Mapping Urban Mobility for Flu Forecasting

Bringing together new spatial datasets to refine and optimise the model for urban forecasting

Influenza infections are major contributors to losses of life, decline of wellbeing, and economic losses worldwide. The influenza season in Australia usually peaks between June and September, but the severity and length of the epidemic varies considerably between years. Understanding, and ultimately predicting the severity, timing and geographic spread of the epidemic is of great importance to better assess the risks to the vulnerable population and effectively manage the response through the public health system.

Effective influenza epidemic prediction systems can save lives and greatly reduce the economic impact of the epidemic. Vast volumes of data are generated within the networked society. Each connected devices transmits data, with many storing their location. This data is enabling new understandings of how people move across the city. This project is investigating how these large-scale tracking data can improve the modelling of influenza infections within metropolitan populations.

Current epidemiological forecasting in Australia and worldwide focuses on the prediction of the timing and possibly the intensity of the epidemic and while great progress has been made to fine-tune the model’s parameters for the capital cities. Such models are, however, still too coarse for effective planning, response and targeted health interventions, such as selective school closure and work-from-home advice.

Incorporating a spatial element to the data allows for the refinement of modes to capture disease transmission across regions. A key challenge in constructing these models and accurately describing the transmission characteristics of a real-world population is to understand how populations in different regions interact and thus facilitate disease transmission.

Until recently, the best source of data about population-size movements in a metropolitan area have been travel surveys, providing detailed information about the movement of a small sample of respondents on a single day. For example, the largest such effort, the Victorian Integrated Survey of Travel and Activity (VISTA) captures only about 14,000 across 5,000 households in Greater Melbourne.

Recent proliferation of apps that track smartphones users resulted in rich data about the mixing of a large population segments at an unprecedented resolution. It is therefore conceivable that these data may be used for spatial models that accurately characterise disease transmission in urban populations.

This project will bring together these new spatial datasets to refine and optimize the model for urban forecasting.

Research Team


Seed Funding 2016

Forecasting flu outbreaks by Liz Wells and Holly Bennett, Pursuit, 13 February 2018.

A week is a long time in politics, Rob Moss, 12 June 2016.

Computing helps the study of infections on a global and local scale by Jodie McVernon, Joshua Ross, Kathryn Glass, Lewis Mitchell, Nicholas Geard and Rob Moss, The Conversation, 6 June 2016.