Atmospheric Extinction and Airmass

Point a telescope anywhere but straight up and you are photographing a star through a dirty, wedge-shaped window. Before a single photon reaches the detector it must run a gauntlet of air: ozone in the stratosphere, water vapour and molecular oxygen lower down, aerosols and dust in the boundary layer. Some photons are absorbed outright; others are scattered out of the beam into the diffuse glow we call skylight. Either way they are lost, and the flux that survives is exponentially smaller than the flux that entered — the Beer–Lambert–Bouguer law, the same exponential that governs light through fog or coloured glass. The bookkeeping variable is not distance in kilometres but the airmass $X$ , the slant column of air divided by the vertical one. At the zenith $X=1$ by definition; tilt toward the horizon and the sightline slices through ever more air, so $X$ grows and the dimming with it. For a thin, flat atmosphere the geometry collapses to a single line, $X=\sec z$ , where $z$ is the zenith angle — the workhorse relation of every observing run. Two facts make this indispensable. First, because magnitudes are logarithmic and Beer's law exponential, the two compose into a perfectly straight line: observed magnitude rises linearly with airmass, and extrapolating to $X=0$ recovers the magnitude a satellite would measure. Second, extinction is not grey — Rayleigh scattering scales as $\lambda^{-4}$ , so blue light is stripped harder than red, dimming and reddening low stars. Every calibrated ground-based magnitude, and every rung of the distance ladder built from it, passes through this correction first.

Key takeaways

Air removes photons from a starbeam by absorption and scattering, attenuating the flux exponentially with the amount of air traversed — Beer's law $F=F_0 e^{-\tau X}$ .
The path is measured by the airmass $X$ , the slant column divided by the vertical column; $X=1$ at the zenith and grows as $X=\sec z$ toward the horizon.
Because magnitudes are logarithmic, the exponential becomes linear: $m(X)=m_0+kX$ , a straight Bouguer line whose slope $k$ is the extinction coefficient and whose $X=0$ intercept $m_0$ is the above-atmosphere magnitude.
Extinction is wavelength dependent — Rayleigh scattering ∝ λ⁻⁴ — so $k$ is largest in the blue and low stars are measurably reddened.
The secant law is a plane-parallel approximation; below about 15° altitude Earth's curvature and refraction cap the real airmass near 38, and formulae like Kasten–Young are required.
Key idea: extinction removes flux exponentially with the air column, so in logarithmic magnitudes it becomes the linear law $m(X)=m_0+kX$ — the intercept recovers the space magnitude and the slope is the band-dependent extinction coefficient.

The core insight

Extinction removes flux exponentially with the air column traversed; because magnitudes are logarithmic, this becomes a linear law $m(X)=m_0+kX$ whose intercept recovers the above-atmosphere magnitude and whose slope is the band-dependent extinction coefficient.

Where this lives today

Extinction correction is a nightly ritual at every ground-based observatory and a live systematic in precision cosmology. On any photometric night an observer images standard stars — Landolt or Stetson fields — at a spread of airmasses, fits a Bouguer line per filter, and folds the coefficients into the transformation of program-star magnitudes onto a standard system. Large surveys automate this: SDSS, Pan-STARRS, and Rubin's LSST monitor extinction continuously, since at millimagnitude precision even the second-order term $k''CX$ and the drift of aerosol through a night become limiting systematics. The stakes reach cosmology directly: Type Ia supernovae used to measure the Hubble constant have their magnitudes corrected for extinction before any distance modulus is derived, so a mis-modelled slope leaks straight into $H_0$ . It is also why the field turns to space — HST, Gaia, and JWST measure $F_0$ directly — and why site testing for the extremely large telescopes obsesses over precipitable water vapour, aerosol loading, and median extinction, the very quantities Bouguer first tried to pin down on a mountainside in Peru.

