What formula does UV Hour use?

Minutes before the first redness = 0.667 × MED ÷ effective UV index. The effective UV index is the clear-sky forecast multiplied by a cloud factor, an altitude factor and a surface reflection factor. That is the whole model.

Where does the 0.667 constant come from?

From the definition of the UV index: one point equals 25 mW/m² of erythemally weighted irradiance. Time in seconds is MED ÷ (UVI × 0.025); dividing by 60 to get minutes gives MED ÷ (1.5 × UVI), which is 0.6667 × MED ÷ UVI. It is not a fitted parameter.

Yes — a simple deterministic one, and saying otherwise would be dishonest. It applies a fixed minimal erythemal dose per Fitzpatrick type, a standard cloud attenuation curve, and corrections for altitude and ground reflection, to a public weather forecast. What it is not is a machine-learning system: no training data, no fitted parameters, no proprietary dataset. Every figure is reproducible with a calculator from the constants on this page.

What does the model deliberately not do?

It never certifies a UV level as safe, and it publishes no threshold of its own. The only protection threshold it shows is the WHO one at UV 3. It reports the estimated time to the first visible redness as a point to stay under, never a target to reach — the biological damage starts below it.

How fresh is the data?

Every location is refreshed at most every 15 minutes and cached in between — which is both what MET Norway requires of API consumers and more than enough for a quantity that changes on the scale of the sun's elevation.

Method & sources

UV Hour is a simple deterministic model: the standard erythemal formula, 0.667 × MED ÷ effective UV, applied to a public weather forecast. No training data, no fitted parameters, no proprietary dataset — but a model all the same, with assumptions worth knowing. This page gives the constants, the named sources and the limits, so you can reproduce or challenge any number on the site.

The calculation

effective UV = clear-sky UV
             × cloud factor(cloud cover %)
             × altitude factor(metres)
             × surface factor(sand / water / snow)

minutes before first redness = 0.667 × MED / effective UV

below UV 2  skin produces no melanin — nothing to report
from  UV 3  the WHO recommends sun protection

There is no threshold of our own anywhere in this. An earlier version of the site derived a per-skin-type "safe ceiling" from a 40-minute rule we had chosen ourselves. That was an editorial judgement wearing the clothes of a safety limit, so it has been removed: the only protection threshold published here is the WHO one.

Minimal erythemal dose, by Fitzpatrick type

The doses this site uses, in joules per square metre of erythemally weighted radiation — 2 to 6 standard erythema doses (1 SED = 100 J/m², ISO/CIE 17166).
Skin type	MED used here	In SED	Reaction
Type I	200 J/m²	2 SED	Always burns, never tans.
Type II	250 J/m²	2.5 SED	Burns easily, tans minimally.
Type III	300 J/m²	3 SED	Burns moderately, tans gradually.
Type IV	400 J/m²	4 SED	Burns rarely, tans easily.
Type V	500 J/m²	5 SED	Burns very rarely, tans very easily.
Type VI	600 J/m²	6 SED	Almost never burns, always tans.

This table is the weakest link in the model, and it should be named as such. The 2–6 SED progression is a convention widely used in UV calculators, not a measured table from a single study. Fitzpatrick's own paper classifies how skin reacts; it does not assign a dose in J/m² to each type. Measured erythemal doses vary substantially between studies, between populations and between individuals of the same phototype — Valbuena Mesa and colleagues (2020), cited below, found correlation coefficients of only 0.5 to 0.69 between measured MED and Fitzpatrick type. Treat every minute figure on this site as an order of magnitude for a population, never a threshold for your skin.

