Two questions get asked about the moon’s color, and almost nobody notices they have the same answer. The first is what color the moon actually is. The second is why it sometimes comes up orange. The answer to the second one is that it never does, and the answer to the first one is a grey you would walk past in a car park.
I wanted numbers rather than adjectives, so I built it. Nineteen Apollo soil samples with laboratory reflectance spectra, the current best absolute measurement of the sun’s spectrum, the CIE observer, and a spherical atmosphere with Rayleigh scattering, aerosol and ozone in it. Everything below is measured out of that stack, and the sanity checks at the end land on published values.
The moon is #6A6057
Start with the rock. The Lunar Soil Characterization Consortium published bidirectional reflectance spectra for a suite of returned Apollo soils, nine from the maria and ten from the highlands, measured from 300 to 2600 nm. Multiply the bulk soil reflectance by the TSIS-1 solar spectrum, integrate against the CIE 1931 observer, and compare it to a perfect white diffuser lit by the same sun.
Weighting the maria at 31 percent, which is roughly what they cover on the near side, and anchoring the overall level to the moon’s accepted geometric albedo of 0.12, the disc comes out here.
#6A6057 · L* 41.22 · hue 80.7° · chroma 5.13 · 12% reflectance
That is darker than the middle grey on a photographer’s card. At 12 percent reflectance the moon sits between fresh asphalt and worn asphalt, well below dry concrete, and about a seventh as bright as office paper. The thing that lights up your garden is roughly the color of a road surface.
The lean is real but small. Hue angle 80.7 degrees is in the yellow, and chroma 5.13 puts it 4.67 CIEDE2000 units away from a perfectly neutral grey of the same lightness. Set against the roughly 2.3 unit threshold I measured in the piece on just noticeable difference, that is about two JNDs. Enough to see if you had a neutral grey card next to it. Not enough to call it brown, which is the word that usually turns up in articles on this.
The yellow is not an artefact of one rock
The obvious objection is sample choice, so I ran all nineteen soils separately with their brightness normalised away, leaving only the shape of the spectrum.
- Hue angle range across all nineteen: 72.5 to 83.8 degrees.
- Mean 79.4 degrees, standard deviation 2.5 degrees.
- Samples landing outside the yellow quadrant: none.
Nine mare soils and ten highland soils, collected by four different missions from sites thousands of kilometres apart, agree on the hue inside a twelve degree window. The moon is uniform in color to a degree that almost nothing on Earth is.
The cause is space weathering. Micrometeorite impacts and solar wind deposit nanophase metallic iron on the surface of every grain, which darkens the soil and tilts its reflectance upward toward the red. In the disc average the reflectance climbs from 11.75 percent at 400 nm to 18.07 percent at 700 nm, a ratio of 1.373 across the visible. A gently rising line like that reads as yellow rather than red, which is exactly where the measurement lands.
Two moons, one disc
The disc average hides something, though, which is that the moon is visibly two different greys and you can see the boundary with your eyes.
#564D45 · 7.59% at 550 nm · L* 33.28
#83776D · 18.96% at 550 nm · L* 50.78
The maria reflect two and a half times less light than the highlands. That gap of 17.5 points of L* is the man in the moon, the rabbit, and every other figure people have read into the disc for as long as there have been people. It is a difference in iron and titanium content between ancient basalt floods and the older crust around them, and it is wide enough that no amount of atmosphere or eyesight will hide it.
So why does it look white
A 12 percent grey that reads as brilliant white is a genuinely interesting perceptual problem, and the answer is that your visual system has nothing to compare it against.
Lightness is not read off absolute luminance. It is assigned relative to the brightest thing in the field of view, which then gets called white and anchors everything else. This is the anchoring account of lightness perception, and the moon is the cleanest demonstration of it anyone could ask for. A full moon in a dark sky is, by an enormous margin, the brightest object present. There is no reference white in the frame, so the moon becomes the reference white, and a road surface gets promoted.
The same machinery is what color constancy runs on during the day, and the same machinery is why the dress splits people. Photograph the moon next to something genuinely white and the illusion dies instantly. The Apollo surface photographs are the proof: astronauts in white suits standing on ground that looks like wet ash, because the suit supplied the missing anchor.
The moon is bright, the moonlight is not
There is a second split here worth putting numbers on, because it explains something people notice without being able to name it. You can see the moon’s color. You cannot see the color of anything it lights.
