(This article should give you a very enjoyable insight into my brain)

Some days, I get overwhelmed by how bright it is outside. This happens when the weather meets one of two conditions: 1) the sun is very intense, and/or 2) the sun is low-intensity but the cloud cover means that there is lots of glare. The latter condition is met at intermediate levels of cloud cover.

Chauvel et al (1982) reports on prior work conducted in France during the 1970s by Aubrée and Chauvel. This prior work defined the rough luminance values for three categories of skies.

  • “dark overcast skies”: 1100—2200 cd/m²
  • “blue skies”: 1900—3700 cd/m²
  • “bright overcast skies”: 2200—5000 cd/m²
    (Though the authors note in the introduction that the luminance of the sky may rise to “several times greater than” 10,000 cd/m² where there are bright sunlit clouds.)

There’s actually a standardised set of 15 different sky types published as a standard by the International Commission on Illumination (CIE), and also visualised in this paper. (Between the fact that this standard exists and my recent publication that leveraged UV Index data, I really should’ve been a sunlight physicist!)

Thus, brightness is not simply a consequence of sunlight—the specific characteristics of local cloud cover are very important. This lines up with my own expertise—my preference is for very dark cloudy skies, but I’d rather have no clouds than a thin wisp of clouds since an intermediate amount of cloud cover merely increase the brightness.

It’d be very straightforward to derive a “glare annoyingness index”. All I would have to do is record how annoying I think the sky brightness is on a number of days, then generate a model that links the “glare annoyingness” outcome variable to proxies for sun intensity (e.g. UV index, which is freely available and easy to access for Australia on the ARPANSA website) and cloud cover (also easy to access). For the latter, you could also set up a real-time monitoring station that is pointed at the sky and automatically calculates the dark channel (as in the excellent “dark channel prior” method for haze removal in photographs created by He et al), which might be a good proxy of the sky’s ability to create glare given some amount of sunlight.

In particular, glare annoyingness would correlate with sun intensity in a positive, linear fashion, and it would correlate with cloud cover in a non-linear fashion (intermediate values of cloud cover seem to cause more glare).

If I had digital photos as an input rather than simply weather data (e.g. if I wanted to set up a camera permanently trained on an outdoors scene, or even temporarily to generate data to validate a relationship with weather variables), then there is software used for measuring glare given some photo, e.g. evalglare described here and available for download here.

Fortunately, I can save myself the time! AccuWeather has this product called the “AccuLumen Brightness Index”. This product is proprietary. However, you can sign up for a free API account and then use the API to query the Brightness Index for a particular location.

For example, for the city Adelaide (ID = 25257) and the brightness index (ID = 47), if you have an API key, you can simply run in your computer’s terminal:

curl -X GET http://dataservice.accuweather.com/indices/v1/daily/1day/25257/47?apikey=YOUR_API_KEY

Right now, it tells me that the brightness index is 6.0 (Medium). Neat!

Edit 20241006:
Looking at the record of brightness values over time, I get the feeling that these values don’t represent anything meaningful.

So, my current system is not to use the Accuweather product. Rather, I have a simple algorithm based on the UV index.

If three conditions are met, then it is too bright outside. These are:

  1. The predicted UV (i.e. UV for a clear sky) is above 2.5.
  2. The observed UV (i.e. actual UV on the ground) is above 1.
  3. The predicted UV is above 4 OR the the difference between the predicted UV and observed UV is above 1.

The first criterion checks whether the sun is out. The second checks whether the clouds haven’t totally blotted out the sun. The third checks whether the absolute brightness of the sun is very high or alternatively whether there is a non-zero amount of cloud.

I’ve defined the numbers based on my gut feel to capture what I think of as an annoyingly bright day outside, and I may adjust these numbers over time.