How the Stargazing Score Works

The Four Weather Factors

Every hour of darkness is scored from 0 to 100 based on four weather measurements, each weighted according to how much it affects observing conditions:

Cloud cover (55% of the score): the dominant factor. ClearSkys uses three-layer cloud data from Open-Meteo: low cloud (below ~2,000m) is weighted at 50%, mid-level cloud (2,000–6,000m) at 35%, and high cloud (above 6,000m) at 15%. This layered approach matters because low cloud is opaque and blocks everything, while high cirrus is translucent and primarily affects contrast on faint objects. A sky with 20% high cirrus but no low cloud scores much better than 20% low stratus.

The cloud score uses a stepped scale: 15% or below cloud scores 100 (essentially clear), 15–30% scores 85 (minor obstruction), 30–50% scores 60 (half the sky compromised), 50–70% scores 30 (mostly cloudy, limited windows), 70–85% scores 10 (brief gaps at best), and above 85% scores 0.

Wind speed (20% of the score): affects telescope stability and atmospheric turbulence. Below 5 km/h scores 100 (calm, excellent seeing potential). 5–10 km/h scores 80 (manageable, minor vibration). 10–15 km/h scores 50 (challenging for telescopes). 15–20 km/h scores 20 (binoculars only). Above 20 km/h scores 0 (stay indoors).

Humidity (15% of the score): drives dewing and atmospheric transparency. Below 70% scores 100 (dry, good transparency). 70–80% scores 70 (dew possible on unheated optics). 80–90% scores 40 (dewing likely, reduced transparency). Above 90% scores 10 (heavy dewing, poor sky quality).

Rain probability (10% of the score): the final backstop. Below 10% scores 100. 10–30% scores 70. 30–60% scores 30. Above 60% scores 0. Rain probability is weighted lowest because the other factors (especially cloud) already capture most of the information that rain probability correlates with.

Two refinements sharpen the picture further. The humidity factor is surfaced in the breakdown as Dew risk, scored from the dew-point spread, the gap between air temperature and the dew point, which predicts dewing on optics more directly than relative humidity. And two conditions adjust how good a clear sky actually looks: transparency (haze), read from visibility, trims the score when scattered light washes out fainter targets even under cloudless skies; and seeing, estimated from high-altitude jet-stream wind, reflects atmospheric steadiness, which governs how much fine planetary and lunar detail you can resolve. These don’t stop you observing the way cloud or rain do, but they’re the difference between a merely clear night and a genuinely good one.

The Moon Penalty

The moon is charged hour by hour. For every hour of your observing window, ClearSkys works out how bright the moon is and how high it stands, checking four times across the hour, and takes off up to 35 points. An hour with the moon below the horizon costs nothing.

The charge for an hour is 35 × (illumination / 100)^1.5 × height factor. The exponent of 1.5 makes the charge climb steeply towards full moon: a half-lit moon high in the sky costs about 12 points an hour against 35 for a full one, because a full moon is around ten times brighter than a quarter moon. The height factor runs from 0.35 for a moon just above the horizon to 1.0 for a moon 45° up or higher, because a low moon brightens far less of the sky than a high one.

The night's moon penalty is the average of those hourly charges. So a full moon that sets two hours after darkness begins costs much less than one that is up all night, and the hours it is up cost more than the hours after it has gone.

The profile multiplier then scales each hour's charge. The default (General) profile uses 1.0. Astrophotography uses 1.3. Long-exposure images are severely affected by sky brightness. Visual/binoculars uses 0.8, wide-field views tolerate some moonlight. Planetary/lunar uses 0.3. The moon is the target, not the problem. Aurora uses 0.25. Aurora is bright enough to compete with moonlight.

Because each hour carries its own moon, the best viewing window is chosen on the score with the moon included, and it moves to the moonless hours when they are clear. If no hour is good enough once the moon is counted, the forecast still shows the clearest stretch and marks it "moon up", and the night is not rated good.

Impact labels: a penalty of 25+ points is "severe" (deep-sky is a write-off), 12–24 is "moderate" (stick to bright objects), 4–11 is "mild" (noticeable but manageable), and below 4 is "minimal" (effectively no impact).

Observer Profiles

The same night scores differently depending on what you're observing. A windy but clear night is excellent for aurora watching but poor for astrophotography. A bright moon ruins deep-sky work but doesn't affect planetary observation at all.

ClearSkys offers five profiles that reweight the scoring factors:

The General profile (default) uses the standard weights: cloud 55%, wind 20%, humidity 15%, rain 10%, with a standard moon penalty. Good threshold is 65. This suits most casual stargazers.

The Astrophotography profile increases cloud weight to 60% and wind to 25% (long exposures demand still air), reduces humidity and rain weight, and applies a 1.3× moon penalty multiplier. Good threshold is 70, astrophotographers need better conditions than visual observers.

The Visual / Binoculars profile is the most forgiving. Cloud drops to 50%, rain increases to 15% (relevant for binocular observers who can't easily shelter equipment), and moon penalty is reduced to 0.8×. Good threshold is 55.

The Planetary / Lunar profile prioritises steadiness over darkness. Wind weight increases to 25% (seeing stability is critical for planetary detail). Moon penalty drops to 0.3×. Moonlight is irrelevant when you're looking at planets. Good threshold is 60.

The Aurora profile maximises cloud weight at 65% (you need clear northern horizon) and minimises wind (10%) and moon penalty (0.25×). Good threshold is 50. Aurora is bright enough that moderate conditions are still worth going out for.