Running the photometry off the same spectra: the sun delivers 134,643 lux above the atmosphere, so a 12 percent Lambertian disc has a luminance of about 5,143 cd/m². That is a daylight number. Meanwhile the disc subtends a very small solid angle, so what reaches the ground is 0.33 lux, and grass under a full moon sits at roughly 0.0105 cd/m².
- Cone vision, which is the only kind that carries color, needs something above roughly 3 cd/m² to work properly.
- The moon’s disc clears that by a factor of about 1,700.
- The moonlit landscape misses it by a factor of about 300, which puts it in rod territory.
The ratio between the two is around 49,000 to 1. You are looking at a photopic object suspended in a scotopic scene, which is why a moonlit field is silver and colorless while the moon itself has an obvious warm tint. Both facts are true at the same time and they are not in conflict.
Every orange moon is made of air
Now the question people actually type at two in the morning. The moon turns orange for precisely the reason the sun does, and I can show that they are the same effect rather than two similar ones.
The model is the one from the sunset piece: Rayleigh optical depth from the Bodhaine formulation, an Angstrom aerosol term at 0.10 optical depth and exponent 1.3 for clear continental air, ozone from the Serdyuchenko cross sections at a 300 Dobson column, and Kasten-Young air mass so the horizon geometry is right. The swatches below are exposure normalised so you can see hue rather than brightness, with the surviving light reported separately.
The striking column is the middle one. A moon on the horizon is delivering one thousandth of the light it delivers overhead. Thirty eight air masses will do that. And the color runs from a barely warm cream to a saturated red orange, 33.44 CIEDE2000 units apart end to end, with excitation purity climbing from 18 percent to nearly 99.
Notice how little the dominant wavelength moves until very late. From the zenith down to 20 degrees it does not shift at all, sitting at 579 nm while purity nearly doubles. The atmosphere does not change the hue of the moon so much as it strips the hue out of everything that is not already yellow. That is the same signature I found for the sky, where the hue pins and the purity varies, and it is a useful thing to be able to recognise.
The altitude where it becomes visible
Because the ladder is continuous, “the moon is orange tonight” needs a threshold to mean anything. Using a zenith moon as the reference and stepping down in quarter degrees:
- The shift passes one JND, 2.3 units, at 38.75 degrees elevation.
- It passes 5 units, which is hard to miss, at 24.50 degrees.
- It passes 10 units, which nobody would argue about, at 13.00 degrees.
Thirteen degrees is about an outstretched hand and a bit above the skyline. So the honest answer to why the moon is orange tonight is that it is low, it is orange every night that it is low, and what changed was not the moon or the air but whether you happened to be outside and facing that way. A rising full moon is the common case because a full moon rises around sunset, when people are out and looking at the horizon.
Haze and smoke move these numbers, but less than the folklore suggests. They add to the aerosol term, which shifts the whole ladder up a rung or two. Elevation is doing the heavy lifting, and I found the same thing when I took the sunset apart: Rayleigh scattering in clean air does most of the reddening, and the dramatic causes people reach for are a small correction on top.
At the horizon, the moon and the sun are the same color
This is my favourite result out of the whole build, and it settles the question cleanly.
Run the sun and the moon through the identical atmosphere at identical elevations. The only difference between them is that the moon’s light has bounced off regolith first, picking up that 1.373 red slope. So the moon should be measurably warmer than the sun at every altitude. It is. The interesting part is what happens to the gap.
The gap collapses from 6.71 units to 0.92 units. At the horizon the moon and the sun are the same color to well inside one JND, which means no observer could tell them apart on hue if you removed the brightness difference and the context.
That is the proof that an orange moon has nothing to do with the moon. Thirty eight air masses of atmosphere overwrite whatever spectrum you send into them. High up, you can still read the regolith in the color. Down at the skyline the air has erased the object and left only itself, and the moon has become a slightly dimmer report on the state of the air between you and it.
A blood moon is every sunrise on Earth at once
Total lunar eclipses are the one case where the moon genuinely changes color rather than having a color imposed on it in the last few hundred kilometres, and it is worth doing properly because the usual explanation stops halfway.
Inside totality the moon is out of direct sunlight. The only light reaching it has passed through the ring of atmosphere around the Earth and been bent inward by refraction. So the relevant calculation is not a vertical path through the air, it is a tangent path that grazes the limb at some height and comes out the other side.
I integrated those tangent paths numerically through a spherical atmosphere: air on an 8 km scale height, aerosol on 1.5 km, ozone as a layer centred at 22 km, each ray sampled along its full chord. A ray grazing at 5 km traverses 37.9 times the vertical air column. The result is not one color, it is a ladder.
