V41B-4814:
Expert Systems for Real-Time Volcano Monitoring

Thursday, 18 December 2014
Carmelo Cassisi1, Flavio Cannavo1, Placido Montalto1, Pietro Motta2, Giovanni Schembra2, Marco Antonio Aliotta1, Andrea Cannata1, Domenico Patanè1,3 and Michele Prestifilippo1, (1)INGV National Institute of Geophysics and Volcanology, Catania, Italy, (2)University of Catania, Dipartimento di Ingegneria Elettrica, Elettronica ed Informatica, Catania, Italy, (3)University of Granada, Instituto Andaluz de Geofísica, Campus Universitario de Cartuja s/n, Granada, Spain
Abstract:
In the last decade, the capability to monitor and quickly respond to remote detection of volcanic activity has been greatly improved through use of advanced techniques and semi-automatic software applications installed in most of the 24h control rooms devoted to volcanic surveillance. Ability to monitor volcanoes is being advanced by new technology, such as broad-band seismology, microphone networks mainly recording in the infrasonic frequency band, satellite observations of ground deformation, high quality video surveillance systems, also in infrared band, improved sensors for volcanic gas measurements, and advances in computer power and speed, leading to improvements in data transmission, data analysis and modeling techniques. One of the most critical point in the real-time monitoring chain is the evaluation of the volcano state from all the measurements. At the present, most of this task is delegated to one or more human experts in volcanology. Unfortunately, the volcano state assessment becomes harder if we observe that, due to the coupling of highly non-linear and complex volcanic dynamic processes, the measurable effects can show a rich range of different behaviors. Moreover, due to intrinsic uncertainties and possible failures in some recorded data, precise state assessment is usually not achievable. Hence, the volcano state needs to be expressed in probabilistic terms that take account of uncertainties. In the framework of the project PON SIGMA (Integrated Cloud-Sensor System for Advanced Multirisk Management) work, we have developed an expert system approach to estimate the ongoing volcano state from all the available measurements and with minimal human interaction. The approach is based on hidden markov model and deals with uncertainties and probabilities. We tested the proposed approach on data coming from the Mt. Etna (Italy) continuous monitoring networks for the period 2011-2013. Results show that this approach can be a valuable tool to aid the operator in volcano real-time monitoring.