MIT is delving into data mining to tackle air pollution. They are now employing movements from cellphones to estimate exposure to PM2.5 and other air pollutants. The project is called Senseable led by Marguerite Nyhan, whose Tedx Talk I just saw (linked below). Consider the equation Risk = Exposure x Hazard, which requires both the hazard (PM2.5) plus exposure (activity data) to be known.
I didn't know you obtain cellphone data this way; it seems there exists a thing called Call Detail Records (CDRs) providing millions of anonymized records of mobile phones and wireless devices. From Dr. Nyhan's publication, what caught my eye was "To the author’s knowledge, a study applying extensive spatiotemporal population mobility estimates from mobile phone data in the assessment of population exposure to environmental pollution over a substantial study domain has not been conducted previously".
Still not quite sure where the CDR data comes from initially, but I think this may be a good starting point.