A few weeks back I animated 24 hours of Twitter data, mapping every tweet with geographic coordinates. I found it interesting to observe not only where geo-tweets occurred but also where they were absent. The video below tentatively explores the relationship between population density and where people are tweeting.
This animation begins with a view focused on Wellington, New Zealand at midday on November 19, 2010 (NZDT time). The camera moves west across the Earth’s continents before ending at Honolulu, Hawaii.
Things to keep in mind:
- The more red a place appears, the greater the population density.
- Each transparent blue circle is a single tweet. The darker the blue, the greater the number of tweets.
- The camera moves west at a constant rate.
- At any given moment, the local time at the place in the center of the map is always roughly midday on November 19, 2010.
- The further west a place is from the map’s center, the earlier in the morning it is there. The further east, the later in the afternoon. (I really should write some code to map areas of daylight and darkness.)
[vimeo width=”680″ height=”510″]http://vimeo.com/17741598[/vimeo]
To create the video I downloaded the 2010 Gridded Population of the World dataset from Columbia University’s Socioeconomic Data and Applications Center. I imported the population density data into a ArcMap (a commercial GIS package), applied some symbology and exported the dataset as a PNG file. I used Tom Patterson’s Natural Earth data to fill in landmasses with minimal population that are not represented in the Gridded Population of the World (e.g. Greenland, Antarctica). I wrapped the density image around a map with an orthographic map projection using Matplotlib’s warpimage() function.