Which Part of Your Sky Is Clear

One Cloud Number Can't Answer a Directional Question

Cloud cover is reported as a single figure: the fraction of the sky covered. Forty percent means roughly two fifths of the sky is blocked. It does not say which two fifths.

For a target sixty degrees up, that distinction rarely matters. The sky overhead dominates, and the total is a fair guide to it.

Low targets are different. A satellite pass rises near the horizon, usually peaks below forty degrees, and crosses a wide arc in a few minutes. A front to the west can leave the zenith clear while blocking everything below twenty degrees in that direction. The forecast reads forty percent. You see nothing.

Same problem for a low moon, a planet near the horizon, a comet at dusk, a partial eclipse at sunset. Anything you look sideways at rather than up at.

The Geometry

Looking straight up, you see through about ten kilometres of atmosphere. At ten degrees above the horizon, nearly six times as much, and through a column of air sitting kilometres away rather than overhead.

A cloud layer at 1,500 metres blocks a target ten degrees up if it sits around nine kilometres away on that bearing. Conditions where you are standing are close to irrelevant. What matters is the weather nine kilometres downrange in that direction.

This is why observers report the forecast was wrong when it was not. It described the sky correctly. It answered a different question.

How Clear Skys Estimates Direction

No weather service publishes cloud cover by compass direction. It does not exist as a product. Gridded cloud cover does, and so does trigonometry.

Given a target's altitude and bearing, the horizontal distance at which your line of sight reaches each cloud layer follows directly. Clear Skys samples the forecast grid at a ring of points around your location, out to forty-five kilometres, and reads the cloud where the sight line crosses each layer rather than where you are standing.

For a satellite pass this is applied at three points along the track, so the answer can differ between the start, middle and end.

The correction stops above forty degrees. That is a design choice, not a law of physics: by then the sample points sit within a few kilometres of you, usually inside the same forecast grid square, so a directional figure would be the ordinary one wearing a compass point.

It also stops at the other extreme. A target a few degrees up needs sample points further out than the forecast is sampled, and rather than assume conditions forty-five kilometres away carry on indefinitely, the ordinary figure is used for that layer. High cloud hits this on every satellite rise and set.

What It Cannot Tell You

A grid square reporting forty percent low cloud says roughly forty percent of that square is covered. It does not say whether your line of sight passes through the covered part. Two people side by side, watching the same satellite, can get different answers.

Cloud is not a flat sheet at a fixed height. The layer heights used to calculate crossing distances are representative values, not measurements. Low cloud at five hundred metres puts the relevant point three times closer than sampled.

Where the model grid is coarse, the correction does nothing. If grid squares are larger than the sampling distance, every sample lands back in your own square. Clear Skys detects this and falls back rather than presenting an unchanged number as a directional one.

Terrain beats all of it. Clear sky ten degrees above the south-west horizon is no use behind a hill.

Where You'll See It

On satellite passes, and only when it disagrees with the ordinary forecast.

Most nights the sky is roughly uniform and the two agree. Nothing is shown. A permanent row repeating the main score is clutter.

When they disagree, one line appears under the pass: which end of the track looks better, and roughly when to look. For example, clearer at the start toward the north-west, with more cloud by the time it sets in the south-east.

This is not part of the night's score. The score describes the whole night. The pass annotation is a narrower claim about a few minutes of sky in one direction. Keeping them separate means neither has to be hedged for the other.

How to spot the ISS and other satellites →

Is It Actually Better?

Not established yet.

Clear Skys checks its forecasts against ERA5 reanalysis, which is how the figures on the accuracy page are produced. The directional estimate has not been running long enough to go through that process.

Measuring it is awkward. Reanalysis is gridded at your location the same as the forecast, so it records what the sky did overhead, not along a bearing. The useful signal is in nights where the two methods disagree, and those are rare by design.

When there is enough data, the result goes on the accuracy page whichever way it falls.

See our verified forecast accuracy →