The UV sensor aims to measure the intensity of the ultraviolet part of the sun’s spectrum (as opposed to the ‘solar’ sensor which measures the visible and near IR part of the spectrum).
UV readings are also expressed in a different format from the ‘solar’ sensor readings: For the solar sensor, straightforward irradiance values in units of W/sqm are recorded. But the prime practical reason for monitoring UV levels is to guide human skin exposure, preventing sunburn and possible skin cancer risks.Therefore UV readings are expressed on a UV Index scale of 0-16 which weights the intensity by the wavelengths of the UV light most likely to cause skin damage. This seems to be explained reasonably well on the relevant Wikipedia page, where one index unit corresponds to an irradiance value of about 25W/sqm. Maximum midsummer UVI readings in the UK rarely exceed 7-8,although substantially higher readings are commonly registered at lower latitudes than the UK.
Davis Application Note #6 also provides further information on interpreting UV readings.
UV readings are technically difficult to make – the sensors need extensive individual calibration and the signal is low in amplitude hence causing noisy readings. This has two consequences. First, UV sensors are expensive because of the prolonged production and calibration process. (You may think that the Davis UV sensor is expensive, but sensitive high-end UV sensors are considerably more so.)
The second consequence is that, without moving to a still more sophisticated and costly sensor design, there is unavoidably some noise and short-term fluctuation in the sensor output, even under dark conditions when of course there should be zero UV readings. When the UV sensor was first introduced, Davis was troubled by support calls reporting overnight UV readings of 0.1 or 0.2 UVI, even occasionally higher, which were simply a consequence of noise in the dark current circuitry. The consensus was that these very low but false readings overnight were more of a concern to users than having good sensitivity at low UV levels. So the decision was taken that there would be a threshold in the displayed UV reading of 0.4 and below which any reading would show as zero and this is the solution that remains in place.
The result is of course that wintertime readings in higher latitudes such as the UK will typically show as zero unless the UVI value is above 0.3.