Forecasting · 7 min read
How to read a cloud-cover forecast for stargazing
Every weather app shows one number for cloud cover. For stargazing, that number is the wrong one — or at least only a third of the answer. Here's what actually matters and how to read it.
"Cloud cover" is really three forecasts stacked together
The single percentage most apps show is total cloud cover: the fraction of sky covered by cloud at any height. But cloud lives at three broad levels, and they don't affect the stars equally.
- Low cloud (below ~2 km) — stratus, fog, fair-weather cumulus. Thick and opaque. This is the stargazer's enemy: even a thin deck of low cloud blots out the sky completely.
- Mid cloud (~2-7 km) — altostratus and altocumulus. Often grey and fairly opaque; a broken mid deck gives you gaps but unreliable viewing.
- High cloud (above ~7 km) — cirrus, made of ice crystals. Thin and wispy. You can often see bright stars and planets straight through high cloud, though it softens faint objects and the Milky Way.
A night forecast to "40% cloud" tells you nothing about which of these you'll get. Forty percent low stratus can mean a wall of grey; forty percent high cirrus can still be a perfectly usable night for the Moon, planets and brighter targets.
How Stellarhound uses the three layers
This is exactly why Stellarhound pulls the low, mid and high layers separately from Open-Meteo and weights them by how much each actually blocks the sky, rather than trusting a single total. The score you see reflects opaque low cloud far more heavily than thin high cloud — so a night that a generic app writes off as "cloudy" can still score well if that cloud is high and thin, and vice versa. The per-night card shows the weighted cloud figure it used, so you can see the reasoning rather than take it on faith.
Read the hours, not just the night
Cloud moves. A night that's overcast at 9 pm can clear by 1 am as a front passes. What matters is the cloud during your dark window — the hours after twilight ends, and specifically the hours you'll actually be out. Stellarhound scores across the astronomical-dark hours of each night for this reason; if you're targeting a specific object that's only up early or late, check when it's above the horizon and weight the forecast toward those hours.
Trust the forecast less the further out it goes
Cloud-cover forecasts are genuinely skillful out to about three or four days, degrade through the end of the first week, and past roughly seven to ten days are closer to climatology than to a real prediction. That's not a flaw in any one model — it's the nature of the atmosphere. Use the far-out end of a 16-day forecast to spot candidate nights, then re-check as they approach. When a forecast can't resolve a night at all, Stellarhound marks it unavailable instead of inventing a plausible-looking score.
A practical routine
- Days out: use the forecast to shortlist the clearest-looking nights and note the Moon.
- Two or three days out: re-check — this is when cloud detail becomes trustworthy.
- On the day: look at the hourly picture across your dark window, not the daily summary.
- Always: prefer a night where the clear hours line up with when your target is high.
Cloud data: Open-Meteo.com (CC BY 4.0). Updated July 15, 2026.