-
Notifications
You must be signed in to change notification settings - Fork 3
February 26, 2018
Can we identify evacuation / shelter-in-place behavior easier with geo-tagged Twitter activity.
Of all the data we've looked at, I've identified a trend change in who is posting geo-located information (and how). Hurricane Sandy saw mostly younger people on Twitter posting from within Twitter with the geotag.
Hurricanes Matthew and Harvey/Irma/Maria (HIM) see a different practice altogether: geolocation. I define this as the difference between having a "geotag" like a latitude and longitude and instead locating to the nearest location that means something to people. This makes sense from an indexing/marketing perspective for textual searches as oppose to spatial queries. As such, we have a lot of tweets from "Miami Beach" in the Irma collection or "Pearland, TX" in the Harvey collection.
- People geo-tag differently, the platforms have changed dramatically.
- Geotags are no longer trustworthy
http://tweetsonamap.com/evacuation/events/#3.42/26.94/-81.47
This user's geolocated activity appears to be exclusively cross-posted from the untapped app which allows people to log the beers that they drink. Therefore, he has discrete clusters at specific places (like 4square). I believe, however, that he has tagged his house as "The Fishtank" He logs beers he drinks at home there and the clustering algorithm agrees.
This user's home location is not actually a home, but rather the town called "Iowa Colony, Texas". This is IN the evacuation zone, however. They appear to shelter-in-place, but it is difficult to tell, they may just be tagging things in the area.
The evacuation inspection tool, tweetsonamap.com/evacuation/ is the 4th iteration of an exploratory tool that I've worked on for this work.
Getting the tweets into this tool goes like this:
- Get Tweets
- Organize tweets into a
geojson. This scales for both contextual streams as well as single-user information. - Using Node to do the geospatial clustering calculations to clean the tweets.