A star that slowly fades as it sinks toward the horizon might not be changing at all. Earth’s atmosphere absorbs and scatters some of its light, and the amount depends on the viewing path. Astronomers use air mass, atmospheric-extinction coefficients and photometric zero points to separate an observing effect from a real change in the source.
This guide works through two observations of one fictitious star, derives an extinction slope of 0.20 magnitude per air mass, calibrates a nightly zero point and explains the limitations of differential photometry. It also shows how to recognise when weather, colour or geometry invalidates an otherwise neat straight-line correction.
eduKateSG · The Gold Standard of Astronomy
Choose your reading route
Observe the cause · Chapters 1–3: light path, air mass and extinction slope.
Calibrate carefully · Chapters 4–6: zero point, differential measurement and testing.
Verify the claim · Chapters 7–8: FAQs, linked guides and sources.
Chapter index
Observe the cause · Chapters 1–3
Calibrate carefully · Chapters 4–6
Verify the claim · Chapters 7–8
CHAPTER 1 OF 8
1. A star does not become intrinsically fainter when it approaches the horizon
Watch a familiar constellation climb through the sky. Its stars may appear clearer and brighter at high altitude than near the horizon. Much of that difference is caused not by a change inside the stars, but by the changing length of atmosphere their light travels through before reaching a telescope. Gases, aerosols and other atmospheric constituents remove light from the measured beam through scattering and absorption.
In precision photometry, this effect is atmospheric extinction. It is separate from the interstellar dust extinction that starlight may experience long before arriving at Earth. Interstellar dust depends on the distant path to the star. The Earth’s atmosphere adds a changing local path dependent on the observing site’s weather and where the object sits in the sky.
Air mass expresses the relative thickness of atmosphere through which a line of sight passes, usually normalised to approximately one straight-down atmospheric column at the zenith. Higher air mass generally means a longer path and more atmospheric dimming. Near the horizon the geometry, refraction and actual atmospheric profile make naive approximations less reliable.
Astronomers therefore record when and where images were taken, how high the object was above the horizon and what standard-star observations reveal about the night’s transmission. Without this context, an instrument can interpret a changing atmosphere as a changing star.
The core question is deceptively simple: did the star change, or did the amount of atmosphere between us and the star change? A good observing programme is designed so those explanations can be distinguished.
For moderate zenith angles, a useful approximation is X ≈ sec(z), where z is the angle from the zenith. Since altitude h = 90° − z, the same approximation is X ≈ 1/sin(h). The Astropy AltAz coordinate class exposes secz, explicitly describing it as a common estimate of air mass.
At an altitude of 90° the star is overhead: z=0°, so X≈1. At an altitude of 60°, z=30°, so X≈1/0.866≈1.155. At an altitude of 30°, z=60°, so X≈2. A star 30° above the horizon has roughly twice the zenith-column air mass in this simplified geometry.
That does not mean every molecule of air has exactly doubled, nor does it imply a constant extinction coefficient through every layer. It is a geometric approximation, helpful for first calculations and moderate air masses. Close to the horizon, curvature, vertical variations, refraction and increased atmospheric variability demand a more suitable model.
Astropy distinguishes geometrical coordinates from refraction-aware transformations. Its docs warn that refraction computations can become inaccurate at very low altitudes. Do not use a singular secant formula at h≈0° to claim a physical infinity of air.
In practice astronomers generally prefer well-elevated targets and record the chosen air-mass calculation method, rather than relying on visually guessed altitude. A difference between X=1.2 and X=2.0 is enough to produce a measurable photometric change on an otherwise clear night.
| Altitude | Zenith distance | Simple X = sec(z) |
|---|---|---|
| 90° | 0° | 1.000 |
| 60° | 30° | ≈1.155 |
| 30° | 60° | 2.000 |
Suppose a fictitious standard star is observed through the same filter on a stable, photometric night. Its measured instrumental magnitude is 12.40 at air mass X=1.2 and 12.56 at X=2.0. The second measurement is numerically larger, so the star appears fainter as the light path lengthens.
Let a simple first-order model be m_inst = m_top + kX, where k has units of magnitude per air mass and m_top includes the instrument’s arbitrary photometric scale but extrapolates to zero atmospheric column. The change in instrumental magnitude is 12.56−12.40=0.16 across ΔX=0.8. The slope is therefore k = 0.16/0.8 = 0.20 magnitude per air mass.
Insert the first observation into the model: m_top=12.40−0.20×1.2=12.16. Check the second: 12.56−0.20×2.0=12.16 again. Both points agree with the same intercept and slope because the toy data were deliberately constructed that way.
