How dark is dark?

It is a description, not a measurement

John Bortle published the scale in Sky & Telescope in 2001. He was not trying to build an instrument. He was trying to give observers a shared vocabulary, because "pretty dark out here" means nothing to somebody who has never left a city and everything to somebody who has driven eight hours to stand in the desert.

So every class is written as a list of things you can or cannot see. Can you find the Andromeda Galaxy's smaller neighbor without a telescope? Does the Milky Way have structure in it or is it a smudge? Is your own shadow on the ground being cast by starlight? Those questions do not need equipment. They need dark-adapted eyes and about half an hour of patience.

That is also the weakness. Two people can stand in the same field on the same night and disagree by a full class, because one of them has better eyes, or more experience knowing what to look for, or simply arrived twenty minutes earlier and adapted properly. Add humidity, altitude, snow on the ground, and where the Milky Way happens to be sitting, and the same site can honestly read differently from one visit to the next.

Which is why I carry a meter. Not to replace the judgment, but to have something that does not care how tired I am.

What to look for

One entry per class, brightest first, the same direction as the scale above. These are the checks I actually run when I get out of the car and want to know where I am standing: no equipment, no counting stars, just the one or two things that give a class away. Classes 8 and 9 share a line because in the field they behave the same way. The 4.5 and the 10 are mine, as above.

• Hard shadows at midnight, thrown by the ground No moon, and you still have a sharp-edged shadow. The brightest thing anywhere in view is below the skyline — signs, facades, a beam. This is not a compromised night sky, it is the absence of one.
• You can read newspaper headlines by the sky At 8 the glow is whitish or orange and familiar constellations are missing stars. At 9 it is lit all the way to the zenith, the fainter constellations are gone outright, and the Little Dipper is down to three stars, four if you are lucky. The Pleiades are the last deep-sky object standing.
• The whole sky reads light grey, right up to the zenith The Milky Way is gone, or so nearly gone it does not matter. Andromeda and the Beehive can both be glimpsed if you already know where they sit, and neither shows any shape at all. Clouds are plainly brighter than the sky behind them.
• The Milky Way survives only straight overhead Directly up there is a faint band most people would walk under without noticing. Everything within about 35° of the horizon glows greyish white, and there is no zodiacal light at all, ever. M33 has dropped out of naked-eye reach.
• The Milky Way is overhead but flat, and gone near the horizon No dust lanes, no rift, no core — a pale smear rather than a band you can read. Hints of zodiacal light on the best spring and autumn nights and not otherwise. Clouds are noticeably brighter than the sky.
• The Milky Way lifts 10 to 15° and no further You get the overhead band and nothing under it, because the glow eats it before it reaches the ground. Clouds have started going grey at the zenith instead of staying black.
• Light domes with height to them, and a Milky Way without detail The domes stand up off the horizon in several directions rather than hugging it. The band overhead is worth looking at but has lost its texture. Zodiacal light is clearly there and never reaches halfway to the zenith. M33 is a hard averted-vision object and only turns up at all above about 50° altitude.
• M33 comes with averted vision, and the globulars go naked-eye Light pollution is back in domes near the horizon and the Milky Way is complex again. M4, M5, M15 and M22 turn into distinct little smudges without optics. M33 is what pins the class though: easy with averted vision here, a fight at 4, and a straight-on object at 2.
• M33 held straight on, and clouds that read as black You look directly at the Triangulum Galaxy and it stays there, with no sliding your eye off to one side. Check the clouds while you are at it: under a class 2 sky they are black holes punched in the stars, not grey shapes. The summer Milky Way has real structure, and the zodiacal light is strong enough to throw a weak shadow of its own after dusk and before dawn.
• Your own shadow, cast by the Milky Way Not the zodiacal light’s weak one — the Milky Way’s own. On pale ground, with no moon, the Sagittarius and Scorpius stretch throws a diffuse shadow you can actually see. M33 has stopped being a test and become obvious. The gegenschein is there if you know where to look, and Jupiter or Venus is bright enough to wreck your dark adaptation.

Two more that Bortle built in, and almost nobody uses

Both are in his original article and both need nothing but your eyes. The first is the clouds, and it is the most reliable check on this page: cloud is a screen, and what it shows you is whatever the ground is throwing upward. The second is to stop looking at the sky at all and look at what is around you — your own hands, the ground, your gear on the table.

Where a cell is blank, Bortle left it blank. He pinned some classes on both counts and not others, and I would rather show the gap than invent a line to fill it.

Why I publish a number too

Every dispatch I write carries a reading in mag/arcsec², taken with a sky quality meter pointed straight up. It is a measure of surface brightness, and it runs backwards from how you would expect: higher is darker. Around 16 is a bright city. Around 22 is about as dark as the sky gets anywhere on Earth, because at that point what is left is airglow and starlight rather than anything humans did.

The two scales line up roughly, and only roughly. Published conversion tables disagree with each other by a tenth of a magnitude in places, which sounds trivial until you notice that a tenth is enough to move a site across a class boundary.

These are the bands I use. They are not identical to every table you will find, and the difference that matters most is the half step at 4.5. Bortle's 4 and 5 are a long way apart in practice, and an enormous number of the places people actually drive to fall in the gap between them. Calling all of that "a 4" flatters it. Calling it "a 5" is unfair to it. So I split it.

The reading is not the whole story

Here is a real example. At Manzanita Lake in Lassen Volcanic I measured 21.45. Read straight off a conversion table that is a Bortle 3, and I called it a 2.

The reason is that it was August, with the galactic core out at full strength and sitting in exactly the part of sky I was metering. The Milky Way itself brightens the reading. My estimate is that it cost me 0.3 mag/arcsec² or more, and correcting for that puts the true baseline darker than the number says. That site is a solid class 2, and on the right night I think it touches class 1.

A meter cannot make that call. It does not know it is looking at the core. That is the part where ten years of standing in fields still matters, and it is why I report the raw number, the class I am giving it, and my reasoning when the two disagree.

Anyone who gives you a class without a number is guessing. Anyone who gives you a number without context is not thinking.