Collaborating with the Center for Geographical Analysis, I collect about 9.5 million geo-tagged tweets around the world each day. Using the data, I examined human mobility perturbation across eleven tropical cyclones and, later on, twenty four different natural disaster events, including earthquakes, snowstorms, wild fires, that affected thirty major population centers (New York City, Tokyo, Manila, Boston, etc.) across the world from 2012 to 2015. My research found that human mobility possesses inherent resilience. I discovered that there was a high degree of correlation between human mobility during the attack of natural disasters (i.e. the perturbation state) and during normal days (i.e. the steady state), suggesting the potential predictability of human mobility perturbations.
Figure 1. Urban Travels of New Yorkers during Hurricane Sandy.
Each dot is a place visited by Twitter users, each line is a trajectory of an individual. The green dots show where fatalities happened, and red areas are the evacuation zones.
Figure 2. Human Mobility Distributions in Multiple Cities under the Influence of Extreme Events.
Human mobility follows the truncated power-law distribution in the 20 cities we examined before (green), during (red) or after (blue) tropical cyclones.
Figure 3. Commute Simulation around NYC based on Twitter Data and Google Maps.
A data-driven simulation test platform. All the locations in the NYC are retrieved using Twitter data and all the commuters' locations outside of NYC are from the U.S. Census data. The commute routes are generated by using Google Maps Distance Matrix API. The colored areas are the six evacuation zones designed by NYC government.