The intercept is not a directly measured star above Earth’s atmosphere. It is a fitted extrapolation within an assumed, stable atmospheric model and one instrumental passband. Nor is 0.20 an atmospheric constant valid for every observing location, night or wavelength. The AAVSO’s observing guidance explains that atmospheric extinction coefficients vary and that bluer bands are generally more strongly affected.
With only two points you can calculate an exact slope, but you cannot demonstrate that the sky remained transparent, the camera behaved linearly or the assumed model was adequate. Real calibration uses repeated appropriate standards and checks residuals and changes over time.
| Air mass X | Instrumental magnitude | Extinction kX | Corrected instrumental intercept |
|---|---|---|---|
| 1.2 | 12.40 | 0.24 mag | 12.16 |
| 2.0 | 12.56 | 0.40 mag | 12.16 |
Atmospheric extinction and instrumental zero point answer different questions. The extinction coefficient tells us how the measured magnitude changes with air mass. The zero point aligns the corrected instrumental magnitude with a specified standard magnitude scale. They can be related in a fitted calibration equation, but treating them as one fixed adjustment hides information.
For the model used here define m_standard = m_inst − kX + Z. Suppose the catalogue’s trusted standard magnitude for the fictitious star is 11.20. We already estimated its extinction-corrected instrumental value as 12.16. Thus Z=11.20−12.16=−0.96. At X=1.2 the prediction is 12.40−0.24−0.96=11.20. At X=2.0 it is 12.56−0.40−0.96=11.20.
That makes the separation visible. The same star’s apparent instrumental magnitude changes as it moves through different atmospheric columns; its correctly calibrated standard magnitude remains unchanged under our simplified assumptions. The zero point itself depends on exposure scaling, throughput, detector and the chosen magnitude convention, among other practical details.
Now suppose a target star has an instrumental magnitude 13.70 at X=1.8 in the same verified instrumental conditions. The model gives m_standard = 13.70−0.20×1.8−0.96 = 12.38. Report this as an example of an inferred calibrated magnitude, not as a source’s physical brightness or intrinsic luminosity.
Real transformations may need colour terms and interactions with air mass because filters, detectors and atmospheric transmission are wavelength dependent. The AAVSO photometry guidance advises using standard photometric filters, similar-colour comparison stars, careful flat-fielding and measured uncertainty. One simple zero-point correction cannot guarantee all these requirements.
| Input | Toy value | Role |
|---|---|---|
| Catalogue standard magnitude | 11.20 | Photometric standard reference |
| Extinction-corrected instrument magnitude | 12.16 | Instrument-only reference on toy scale |
| Zero point Z | −0.96 | Aligns instrument to standard system |
| Target m_inst at X=1.8 | 13.70 | Uncalibrated target measurement |
| Target m_standard | 12.38 | Corrected using k=0.20 and Z=−0.96 |
CHAPTER 5 OF 8
5. Differential photometry cancels some atmosphere—but not everything
Many variable-star measurements compare a target to nearby non-variable stars in the same image rather than calibrating every observation as an independent all-sky standard measurement. If the target and comparison stars are close together and observed simultaneously, much of the common atmospheric transparency change can cancel in the magnitude difference.
Suppose both target and comparison star experience exactly the same extinction offset of +0.30 magnitude in an image. Subtracting their measured magnitudes cancels that common term. That is the appeal of differential photometry: stable reference stars can isolate a target’s change despite some conditions shared across the frame.
The cancellation is not unlimited. A wide field can contain a meaningful air-mass gradient; stars of different spectral colour can be extinguished differently because atmospheric transmission depends on wavelength. The AAVSO’s DSLR observing manual specifically discusses primary extinction, colour-dependent secondary extinction and differential extinction across a wide field.
Imagine a target and comparison separated enough to experience X=1.6 and X=1.3, with equal k=0.20. Their differential atmospheric term is k(1.6−1.3)=0.06 magnitude. If it is omitted, the target–comparison difference includes this spurious offset. A genuine variable’s small-amplitude light curve may be distorted by a contribution of that size.
