Do Clouds Make Better Sunsets? The Science
Some clouds make a sunset and others kill it. The difference is which layer, and it is in every hourly forecast you can already read.
A clear sky gives you a clean sunset and not much else. The colour that makes people stop and take a photograph is light hitting cloud from underneath, and whether that happens depends on which clouds are where.
The physics is covered well in a dozen places. What is not published anywhere is a model you can actually apply to a forecast, so here is ours, thresholds and all.
Why clouds matter
At sunset the light reaching you has travelled a long, shallow path through the atmosphere. Rayleigh scattering has removed most of the short wavelengths, so what emerges is red and orange.
A clear sky has nothing for that light to land on. You see a warm gradient near the horizon and it fades.
High cloud — cirrus, cirrostratus, altocumulus, anything from about 5 km up — is still in direct sunlight after the sun has set at ground level. It is lit from below by light that has already been reddened, which is why it goes pink and orange while the ground is in shadow. This is the mechanism behind every sunset worth photographing.
Low cloud is the enemy, and not because it is ugly. It sits between the setting sun and everything above it. If there is a solid deck on the western horizon, the reddened light never gets out to illuminate the high cloud, and the sky above stays grey. A beautiful cirrus display can be completely wasted by a bank of stratus fifty kilometres to the west that you cannot even see.
So the ideal is specific: cloud high up, clear air low down, especially towards the west.
The model
This is what runs behind the conditions widget on every city page. It reads the hourly forecast for the hour of sunset and scores it.
| Condition at the hour of sunset | Score | Why |
|---|---|---|
| High cloud 20–70% with low cloud under 15% | +20 | Cloud high enough to be lit from beneath, with a clear path to it |
| Mid cloud 20–60% with low cloud under 25% | +10 | A second deck to catch colour |
| Visibility above 15 km | +5 | Clean air holds saturation |
| Low cloud above 25% | −15 | Starts blocking the western horizon |
| Low cloud above 50% | −40 | The light cannot reach anything above it |
| Any layer above 90% | −50 | Overcast; no direct light gets through |
| Visibility below 5 km | −25 | Haze mutes the colour |
| Rain probability above 50% | −20 | Usually arrives with thick low cloud |
| Humidity above 85% | −5 | Softens contrast |
| Total cloud under 5% | −10 | A clean sky, but nothing to catch the colour |
The score starts at 55 and is clamped to 0–100. Above 75 is excellent, 55–75 good, 35–55 fair, below that poor.
None of these numbers are laws of physics. They are thresholds chosen to match what the cloud layers actually do, and they are stated openly here so you can disagree with them or apply them yourself to any forecast that reports cloud by level.
The key term is the first: high cloud between 20 and 70 per cent with low cloud under 15 per cent. Both halves matter. High cloud alone is not enough if something is blocking the light path, and a clear horizon alone gives you nothing to light.
Reading a forecast yourself
Most weather services publish total cloud cover as a single number, which is nearly useless for this. What you want is cloud broken out by level — low, mid and high — which is available from Open-Meteo, from many aviation forecasts, and from the model output most national services publish.
Then:
- Look at the hour of sunset, not the daily summary. Conditions at 15:00 tell you very little about 20:00.
- Check low cloud first. Above 50 per cent, stop; it is unlikely to be worth the trip.
- Then check high cloud. Somewhere between a quarter and two-thirds is the sweet spot. Fully overcast at high level is nearly as bad as low cloud, because it is a lid rather than a set of surfaces.
- Check visibility. Below about 5 km, haze will mute whatever happens.
- Look west, physically. The forecast is for your location, but the light comes from a point on the horizon that may be 50 to 100 km away. A clear sky overhead and a wall of cloud on the western skyline is a common and disappointing combination.
What the model cannot tell you
Aerosols. Fine particles high in the atmosphere — from volcanic eruptions, large wildfires, or dust transported thousands of kilometres — produce the most intense sunsets there are, and no ordinary forecast reports them. The Krakatoa eruption in 1883 produced sunsets vivid enough that fire brigades were called out. Wildfire smoke aloft does a smaller version of the same thing several times a year in many places.
Timing within the hour. Cloud moves. A forecast for the sunset hour is an average over sixty minutes, and the arrangement at the moment the light peaks — usually a few minutes after the sun has gone — can be quite different.
Local geography. Coastal fog, valley inversions and lake effects all operate at a scale finer than any global model resolves.
The honest summary
Clouds do make better sunsets, but only some clouds, and the ones that ruin it are often the ones you cannot see from where you are standing.
The single most useful habit is to stop looking at "cloudy or clear" and start looking at cloud by level. A day forecast as 80 per cent cloudy can produce the best sunset of the month if that cloud is all at 8 km, and a day forecast as 30 per cent cloudy can produce nothing at all if the 30 per cent is a stratus deck sitting exactly where the sun is going.
Every city page here runs this model against the live forecast and gives you a one-line reason for the verdict. It is a forecast, and it is wrong sometimes. The timing on the same page is not — that part is arithmetic.
Which clouds, by name
If you want to recognise the setup on sight rather than in a forecast, a few types do most of the work.
Cirrus and cirrostratus are the best of them. Ice crystals at 6 to 12 km, thin enough to transmit light rather than block it, and high enough to stay in sunlight long after the ground is dark. A sky with scattered cirrus and a clear western horizon is the classic arrangement.
Altocumulus at 2 to 6 km gives the mackerel sky that breaks the light into hundreds of separately lit elements. When it works it is the most dramatic of all, because each cloudlet is individually rimmed.
Cumulus is a mixed blessing. Fair-weather cumulus often dissipates towards evening as surface heating stops, which can clear a sky that looked hopeless at four o'clock. Larger cumulus stays and blocks.
Stratus and stratocumulus are the ones that ruin an evening. Low, flat, and usually extensive enough that there is no gap at the horizon for light to come through.
Contrails count as high cloud and behave like cirrus, which is why sunsets near busy air corridors sometimes light up in straight lines.
Why the best sunsets follow weather fronts
A useful pattern: the evening after a front passes is often excellent.
A departing front leaves high cloud behind while the low cloud clears from the west first, which is precisely the geometry the model rewards — a lit ceiling and an open horizon. The air behind a front is also usually cleaner, which helps saturation.
The corresponding rule is that the evening before a front arrives is usually poor. The cloud thickens and lowers from the west, closing the horizon exactly where the light needs to come from.
Neither is reliable enough to plan a trip around, but if you are deciding whether to go out tonight, "did a front just clear" is a better question than "is it cloudy".