Correction factors

Cloud — 1 − 0.75 × (cover/100)^3.4. An empirical curve, chosen to sit inside the range the measurement literature reports (Calbó and colleagues, 2005, below): half cover removes about 7% of the UV, full overcast about 75%. Real skies vary with cloud type and thickness in ways a single cover percentage cannot capture.
Altitude — 1 + 0.12 × (metres/1000), i.e. +12% per 1,000 m.
Surface reflection — grass ×1.03, beside water ×1.2, dry sand ×1.15, snow ×1.8. Standard albedo approximations, applied as a flat multiplier.
Sunscreen — deliberately not turned into a published duration. The SPF multiplier is measured at 2 mg/cm², an application rate field studies rarely find; protection falls roughly with the square root of the shortfall, and reapplication governs the rest.

Sources

Phototype classification. Fitzpatrick TB. "The validity and practicality of sun-reactive skin types I through VI." Archives of Dermatology, 1988;124(6):869–871. doi:10.1001/archderm.1988.01670060015008 · PMID 3377516. Defines types I–VI by how skin reacts. It does not assign a dose in J/m² to each type — the table above is a convention, not this paper's content.
Erythemal weighting and the SED unit. ISO/CIE 17166:2019, Erythema reference action spectrum and standard erythema dose (originally CIE S 007/E-1998). iso.org/standard/74167. Fixes the action spectrum behind every "erythemally weighted" figure here, and defines 1 SED = 100 J/m².
UV index definition, bands and the protection threshold. World Health Organization, World Meteorological Organization, UN Environment Programme and ICNIRP. Global Solar UV Index: A Practical Guide. Geneva: WHO, 2002 (WHO/SDE/OEH/02.2). who.int/publications/i/item/9241590076. Source of the 25 mW/m²-per-point definition, the five bands, and the UV 3 protection recommendation.
Cloud attenuation of UV. Calbó J, Pagès D, González JA. "Empirical studies of cloud effects on UV radiation: a review." Reviews of Geophysics, 2005;43:RG2002. doi:10.1029/2004RG000155. The review our cloud curve is checked against.
How much measured MED actually varies. Valbuena Mesa MC, Nova Villanueva JA, Sánchez Vanegas G. "Minimal erythema dose: correlation with Fitzpatrick skin type and concordance between methods of erythema assessment." Actas Dermo-Sifiliográficas, 2020;111(5):390–397. doi:10.1016/j.ad.2019.12.003. Correlation between measured MED and phototype was 0.5–0.69 — the reason the figures here are estimates for a population, not thresholds for a person.
Live forecast. MET Norway Locationforecast 2.0, clear-sky UV index and cloud cover, under CC BY 4.0. api.met.no/doc/TermsOfService.

Weather data from MET Norway (CC BY 4.0). Requests identify themselves as uvhour.com bonjour@agencedebord.com and every location is cached for 15 minutes, as MET Norway's terms require. Your location comes from coarse network-edge geolocation and is not stored — privacy.

What this model cannot know

Your actual erythemal dose. It uses a population figure for your phototype; individual sensitivity varies widely within every type.
Your medication and your history. It assumes a healthy adult on no photosensitising drugs. Several common prescriptions lower the erythemal threshold substantially, and nothing here accounts for that.
What you are actually doing. Clothing, shade, sunscreen as actually applied, posture and the angle of your skin to the sun all change the dose. The model sees a horizontal surface.
Cumulative risk. The formula estimates one threshold on one day. Long-term skin cancer risk is built from lifetime dose, which no page here calculates.

What this site deliberately does not do

It never certifies a UV level as safe, and publishes no threshold of its own. The one protection threshold shown is the WHO's, at UV 3.
It never recommends staying out beyond the erythemal threshold. Every number is a point to stay under, not a target to reach — and the damage starts below it.
It publishes no "time × SPF" duration, because such figures read as permission regardless of the caveat attached.
It makes no health claim about sun exposure, in either direction, and it is not medical advice.
It knows nothing about you. There is no account and no cookie; visits are counted by cookie-free analytics that keep no identifier — the privacy page has the detail.

Corrections

If a constant here is wrong, or a factor is applied in a way that does not match the cited literature, write to bonjour@agencedebord.com and it will be fixed. The arithmetic is the product; an error in it is a bug, and so is a source that does not say what we claim it says.