VULCAN.ears · Volcano-seismic Unsupervised Labelling and ClAssificatioN Embedded in A Real-time Scenario

Volcano-seismic Unsupervised Labelling and ClAssificatioN Embedded in A Real-time Scenario

Active volcanoes have a big impact on the global economy and society. Current Volcano Monitoring (VM) is mainly based on the evaluation on the seismic activity, the main input used for eruption forecasting and early warning systems. The usual way to track this activity is to detect relevant events in a continuous data stream and to classify those events into groups or classes according to their physical origin, paying special attention to those classes acting as eruption precursors or involving population safety. In most Volcano Observatories (VOs) this Volcano-Seismic Recognition (VSR) is manually carried out by expert technicians, but, during a crisis, the VSR cannot be performed neither fast enough nor in a reliable way in order to quickly evaluate possible hazards. Even if is estimated that exist around 800 active volcanoes in the world, very few observatories have their own VSR system. Most VOs demand an automatic VSR solution to monitor the seismic precursors and classes which compromise casualties as lahars or phyroclastic flows. However, to deploy a customised VSR system takes time, requires qualified operators and labelled databases (DBs), which is not easily affordable. Even more, the built system won't be exportable to other volcanoes, and can become inefficient if the seismic activity patterns change. The ultimate goal of the VULCAN.ears proposal is to provide an automatic, universal, Volcano-Independent Seismic Recognition (VI.VSR) system, easily integrable into any VO to allow online hazard assessment by real-time analysis of the seismicity. To fulfill this aim, two main objectives have to be achieved: (i) To design robust, VI.VSR recognition models, by gathering lots of events from different volcanoes, efficiently describing each event class. (ii) To maximise the system applicability: disseminating it and integrating it into several VOs and eruption forecasting (EF) frameworks to obtain useful feedback from partners, encompassing the system improvement and usability. At the end of the -administrative- lifetime of VULCAN.ears project, evaluating VI.VSR technologies in real-case scenarios, we can draw these conclusions: 1. Volcano-Independent VSR is actually a breakthrough technology aimed to become the next step of seismic monitoring. The results achieved so far prove that our VI.VSR system is able to operate in real-time, recognising events of a given volcano using universal VSR models, built by universal DBs obtained in other volcanoes. 2. Modern VOs have a real need and interest in installing VI.VSR systems. Built from a scratch their own systems requires an effort that not all the VOs can afford. Therefore, the dissemination of a universal VI.VSR framework easy to be embedded in the VOs is a must in order to find a solution. The Open Access software distributed by VULCAN.ears solves this ‘must’. 3. The improvement of the VI.VSR system incorporating new universal VSR models depends on the international collaboration and transfer of knowledge among partners. This is an ongoing task, beyond the duration of this project, with a direct social impact, specially in developing countries with active volcanoes but few resources.

Текст от CORDIS, на английски · Данни: CORDIS, © Европейски съюз