CL system for UV explained



CL System for UV calibration explained

During early October 2018, we quietly launched our new Continuous Learning (CL) auto-calibration system for UV. The system is designed to evaluate the raw UV sensor data from your SKY, and apply calibration corrections if needed. The data science and back-end techniques involved are fairly sophisticated; here’s a simplified explanation:

The UV CL system begins by considering a daily upper threshold UV index based on your precise location, the time of year, and the satellite observed global ozone concentration. Each day of your SKY’s data is then quality controlled and analyzed by smart algorithms to determine if the sky was clear around local solar noon, and therefore if your location would have observed close to the possible upper limit. When a sufficient amount of data has been analyzed, a calibration is then computed by comparing your observed UV to the threshold value on those days. The calibration is unique to each station, stored locally on your HUB, and applied automatically to each subsequent observation keeping your data in alignment forever.

Fortunately, UV sensors do not tend to drift much but we still run the UV CL system for each station once per day. This ensures that your station will remain in near perfect calibration on an ongoing basis (unlike any other weather station on the market!).

If your UV calibration still seems suspiciously high or low, please let us know directly and we’ll dig into your data.

Continuous Learning - which variables?
UV seems too high or too low (solved with auto-calibration CL system)
UV seems too high or too low (solved with auto-calibration CL system)
New WF station Knjazevac id 5873
Barometric Pressure way off (corrected once CL system is deployed)
pinned #4


Today a peak of UV 3.
Normally it should be a maximum of 1 to 1.5

Could it be reflectation of clouds? :slightly_smiling_face:

The comming days are going to be quiet cloudy, so I’m gonna check the Sky daily for UV readings.

The days before the UV readings where perfect.


Hi @poider1982 . Thanks for the report, we really appreciate it. The CL system for UV is in it’s early days and can certainly benefit from input from the field. We took a look at your data. The CL system benchmarked a reference UVI value of 1.6 on Oct 20th which helped calibrate your sensors over the last couple days. We’re not sure what happened today. Let us know what the next few days report. Thanks!


I’ll keep You up to date.
Furthermore a big thumbs up for Weatherflow.
UV readings and especially air pressure doing great now.
Also Lightning detection works perfect.
Never had any false strikes.

Oké it’s going off topic, but all the efforts of Weatherflow deserves a big compliment. :ok_hand:



My UV readings are approximately half of what they should be. Currently my Sky is reading 4.4. The forecast for today suggests the UV should above 8. The sky is clear and I have rest the device to ensure all updates are active.

Any suggestions about why the reading is so far off?


Without knowing your station ID, location and how your Sky is mounted it’s impossible to say.


Sorry Gary,

Station ID: 4581

A clear view of the Sky in all directions.


Hi Geoff. We just took a look at the CL analysis for your SKY and it doesn’t have enough data to confidently suggest a change in calibration. However, a manual inspection reveals that it’s indeed off by a factor of about 2.1. We’ll apply that adjustment today and the CL system will update it over time as more data is collected.


Thanks David,

As you would expect, today has been overcast. However, as the UV levels recorded today are about the same as those recorded yesterday in full sun, it looks hopeful.

Out of interest, how many days worth of data is required to make an adjustment?


We need at least 4 days of clear days, but that’s just one condition that must be met so it can often take more.

pinned #14