Refraction is what sorts these onto the moon. Bending falls off with height, so the low rays, which are also the reddest, are the only ones turned sharply enough to reach deep inside the shadow. The high rays are barely deflected and mostly miss. That is why the umbra is a gradient: deep red at the centre where only the 5 to 10 km rays arrive, brightening to orange toward the edge.
The copper color everybody photographs is the 8 to 12 km band, somewhere around #FFAC4B, carrying about 5 percent of the beam that entered. Sunlight that has travelled the long way through twenty air masses of Earth’s lower atmosphere, which is the same thing as saying it is sunset light. Every sunrise and sunset happening on Earth at that moment, collected into a ring and projected onto a rock.
The turquoise rim is ozone, and only ozone
Look again at the 20 km rung. It is blue. That is not a bug, and it is not an artefact of the exposure normalisation. Observers have long reported a turquoise band around the outer umbra during totality, and it falls straight out of the model.
Here is the same ray with the ozone layer deleted and nothing else changed.
with ozone · 479.5 nm
without · 578.5 nm
That is a swing of 45.68 CIEDE2000 units and 99 nanometres of dominant wavelength, from a warm cream to a proper blue, caused by one trace gas present at a few parts per million. Ozone’s Chappuis band absorbs broadly across the middle and long visible, peaking near 600 nm. Along a tangent path at 20 km the ray passes lengthwise through the heart of the ozone layer, roughly 50 times the vertical column, while it is high enough that Rayleigh scattering has stopped mattering much. Ozone gets to work on light that air has left alone, and it eats the red.
Run the same comparison down the ladder and the effect is strongly height dependent: 3.44 units at 5 km, 6.28 at 10 km, 22.92 at 15 km, 45.68 at 20 km, then falling away again above 25 km as the ozone thins out. Deep in the umbra ozone barely registers, because Rayleigh scattering has already removed everything except the far red and there is nothing left for the Chappuis band to take. The blue rim can only exist in the narrow band where one mechanism has finished and the other has not started.
I got the same shape of answer when I worked out why twilight goes blue after the sun has gone: ozone doing visible work in a window where scattering has run out. It is the most underrated pigment in the sky.
What this model leaves out
The soil spectra are laboratory measurements on sieved samples at fixed illumination and viewing angles. The real lunar surface is porous and strongly backscattering, which is why the full moon is more than twice as bright as two half moons rather than exactly twice. I sidestepped that by taking the spectral shape from the samples, which is robust, and the absolute level from the published geometric albedo, which folds the opposition effect in already.
The atmospheric model is single scattering with no refraction in the elevation ladder, so the very lowest rungs are approximate. Real refraction lifts an apparent horizon moon above the geometric horizon and squashes it into the oval everyone recognises, and it shortens the path slightly compared to what I computed. The direction of that error is toward slightly less extreme reddening at the last degree.
The eclipse calculation assumes a clean, cloud-free limb. It is not. Cloud tops block the lowest tangent rays on most of the ring, and after a major volcanic eruption stratospheric aerosol can darken totality enough that the moon nearly vanishes. That variability is what the Danjon scale was invented to record, and it is why no two blood moons photograph the same.
None of that touches the central results. The moon’s hue is stable across nineteen samples and four landing sites, the elevation ladder is dominated by air mass rather than by anything uncertain, and the ozone result is a difference between two runs of the same model.
The part that matters if you care about color
Three things I keep turning over after building this.
The first is that the moon is the best available demonstration that lightness is relative and not absolute. A surface that would look like dirty pavement in daylight becomes the definition of white when it is the brightest thing you can see. Nothing about the surface changed. The comparison set changed.
The second is that the horizon result generalises. At 38 air masses the atmosphere does not tint the moon, it replaces it, which is why the moon and the sun converge to within 0.92 units. Anything you look at through enough of a medium stops reporting on itself. I found 83 percent of the sea’s light is reflected sky, and a mirror is measurably green because of its glass rather than its coating. Objects own their color far less often than the language suggests.
The third is how small the moon’s own signal is. Chroma 5.13 is two JNDs off neutral. It survives at the zenith, weakens by 10 degrees and is gone at the skyline. Reading it at all means holding a grey in your head accurately enough to notice a difference of a few units, which is exactly the skill the color memory game drills. If you want the version that works on this specifically, the gradient mode puts near neighbours of one hue side by side and asks you to carry the difference, and training your eye for color covers how to practise it deliberately. Next clear night, find the moon when it is high, then again when it is low, and see whether you can put a number on the gap.