How to measure UVA yourself: meters, cards and phone myths

This site models UVA rather than measuring it, for a reason worth being upfront about: measuring ultraviolet properly is genuinely hard, and the instruments that do it well cost more than a car. What follows is an honest survey of what is available, what each class of device actually tells you, and how to tell a real reading from a plausible-looking number.

The instrument classes

The reference: spectroradiometers

A double-monochromator spectroradiometer disperses incoming light and measures each narrow wavelength slice separately, typically from around 290 nm to 400 nm or beyond. Because it produces the whole spectrum, any weighting can be applied afterwards in software: erythemal for a UV Index, vitamin D for a synthesis rate, or no weighting at all for the unweighted 315–400 nm integral this site calls UVA.

That flexibility is exactly why it is the reference. Every other device bakes one weighting into hardware and cannot be re-purposed. The drawbacks are cost, size, the need for regular calibration against traceable standard lamps, temperature stabilisation, and scan times long enough that a passing cloud can corrupt a measurement.

Broadband radiometers

These use a stack of optical filters and a detector engineered so the combined response approximates one target curve. Two kinds matter here:

• Erythemal biometers approximate the CIE sunburn action spectrum. These are what most national UV monitoring networks deploy, and what most published "UV Index" observations come from.
• UVA radiometers aim for a flat response across roughly 315–400 nm. These measure the quantity this site models, and are the only affordable instruments that do.

They are robust, weatherproof and log continuously. Their weakness is that the response curve is only ever an approximation, and it drifts as filters age — which is why serious deployments recalibrate annually against a spectroradiometer.

Consumer meters: what you are actually buying

A pocket UV meter is a photodiode, a filter, and a microcontroller applying a conversion factor. Whether the number it shows means anything depends on three things that datasheets rarely mention.

• Detector chemistry. Plain silicon photodiodes respond strongly across the visible and near-infrared, so they depend entirely on filtering to reject perhaps 99.9% of the light actually hitting them — a demanding requirement met unevenly at low cost. Silicon carbide and gallium nitride detectors are intrinsically visible-blind: their bandgap means they barely respond above the ultraviolet at all. A meter specifying SiC or GaN is a substantially better instrument than one that does not specify.
• Cosine response. A flat sensor must count light arriving at 60° from vertical at half the weight of light arriving straight down, because that is how irradiance onto a horizontal surface works. Achieving this requires a properly shaped diffuser. Cheap meters with a bare window under-read low-angle light badly — precisely the diffuse skylight that carries a large share of UVA.
• Spectral assumption. Most consumer meters display a "UV index", derived from a single broadband reading by assuming a standard clear-sky solar spectrum. Change the spectrum and the assumption fails. Under cloud, at low sun, behind glass, or under a nail lamp or sunbed, the true spectral shape is nothing like the assumed one, and the displayed index can be wrong by a wide margin — in either direction.

Cards, beads and badges

Photochromic UV cards and the beads sold for classroom demonstrations contain spiropyran or spirooxazine dyes that change colour on ultraviolet exposure. They are excellent for making an invisible thing visible — put a bead behind a window and behind sunscreen and the difference is immediate and convincing.

What they cannot do is quantify. The response is non-linear, saturates quickly, resets over time, drifts with temperature, and its spectral sensitivity is a property of the dye rather than anything standardised. Two cards from different manufacturers will disagree, and neither is weighted to any recognised curve.

Polysulphone film dosimetry is the serious version of the same idea. A thin film worn as a badge changes optical absorbance in proportion to cumulative UV dose, read out afterwards in a spectrophotometer. Its response is a rough approximation of the erythemal curve, which makes it the classic tool for personal sunburn-dose studies — and, for the same reason, a poor instrument for UVA specifically.

Why your phone cannot do this

This deserves stating plainly because the app stores suggest otherwise. No mainstream smartphone contains a UV sensor. Three independent barriers stand in the way:

• The ambient-light sensor is designed to track perceived brightness for screen dimming. It is filtered to approximate the human photopic response and to reject infrared. It has no ultraviolet channel.
• The camera sensor sits behind a UV/IR-cut filter, fitted precisely to stop non-visible light from degrading colour accuracy.
• The lens is glass or optical plastic, which absorbs strongly through the UVA range on its own — the same physics covered in does glass block UVA.

So any app showing a UV number is doing what this site's calculator does: taking your coordinates and retrieving or computing a value. That is a legitimate and useful thing to do — it is simply not measurement, and an app that implies otherwise by asking you to point your phone at the sky is misrepresenting itself. Clip-on and Bluetooth UV accessories are a different matter: those contain real photodiodes and are subject to the same caveats as any other consumer meter.

Sanity-checking any reading

Whatever device you have, these five checks will expose most problems without any reference equipment:

1. Night reads zero. A non-zero night reading means stray light, an offset error or infrared leakage.
2. The peak lands at solar noon, which is often an hour or more from clock noon — see what time UVA is highest. A peak at 12:00 sharp during daylight saving is a sign the device is reporting a modelled value rather than sensing one.
3. The daily curve is a smooth broad arc, not a plateau or a step. Steps suggest coarse internal quantisation.
4. Behind a closed window, the reading behaves diagnostically. An erythemally weighted device should collapse toward zero, because glass removes nearly all UVB. A true UVA device should retain a large fraction. If your "UVA meter" reads near zero indoors by a sunlit window, it is not measuring UVA.
5. Shade drops it, but not to zero. Step into full shade with open sky above and a correct instrument still reads substantially — that is diffuse skylight. A device that reads near zero in open shade has a broken cosine response.

Comparing a meter against this site

If you want to check a UVA meter against the modelled value here, three things have to line up first:

• Match the quantity. The model outputs unweighted irradiance over roughly 315–400 nm in W/m², and the UVA Index is that figure divided by six. A meter reporting an erythemal "UV index" is measuring a different physical quantity and no agreement should be expected.
• Match the geometry. The model assumes irradiance onto a horizontal surface. Hold the sensor level, facing straight up, away from walls and reflective ground.
• Choose the right conditions. Compare under clear sky with the sun well up. Under broken cloud the model is attenuation-only and cannot reproduce cloud enhancement, so real readings can legitimately exceed it.

Under those conditions, agreement within roughly 20% would be a good result for a modelled value against a consumer instrument — and if you find a systematic offset with a properly calibrated device, that is genuinely useful information. The model's coefficients all live in one block of js/uva.js specifically so they can be recalibrated against real measurements, and the project welcomes that kind of contribution.