Further Refinements

2013-2014 Analysis

2015-2018 Updates

1. Text to add:
   8,601 geotagged tweets by 34 bots were removed using lists of known-bots. Add to our bot list on Github
2. Included (add update): "Movement patterns are confined to a circle encompassing city boundaries based on the city center provided by a Google Search. The analyzed data does not include long-distance travel steps that start or end outside the urban boundary radius."
3. Added correlation to Hypothesis 2:
   "Deviations from normal daily power-law mobility patterns correlate to wind intensity."
4. Added correlation to Hypothesis 3:
   "Disruptions to normal daily radii of gyration distances correlate to wind intensity."
5. "Loss of Mobility Resilience" is a more accurate description that eliminates the barage of questions one encounters with simply "Loss of Resilience". "Mobility Resilience" is the term used in the recent 2019 study.
6. Since Typhoon Haiyan's impact on Tacloban is the basis for the "Loss of Mobility Resilience" finding, more attention could be given to telling the story of this storm's impact. Compared to one year later, how did mobility patterns and Twitter usage change? (Use Rammasun Philippines data in table ph2014ram.)  Where did the residents relocate to? (Fisherman were given new jobs inland.)

Future papers could include...

1. Additional areas for measuring transitions to a net positive norm:
   
   A. Reduced driving distances, more walking and localized transactions
   B. Reduced use of vehicles that emit carbon
   C. Reduced use of petrochemicals in production
   D. Reduced pollution and carbon emissions
   E. Reduced use of vehicles with large passenger compartments
   F. Reduced use of facilities in harm's way (near low shorelines or within tight quarters)
   

2. Tweets captured directly to a NoSQL database with relational query support, such as Spark (an extension of Hadoop).
3. Use NOAA weather data API or other data service for city wind data. (Online news reporters do not always distinguish between sustained winds and gusts.) Sample using NOAA 3-hour wind data: model.earth/storm/tracker/
4. Track 10-minute sustained, 2-minute sustained and gusts using weather data feeds.
5. Generate city boundaries dynamically using city shape files, and/or use cones for storm paths similar to Yan's study.
6. Use 3-hour time frames to match storm movement data.
7. Compare impacts using 3-hour NOAA time periods to distinguish movement before and after strong winds.
8. Analyse data using shorter time-frames coinciding with peak winds.
9. Limit origins of travel steps to smaller communities closer to the shore to see if hypothesis 2 is then supported.
10. Include effect of steps entering and exiting city boundary areas.
11. Investigate effect of travelers that travel through cities (like Tampa and Jacksonville) and travelers that travel to cities during storms (like San Juan).
12. Add granular tracking of smaller communities to capture areas closer to coast, but allow for long-steps leaving and entering smaller boundary areas.
13. A more robust analysis of Tacloban could compare tweet levels across multiple years while documenting how the end of automatic geotagging in 2015 impacted changing geotag levels across multiple cities using data from the Philippines, Japan and Southeast US.
14. Add more comprehensive mobility data from a cell phone company or popular mobile app.
15. Establish the approximate home location of individuals (using weekend and evening posts) to include steps when individuals did not tweet from home before traveling to work and other destinations.
16. Include longer distance travel step segments by including 10 to 50km, 50km to 200km, and over 200 km steps to distingish commutes from regional long step travel. Experiment with different split points (between 40km-100km and 200km-300km) to find points that evenly cluster normal step distances.
17. Anonymize using zipcodes to provide real-time reporting with demographic details.
18. Create data exchanges with other university departments, power companies, volunteer organizations, and sustainable development initiatives.