The physical picture

Think of the atmosphere as a stack of thin, semi-transparent sheets. Each sheet removes a fixed fraction of whatever light reaches it — not a fixed amount — so the survivors compound multiplicatively, sheet after sheet, exactly as compound interest compounds a balance. That is the definition of an exponential, which is why the transmitted flux is $F_0 e^{-\tau_s}$ with $\tau_s$ the total optical depth along the slant path. Now tilt the stack: at zenith angle $z$ the sightline crosses each sheet along a longer diagonal, meeting more material, so the optical depth scales up by the geometric factor $X$ . That factor is the airmass, and for a flat atmosphere it is just $\sec z$ .

Figure

Airmass geometry for a plane-parallel atmosphere: a sightline at zenith angle z crosses a slant path h over cos z, longer than the vertical column h, so the airmass is X = sec z. At z = 60° the slant path is exactly twice the vertical one, giving X = 2

The dimming is not colour-blind. Molecular (Rayleigh) scattering rises steeply toward short wavelengths as $\lambda^{-4}$ , so blue photons are flung out of the beam far more readily than red ones — the same physics that paints the daytime sky blue and the setting Sun red. A star viewed through two airmasses therefore loses relatively more blue flux, arriving fainter and redder than it truly is. When we plot measured magnitude against airmass we get a straight line, and the whole trade of a photometric night is reading that line's slope and intercept. The second figure shows such a Bouguer line for a star tracked from the zenith toward the horizon.

Figure

A Bouguer line: the measured magnitude of one star rises linearly with airmass, m of X equals m sub zero plus k X. Three nightly points at X = 1.0, 1.5, 2.0 give magnitudes 12.30, 12.45, 12.60; the fitted slope is k = 0.30 mag per airmass and the extrapolated intercept at X = 0 is the space magnitude m sub zero = 12.00

The core equations

The physics starts with attenuation. Each thin layer removes a fraction of the flux proportional to its optical depth, and integrating that loss over the slant column gives an exponential. Writing the slant optical depth as the vertical depth $\tau$ times the airmass $X$ puts the law in observing form.

F(X)=F_0\,e^{-\tau X} \qquad (1)

where: $F(X)$ — flux received at the ground through airmass $X$ ; $F_0$ — flux above the atmosphere, the space value; $\tau$ — vertical optical depth, dimensionless; $X$ — airmass, the ratio of slant to vertical path. Reading it: at $X=0$ nothing is lost and $F=F_0$ ; each added airmass multiplies the surviving flux by $e^{-\tau}$ . Astronomers measure magnitudes, which are $-2.5\log_{10}$ of flux. Feeding the exponential through that logarithm turns multiplication into addition and produces the single most useful equation in practical photometry — a straight line in airmass.

m(X)=m_0+kX,\qquad k=\frac{2.5}{\ln 10}\,\tau\approx 1.0857\,\tau \qquad (2)

where: $m(X)$ — apparent magnitude through airmass $X$ ; $m_0$ — above-atmosphere magnitude, the $X=0$ intercept; $k$ — extinction coefficient in magnitudes per airmass; $1.0857=2.5/\ln 10$ converts optical depth to magnitudes. Reading it: the slope is the atmosphere's fingerprint for that filter and night; the intercept is what you want. The geometry that sets $X$ is, in the simplest model, pure trigonometry: treat the atmosphere as a flat slab of thickness $h$ ; a ray at zenith angle $z$ crosses it along a path $h/\cos z$ , so the airmass is the secant of the zenith angle.

X=\sec z \qquad (3)

where: $z$ — zenith angle of the source, measured from straight up; $\sec z=1/\cos z$ . Reading it: $X=1$ at the zenith, $\sqrt{2}$ at $z=45°$ , and $2$ at $z=60°$ ; it climbs without bound as $z\to 90°$ , precisely where the flat-slab model must fail. To use equation (3) you need the zenith angle, which you rarely observe directly. It follows from the spherical triangle joining the pole, the zenith, and the star, giving $z$ from the observer's latitude and the source's equatorial coordinates.

