Observational methods

Astronomy is an observational science with no experiments. Everything is inferred from radiation that arrives, so the methods of measurement and the handling of their systematics are the substance of the field rather than a preliminary to it.

Magnitudes

The magnitude system is logarithmic, inverted and inherited from antiquity, which makes it awkward and unavoidable.

A difference of five magnitudes is exactly a factor of 100 in flux, and one magnitude is a factor of about 2.512. Brighter objects have smaller magnitudes, and the brightest are negative.

Apparent magnitude $m$ is how bright an object appears. Absolute magnitude $M$ is how bright it would appear at 10 parsecs. The difference encodes distance:

This distance modulus is the basic tool of extragalactic distance measurement: if the intrinsic brightness of a class of object is known, its observed brightness gives its distance.

Magnitudes are defined within a photometric system specifying filter bandpasses. The Johnson UBVRI system and the SDSS ugriz system are the common ones, and they are not interchangeable. AB magnitudes are tied to a physical flux density rather than to a reference star, which is why they are preferred for anything requiring absolute calibration.

The distance ladder

Distance is the hardest quantity to measure in astronomy, and it is obtained by a chain of methods each calibrated on the one below.

Only the first two rungs are geometric and free of astrophysical assumptions. Everything beyond depends on the calibration of the rung below, so a systematic error at any level propagates all the way out.

Parallax is the only direct method beyond the solar system, and Gaia transformed it. Measuring positions to tens of microarcseconds gives useful distances across much of the Galaxy, and critically it recalibrated the Cepheid scale that everything above rung 3 rests on.

This structure is exactly why the Hubble tension described in Cosmology is hard to resolve. A discrepancy between the ladder-based and CMB-based values could be new physics or could be a systematic in one rung, and distinguishing them requires independent routes such as the tip of the red giant branch or gravitational wave standard sirens.

Spectroscopy

A spectrum contains more information than an image of the same object. Line identifications give composition, line widths give temperature and turbulent or rotational velocity, line ratios give density and ionisation state, and the overall shift gives velocity.

For small $z$ this corresponds to a recession velocity $v \approx cz$. At cosmological distances it is not a velocity at all but a measure of how much the universe has expanded since the light was emitted: a factor $1+z$.

Resolving power $R = \lambda/\Delta\lambda$ sets what is achievable:

Astrometry

Measuring positions precisely yields three things at once: parallax gives distance, proper motion gives transverse velocity, and repeated measurement reveals unseen companions through reflex motion.

Gaia measures nearly two billion sources at microarcsecond precision. Its catalogue underpins essentially all modern Galactic astronomy, because knowing distance converts every apparent quantity into a physical one. It is used by more papers than almost any other dataset in astronomy, largely as an input to work that is nominally about something else.

Time-domain surveys

Repeated wide-field imaging turns the sky into a time series and detects anything that moves or varies: supernovae, variable stars, asteroids, tidal disruption events, and the electromagnetic counterparts of gravitational wave events.

The scale is now a data engineering problem in its own right. The Rubin Observatory’s survey will produce on the order of 20 TB per night and issue millions of alerts nightly. No human examines these; the alerts feed automated brokers that filter, classify and distribute candidates to whoever has registered interest in that class.

The bottleneck has moved from detection to follow-up. Finding a transient is now easy; obtaining a spectrum to determine what it is remains expensive and slow, so classification increasingly relies on photometric machine learning with spectroscopic confirmation reserved for the interesting minority.

Calibration and reduction

The distance between raw detector output and a scientific measurement is large, and most systematic errors live in it.

Two of these steps deserve emphasis because they are where results go wrong.

Flat fielding corrects for the fact that pixels differ in sensitivity and that the optical path vignettes. An error here produces a smooth spurious gradient across the field, which looks like a real astrophysical signal rather than like noise.

Extinction correction accounts for dust absorbing and reddening light. Since dust reddens as well as dims, an uncorrected object appears both fainter and redder, which can be mistaken for a more distant or intrinsically redder object. Dust maps are themselves uncertain, and this is a leading systematic in supernova cosmology.

The general lesson is that in astronomy the statistical uncertainty is usually the easy part. Systematic uncertainty dominates most modern measurements, and the credibility of a result rests on how thoroughly the systematics were enumerated rather than on the size of the error bar.

References

• Howell, S. B. Handbook of CCD Astronomy, 2nd ed., Cambridge University Press, 2006.
• Bessell, M. S. “Standard Photometric Systems”, Annual Review of Astronomy and Astrophysics, 43:293, 2005.
• Freedman, W. L. and Madore, B. F. “The Hubble Constant”, ARA&A, 48:673, 2010.
• Gaia Collaboration, “The Gaia mission”, Astronomy and Astrophysics, 595:A1, 2016.
• Ivezić, Ž. et al. “LSST: From Science Drivers to Reference Design and Anticipated Data Products”, The Astrophysical Journal, 873:111, 2019.
• Ivezić, Ž. et al. Statistics, Data Mining and Machine Learning in Astronomy, Princeton University Press, 2020.