---
title: "Read Gaia’s parallax zero point and scanning errors in the original papers"
description: "Follow fixed Gaia papers through source models, parameter correlations, scanning patterns and the range of parallax zero-point corrections."
canonical_url: "https://sandboxphysics.com/stories/astrometry/reading-gaia-systematics/"
language: "en"
page_type: "story"
alternate_languages:
  zh-CN: "https://sandboxphysics.com/zh/stories/astrometry/reading-gaia-systematics/"
last_modified: "2026-10-28"
---

[Measuring the depth of the sky](https://sandboxphysics.com/stories/astrometry/)

L4 · Stellar astrometry

# Read Gaia’s parallax zero point and scanning errors in the original papers

Sandbox Physics · 2026-10-28

Open Lindegren and colleagues’ arXiv:2012.03380v1 and arXiv:2012.01742v2. The first explains the astrometric solution; the second investigates its parallax bias. The sections and figures below help you distinguish formal uncertainty from calibration error.

## Before you begin

Complete the Gaia sample tutorial first. These are fixed versions uploaded in 2020 and subsequently published. They provide a reproducible reading target, without claiming that later releases have removed every issue.

## Begin with the star assumed by section 3.1

In section 3.1 of The astrometric solution, identify the source direction, its spatial motion and the observer’s position. The source model connects many scans into one trajectory. The annual observer term and the star’s motion enter the same calculation.

Compare it with the short five-parameter teaching equation. A circular observer orbit and omitted radial acceleration help make that equation linear and compact. The real solution also needs reference frames, ephemerides and light-propagation effects. This comparison shows where an unmodelled trajectory could enter the residuals.

An unresolved binary adds another complication: the measured light-weighted centre can move around the system’s centre of mass. Motion near an annual timescale can interfere with parallax. Unusual residuals supply a clue. Determining a binary orbit requires fitting its motion and testing other error sources.

Reading: Lindegren et al.

## The ten correlations in section 5.5 carry information

Continue to section 5.5, equation 23 and figure 10 of the same paper. Five parameters have ten distinct pairs. The equation reconstructs their covariance from standard errors and correlations. Figure 10 maps those correlations across the sky.

Read the sample conditions in the caption before interpreting the colours. Each map shows correlation between parameter errors. It does not show arrows of stellar motion or a necessary physical relationship between two stellar properties. Figure 12 provides a useful comparison in ecliptic coordinates, showing how the coordinate choice alters the pattern.

Return to ra\_pmra\_corr in the tutorial. Position and proper motion were fitted from the same observations and can compensate for each other. Their covariance contributes when uncertainty is propagated from one epoch to another.

Reading: Lindegren et al.

## Why the scanning pattern appears in error maps

Turn to section 2.3 and figure 2 of Parallax bias versus magnitude, colour, and position. Gaia’s scanning schedule supplies different directions, dates and numbers of observations across the sky. First distinguish the panels showing precision, parameter correlation and quasar parallax: their colours represent different quantities.

Nearby catalogue sources can also share calibration errors. Independent random errors diminish when many measurements are averaged; a shared offset remains. The tutorial reconstructs covariance within each source’s five fitted parameters. It does not supply covariance between different sources.

For a comparison between sky regions, inspect their observation geometry and error structure. Comparing mean formal errors alone can miss the calibration difference most relevant to the result.

Reading: Lindegren et al. · Lindegren et al.

## Figure 5 does not reduce to one horizontal correction

Read section 4.1 and figure 5 of the bias paper. Quasars are so distant that their parallaxes are effectively zero, making them useful calibration references. Figure 5 plots mean parallax against magnitude, effective wavenumber and ecliptic latitude. Effective wavenumber describes spectral colour entering the instrumental calibration; it is not an elemental absorption-line wavelength.

A common additive error for every source would give nearly horizontal trends. The plotted dependence instead motivates a calibration that varies with brightness, colour and sky position. Section 4.1 also shows limited quasar coverage at bright magnitudes and at colour extremes. The calibration is more directly constrained where its reference sample actually exists.

This is why the exercise script does not add one constant to every parallax. It retains uncorrected values to illustrate inference. A scientific analysis needs suitable source-specific calibration inputs and uncertainty.

Reading: Lindegren et al.

## Check section 8 against figures 21 and 22

Section 8 gives different bias functions for five- and six-parameter solutions. Start with astrometric\_params\_solved: all tutorial sources have value thirty-one, indicating five-parameter solutions. Value ninety-five indicates six-parameter solutions with a fitted pseudocolour. These cases require different functions.

Figures 21 and 22 show predicted bias at several ecliptic latitudes. Follow the dependence on colour and magnitude, then inspect tables 9 and 10 and their validity ranges. With the paper’s sign convention, $\varpi_{\rm corrected}=\varpi_{\rm catalogue}-Z$. Subtracting a negative bias increases the parallax.

The authors describe the corrections as provisional and encourage targeted bias estimates for particular applications. Before using a recipe, record the release, solution type, availability of calibration inputs, location within the calibration range, and possible binary or between-source effects.

A three-dimensional Galactic model eventually needs distance distributions, observational selection and calibration errors together. The small public sample is enough to practise those steps. Large-scale structural conclusions require broader data and a tested error model.

Reading: Lindegren et al. · Bailer-Jones et al.

## Try the geometry

### [Distinguish noise from projection geometry](https://sandboxphysics.com/labs/stellar-parallax-proper-motion/)

1. Hold distance fixed and raise the position uncertainty. Inspect the random scatter and fitted uncertainty.
2. Change ecliptic latitude and inspect the projected shape. Gaia’s scanning law adds another change in measurement weights; keep these effects distinct while reading.

## Exercises and explained answers

Does correlation near minus one mean the two physical parameters must be anticorrelated in a stellar population?

These maps describe errors in fitted parameters for individual sources. One parameter can compensate for another under the observation geometry. That does not establish a physical population relation.

Why not apply the six-parameter bias function to this five-parameter sample?

The solution types handle colour information differently and have different calibration inputs and fitted dependencies. Identify the solution type before choosing the function and checking its valid range.

[Previous: Compare two distance estimates for twenty-four Gaia stars](https://sandboxphysics.com/stories/astrometry/gaia-distance-tutorial/)

## Sources and calculation scope

1. [Lindegren et al., Gaia EDR3 — The astrometric solution, arXiv:2012.03380v1](https://arxiv.org/html/2012.03380v1)

   Sections 3.1, 5.3–5.7; equation 23 and figures 10–12. The fixed single-source model and covariance conventions apply to the basic DR3 astrometry inherited from EDR3.

2. [Bailer-Jones et al., Estimating distances from parallaxes V, arXiv:2012.05220v1](https://arxiv.org/html/2012.05220v1)

   Sections II.1–II.3, V.2 and V.5. The published direction-dependent Galactic prior differs from this tutorial’s simple EDSD prior.

3. [Lindegren et al., Parallax bias versus magnitude, colour, and position, arXiv:2012.01742v2](https://arxiv.org/html/2012.01742v2)

   Sections 2.3, 4.1 and 8; figures 2, 5, 21–22 and tables 9–10. The tutorial applies no zero-point correction and does not claim calibrated catalogue distances.