How UV Forecasts Work, and How Far Ahead to Trust Them

When you check tomorrow's UV index, part of that forecast was settled centuries ago and part of it was decided this morning. The predictable part is the sun, whose position over any city on any future date can be worked out exactly. The uncertain part is the weather, and specifically the clouds. Understanding which is which is the whole answer to how far ahead a UV forecast can be trusted.

The chart below shows both at once. It is a five-day UV index forecast for one city, with the cloud-free ceiling drawn separately from the value you would actually stand under.

Five-day UV index forecast for Berlin, retrieved on 2 July 2026. The dashed line is the clear-sky ceiling set by the sun, the ozone layer and haze; the solid line is the forecast once predicted cloud cover is included; the shaded area is the UV that clouds are expected to remove. Source: CAMS via Open-Meteo.

A UV forecast is two forecasts in one

Every UV index forecast comes out of a computer model that combines a short list of ingredients: how high the sun sits, how much ozone lies overhead, how much cloud is in the way, and smaller contributions from altitude, haze and reflective ground. The World Health Organization lists these same factors as the things that set solar UV levels, and the ICNIRP notes that the UV Index is normally calculated with models that account for ozone and cloud cover. The CAMS model behind every forecast on this site, run by Europe's ECMWF, works exactly this way. Its documentation of how the UV index is computed describes the calculation: the model works out the UV irradiance arriving at the surface, taking into account the surface's reflectivity, airborne particles, clouds and the ozone overhead, and then weights it by the erythema spectrum, the standard curve for how strongly each wavelength reddens skin. The next-day forecast published by the US Environmental Protection Agency is built from the same ingredients, relating ground-level UV to forecast ozone, forecast cloud, the sun's angle and elevation, as the EPA's own description of the calculation sets out.

What matters for a forecast is that these ingredients are not equally predictable. Some are fixed years ahead. One of them, cloud, is only knowable for a few days. Separating the reliable part from the uncertain part is what tells you how far ahead the number can be trusted.

The sun's contribution is fixed centuries ahead

The single largest driver of the UV index is how high the sun climbs. The higher it is, the shorter the path its rays take through the atmosphere, and the more UV reaches the ground; the WHO puts it plainly, that the higher the sun in the sky, the higher the UV level. That height is pure geometry. It depends only on latitude, the day of the year and the time of day, which is exactly how the EPA describes the sunlight-angle step of the calculation. None of those inputs cares what the weather is doing.

Because it is astronomy, this part is knowable without limit. The sun's position over Berlin at 1 p.m. next Tuesday, or on the same date in 2050, can be computed today to the minute. That is why the dashed line in the chart is so smooth, tracing the same arc day after day; the modest shifts in its height, like the step down after the first day, come from ozone and haze, the slower-moving parts of the atmosphere, not from the sun. It is the clear-sky UV index, the value you would get under a perfectly cloudless sky, and both CAMS and the Open-Meteo air-quality API publish it as a variable in its own right. It sets a ceiling that the forecast can approach but never beat. The shape of that daily arc, and why its height changes so much with latitude and season, is the subject of when is UV strongest.

Ozone changes slowly, and it is watched daily

The second ingredient is the ozone layer, which absorbs much of the sun's UVB before it reaches the surface. Less ozone means more UV gets through, as the WHO notes. Ozone shifts more than the sun's geometry does, but it moves slowly and over large areas, and it is measured continuously from orbit.

Each CAMS run starts by folding fresh satellite observations of ozone, along with other measurements of the atmosphere, into its previous forecast through data assimilation, as the ECMWF documentation describes. The American system rests on the same principle: satellite ozone readings are projected a day forward per the EPA, and the operational NOAA model assimilates them too, as its technical note on how the index is computed sets out. Because ozone drifts gradually, this part of the forecast is nearly as dependable as the sun on the timescale of a daily forecast. It rarely produces large surprises from one day to the next.

Clouds are what the forecast is really predicting

Then there is cloud, and this is where the certainty runs out. A thin veil can leave UV almost untouched, while a thick overcast can pull it down hard. Under heavy cloud, the EPA notes that ground-level UV can fall to roughly a third of its clear-sky value; the full picture of what different skies let through is in does UV go through clouds, windows and water.

The model therefore computes the number both ways. CAMS carries its forecast clouds directly into the UV calculation and publishes the result alongside a clear-sky companion value with the clouds left out, which is exactly the pair of lines in the chart; the US system reaches the same split by working out the clear-sky UV first and multiplying it by a cloud transmission factor, as the NOAA method spells out. Either way, the cloud part is genuinely hard to pin down. NOAA points out that because the opacity of clouds varies so much, there is a large range in how much UV gets through for a given amount of cloud. Everything between the two lines in the chart is this cloud effect. On Berlin's clear first day the forecast nearly touches the ceiling; across the cloudier middle of the week it drops by half or more, even though the sun and the ozone above barely moved.