\cos z=\sin\phi\,\sin\delta+\cos\phi\,\cos\delta\,\cos H \qquad (4)

where: $\phi$ — observer's geographic latitude; $\delta$ — declination of the source; $H$ — hour angle, the source's angular distance west of the meridian. Reading it: at transit $H=0$ and $\cos z$ is maximal, so airmass is least — the reason schedulers observe near the meridian. The secant law is only the leading term. Because the real atmosphere is a curved shell, the true airmass grows more slowly than $\sec z$ near the horizon; Hardie's cubic polynomial in $(\sec z-1)$ tracks the departure to moderately large zenith angles.

X\approx\sec z\,\big[\,1-0.0018167(\sec z-1)-0.002875(\sec z-1)^2-0.0008083(\sec z-1)^3\,\big] \qquad (5)

where: $\sec z$ — the plane-parallel estimate being corrected; $X$ — the refined airmass, valid to larger zenith angles. Reading it: a small negative nudge that vanishes at the zenith, where $\sec z-1=0$ . For work to the horizon, Kasten and Young's empirical fit is the modern standard; it stays finite as the secant diverges.

X=\frac{1}{\cos z+0.50572\,(96.07995-z_{\deg})^{-1.6364}} \qquad (6)

where: $z_{\deg}$ — the zenith angle in degrees; $X$ — airmass accurate to the horizon. Reading it: at $z=90°$ the added term keeps the denominator nonzero, capping $X\approx 38$ instead of letting it blow up. Extinction depends on colour because its dominant molecular component, Rayleigh scattering, is a steep power law in wavelength.

\tau_{\text{Ray}}(\lambda)\propto\lambda^{-4} \qquad (7)

where: $\tau_{\text{Ray}}$ — the Rayleigh (molecular) contribution to the vertical optical depth; $\lambda$ — wavelength of light. Reading it: halving the wavelength multiplies the Rayleigh optical depth sixteenfold, so blue extinction vastly exceeds red. Because a filter passes a band of wavelengths, the effective extinction depends slightly on where within that band a given star radiates — that is, on its colour. Precise reductions add a second-order term proportional to a colour index.

m_\lambda(X)=m_{\lambda,0}+k'_\lambda\,X+k''_\lambda\,C\,X \qquad (8)

where: $k'_\lambda$ — first-order (principal) extinction coefficient; $k''_\lambda$ — second-order coefficient, a few hundredths of a magnitude; $C$ — a colour index such as $B-V$ . Reading it: the second-order term corrects the systematic reddening of low stars when transforming to a standard system.

Derivation AFrom Beer's law to the linear extinction law

The goal is to show, rigorously, why an exponential attenuation of flux becomes a perfectly straight line in magnitude versus airmass — the fact that makes atmospheric correction a matter of drawing one line and reading its slope and intercept. We start from the differential Beer's law, integrate to the exponential, insert the airmass as the geometric multiplier of optical depth, then push the result through the logarithmic Pogson magnitude scale, carrying the conversion factor exactly so the slope's units are fixed.

Step 1. Along the slant path, each infinitesimal layer of optical depth $\mathrm{d}\tau_s$ removes a fraction of the current flux: $\mathrm{d}F=-F\,\mathrm{d}\tau_s$ . Separating variables and integrating from the top of the atmosphere ( $F=F_0$ ) to the ground gives $F=F_0\,e^{-\tau_s}$ . Why: a fixed fractional loss per layer is a first-order linear ODE, whose only solution is an exponential — the mathematical signature of multiplicative compounding.

Step 2. The slant optical depth is the vertical optical depth scaled by how much longer the slant path is: $\tau_s=\tau X$ . Substituting gives the observing form $F(X)=F_0\,e^{-\tau X}$ . Why: separating the intrinsic material property $\tau$ from the pointing-dependent geometry $X$ is exactly what lets one measured slope characterize the whole sky at once.

