Sunday, August 02, 2015

Crowdsourcing Air Quality


Last year as I was contemplating a Silicon Valley job, I was brainstorming possible data science projects. I thought it would be cool to try to estimate air pollution levels from existing data sets. (Growing up in the Central Valley, I think conventional air pollution is a huge looming environmental health problem that doesn't get enough attention.) I figured you might be able to extract at least some air quality information from geo-tagged and time-stamped online photos of the sky. I ended up getting a different job, but found some pretty sweet ongoing projects.

Naturally, a quick search showed that someone had already thought of this idea. SkySnapper is a fairly recent project that enables people to upload images of the sky with the goal of estimating AQ, although they don't seem to have done much with the concept so far. More fully developed is this project from a research team at USC. They have already developed an Android app and done the hard work of building a mathematical model in order to correctly extract air visibility information from a photograph (pdf). Although their concept seems sound, it doesn't appear that many people have submitted data through their app.

Smartphone photos aside, the broader world of air pollution monitoring is also being transformed by big data and citizen science concepts. A group of Berkeley undergrads are developing personal PM2.5 monitors -- a very cool idea, but apparently still in development. Similarly, Smart Citizen helps people set up environmental monitoring networks in their cities using Arduino-controlled hardware. And the startup Aclima has been getting some good press lately collaborating with Google to study indoor air pollution and put monitors on the Google Street View cars. Cool stuff, but of course the big challenge will be scaling up these concepts and understanding the data well enough to make meaningful contributions to science or policy. Something for when I get a chunk of free time...

Update: also this.