Spatio-temporal Epidemiology of Emerging Viruses

Infectious diseases are increasing becoming a problem of global concern. It is important to recognise spatial and temporal clusters of diseases and to predict their spread. This project aims to generate new insights into the spatio-temporal epidemiology of novel viruses by combining various freely available data from social media and official health data. Diseases exhibit fine-grained distribution patterns both spatially and temporally. The analysis of the geographical distribution of infectious diseases has so far mostly only been carried out over a large area (e.g. at the country level), with the focus more on temporal snapshots than on continuous monitoring over time. This has made it difficult to understand drivers of disease spread or to predict future changes in the spatial distribution of diseases. A finely resolved spatio-temporal spread of diseases on the level of urban agglomerations is still greatly lacking in scientific research. Therefore, this project analyses various freely available social media data (Twitter posts, Google searches, etc.) and official health data (e.g. COVID case numbers from authorities). With the development of new artificial intelligence (AI) algorithms that result in a higher information quality, knowledge can be gained that allows more precise predictions about the spatio-temporal spread of diseases. These methods are validated with infection data in order to analyse diseases such as yellow fever, dengue, corona, Zika or Chikungunya. The novelty of this approach is the inclusion of the temporal dimension into spatial epidemiological analysis. So far, maps have been important for visualising and understanding infectious diseases because information such as the risk of transmission, information about the population at risk of infection, or economic effects can be made visible. The spatio- temporal mapping of infectious diseases can also serve to better understand the research and observation of disease spreads and thus provide decision-makers with information on the basis of which risks can be visualised and decisions for the containment of the spread of the disease can be communicated.

The spread of infectious diseases is shaped by living conditions, environmental influences, and human behavior rather than occurring at random. In the GeoEpi project, we explored these relationships by systematically analysing geospatial data such as geo-social media traces and satellite imagery. The aim was to develop new methods and algorithms that use spatio-temporal patterns to detect infectious disease outbreaks early and in a transparent way. A key outcome of our work was demonstrating that geographic and digital data can act as early indicators of emerging hot spots of disease. Across several studies, we showed that geo-social media activity can provide indications of local outbreaks weeks before official surveillance data become available. We also found that these signals depend strongly on spatio-temporal context. For example, regional political attitudes and pre-existing socioeconomic conditions influenced how early or which symptoms or preventive measures people discussed online, which in turn affected how reliable geo-social media-based early detection of COVID-19 was. In addition, we demonstrated how satellite images can be used to identify water storage containers in urban environments, enabling insights into the spread of diseases such as dengue fever. Our results have been published in journals including Scientific Reports, the International Journal of Applied Earth Observation and Geoinformation, JMIR Infodemiology, and Frontiers in Public Health, gaining international visibility. These findings were presented and discussed with both researchers and the public at workshops held during the Harvard CGA Conference: From Geospatial Research to Health Solutions (2023) and The Geography of Digital Twins & Symposium on Spatiotemporal Data Science (2025). Building on this work, the follow-up projects Climate Agents and Health Detectives now continue to use and further develop the methods established in GeoEpi.