Build your own. If none of the presets match how you observe, ClearSkys lets you create a custom profile. You set the weight of every factor yourself. Cloud, wind, humidity, rain and seeing, and they simply have to add up to 100, so dropping a factor to zero removes it from your score entirely. You can go further and control the cloud-layer split: how much low, mid and high cloud each count toward the cloud score. Weight high cloud heavily and thin cirrus will tank a deep-sky night; weight it near zero and the same cirrus barely dents a planetary score. Moon and sky-brightness sensitivity are separate dials too. Your custom profile saves to your device, or to your account if you're signed in, so it follows you across visits.

Twilight Scoring for High Latitudes

At latitudes above ~50°N, summer nights don't reach full astronomical darkness (sun below −18°). Rather than showing no data for weeks, ClearSkys uses twilight scoring: the observing window becomes the civil twilight window (sun below −6°), and every hour in it pays for how bright the sky is in that hour.

The charge follows the sun's depth below the horizon. On the General profile, an hour with the sun 16° below the horizon costs about 3 points, 14° below about 7, 12° below about 12, 9° below about 18, and 6° below, the end of civil twilight, about 27. At 16° below you can still see down to roughly magnitude 5.5 to 6.0; at 9° below only bright objects are realistic. The sun's depth is checked four times in each hour, so the hour after sunset pays more than the hour around midnight, and the night's twilight penalty is the average.

So the best viewing window moves towards the darkest part of the night. If no hour is good enough once the twilight is counted, the forecast still shows the clearest stretch and marks it "twilight", and the night is not rated good. Profiles feel twilight differently: the charge is multiplied by 1.0 on General, 1.3 on Astrophotography, 0.8 on Visual, 0.2 on Planetary and 0.3 on Aurora.

The same charge applies on the first and last nights of the year with real darkness. When astronomical darkness lasts under three hours, ClearSkys widens the window into the surrounding twilight so the score doesn't jump from one night to the next, and those twilight hours pay for their brightness too.

If the sun never drops below −9°, the night is unscored. Conditions are too bright for meaningful stargazing beyond planets. For the UK and similar latitudes (50–59°N), the sun reaches at least −12° on every night of the year, so every UK night gets a score even at the solstice.

Where the Twilight Numbers Come From

The twilight curve runs from nothing at full astronomical darkness up to 45 points at civil twilight. Those values were set by judgement when the scoring was first written, so it is worth setting out how they hold up against published measurement.

Two independent sources agree on the underlying physics. The US Naval Observatory's Explanatory Supplement to the Ephemeris tabulates illumination on a horizontal surface against how far the sun has dropped below the horizon. The Liverpool Telescope publishes a sky brightness model built from several years of observations at its site on La Palma, expressed in magnitudes above the darkest sky it records. They measure different quantities by different methods, decades apart, and agree to within about a magnitude across the whole twilight range.

The pattern is simple. The sky brightens by roughly one magnitude for every degree the sun climbs back toward the horizon, from 12 degrees below up to about 4 degrees below. The curve then flattens as it approaches the natural airglow floor, the faint permanent glow that remains even on the darkest nights. From 18 degrees to 6 degrees that is 8 to 9 magnitudes, a factor of several thousand in brightness.

Turning that into a score penalty needs a decision the physics cannot make. Does a point represent how much brighter the sky has become, or how many stars you can no longer see? The two diverge because star loss saturates: by nautical twilight most naked-eye stars have already gone, and further brightening takes away progressively fewer. Scored on brightness, 12 degrees below the horizon would cost about 11 points. Scored on stars lost, about 34. The curve takes 20 there, and each hour pays 0.6 of the curve, so an hour spent at 12 degrees below costs 12 points, close to the brightness reading.

That sits closer to the lenient end deliberately. At UK latitudes the sun never reaches 18 degrees below the horizon between late May and late July, so this penalty applies to every night of the summer. The harsher reading would score every clear June night as poor, when a clear June night still gives you the planets, the moon, noctilucent cloud and the best ISS passes of the year.

The 0.6 is there because every hour of a twilight night is at least as bright as its darkest moment. Until October 2026 ClearSkys charged the whole night at that darkest moment; charging every hour on the full curve would have cost about 1.65 times as much. Across the summer of 2026 at fifteen places from 40° to 64° north, 0.6 keeps the average twilight penalty where it was, so the change moved the best viewing window into the darker hours rather than making summer harsher.

There is a limit to how precise any of this can be. Twilight brightness at a given sun altitude is not a fixed quantity. Measurements from Abastumani Observatory show it moving by about a magnitude between seasons in ordinary conditions, driven by haze in the lower atmosphere, and by considerably more after a major volcanic eruption. The anchors are already specified more finely than that variation, so tuning them further would add precision without adding accuracy.

Liverpool Telescope sky brightness model ↗ Mateshvili et al. (2005), twilight sky brightness measurements ↗

Data Sources

Weather data comes from Open-Meteo (primary) with MET Norway Locationforecast 2.0 as an automatic fallback if Open-Meteo is unavailable or rate-limited. Both provide hourly cloud (three layers), wind, humidity, and precipitation data.

Astronomy calculations (sun altitude, darkness windows, moon position, planet ephemeris) use the astronomy-engine library, the same engine used in planetarium software, providing sub-arcsecond accuracy.

Aurora data combines the NOAA 3-day Kp forecast (updated every 3 hours) with the NOAA Ovation model (real-time probability map based on current solar wind measurements).

Satellite passes are computed locally using SGP4 propagation with TLE data from Celestrak, refreshed every 12 hours. This covers the ISS, Hubble, Chinese Space Station, and approximately 20 other reliably bright satellites.