Step 3. Apply the Pogson definition $m=-2.5\log_{10}F+\text{const}$ to both the observed and the space flux and subtract, so the additive constant cancels: $m(X)-m_0=-2.5\log_{10}\!\big(F(X)/F_0\big)=-2.5\log_{10}\!\big(e^{-\tau X}\big)$ . Why: taking the difference of two magnitudes removes the arbitrary zero point, leaving a pure ratio of fluxes that Beer's law has already handed us.

Step 4. Convert the natural exponential to base ten using $\log_{10}e=1/\ln 10$ : $-2.5\log_{10}(e^{-\tau X})=\dfrac{2.5}{\ln 10}\,\tau X$ . Identifying the coefficient of $X$ gives $k\equiv 2.5\tau/\ln 10\approx 1.0857\,\tau$ and hence $m(X)=m_0+kX$ . Why: the factor $2.5/\ln 10=1.08573620\ldots$ is what marries the natural-log physics of absorption to the base-ten magnitude convention, and it fixes $k$ in magnitudes per airmass. *As a check: $2.5/\ln 10=1.08573620\ldots$ , and a vertical optical depth $\tau=0.28$ gives $k=1.0857\times0.28=0.304$ mag per airmass — a realistic V-band value at a mediocre site. The relation is exactly linear; no approximation was made after Step 1.*

Go deeper · Derivation B · 1 minAirmass geometry and the breakdown of the secant law

Derivation BAirmass geometry and the breakdown of the secant law

The goal is to derive $X=\sec z$ from nothing but path-length geometry, then expand it as a power series in the zenith angle to expose both why it is superb near the zenith and why it must fail near the horizon. The small-angle expansion pins down the accuracy quantitatively, and evaluating the secant at benchmark angles shows where curvature corrections take over — the boundary between the trigonometric toy model and the empirical horizon formulae.

Step 1. Model the atmosphere as a plane-parallel slab of vertical thickness $h$ . A ray arriving at zenith angle $z$ travels a slant distance $\ell(z)=h/\cos z$ to cross it, while a vertical ray travels $\ell(0)=h$ . Why: in a flat slab every horizontal layer has the same thickness, so the only thing that changes with pointing is the obliquity factor $1/\cos z$ .

Step 2. The airmass is the dimensionless ratio of these paths: $X=\ell(z)/\ell(0)=\sec z$ . This is the boxed result — the airmass is simply the secant of the zenith angle. Why: normalizing to the vertical column makes $X$ a pure geometric multiplier, independent of the actual physical thickness $h$ of the atmosphere.

Step 3. Expand for small $z$ (in radians): $\sec z=1+\tfrac12 z^2+\tfrac{5}{24}z^4+\cdots$ . The leading correction is quadratic, so the airmass departs from unity only slowly. Why: a purely quadratic onset means the error of treating $X\approx 1$ is second order — the reason zenith photometry is so forgiving, and why $\sec z$ tracks the truth to a fraction of a percent out to $z\approx 60°$ .

Step 4. Evaluate at benchmarks: $\sec 45°=\sqrt 2=1.414$ , $\sec 60°=2.000$ exactly, but $\sec 80°=5.759$ against a true airmass of $5.586$ — already a 3 percent overestimate — and $\sec 90°=\infty$ against a real value near 38. Why: the divergence is unphysical because a curved shell has finite thickness along every sightline; past $z\approx 70°$ one must switch to Hardie's polynomial or the Kasten–Young formula. *As a check: $\sec 60°=2$ and $\sec 45°=\sqrt 2$ exactly; at $z=80°$ the Kasten–Young airmass is $5.586$ versus $\sec 80°=5.759$ ; and at $z=90°$ Kasten–Young gives $X=37.9$ — the finite horizon airmass that the secant law cannot reproduce.*

Go deeper · Derivation C · 2 minThe Bouguer line and recovering the space magnitude

Derivation CThe Bouguer line and recovering the space magnitude

The goal is to turn the linear law into a working measurement recipe: given one star observed at several airmasses across a night, extract both the extinction coefficient and the star's above-atmosphere magnitude with nothing more than a straight-line fit. This is the operation that removes the sky from ground-based data, and its power comes from the exact linearity established above — a line is fixed by two numbers, slope and intercept, and here each carries a distinct piece of physics.

