Purdue University, the Department of Homeland Security, and National Weather Service are all working together to use social media to improve the public's response to severe weather.  They hope to develop software to mine tweets for data to be used to warn people of a dangerous storm approaching.

Those working on the project realize the danger of relying on just tweets since not every tweet is accurate.  However, there's an algorithm developed to attempt to weed out bad information by looking how often a Twitter user's messages are retweeted and by determining if this person has been accurate in the past.  

Presently, the current tracking program is still better at after-the-fact analysis than it is at real-time warning.  As it improves, first responders hope they can target areas most in need of help, based on tweet content.  Those in the direct path of the storm could get a few extra seconds to take cover as well.