Fig 1.
System architecture of the ODT Flow platform.
Fig 2.
Illustration of origin-destination-time (ODT) cube for big OD data query analytics.
Fig 3.
Four application scenarios exemplifying how ODT cube coupled with the traditional cube operations and a scalable parallel computing environment can help analyze big mobility data.
Table 1.
Statistics of the derived daily OD flows from Twitter data and SafeGraph data.
Fig 4.
The user interface of the interactive spatial web portal.
The world country and world first-level subdivision boundary data are derived from https://gadm.org.
Fig 5.
SafeGraph-derived county population flows to New York County from (a) 03/08/2020 to 03/14/2020 and (b) for the following week (03/15/2020 to 03/21/2020). Twitter-derived in & out flows between England, UK and other first-level administrative units in the Europe area for (c) 01/01/2020 to 02/29/2020, and (d) 03/01/2020 to 04/30/2020.
Fig 6.
County-level population flows from 01/01/2020 to 01/05/2020 derived from (a) Twitter and (b) SafeGraph. Note: For SafeGraph-derived mobility data, only flows with aggregated device number great than 20 within the selected time period are displayed for performance consideration.
Fig 7.
Daily population movement in different geographic scales.
(a). Intraflow for Spain (top line) and Argentina (bottom line) in 2019 and 2020; (b) Inflow for New York County, U.S. in 2019 and 2020; (c) Intraflow for a census tract in Columbia, South Carolina (mainly located within the University of South Carolina) from 01/01/2019 to 02/24/2021; (d) Interflow (In&Out) for a census tract in a residential area of Columbia from 01/01/2019 to 02/24/2021.
Fig 8.
(a) Extract and download the Twitter-derived mobility data in ODT flow explorer at the world first-level subdivision from 01/01/2020 to 03/31/2020; (b) visualize the data in kepler.gl as a flow map; (c) visualize the origin locations as a point density map.
Fig 9.
The workflow of human mobility intensity trends visualization.
Fig 10.
Visualization results of human mobility trends from the workflow.
Fig 11.
The correlation analysis of human mobility and COVID-19 infection cases in the U.S.
Fig 12.
The correlation coefficients between outflows from New York and COVID-19 infection cases.
Fig 13.
Sample codes of reading the boundary file (for mapping) and obtaining human mobility data using the ODT Flow API.
Fig 14.
Monthly mobility change rates in 13 French administrative regions.
Fig 15.
Daily COVID-19 new cases in France in 2020.
(Data source: World Health Organization Coronavirus Dashboard, https://covid19.who.int/info/. Accessed on March 19, 2021).
Fig 16.
Python code of extracting the world first-level administrative OD flow matrix using the ODT Flow API and visualizing the data as an interactive flow map with kepler.gl.