204.3. Astrometric calibration

204.3. Astrometric Calibration¶

For the Rubin Science Platform at data.lsst.cloud.
Data Release: Data Preview 1
Container Size: large
LSST Science Pipelines version: r29.2.0
Last verified to run: 2025-09-02
Repository: github.com/lsst/tutorial-notebooks

Learning objective: To understand the astrometric calibrations.

LSST data products: visit_image, deep_coadd

Packages: lsst.daf.butler, lsst.geom, lsst.rsp

Credit: Originally developed by Andrés A. Plazas Malagón and the Rubin Community Science team with input from Clare Saunders, Jim Bosh, and Pierre-François Leget. Please consider acknowledging them if this notebook is used for the preparation of journal articles, software releases, or other notebooks.

Get Support: Everyone is encouraged to ask questions or raise issues in the Support Category of the Rubin Community Forum. Rubin staff will respond to all questions posted there.

1. Introduction¶

The World Coordinate System (WCS), or astrometric solution, defines how pixel positions in an image map to celestial coordinates—right ascension and declination (RA, Dec).

The LSST Science Pipelines perform astrometric calibration in two main stages.

1. An initial WCS is estimated using a static camera model and refined per detector and exposure by matching sources to Gaia DR3, achieving an accuracy of approximately 10-20 milliarcseconds—sufficient for associating sources across visits.
2. A more precise solution is derived using overlapping visits with the GBDES (Bernstein, Armstrong, Plazas et al., 2017) algorithm. This model fits two components: a per-detector static map (capturing camera distortions and the physical layout of the camera) and a per-visit dynamic map (accounting for time-varying effects like atmospheric refraction). The solution minimizes positional differences between matched stars and Gaia references, and can optionally include proper motion, parallax, and corrections for differential chromatic refraction (DCR).

Ongoing work aims to better account for atmospheric turbulence, detector-level systematics, and long-term camera evolution, with plans to extend astrometric calibration from isolated calibration stars to the full LSST object catalog.

References

• Data Management Technical Note - 266: Astrometric Calibration in the LSST Pipeline
• PSTN-019: The LSST Science Pipelines Software: Optical Survey Pipeline Reduction and Analysis Environment
• RTN-095: The Vera C. Rubin Observatory Data Preview 1
• AST: A library for modeling and manipulating coordinate systems

Related tutorials: The Monster reference catalog is covered in another tutorial in this series (200).

1.1. Import packages¶

Import numpy, a fundamental package for scientific computing with arrays in Python (numpy.org), and matplotlib, a comprehensive library for data visualization (matplotlib.org; matplotlib gallery).

From the lsst package, import modules for plotting, accessing the butler and the TAP service, and geometry utilities from the LSST Science Pipelines (pipelines.lsst.io).

In [1]:

import lsst.daf.butler as dafButler
from lsst.rsp import get_tap_service
import lsst.geom
import numpy as np
import matplotlib.pyplot as plt
import astropy.units as u
from lsst.utils.plotting import (get_multiband_plot_colors,
                                 get_multiband_plot_linestyles)

1.2. Define parameters and functions¶

Instantiate the butler.

In [2]:

butler = dafButler.Butler("dp1", collections="LSSTComCam/DP1")
assert butler is not None

Instantiate the TAP service.

In [3]:

rsp_tap = get_tap_service("tap")
assert rsp_tap is not None

Get standard colors and linestyles

In [4]:

filter_colors = get_multiband_plot_colors()
filter_linestyles = get_multiband_plot_linestyles()

2. WCS for a visit image¶

The visit_image WCS is the most accurate pixel-to-sky transformation available for single-visit data. It is computed by fitting observed star positions to a reference catalog (e.g., "The Monster" reference catalog which includes Gaia stars), accounting for both telescope pointing and optical distortions.

Define a particular location on the sky and a band. Use the center of the Extented Chandra Deep Field South, in degrees.

In [5]:

ra_cen, dec_cen = 53.13, -28.10
my_band = 'i'

Get one visit_image around that location and band.

In [6]:

query = """visit.region OVERLAPS POINT(:ra, :dec) and band = :band"""
bind = {'ra': ra_cen, 'dec': dec_cen, 'band': my_band}
visitimage_refs = butler.query_datasets('visit_image',
                                        where=query,
                                        bind=bind,
                                        order_by=['detector'],
                                        with_dimension_records=True,
                                        limit=1)

Inspect the information about the visit image.

