Demonstrating how methods used in fields like biomedical research for measuring the randomness of observable data can be used to predict when volcanoes are close to eruption. Developing better forecasting systems is an urgent priority for civil managers and researchers, but these new systems will need to be easily deployed to areas with minimal resources. To accomplish this Dr Kostas Konstantinou, an Associate Professor in the Department of Earth Sciences at National Central University, Taiwan, and his team are exploring ways to extract meaning from large amounts of seismic data. Specifically, Konstantinou is taking a methodology commonly used in several other fields that is known as complexity metrics and applying it to seismic data collected from previous eruptions. In this way, they hope to find a signal in the ambient noise of seismic activity that can better predict eruptions. Seismological data is usually collected by installing sensors at different locations that can record ground motion along three directions (vertical, north-south and east-west). Even if there is no seismic activity, the sensors will still record a signal that is produced by the coupling of atmospheric disturbances to the solid Earth. This signal is called 'ambient noise' and it is a random signal since the position and strength of its source may change erratically in space and time depending on weather conditions. Currently, detecting minute changes in rock properties using ambient noise is one way in which volcanologists have tried to forecast volcanic eruptions. However, there is still much more to be learned about the complexity of this type of signal and this is the realm of complexity metrics. Adding such a tool to the forecasting toolbox is not without challenges and still needs some fine tuning. One such challenge for Konstantinou is to find a physical basis for interpreting the results. Knowing why the signal changes will be crucial for fine tuning the method and the team is employing other traditional tools like spectral and polarisation analysis to understand the physical process behind the observations.