Tracking Patterns of Pandemics from Space: AI & Machine Learning to Help Automate Outbreak Prediction

Data from Earth-gazing satellites is key for scientists hoping to understand and track disease outbreaks including the unfolding coronavirus pandemic.

Why it matters: A satellite’s view can allow health researchers to understand the context of an outbreak in a way other tools cannot, and it has the potential to help scientists predict when and where the next infectious disease outbreak may occur.

What’s happening: Companies like Planet and Maxar are able to track empty parking lots, roads and businesses to understand the economic impacts of the pandemic.

• Satellites are also able to see the flow of people from one place to another, which could allow scientists to trace outbreaks of the disease in densely populated areas.

Details: Scientists are in the midst of collecting data about possible ecological factors that may have led to the current pandemic.

• The data may also feed models that can help explain how the coronavirus transmits between people and spreads around the world.
• “We’ll look at that data, and we’ll really be able to map out, not just the spread of the disease from the migration patterns of people but the recovery process,” public health researcher Timothy Ford of the University of Massachusetts told Axios.

The big picture: Satellite data can also help scientists predict when and how new viruses may jump from animals into human populations.

• Photos from space reveal where people are encroaching on animal habitats.
• Satellite photos are also used to track vegetation, rainfall and other factors in the four-corners region of the U.S. that put the area at a high risk for hantavirus outbreaks.
• Previous studies have also shown how environmental factors that play into outbreaks of diseases like Ebola can be seen and predicted using data gathered from orbit.

Yes, but: Robust models require multiple data points and it will likely take more than a dozen outbreaks of these types of coronaviruses to accurately predict when and where the next might occur.

• “I can definitely say to you that we don’t have enough information about SARS-like coronaviruses yet, and hopefully we won’t because it would take 15 or 20 more such emergence events,” Townsend Peterson, a researcher at the University of Kansas who uses remote sensing data, told Axios.

What’s next: Scientists hope to eventually use machine learning and AI to help automate outbreak prediction by quickly analyzing the wealth of data beamed back from orbit.

• “With the advances in computational capacity and in modeling, people are coming up with new ways to look at these big data sets,” Lindsay Campbell, a professor at the University of Florida, told Axios.
• Governments and commercial companies are also making their data available to researchers so that scientists can find new ways to understand how environments change over time, possibly contributing to health outcomes.