September 21, 2022
Journal Article

Near real time monitoring and forecasting for COVID-19 situational awareness


In the opening months of the pandemic (March - June), the need for situational awareness was urgent. Traditional forecasting models such as the Susceptible Infectious Recovered (SIR) model were hampered by limited testing data and key data on mobility, contact tracing, local policy variations and so forth would not be available with reliability for months. Reliably available were new case counts from John Hopkins University and the NY Times. Using these data, the challenge was to develop a robust monitoring capability in support of U.S. decision making. The Department of Energy’s National Virtual Biotechnology Laboratory responded by developing the COVID County Situational Awareness Tool (CCSAT). The result is significant in three ways. First, we developed a retrospective 7-day moving window map of county level disease magnitude and acceleration that smoothed daily variations and categorized counties with intuitive labels such as “high but decelerating”. Secondly, we developed a Bayesian model that reliably forecasted county level magnitude and acceleration maps for the upcoming week based on population and new case count data. Together these formed a robust operational update delivered weekly to the U.S. government. In this paper, we provide CCSAT details and apply it to a single week in June 2020.

Published: September 21, 2022


Stewart R., S.H. Erwin, J. Piburn, N. Nagle, J. Kaufman, A. Peluso, and J. Christian, et al. 2022. Near real time monitoring and forecasting for COVID-19 situational awareness. Applied Geography 146. PNNL-SA-173905. doi:10.1016/j.apgeog.2022.102759