Model Case Studies About

A Multi-Region SEIR Model with Mobility

This research and the case studies listed below are currently in an early exploratory stage. They are not meant to convey specific predictions and are intended to be a thought exercise in understanding the value of multi-region SEIR models that explicitly model daily travel patterns. We welcome suggestions regarding specific experiments of interest and data to support these experiments.

Case Study: COVID-19 Epidemic and Commuter Travel in the Nashville Metropolitan Area

In our case study, we aim to demonstrate how the commuter Multi-Region SEIR model could help obtain more fine-grained COVID-19 projections in the State of Tennessee, particularly in the Nashville Metropolitan Area. We focus on the census tracts pertaining to seven counties, namely Davidson, Rutherford, Williamson, Sumner, Wilson, Robertson, and Cheatham. These counties surround the city of Nashville, county seat of Davidson, and hence are the most likely to exhibit significant recurrent commuter flows. See below for more details regarding the simulation including main assumptions and input data.


Loading data..
First time loading may take some time. Please wait.

Slide to adjust date


Click to select the population




Click to switch unit

Population Infected

red color scale


Recent research conducted at the Vanderbilt University Medical Center recounts the state of the COVID-19 epidemic in the State of Tennessee. The report makes projections for the state of the epidemic (and the peak number of hospitalizations) under three different scenarios. In the status quo scenario, R0 remains roughly at 1.4 and the hospital capacity becomes stressed at the peak. In the optimistic scenario, R0 falls below 1 and hospital capacity is sufficient to treat all patients needing inpatient level care. In the scenario that lifts social-distancing measures, R0 rises beyond 2 and the healthcare system becomes overwhelmed at the peak. R0 remains around 1 as of April 23, 2020. They emphasize the need to maintain it at that level or below to prevent a second wave of infections.

Main Assumptions and Input Data

We adhere to peer-reviewed estimates of model parameters (i.e., latency period, transmission rate, reporting rate, and recovery rate) from the experience with COVID-19 in Wuhan, China shortly after social-distancing measures were introduced. Admittedly, the real parameters in the case of the Nashville Metropolitan Area may be different, even if similar disease control strategies were put in place. Our case study is meant to provide a rough projection of the state of the COVID-19 epidemic, not a high-fidelity prediction, when mobility patterns are explicitly considered. More specifically, it is meant to highlight the unanticipated network effects that may arise due to recurrent commuter travel.

We obtain commuter flows and census tract data from publicly available US Census Bureau datasets. We assume the working population may become exposed for 8 hours while at work and for 8 hours while in their home area but not at home (e.g., shopping, dining). We assume the non-working population stays in their home area but not at home during the day. There is no exposure for 8 hours during night time.

We scale the transmission rate β of each census tract by the ratio of its population density and that of the entire metropolitan Nashville Metropolitan Area. In our experiments, we will explore the effects of changing the transmission rates at a county level (e.g., reflecting county-level changes in social-distancing measures). We scale the transmission rate of the involved census tracts accordingly.