The use of social media in disaster response is certainly one of the more interesting uses for the social web. Many of these approaches apply either crowdsourcing or big data, or sometimes a combination of the two in order to help emergency response teams do their jobs more effectively.
With text there has been a lot of analysis done by the likes of Google Flu Trends in order to determine what the search queries we enter mean for the spread of disease. CrowdOptic are attempting to do a similar thing with the many photos that are uploaded when a disaster hits.
Their software enables people to automatically zoom in on the area people were looking at when they take photos or record videos on their phones. It does this by triangulating the metadata contained in the media taken by the phone.
With an increasing number of photos and videos taken during and after disasters, it should be possible to upload them to something like CrowdOptic, who could then identify clusters of media based upon the location and timestamps. These clusters could then be identified by crowdsourced humanitarian volunteers to determine whether they highlight areas of disaster damage.
It's an interesting approach that has been trialled at UN level already, so is certainly one to keep an eye out for.