Using several stable comparison stars can help test consistency, but more comparison stars do not automatically correct colour mismatches, instrumental saturation or changing cloud cover. Inspect the light curves of the references themselves, note their filters and colour, and test whether residuals correlate with air mass.
| Comparison | Target X | Reference X | Differential kΔX |
|---|---|---|---|
| Well-matched local sightline | 1.5 | 1.5 | 0.00 mag |
| Moderately separated sightlines | 1.6 | 1.3 | +0.06 mag |
Suppose the true coefficient is 0.20 but an observer fits k=0.12. Their estimated extinction is too small by 0.08 magnitude per air mass. If the zero point was established from a standard star at X=1.2 and the target was measured at X=1.8, the relative correction error after anchoring the standard is 0.08×(1.8−1.2)=0.048 magnitude in magnitude; its sign depends on the convention used for reporting the residual.
This illustrates why it is more informative to know both the standard’s and target’s air masses than to declare that the coefficient was wrong by a certain percentage. Calibration errors partly cancel when calibration and target are at the same air mass, but become more visible when the observations are widely separated.
Check the observing programme for repeat observations of standard stars over a useful range of air mass. If all standards occupy nearly the same air mass, extinction and zero point are difficult to disentangle. If conditions vary between observations, a simple straight line in time or air mass may not represent the night’s transparency.
Plot standard-star residuals against air mass, time, detector position and colour. A gradient in the first may expose a bad extinction slope; a trend with time may show variable transparency; positional structure can indicate flat-field errors; colour-dependent residuals can indicate an inadequate transformation.
A cloud passing through the field is not necessarily represented by the same smooth air-mass law. Likewise, aerosol content may change during the night, so a slope fitted early in the session may not be valid later. The response to an unstable calibration is to restrict the claims, model the variation with sufficient evidence or repeat observations under suitable conditions.
Record image calibration, shutter/exposure convention, photometric filters, standard-star identifiers, calculated air mass, fitted coefficients, zero-point definitions and uncertainties. These details make a claimed two-hundredths-of-a-magnitude variation reproducible instead of merely plausible.
Does a higher air mass always mean a genuinely fainter star?
No. It usually means a longer atmospheric path and more attenuation in the measured beam. The star’s intrinsic luminosity need not change.
Is air mass exactly 1/sin(altitude) at every elevation?
No. That is a useful approximation at moderate zenith angles. Near the horizon real atmospheric structure and refraction demand other models.
Is k=0.20 magnitude per air mass always safe?
No. It is invented for these calculations. Coefficients vary by site, filter and atmospheric conditions.
Does a zero point include extinction?
Different calibration equations combine terms differently. In this guide the zero point and kX are written separately; always read the adopted convention.
Why choose comparison stars with similar colours?
Colour-dependent atmospheric extinction and filter transformations can otherwise leave residual trends even after common transparency variations cancel.
Can a star’s observed colour change with altitude?
Yes, the atmosphere’s extinction is wavelength dependent, so the same source can acquire air-mass-dependent colour shifts in measured bands.
Does observing near the zenith eliminate calibration?
No. Air mass is still about one, not zero, and detector throughput, standards, flat-fielding and noise still matter.
Are atmospheric and interstellar extinction interchangeable?
No. They occur along different parts of the light path and are inferred or corrected through different evidence and models.
Reliable stellar observations require a chain of measurements rather than a single attractive graph. Begin with the detector values and sky subtraction, control the atmosphere, anchor the photometric scale, then investigate whether the source genuinely varied. Later inferences about stellar luminosity, transit depths or periods add further models and uncertainties.
To test your understanding, repeat the standard-star experiment with an assumed k=0.25 mag per air mass while retaining the same atmospheric-free instrumental intercept 12.16. At X=1.2 you would predict m_inst=12.46; at X=2.0 you would predict 12.66. Explain why the increase in slope does not require a change in the star itself.
Return to the subject hub
The Gold Standard of Astronomy — navigate the subject and its specialist investigations.
The Gold Standard of Education: Learning Transfers into Life — practise checking whether a method survives changed conditions.
How X Works — wider explanatory pathways.
Build the observation chain
Aperture photometry and sky counts — measure star and sky consistently.
Stellar reddening and interstellar extinction — distinguish the faraway dust path from the local atmosphere.
Difference-image flux and reference noise — identify a real change relative to a baseline.
Authoritative sources and model scope
Sources reviewed on 22 October 2026. The star magnitudes, k value, zero point and differential measurements here are entirely original teaching data; they are not a calibrated observing log. The air-mass expression is a moderate-zenith-angle approximation and the linear extinction law is a deliberately restricted model.
AAVSO: Stellar Photometry — data-quality advice, comparisons, filters and photometric practice.
AAVSO DSLR Observing Manual, Appendix C — primary, secondary and differential atmospheric extinction.
Astropy: AltAz — geometric air-mass estimator and atmospheric refraction constraints.