In [7]:

print(visitimage_refs)

[DatasetRef(DatasetType('visit_image', {band, instrument, day_obs, detector, physical_filter, visit}, ExposureF), {instrument: 'LSSTComCam', detector: 0, visit: 2024110800245, band: 'i', day_obs: 20241108, physical_filter: 'i_06'}, run='LSSTComCam/runs/DRP/DP1/DM-51335', id=1e1cd322-4ddd-4918-b2ac-c38e94fc8af5)]

In [8]:

visit_image = butler.get(visitimage_refs[0])

Get the WCS object.

In [9]:

wcs = visit_image.getWcs()

2.1. Pixel scale at a particular point.¶

Use the bounding box (bbox) information for the visit_image to get the pixel scale, in arcseconds, at the center of the detector. The pixel scale can be calculated at different points, and it may differ across the focal plane.

Warning: Avoid calling wcs.getPixelScale() without an argument; the implicit reference point may not suit your use case.

In [10]:

bbox = visit_image.getBBox()

In [11]:

wcs.getPixelScale(lsst.geom.Point2D(
    bbox.centerX, bbox.centerY)).asArcseconds()

Out[11]:

0.20034255528302644

Warning: The visit_image WCS has a getFitsMetadata() method that returns metadata information. For a visit_image, this metadata should not be used to reconstruct a WCS (e.g., with astropy) for precision astrometric analysis, as it only provides an approximation to the full WCS.

Clean up.

In [12]:

del query, bind, visitimage_refs
del visit_image, wcs, bbox

3. WCS from a coadd image¶

Query the butler for an i-band deep_coadd image that is near the center of the ECDS field.

In [13]:

center_coadd_datasetrefs = butler.query_datasets("deep_coadd",
                                                 where=f"band.name='{my_band}' AND\
                                                 patch.region OVERLAPS POINT(ra, dec)",
                                                 bind={"ra": ra_cen, "dec": dec_cen},
                                                 with_dimension_records=True,
                                                 order_by=["patch.tract"])

In [14]:

center_coadd_datasetrefs

Out[14]:

[DatasetRef(DatasetType('deep_coadd', {band, skymap, tract, patch}, ExposureF), {band: 'i', skymap: 'lsst_cells_v1', tract: 5063, patch: 14}, run='LSSTComCam/runs/DRP/DP1/DM-51335', id=63cf9a40-44ce-40d8-b1a9-827f236c3a8a)]

Retrieve the deep coadd image from the butler.

In [15]:

coadd = butler.get(center_coadd_datasetrefs[0])

Get the WCS.

In [16]:

wcs_coadd = coadd.getWcs()

3.1. Convert pixel to sky coordinates¶

The WCS methods pixelToSky and pixelToSkyArray can convert pixels to sky coordinates. The latter is faster.

Warning: The arguments passed to pixelToSkyArry must be double-precision floats, to avoid call errors.

In [17]:

wcs_coadd.pixelToSkyArray(300.4, 5000.6)

Out[17]:

(array([0.94270351]), array([-0.48996641]))

Warning: A WCS object can also be obtained from raw images. However, it is a rough estimate based on where the telescope thought it was pointing. It does not account for distortions or actual star positions and is only intended for internal use during initial processing. It should not be used for scientific measurements.

Clean up.

In [18]:

del center_coadd_datasetrefs
del coadd, wcs_coadd

4. Sky coordinates, errors, and astrometric accuracy¶

In the source and object catalogs, centroids are measured using the SdssCentroid algorithm, which applies a Gaussian weight function initially matched to the PSF size and then increases it by factors of two to match the weighted size of the object.

This algorithm operates entirely in pixel coordinates, not sky coordinates. Once the centroid and its uncertainty are determined in pixels, the WCS is used to transform these measurements into right ascension and declination. The uncertainties in RA and Dec are derived by propagating the pixel-level errors through this transformation.

4.1. Centroid uncertainties¶

Use the TAP service to retrieve from the object table the centroid uncertainties and magnitudes for PSF stars used in modeling.

