SIParCS 2022 - Shaniah Reece
Forecasting the COVID-19 Pandemic: Using Ensemble Data Assimilation to Enhance and Guide Models to More Reliable Predictions
Data modeling is a common approach employed by researchers and scientists to analyze and describe the behavior of a physical system. However, model, observational error and gaps in prior data are major limitations of this approach to accurately representing the state of a system and generating reliable forecasts. Data assimilation is a widely implemented computational technique in the Earth Sciences that can enhance model predictions by combining observations with prior forecasts to generate more reliable results. Most notably, it is used to initialize numerical atmospheric models for weather forecasting. This research demonstrates how the use of NCAR’s Data Assimilation Research Testbed (DART) on an extended SEIR model generates a better forecast of the COVID-19 pandemic than the SEIR model alone. We compared the outcome of data assimilation performed on data for four different countries around the world with different levels of COVID preparedness and responses to demonstrate its effectiveness. The findings of this research denotes that even for datasets with sparse or missing values, data assimilation can still effectively compensate for these gaps and create a fairly informative picture of the pandemic. It is hoped that this study will motivate the adoption of data assimilation to aid numerical models.
Mentors: Ben Johnson, Moha Gharamti
Slides and poster