Step 1. Track the same star as it rises or sets, recording pairs $(X_i,\,m_i)$ at several zenith angles, converting each epoch's hour angle to an airmass via $\cos z=\sin\phi\sin\delta+\cos\phi\cos\delta\cos H_i$ and then $X_i=\sec z_i$ . Why: sampling a range of $X$ is what gives the line leverage — a single point could never separate the intrinsic brightness from the atmospheric loss.

Step 2. Fit the model $m_i=m_0+kX_i$ by least squares. The best-fit slope is $k=\big[\sum (X_i-\bar X)(m_i-\bar m)\big]/\big[\sum (X_i-\bar X)^2\big]$ and the intercept is $m_0=\bar m-k\bar X$ . Why: least squares is the maximum-likelihood line under Gaussian photometric errors, and it uses every point to beat down the noise on both parameters.

Step 3. Read the physics off the two fitted numbers: the slope $k$ is the extinction coefficient for that filter and night, and the intercept — the extrapolation to $X=0$ , i.e. no air — is the above-atmosphere magnitude $m_0=m(X)-kX$ . Why: $X=0$ is the mathematical stand-in for a detector in space; the line lets ground data reach a point no ground telescope can occupy.

Step 4. Sanity-check with a known standard: since $m_0$ is catalogued for standard stars, comparing the fitted intercept against the catalogue value validates the night's photometric quality, and the residual scatter about the line flags thin cloud or variable transparency. Why: a standard star turns the fit into a self-diagnostic — a good straight line certifies a photometric night; a curved or noisy one warns you off. *As a check: for the points $(1.0,12.30)$ , $(1.5,12.45)$ , $(2.0,12.60)$ the slope is $(12.60-12.30)/(2.0-1.0)=0.30$ and the intercept is $12.30-0.30(1.0)=12.00$ ; at $X=1$ the fit reproduces $m_0+k=12.30$ , as it must.*

Limiting cases

Airmass and extinction are easiest to grasp at the extremes of zenith angle and wavelength:

Case	Result	Meaning
Zenith	$z=0\;\Rightarrow\;X=1,\ m=m_0+k$	Minimum path; the star still suffers one coefficient's worth of extinction — even overhead you are not above the air.
Small zenith angle	$X=\sec z\approx 1+\tfrac12 z^2$	Plane-parallel law; accurate to better than 1% for $z\lesssim 60°$ , i.e. $X\lesssim 2$ .
Near the horizon	$z\to 90°:\ \sec z\to\infty,\ X_{\text{true}}\to\approx 38$	Curvature and refraction cap the real airmass; the secant law overestimates badly and must be replaced.
Red / infrared limit	$k(\lambda)\to k_{\text{aero}}+k_{\text{abs}}$	The Rayleigh $\lambda^{-4}$ term becomes negligible; residual extinction is small, dominated by aerosols and molecular bands, and nearly grey.

Common trapThe secant law works everywhere, so airmass just keeps growing toward the horizon

It does not. $X=\sec z$ is a plane-parallel approximation that assumes a flat, thin slab of air, and it diverges to infinity as $z\to 90°$ — which the real atmosphere never does. Earth's curvature gives every horizontal sightline a finite length, and differential refraction bends low rays, so the true airmass saturates near 38 at the horizon rather than blowing up. The secant already overshoots by a few percent by $z\approx 80°$ (5.76 versus 5.59) and by tens of percent lower down. Never apply $\sec z$ blindly below about 15–20° altitude; switch to Hardie's polynomial or the Kasten–Young formula, both of which stay finite. A subtler trap: even at the zenith the extinction is not zero — $X=1$ , not $0$ — so the overhead star is already dimmed by one full coefficient $k$ , and only the extrapolated intercept $m_0$ is truly atmosphere-free.