In [19]:

radius = 2.0

In [20]:

query = (
    "SELECT objectId, "
    f"{my_band}_raErr, {my_band}_decErr, "
    "refExtendedness, "
    f"scisql_nanojanskyToAbMag({my_band}_psfFlux) as"
    f" {my_band}_psfmag, "
    f"{my_band}_calib_psf_used, "
    f"{my_band}_pixelFlags_inexact_psfCenter, "
    f"{my_band}_calibFlux "
    "FROM dp1.Object "
    "WHERE refExtendedness = 0.0 "
    f"AND {my_band}_calib_psf_used = 1 "
    f"AND {my_band}_pixelFlags_inexact_psfCenter = 0 "
    f"AND {my_band}_calibFlux > 360 "
    "AND CONTAINS(POINT('ICRS', coord_ra, coord_dec), "
    f"CIRCLE('ICRS', {ra_cen}, {dec_cen}, {radius})) = 1 "
    "ORDER BY objectId"
)
print(query)

SELECT objectId, i_raErr, i_decErr, refExtendedness, scisql_nanojanskyToAbMag(i_psfFlux) as i_psfmag, i_calib_psf_used, i_pixelFlags_inexact_psfCenter, i_calibFlux FROM dp1.Object WHERE refExtendedness = 0.0 AND i_calib_psf_used = 1 AND i_pixelFlags_inexact_psfCenter = 0 AND i_calibFlux > 360 AND CONTAINS(POINT('ICRS', coord_ra, coord_dec), CIRCLE('ICRS', 53.13, -28.1, 2.0)) = 1 ORDER BY objectId

In [21]:

job = rsp_tap.submit_job(query)
job.run()
job.wait(phases=['COMPLETED', 'ERROR'])
print('Job phase is', job.phase)
if job.phase == 'ERROR':
    job.raise_if_error()
assert job.phase == 'COMPLETED'
psf_used_table = job.fetch_result().to_table()

Job phase is COMPLETED

Plot the distribution of the propagated uncertainties on the centroids. Convert the errors from degrees to miliarseconds.

In [22]:

ra_err = psf_used_table[f"{my_band}_raErr"].to(u.mas)
dec_err = psf_used_table[f"{my_band}_decErr"].to(u.mas)

In [23]:

max_err = max(ra_err.max(), dec_err.max())
error_bins = np.linspace(0, max_err, 50)

fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(8, 3), sharey=True)

for ax, err, label in zip((ax1, ax2), (ra_err, dec_err), ('RA Error', 'Dec Error')):
    _, _, patches = ax.hist(err, bins=error_bins,
                            histtype='step',
                            linewidth=2,
                            color=filter_colors[my_band],
                            label=label)
    for patch in patches:
        patch.set_linestyle(filter_linestyles[my_band])
    ax.set_xlabel(f'{label} [mas]')
    ax.set_title(f'{label} Distribution')
    ax.legend()

ax1.set_ylabel('Count')
plt.tight_layout()
plt.show()

Figure 1: Right Ascension and Declination SdssCentroid error distributions for a subset of stars used in PSF modeling from the object table.

4.2. Centroid error vs magnitude¶

Visualize the centroid error as a function of magnitude. The centroid error is proportional to the seeing and inversely proportional to the signal-to-noise ratio of each object.

Compute total centroid error in quadrature.

In [24]:

centroid_err = np.sqrt(ra_err**2 + dec_err**2)

Get the corresponding PSF magnitude.

In [25]:

mag = psf_used_table[f"{my_band}_psfmag"]

Display the centroid error as a functionn of magnitude.

In [26]:

plt.figure(figsize=(8, 5))
plt.scatter(mag, centroid_err, s=10, alpha=0.7)
plt.xlabel(f"{my_band} PSF Magnitude", fontsize=12)
plt.ylabel("Centroid Error (mas)", fontsize=12)
plt.title(f"Centroid Error vs. Magnitude in {my_band}-band", fontsize=14)
plt.grid(True)
plt.tight_layout()
plt.show()

Figure 2: SdssCentroid errors as a function of i-band magnitude.

5. Notes on astrometric accuracy¶

The plots in Sections 4.1 and 4.2 use the propagated uncertainties on the centroids from the fit with the SdssCentroid algorithm, but they are not indicative of the overall astrometric accuracy. To assess the performance of the complete astrometric solution, internal and external consistency tests are performed.

For internal consistency, the repeatability of position measurements for the same object (calculated as the RMS of the fit positions per object) shows a mean per-tract astrometric repeatability of approximately 10 milliarcseconds.

For external consistency, the median separation between sources not included in the astrometric fit and their associated counterparts in Gaia DR3 is within 5 milliarcseconds—well below the main survey’s design requirement of 50 milliarcseconds. Remaining residuals are due to distortions not yet included in the LSSTComCam astrometric model (but planned for future inclusion), such as atmospheric, camera, and detector-level distortions.

For more details, see Section 5.3 on Astrometry in the DP1 paper.