The gap is not always wide. Retrieved on the same afternoon, the five-day forecast for Madrid sat under a settled summer high, and its forecast line ran almost exactly on top of its clear-sky ceiling, peaking near 10 every day. The size of the gap is decided by the weather, which is precisely why it is the uncertain half of the forecast.

So how far ahead is a UV forecast reliable?

This split explains the horizons the official services actually use. The EPA and the National Weather Service publish a next-day UV index for locations across the United States, built on that next-day ozone forecast. The CAMS system run by ECMWF, which feeds a great many UV apps and websites, issues a five-day forecast twice a day, at 00:00 and 12:00 UTC, per its own documentation; the European climate and health authorities present the CAMS clear-sky and total-sky UV index for the next four days. Those horizons are set by clouds, not by the sun.

A weather forecast loses skill with each extra day. The long-recognised limit of day-to-day atmospheric predictability is about two weeks, and beyond it even the best models struggle to do better than the seasonal average, as ECMWF discusses in its work on the forecast skill horizon. Useful cloud detail fades out well before that. So the honest answer has two halves. The clear-sky ceiling can be given, near enough, for any date you like, this year or next. The UV you will actually stand under is only as good as the cloud forecast, which means today and tomorrow are dependable, the next few days are a solid guide, and anything past about five days is really a statement about the sun rather than the sky.

Why the number changes when you check again

If tomorrow's UV index reads 7 today and 5 when you look in the morning, the sun did not move and the ozone layer did not collapse. The cloud forecast was revised. CAMS reruns its global forecast every twelve hours, and each run begins by assimilating the newest satellite and weather observations into its starting state, per the ECMWF documentation, so the predicted cloud cover for a given afternoon comes into focus as that afternoon draws closer. A UV forecast that changes is usually a cloud forecast that got better. Look at the clear-sky value on the same page and it will have barely shifted.

What this means in practice

Trust the near days, treat the far days as provisional. Today's and tomorrow's UV index carry the reliability of a short-range weather forecast. Read the fourth or fifth day as a rough guide that will sharpen as it arrives.

A changed forecast is a changed cloud outlook. If the number moves between checks, it is the weather firming up, not the sun. For anything you are planning around, look again the morning of.

Watch the clear-sky value for the ceiling. On a day that turns out sunnier than expected, UV can climb toward the clear-sky figure, so the higher of the two numbers is the safer one to plan protection around. The WHO advises sun protection once the UV index reaches 3, and the clear-sky value tells you the most the day could reach. What each level means is covered in what is a safe UV index.

Methodology and sources

The two forecasts referenced on this page are the UV index and the clear-sky UV index from the CAMS model, retrieved from Open-Meteo on 2 July 2026 for Berlin and Madrid, in each city's local time and rounded to one decimal. CAMS provides the same two quantities that the chart separates: the clear-sky ceiling and the total-sky value that includes cloud. The index itself follows the WHO definition: CAMS computes the biologically effective UV irradiance at the surface and divides it by 0.025 W/m², so one step on the index corresponds to 25 mW/m² of skin-damaging UV. Every city page on UVI.today draws its hour-by-hour UV forecast from this same CAMS data; how those forecasts are produced is described on the methodology page, and the bodies behind UV measurement and sun-safety guidance are collected in UV index official sources. The other factors that shape the numbers, beyond sun, ozone and cloud, are covered in what affects the UV index.

• World Health Organization: Ultraviolet radiation fact sheet — factors that determine solar UV levels; sun protection advised at UV 3 and above.
• ECMWF Knowledge Base: CAMS global atmospheric composition forecast documentation — the "UV Index definition and computation" section: surface UV irradiance computed from surface albedo, aerosol, clouds and ozone, weighted by the erythema spectrum, to the WHO definition; the twice-daily five-day forecast cycle; satellite ozone and aerosol observations assimilated at the start of each run.
• US EPA: Learn About the UV Index — how the National Weather Service calculates the next-day UV index from forecast ozone, sun angle, elevation and cloud.
• NOAA Climate Prediction Center: UV Index, How It is Computed — the radiative-transfer model, satellite ozone assimilation, and the clear-sky-times-cloud-transmission step.
• ICNIRP: The Global Solar UV Index — models account for ozone and cloud cover; clear-sky peaks near midday.
• Copernicus Atmosphere Monitoring Service (CAMS) and Open-Meteo Air Quality API — the UV index and clear-sky UV index products, five-day global forecast refreshed twice daily.
• Climate-ADAPT: Four-day forecast of UV index (CAMS) — clear-sky and total-sky UV forecast to four days, using WHO methodology.
• ECMWF: The Forecast Skill Horizon — the classical two-week limit of atmospheric predictability and how forecast skill falls with lead time.