This project tests the potential of geo-located data from Twitter as a covariate in the production of population maps in Indonesia. Indonesia showed one of the highest Twitter usage rates worldwide, thus the data potentially provides for a high spatial and temporal resolution that is made visible in 1x1 kilometre grid cells over a time period of 2 months in 2013. This data was used as a covariate in a previously forests-based census disaggregation method.
The comparison of internal measures of accuracy and external assessments between models showed that population mapping gained in accuracy using the geo-tweet densities as a covariate layer. The work highlights the potential for such social media-derived data in improving the understanding of population distributions and offers potentials for more dynamic mapping with this new data source. Potentially, this approach can be used with further social media data – such as from Facebook, Instagram, and LinkedIn –, and in particular in large areas with limited available census data.
(Picture: © Jean-Marie Micaud, IOM)