Large-scale measurement of aggregate human colocation patterns for epidemiological modeling

Area of Science:

• Epidemiology
• Network Science
• Computational Social Science

Background:

• Epidemiologists require high-resolution spatial-temporal data on human movement and interaction for public health emergency modeling.
• Traditional data sources often lack the necessary geographic breadth and temporal granularity for studying phenomena like global pandemics.

Purpose of the Study:

• Introduce Colocation Maps, a novel spatial network dataset from Meta's Data For Good program.
• Describe the methodology behind constructing Colocation Maps and their unique features.
• Illustrate the utility of Colocation Maps in disease spread modeling and analyzing inter-regional interactions.

Main Methods:

• Colocation Maps estimate the co-occurrence rate of individuals from different geographic regions within the same physical space per minute, per week.
• The study details the data construction process, addressing assumptions on representativeness and contact heterogeneity.
• Demonstrates application in compartmental modeling for disease transmission.

Main Results:

• Colocation Maps provide granular insights into human spatial interactions, crucial for epidemiological studies.
• A key finding is that high colocation between regions can occur independently of direct travel volume between them.
• The datasets have been successfully applied during the COVID-19 pandemic for modeling and analysis.

Conclusions:

• Colocation Maps offer a valuable new data resource for understanding population dynamics in public health.
• The findings highlight the nuanced relationship between physical proximity and population movement.
• Further development of these datasets holds significant potential for advancing epidemiological research and preparedness.