Worked examples

1 · Reducing a Bouguer line to a space magnitude (fully worked)

A standard star is measured three times through the night as it descends from the meridian, giving apparent magnitudes 12.30, 12.45, and 12.60 at airmasses 1.0, 1.5, and 2.0. Find the extinction coefficient $k$ and the above-atmosphere magnitude $m_0$ , then predict the magnitude you would measure at $z=70°$ .

Step 1. Confirm linearity and take the slope. The magnitude rises by 0.15 for each 0.5 step in airmass, a constant increment, so the points lie on a line. The slope is $k=\dfrac{12.60-12.30}{2.0-1.0}=\dfrac{0.30}{1.0}=\mathbf{0.30}$ mag per airmass — a typical V-band value at a good site.

Step 2. Read the intercept. Using the first point, $m_0=m-kX=12.30-0.30\times1.0=\mathbf{12.00}$ . Cross-check with the third point: $12.60-0.30\times2.0=12.00$ . The above-atmosphere magnitude is 12.00; the atmosphere has stolen 0.30 mag even at the zenith and 0.60 mag at $X=2$ .

Step 3. Predict at $z=70°$ . The airmass is $\sec 70°=1/0.342=2.924$ , and since this is beyond the safe secant regime the Kasten–Young value is a slightly smaller $2.90$ . Then $m=m_0+kX=12.00+0.30\times2.90=\mathbf{12.87}$ . The same star that reads 12.30 near the zenith fades by more than half a magnitude at $70°$ — a factor $10^{0.6/2.5}\approx1.7$ in flux lost to the extra air.

2 · Differential extinction reddens a low star (fully worked)

An observer measures a star simultaneously in $B$ and $V$ , with extinction coefficients $k_B=0.25$ and $k_V=0.15$ mag per airmass. The star's true colour above the atmosphere is $(B-V)_0=0.60$ . What colour index is measured at the zenith ( $X=1$ ) and at $z=60°$ ( $X=2$ ), and how much has the atmosphere reddened it?

Step 1. Write each band's extinction. In $B$ the star dims by $k_B X=0.25X$ and in $V$ by $k_V X=0.15X$ . The measured colour is $(B-V)_{\text{obs}}=(B-V)_0+(k_B-k_V)X$ , because the colour index is the difference $B-V$ and the space colour $(B-V)_0$ carries over the intercepts.

Step 2. Evaluate the differential coefficient: $k_B-k_V=0.25-0.15=0.10$ mag per airmass. At the zenith, $(B-V)_{\text{obs}}=0.60+0.10\times1=\mathbf{0.70}$ ; at $z=60°$ , $(B-V)_{\text{obs}}=0.60+0.10\times2=\mathbf{0.80}$ .

Step 3. Interpret. The atmosphere has reddened the star by 0.10 mag at the zenith and 0.20 mag at $X=2$ — blue light is preferentially scattered out, and the effect grows linearly with airmass. If left uncorrected this bias would masquerade as a cooler, redder star; the fix is to fit and remove $k_B$ and $k_V$ separately (and, at high precision, the second-order term $k''C X$ ) before quoting any colour.

3 · A Rayleigh sky, in numbers (estimate)

Estimate the V-band ( $\lambda\approx 550$ nm) Rayleigh extinction at sea level ( $\tau_{\text{Ray}}\approx 0.10$ ), and at a high site at 600 mbar.

Step 1. Convert to a coefficient with equation (2): $k_{\text{Ray}}=1.0857\times0.10\approx\mathbf{0.11}$ mag per airmass at sea level — the floor set by molecules alone.

