Background
“The Cosmology group at the University of Auckland started working on this during the first Level 4 Covid-19 lockdown in 2020 as a way of contributing to the national pandemic response. This work was largely completed while Aotearoa New Zealand was at lockdown levels 3 and 4 and mobility restrictions were in place. A large amount of our day-to-day work is computational, so we were well placed to contribute to solving Covid-related problems using numerical methods.
As Covid-19 was still in its early days, little was known about how it was spreading and there was a lack of detailed tracking data available. We therefore sought to develop a tool which could be used to predict which areas of New Zealand may be at high risk for infection.
The spread of transmissible disease is known to be tightly correlated with population movements, so we wanted to develop a means by which to infer daily travel patterns and thereby predict which people were likely to come into contact with one another during the course of a normal working day.
We made use of aggregate spatio-temporal data collected by telecommunications networks. This data provides total counts of contactable cell phones in over 2000 small geographical regions per hour, and cannot reveal any personal information about individual users.
Using the telecommunications data, we implemented an algorithm to infer the numbers of people who moved into and out of any given region per hour, and to determine which region they were likely to have travelled to or from. We found that this algorithm can yield useful predictions regarding population movements where precise data is unavailable.
As it turned out this work was not needed as part of our eventual national response. However we significantly improved the performance of existing algorithms used to analyse this data and developed a better understanding of their properties. We anticipate that our implementation may serve as a useful public health tool in the future, and we have identified various avenues for further improvement.”
— Professor Richard Easther & Dr Emily Kendall.