You've got to get the secret watchlist. The Environmental Protection Agency for years has maintained a secret watch list of Clean Air Act violators — "recidivist and chronically noncomplying facilities whose violations have not been formally addressed by either the state or EPA." For years, EPA has kept it under wraps.
There is a news story behind many of the 464 facilities on the list. This year, the Center for Public Integrity's iWatch News and National Public Radio got the list using the Freedom of Information Act and during November 2011 they published a feature package: "Poisoned Places: Toxic Air, Neglected Communities," which seems sure to win major prizes. But they did not tell all the stories. They left some of them for you.
The secret watchlist is merely a distillation of purported bad actors from a much larger data set. You can mine further noncompliance stories from some bigger EPA databases. You can localize the toxic air impunity story to your own area.
- For example, you could go to EPA's ECHO  (Enforcement and Compliance History Online) database, and do a query on Clean Air Act violators just in your own state, restricting the search only to the air program and checking boxes to focus only on high priority violators who are still in violation and whose compliance status is unknown. If you are interested in environmental justice, you could restrict your search to show only facilities within a 3-mile radius of communities with minority populations of 50% or more.
- There are plenty of heart-rending stories that involve pollution media other than air, but toxic air emissions often capture reader and viewer imaginations. Another data tool that can add deeper insight into actual pollution hazards is the Risk Screening Environmental Indicators  model, which you have to download and run on your own computer. It is based primarily on the Toxics Release Inventory. 
A few caveats: Don't assume what's in the databases is accurate or current — check it out. Don't assume that all violations are substantial or that they all endanger people. Databases are good tools for helping you find stories, but they are no substitute for shoe-leather reporting, which must be part of the project.