Abstract.  Microblogging  is  an  important  source  of  information about  what  is happening in the real world. In this work, we propose a novel approach for real-time  event  detection  targeting  accident  and  disaster  events  (ADEs)  using microblogs from Sina Weibo. Our aim is to detect out every microblog which reports a real-world occurrence of a target event from the microblog stream. We formulate  the  event  detection  problem  as  a  classification  problem  using microblog-based  features,  linguistic  features,  content  features,  and  event  features. We propose a street-level location extraction method based on the textual content  to  cooperate  geo-information  extraction.  In  order  to  deliver  fresh events, we use a temporal analysis method to filter away past events. We compare  our  method  with  two  state-of-the-art  baselines  on  event  detection,  and achieve improvements in both precision and recall.