why number of blue sky days is a terrible metric

In my post summarizing Beijing’s 2009 air quality, I noted that although the number of Blue Sky Days increased from 2008 to 2009, there was a very minimal reduction in annual average particulate matter concentration (the reduction has now been confirmed by the Beijing EPB to be just 0.8%). I then called the “number of Blue Sky Days” metric scientifically meaningless.

Here is a simple example showing clearly how and why the use of number of Blue Sky Days can distort the reality of air quality:

Suppose you take two sets of two days, and you wish to evaluate which period had better air quality. Here are the data you have:

Day 1 and Day 2: API* is 100 on both days.
Day 3: API is 101.
Day 4: API is 1.

Over the period Day 1-2, we have an average API of 100, and a total of 2 Blue Sky Days.
Over the period Day 3-4, we have an average API of 51, and a total of 1 Blue Sky Day.

If you are judging air quality by “number of Blue Sky Days,” you would conclude that the air quality was better on Days 1-2. On the other hand, if you are judging air quality by average concentration of pollutants people are exposed to, you would judge that Days 3-4 were much better. Actually, because the normalization from pollutant concentration to API is non-linear, in this example, the average pollution level of Days 1-2 could be up to three times higher than the pollution on Days 3-4, and yet this period is judged as being better, because it has more Blue Sky Days.

This is what I mean by a scientifically meaningless metric.

*Reminder: API (air pollution index) is a 0-500 normalized measure of the pollution people are exposed to; a Blue Sky Days is a day with API of 100 or below.

Leave a Reply