite is most likely at the ellipsoid's centre, with likelihood falling off outward. A one-sigma ellipsoid contains the true position roughly 20% of the time in three dimensions, so screening work typically thinks in two- or three-sigma volumes — and in the overlap of two objects' volumes rather than either alone.
Because Pc is a ratio of overlap. Shrink the covariance and a near miss looks laser-precise: either clearly safe or clearly dangerous. Inflate it and the probability mass spreads so thin that even a genuinely close pass yields a small Pc — the notorious "dilution" regime, where a scary conjunction can show a reassuring number precisely because the data is poor.
From the orbit-determination process itself: the same least-squares or Kalman filter that estimates the state also estimates its error statistics, based on observation noise and force-model assumptions. Screening messages such as the CDM carry each object's covariance so that operators can recompute and combine risk with their own tools.

Sources & References

Definitions are reviewed against primary sources. Last reviewed: 2026-09-07.