Air Quality Algorithms (Part I)

I’ve had the Air Quality Index forecast and current reports open in a browser continuously for the past few days: Canada is on fire again and the smoke is blowing slowly across the continent toward the east coast. But the air quality reports from different organizations are so varied that it’s hard to know what you’re looking at.

wunderground.com

Catherine and I keep getting quite different numbers from Apple, which gets its data from Google via Breezometer, and Weather Underground (owned by the Weather Channel, in turn owned by … well, more on that another day). 

At various websites shortly after noon today, I found readings ranging from 43 to 107: Weather Underground posted an AQI of 99, “moderate” per the EPA, while its parent company had 107, “unhealthy for sensitive groups.” IQAir similarly reported 104. Meanwhile, Weatherbug, part of the US-based nonprofit Ground Truth Project, posted a reading of 43 on a 100-point scale, corresponding to a rating of “moderate,” while AccuWeather reported 64 on a scale of zero to 250, and called it “poor.”

When I tried to figure out why, I learned that different governments, regional agencies, and commercial weather organizations measure different data, consider varying pollutant levels dangerous, and use different algorithms to set the air quality index. It was a long trawl through various kinds of documentation at different levels of specificity and clarity.

The EPA — which provides clear and detailed documentation not yet censored by the current administration — tracks ozone, particulate matter (both PM2.5 and PM10), carbon monoxide, sulfur dioxide, and nitrogen dioxide, and indexes them against a scale of zero to 500. Below 50 is good, more than 150 ranges from unhealthy to hazardous, and between 50 and 150 is moderate (“acceptable,” though potentially risky for “unusually sensitive” people) or “unhealthy for sensitive groups.” The AQI is based on the highest-level pollutant at any given time. 

But neither Weatherbug nor AccuWeather uses the EPA scale.

Weatherbug lists levels of each pollutant it tracks: the same as those monitored by the EPA. Tapping on “i” above current conditions yields a pop-up window with a 100-point. At “moderate” air quality, Weatherbug advises that “unusually sensitive people” should take precautions; at “low” and “poor” levels, ‘people with lung disease …, children, older adults, and outdoor workers” should limit outdoor time and exertion.

AccuWeather uses data from Plume Labs, which has a lot to say about how scientific and accurate its weather forecasting is, but doesn’t provide much in the way of detail. Plume says its scale is based on World Health Organization recommendations about length of exposure:

Each category represents the amount of time it is safe to spend in that level of pollution. For example: one year (PAQI < 20), one day (PAQI <50), one hour (PAQI <100).

But the actual scale doesn’t mention time: instead stating that “sensitive groups” and “healthy individuals” are at risk at different levels of pollution. 

It’s hard to know what substances Plume includes in its index: advertising for the company’s app mentions particulate matter smaller than 10 and 2.5 micrograms, ozone, and nitrogen dioxide. An article explaining Plume’s AQI refers to volatile organic compounds, but not ozone. Plume’s air quality sensor, according to promotional material, also measures particulate matter smaller than 1 microgram.

Meanwhile, different countries and regions consider widely different levels to pollutions to be a problem. More on that in a future post.