This one is going to be a little nerdy.
It's been a while since I've written about emerging mobile air quality sensors such as the Air Quality Egg or the AirBeam. To recap what I've written previously:
1. Regulatory sensors cost a lot of money to install and operate, about 500K for the hardware plus operating costs (~50K/year). They are hard to site and typically fill up a small ATCO trailer, and so are not easily portable.
2. On the up side, they are sensitive and the standard to which all other, non-regulatory sensors are compared.
3. In contrast, new emergent sensors such as the Air Quality Egg and the AirBeam offer a much less expensive alternative (>200$), that are portable and consumer friendly. A citizen could, for instance, set an Air Quality Egg up on their downtown balcony and feed open air quality data to the web.
4. The downsides to emerging air quality devices are (a) we are uncertain about the data quality, and (b) how they will perform in the cold.
Working with the City of Edmonton and the Alberta Central Airshed we tested the AirBeam PM2.5 sensor adjacent to the Woodcroft Air Quality Station neat the Telus World of Science. We ran thew AirBeam sensor for 3 days until the fan started to sound funny, and the BlueTooth connection to the Samsung 5s (the data storage 'module') was lost. I created the following two graphs in Tableau:
1 minute average AirBeam data as compared to a co-located regulatory PM2.5 sensor at the Woodcroft Air Quality Monitoring Station.
1 hour average AirBeam data as compared to a co-located regulatory PM2.5 sensor at the Woodcroft Air Quality Monitoring Station.
Great. The graphs look good.
But I had to run some stats to really draw any conclusions from the data. I normalized the data to minute time-series (the AirBeam records per second) and checked that I had a similar number of data points for both the Woodcroft and AirBeam data (with 16 points of each other), I ran a linear regression (NOTE - **I am not a stats guy. If you have a suggestion, let me know in the comments section**) with an R Squared of 0.621.
Scatterplot of AirBeam vs Woodcroft PM2.5 data collected over 3 days and normalized to 1 minute averages.
AirBeam has some testing results (here and here) suggesting that an R2 0.7 to 0.9 could be expected. The testing that produced an R2 of 0.9 or better (meaning that there is a high correlation between the baseline data from a regulatory (or similar) sensor and the AirBeam data) were conducted under ideal, indoor, conditions. Those of 0.7 were in more real world outdoor conditions.
An email form the manufacturer of AirBeam states that "High humidity and fog will falsely elevate the AirBeam's measurements.". They did not mention cold.
I am optimistic with these results. They indicate that the AirBeam PM2.5 sensor has some potential for citizens to engage in environmental monitoring. More on my concerns with the sensors, citizen science and open data later...