Airmass and Altitude: The Secant Rule, Spherical Geometry and Horizon Limits

What is airmass in astronomy?

Airmass is dimensionless. It is not the mass of all the air above a city, a weather-pressure reading or the distance to the star. The same target has different airmass as its altitude changes, even though its astronomical distance is effectively unchanged during the observation.

This guide derives the simple expression, then compares it with two deliberately idealized spherical layers. The comparison explains the geometry of the approximation. It does not turn either illustrative layer into a calibrated model of the real atmosphere.

Define the column before simplifying the atmosphere

A more general expression uses atmospheric density ρ along the ray: X = ∫ray ρ ds / ∫vertical ρ dz, with both columns referred to the same observer. The numerator and denominator have matching units, which cancel. A vertical ray therefore has relative airmass one by the chosen normalization.

Path length alone equals this density-weighted ratio when the compared paths traverse the same uniform-density material. In a stratified atmosphere, the changing density must be included. For a plane-parallel atmosphere with horizontal layers and a straight ray, the same secant factor applies to each layer, so it can still be taken outside the integral.

Different absorbing or scattering constituents need not have identical vertical distributions. Consequently, one geometric number does not describe every wavelength-dependent transmission effect. The notes by Edward F. Brown provide the atmospheric-column definition; the finite-shell comparison below is an original geometric teaching calculation.

Derive the secant rule from a right triangle

Imagine a horizontal layer of vertical thickness t. A straight ray at altitude h crosses a distance L satisfying L sin(h) = t. Dividing L by t gives X = 1/sin(h). Because z = 90° − h, this is also 1/cos(z), or sec(z).

At h = 90°, sin(h) = 1, so the sloping and vertical paths coincide. At 30°, the sine is one half, so the ray crosses twice the vertical thickness. At 60°, X is about 1.1547; lowering a target from the zenith to 60 degrees does not double the column.

Do not interchange h and z when entering a formula. Taking 1/cos(30°) for a target whose altitude is 30° gives 1.1547 instead of 2. The former calculation describes a zenith distance of 30°, which corresponds to altitude 60°. Labelling the angle prevents a very common error.

Compare thirteen altitudes numerically

The second column is the plane model. The next two columns use the spherical construction derived below, with an illustrative radius of 6,371 km and uniform layers of thickness 5 km or 10 km. Neither thickness is being presented as a real atmospheric boundary.

At altitude 30°, all three values are close to two. At 1°, they differ strongly: approximately 57.30, 32.93 and 26.28. Those disagreements demonstrate sensitivity to model assumptions near the horizon. They do not identify which number would match an observation at a particular site.

The table rounds to four decimal places so that comparisons are reproducible. That display precision is not a claim of atmospheric accuracy. Temperature structure, density changes, aerosols and bending of the ray are absent from these two spherical teaching models.

Derive a finite spherical-layer path

Put the observer on a sphere of radius R and the top of a uniform layer at radius R + H. Here H is layer thickness, not hour angle. Let a straight outward ray make altitude h with the local tangent. If L is the distance travelled until it leaves the layer, its endpoint obeys (R + L sin h)² + (L cos h)² = (R + H)².

Expanding gives L² + 2R L sin h = 2RH + H². The positive solution is L = √(R² sin²h + 2RH + H²) − R sin h. Divide by the vertical path H to obtain the uniform-layer relative path X = L/H.

Subtracting two large, similar numbers can lose numerical precision. An equivalent rationalized form is X = (2R + H) / [√(R² sin²h + 2RH + H²) + R sin h]. It follows by multiplying the original numerator and denominator by the conjugate, not by changing the physical model.

Read the original comparison chart

The vertical scales differ between the panels: the upper plot runs from one to six and the lower from zero to 120. Compare values using the labels, rather than comparing the apparent steepness of a line across panels. The common ten-degree endpoint appears in both views.

The sampled curves are joined to make their trends readable. The graph does not resolve every small angular interval near zero, where the plane-model change is particularly rapid. The equations and the numerical table define the model more precisely than the thickness of a plotted line.

Check the zenith, horizon and thin-layer limits

At the zenith, the ray travels exactly H and all three models give one. At the geometric horizon, the spherical result is X = √(2R/H + 1). With the two illustrative thicknesses it is about 50.4916 and 35.7099. The plane layer instead extends horizontally without end, making its horizon secant undefined.

For a fixed positive altitude, taking H/R toward zero makes the spherical expression approach 1/sin(h). This explains why the curves nearly coincide high in the sky. It does not justify extending the plane formula to exactly zero altitude: the horizon limit has different behaviour.

A thicker uniform layer gives a longer absolute horizon path, yet a smaller relative ratio because its vertical normalizing path is also thicker. For H = 5 km the horizon path is approximately 252.46 km; for H = 10 km it is 357.10 km. Always distinguish a distance from that distance divided by H.

Turn a plane-model airmass limit into an altitude

If an illustrative planning condition is X ≤ Xmax, with Xmax at least one, the plane formula gives h ≥ arcsin(1/Xmax). Thus X ≤ 2 corresponds to altitude at least 30 degrees, while X ≤ 1.5 corresponds to about 41.8103 degrees.

These are conversions within a named approximation, not universal observatory requirements. A telescope, instrument or observing programme can use a different airmass algorithm and additional constraints. Use its documented convention when a numerical limit is part of an actual observation request.

For a simple time-window example, place a fixed-declination-zero target at an equatorial observer. Its geometric altitude above the horizon is 90° minus the absolute hour angle while it is above the horizon. Altitude at least 30° therefore means hour angle between −60° and +60°, an eight-sidereal-hour interval centred on transit.

That example is a geometric interval, not eight hours of darkness on a particular night. Twilight, daylight and the local horizon can remove part or all of it. The hour-angle and altitude guide develops the general latitude and declination calculation.

Airmass, refraction and visibility are separate quantities

The refraction guide discusses apparent versus geometric altitude. Every calculation here uses geometric altitude and a straight ray. Feeding a refracted angle into a model defined for the true angle changes its input contract; it is not an automatic improvement.

The ESO HDRL manual distinguishes airmass approximations and their angle conventions. In particular, its description of Young's 1994 method specifies the true zenith distance. Named formulae should be used with their own stated inputs and validity limits, rather than combining whichever terms happen to be available.

Lower airmass also does not guarantee a clear view: a cloud can block a high target. Conversely, the toy finite horizon value does not promise that a low target can be detected. Horizon colour and field rotation involve related observing geometry but answer other questions.

Why twice the airmass does not mean half the brightness

Under a simple model with a constant extinction coefficient k in magnitudes per airmass, a change in airmass produces Δm = k ΔX. The corresponding flux ratio is 10^(−0.4 Δm). This separates geometric column length from the coefficient describing attenuation in a chosen band.

For an invented teaching coefficient k = 0.20 and a move from X = 1 to X = 2, Δm = 0.20 and the flux ratio is about 0.8318. That is approximately 83.18 percent of the previous flux under those assumptions, not one half. The number 0.20 is illustrative, not a measured value for Astrogyana's users or any named site.

Real calibration requires appropriate observations and a model for the conditions and passband. It is not supplied by a geometric chart. A useful worksheet records altitude convention, airmass method, site, time and any separately measured attenuation information, keeping calculated geometry distinct from observed transparency.

Frequently asked questions