Step 2. Rayleigh optical depth tracks surface pressure, so at $600$ mbar it falls by $600/1013\approx0.59$ to $k_{\text{Ray}}\approx\mathbf{0.064}$ — roughly 40% of the molecular scattering gone once you climb above it, the case for high summits.

Physics of the portrait

Atmospheric extinction is where photometry was born: the problem of how much light a star loses on its way down through the air forced the first quantitative laws of absorption — and the man who framed it lent his name to the line observers still fit every night.

Origins · 1 minPierre Bouguer and how this was discovered

Pierre Bouguer

1698–1758 · the founder of photometry and the extinction line

Pierre Bouguer was a French polymath — hydrographer, mathematician, and one of the great geodesists of the Enlightenment — who essentially invented the science of measuring light. Appointed professor of hydrography at fifteen, he turned to optics and in his 1729 Essai d'optique sur la gradation de la lumière laid down the principles of photometry: that brightness can be quantified by comparison, and that light is progressively weakened as it passes through an absorbing medium. Studying precisely how starlight and moonlight dim on their way through the atmosphere, he formulated what is now called Bouguer's law — the exponential loss of intensity with the amount of matter traversed — the direct ancestor of the Beer–Lambert–Bouguer law and of the linear magnitude-versus-airmass "Bouguer line" that bears his name. He was no armchair theorist: on the French Geodesic Mission to Peru (1735–1744) he measured an arc of the meridian to test the shape of the Earth and took some of the earliest careful measurements of atmospheric transparency at altitude. Lambert's 1760 Photometria systematized and extended his exponential law, and Beer in 1852 tied absorption to the concentration of the absorbing species — but the founding insight, that light fades exponentially with the air it crosses, is Bouguer's. Every time an observer plots magnitude against airmass and extrapolates to zero, they are drawing Bouguer's line.

History

1729

— Bouguer's Essai d'optique founds photometry and quantifies atmospheric dimming

1760

— Lambert's Photometria casts absorption as an exponential in path length

1856

— Pogson fixes the magnitude scale at a factor of 100 per five magnitudes

1871

— Rayleigh derives the λ⁻⁴ molecular-scattering law behind the blue sky and reddened Sun

1962

— Hardie publishes the standard extinction-reduction and refined airmass formulae

1989

— Kasten and Young give an airmass formula accurate all the way to the horizon

Going further

Texts. Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., §3.6 (photometry and atmospheric extinction) · Ryden, Introduction to Cosmology, 2nd ed. (magnitude and flux conventions) · Binney & Tremaine, Galactic Dynamics, 2nd ed. (photometric preliminaries) · Sterken & Manfroid, Astronomical Photometry: A Guide (extinction reduction in practice) · Bessell (2005), ARA&A 43, 293 (standard photometric systems). Key numerical anchors: $2.5/\ln 10=1.08573620\ldots$ ; $\sec 60°=2$ and $\sec 45°=\sqrt 2$ exactly, with $\sec 80°=5.759$ against the Kasten–Young value $5.586$ ; the Beer's-law-to-magnitude derivation gives $k=1.0857\,\tau$ ; the Kasten–Young and Hardie airmass formulae yield a horizon airmass $\approx 38$ ; and the Rayleigh $\lambda^{-4}$ scaling gives a sea-level V-band $\tau_{\text{Ray}}\approx 0.10$ .

Prerequisites. Horizontal and Ecliptic Coordinate Systems · Luminosity and Radiant Flux · The Magnitude System. Leads to. Photometric Systems and Filters · Color Indices and Bolometric Corrections · Standard Candles · The Cosmic Distance Ladder.

Ask the tutor: Ask me to walk you through fitting a Bouguer line from your own multi-airmass measurements, or to show how a mis-estimated extinction slope propagates into a Type Ia supernova distance and the Hubble constant.

What exactly is the airmass, and why is it 1 and not 0 at the zenith?

