ORNL study quantifies effects of pandemic restrictions on population-based activity patterns

ORNL study quantifies effects of pandemic restrictions on population-based activity patterns

May 5, 2022

Using GPS data from location data company SafeGraph, researchers at Oak Ridge National Laboratory (ORNL) have empirically quantified the shifts in routine daytime activities—such as getting a morning coffee or takeaway dinner—following "safer at home" orders during the early days of the COVID-19 pandemic. These insights, published in the Journal of Transport Geography, could help officials better understand traffic patterns and supplement the response to national security situations and other emergencies.

Safer at home orders across the nation at the start of the pandemic significantly changed how people went out into public throughout the day. Schools shifted to virtual classes, employees transitioned to working from home, and many extracurricular activities were cancelled. Yet critical workers across industries continued to report to work. Retail stores, hospitals, restaurants, and central economic businesses remained open to both employees and patrons.

Opportunities for visiting public places are often coupled with times of the day, such as grabbing a coffee while dropping the kids off at school or stopping for groceries on the way home from work. When these time-driven catalysts are removed and people can choose when they want to do these tasks, such as during safer at home orders during the COVID-19 pandemic seen across the nation, researchers have an opportunity to gain a new perspective on how people make decisions with their time and movement.

ORNL researchers studying human dynamics sought to use behavior changes during the COVID-19 pandemic to understand the intersection of human agency, or free choice, and human predictability. Though people have a choice of how they spend their time, their behavior is often driven by convenience and necessity.

“Human agency and behavior impact individual and collective responses to surprising events — things like COVID-19 or a climate hazard like wildfires or hurricanes,” said Christa Brelsford, a research scientist in the Human Dynamics section at ORNL. “The research community is lacking in quantitative methods to think about aggregate responses to these events — from individuals, communities, and societies.”

When planning for response to emergencies, authorities need accurate data on populations at risk. Knowing where people are over the course of a day can improve response efforts to those most impacted. Understanding how people responded when business and school interactions suddenly shifted into their residences gives researchers the opportunity to identify important patterns of human behavior at the country, state, and county levels.

Observable Shifts in Behavior in Public Places

Kevin Sparks, a research associate in ORNL’s Human Dynamics section and first author of the paper, noticed data from various sources stating public places were seeing a decrease in the number of people compared to pre-COVID-19 levels, but he wasn’t seeing a synthesis of how mobility changed. If the dial is turned down on the number of people going out to places, he wondered if their behavior patterns remained the same.

Sparks, together with other researchers across ORNL’s Geospatial Science and Human Security Division, sought to quantify how behavior changed during the unprecedented pandemic. Since the pandemic had a relatively clear beginning in the United States, in March 2020, researchers decided to compare daily movements in 2019 with 2020, with an additional analysis of the initial safer at home period of March and April 2020.

Using publicly released data from commercial data provider SafeGraph showing aggregate, anonymized GPS location markers at millions of points of interest, researchers could graph at what times people were most active during a 24-hour period. In 2019, for example, the peak morning hour was 8:00 a.m. In 2020, the peak morning hour shifted two hours to 10:00 a.m., showing people started their days later when school and work start times didn’t dictate early morning get-ups. Where people in 2019 were still quite active in evening hours and showed a peak evening active time of 6:00 p.m., this peak disappeared in 2020.

“We saw the largest differences in temporal and geographical behaviors during the morning and evening in 2020. With an increase in remote work and virtual schooling, we can see how people’s activities changed in response to the reduction of normal commutes,” Sparks said.

Though Sparks wasn’t surprised to see the changes to start times or routines after the pandemic disrupted lives across the country, he was surprised to see regional differences across the United States. Counties on the east and west coasts experienced the greatest shift in temporal behaviors, where the middle of the country didn’t have so dramatic of a change. The researchers attribute this to several factors including industries where work from home was feasible or impossible, access to reliable internet for virtual work and schooling, and political sentiments on the pandemic.

Building Compute Infrastructure for Research

Commercially available datasets are becoming a common method for scientific research. SafeGraph offers several mobility and point of interest datasets available for purchase. These datasets are large and rich in content, requiring significant compute infrastructure to ingest, catalog, query and analyze the data to support research across the Lab.

“COVID-19 shook the core assumptions in population modeling. We now look to other types of data to tease out what people are doing and where,” said Jessica Moehl, ORNL research scientist and spatial modeler.

The IT resources available to ORNL researchers is built on a system of scalability and connectivity. Across a division with multiple types of research projects of varying sizes lies compute infrastructure capable of offering virtual machines and cloud storage to flex when the unexpected happens. Moehl and her fellow research scientists coordinate resources for quick data ingest and indexing in the compute environment. Using SQL to write algorithms on a remote server, researchers can then virtually connect with the data, run mathematical computations, and get an output for analysis.

The compute environment is the backbone for researchers to make sense out of enormous amounts of data and allows them to answer questions previously unanswerable. The robust, scalable nature of ORNL’s infrastructure allows scientific researchers with a social science interest to challenge assumptions about how people live. Moehl and team leverage that compute power to stand up the database quickly and get to the scientific analysis faster.

Knowledge for Post-Pandemic Planning

When planning for response to natural and man-made emergencies, authorities need accurate data of populations at risk. Knowing where people are over the course of a day can improve response efforts to those most impacted. Understanding how people responded to stay at home or safer at home orders across the country illuminates important patterns of human behavior at the country, state, and county levels.

ORNL’s expertise in human dynamics and scalable infrastructure is enabling new data sets to inform our understanding of human behavior. Collaboration among researchers with diverse backgrounds is helping find meaningful patterns of human behavior. The pandemic has given researchers a unique glimpse into the collective patterns of behavior when behavior changed across the world. Businesses and authorities use ORNL’s research not only to understand patterns of human mobility throughout the course of a day but also to plan better for future pandemics and other unexpected events.

This research was supported by U.S. Government funding sources including DOE Office of Science through the National Virtual Biotechnology Laboratory, a consortium of DOE national laboratories focused on response to COVID-19, with funding provided by the Coronavirus CARES Act.

Read more: https://www.ornl.gov/research-highlight/how-behavior-communities-quantif...

Read the study: https://www.sciencedirect.com/science/article/pii/S0966692322000187