The airmass $X$ is the length of the slant air column along your sightline divided by the vertical column straight overhead. By that definition the zenith is the reference: looking straight up you cross exactly one vertical column, so $X=1$ , not $0$ . It is a common error to think the zenith is "above the atmosphere" — it is not; even overhead your star suffers one full coefficient $k$ . Zero airmass is a mathematical limit, the case of no air at all, which is why the space magnitude is obtained by extrapolating the Bouguer line to $X=0$ rather than by any real observation. The airmass grows as you tilt toward the horizon, roughly as $\sec z$ , reaching about 2 at $60°$ and saturating near 38 at the horizon.

Why does the extinction law come out linear in airmass instead of exponential?

The flux really does fall exponentially — $F=F_0 e^{-\tau X}$ — but astronomers report magnitudes, defined as $-2.5\log_{10}$ of the flux. A logarithm turns an exponential into a straight line: taking $-2.5\log_{10}$ of $e^{-\tau X}$ gives $+1.0857\,\tau X$ , linear in $X$ . So the linearity is not a property of the scattering physics; it is a property of the magnitude scale we measure in. This is a genuine convenience: two numbers, slope and intercept, capture the entire atmospheric correction for a filter on a given night.

What sets the extinction coefficient k, and what is a typical value?

The coefficient $k$ is $1.0857$ times the vertical optical depth $\tau$ in that filter, and $\tau$ sums three contributions: Rayleigh scattering by air molecules (dominant in the blue, scaling as $\lambda^{-4}$ ), aerosol scattering by dust and haze, and molecular absorption by ozone, water vapour, and $\mathrm{O_2}$ . At a good high-altitude site a V-band coefficient is around 0.10–0.15 mag per airmass; in $B$ it is larger, 0.20–0.30, because of the steep Rayleigh term; in the near-infrared it is smaller still. The value is site- and time-specific — aerosol and water content drift through a night — which is why $k$ is remeasured nightly rather than looked up.

Why do observatories go to such lengths to sit on high, dry mountains?

Because both the Rayleigh and the absorptive parts of the extinction scale with the air and water above you, and altitude removes both. Rayleigh optical depth is proportional to surface pressure, so a site at 600 mbar (about 4 km) has roughly 40% less molecular scattering than sea level; going higher also drops you below most of the aerosol boundary layer and the precipitable water vapour that blocks the infrared. Dryness matters as much as height: water vapour carves absorption bands through the near- and mid-infrared, so the driest high sites — Mauna Kea, the Atacama, the Antarctic plateau — are prized because their low $\tau$ opens windows that are closed at wetter locations. Lower $\tau$ means a smaller, more stable correction and more photons on the detector.

Is atmospheric extinction the same thing as interstellar extinction?

No, though the mathematics rhymes. Both dim and redden starlight by scattering and absorption, and both follow an exponential-in-column, wavelength-dependent law — but the medium and geometry differ entirely. Atmospheric extinction happens in the last few kilometres of air, depends on your airmass, can be removed nightly by a Bouguer fit, and vanishes for a telescope in space. Interstellar extinction happens over parsecs of dust between the star and us, does not depend on how you point, cannot be extrapolated away, and follows a characteristic reddening law (parameterized by $R_V\approx 3.1$ ) with features like the 2175 Å bump that have no atmospheric analogue. This article is strictly the atmospheric term; interstellar reddening is corrected separately, after the atmosphere has been removed.

Do space telescopes escape this correction entirely?

Essentially yes, and that is one of their quiet advantages. Above the atmosphere there is no air column, so $X$ is meaningless and $F=F_0$ — Hubble, Gaia, and JWST measure the above-atmosphere flux directly, with no Bouguer line to fit and no nightly coefficient to chase. This is a major reason space photometry holds millimagnitude stability and why Gaia's all-sky magnitudes are such clean anchors for calibrating ground-based systems. What space telescopes do not escape is interstellar extinction, which lies beyond the atmosphere along the line of sight and must still be modelled. So going to space deletes the atmospheric term while leaving the interstellar term untouched.