A large variety of natural and anthropogenic phenomena produces intense low-frequency acoustic waves below the 20 Hz human hearing threshold [1]. These signals, referred to as infrasound, can propagate over large distances through the atmosphere due to low attenuationin acoustic waveguides between the ground and troposphere, in the stratosphere and lower thermosphere [2-5]. Depending on the upper-wind structure, ducting can then be reinforced or reduced [6-9]. Interest in such propagation studies had been revived since the Comprehensive Nuclear Test-Ban-Treaty (CTBT, http://www.ctbto.org/) was adopted and opened for signature in 1996. The International Monitoring System (IMS) is designed to ensure compliance with the CTBT by detecting and locating explosions with a minimum yield of one kiloton of TNT-equivalent anywhere in the world using at least two stations [10,11]. The infrasound component of this network currently consists today of 45 certified stations out of the 60 that are planned to be constructed. Even not yet fully established, this network already allows studies on a global scale as it has demonstrated its capability to detect and locate a large number of geophysical and anthropogenic sources [12,13].

Among numerous naturally occurring geophysical phenomena generating acoustic waves, volcanic eruptions are outstanding sources of repetitive signals [14-16]. As infrasound signals are associated with the massive ejection of material and the release of conduit overpressure, they are a good indicator that an eruption has occurred [17-19]. Moreover, due to the long-range propagation of infrasound, this technique is valuable to remotely monitor volcanoes in regions where ground-based observations are sparse [20-23] and identify potential hazards for aircraft safety [24,25].

Many active volcanoes are permanently detected by the IMS infrasound network [24]. In particular, Mt. Etna in Italy (37.73˚N, 15.00˚E; 3330 m high) is in Europe the highest and most active strato-volcano. It is located on the east coast of Sicily, lying above the convergent plate margin between the African and the Eurasian plates. This volcano has experienced a variety of eruption styles. Its volcanic activity can be divided into two main types: effusive flank eruptions, mainly characterized by the opening vents or fractures to feed voluminous lava flows, and persistent explosive summit activity, including mostly violent Strombolian and phreatomagmatic explosions, lava fountaining and persistent degassing [26,27]. Its current activity is typically effusive with explosive episodes and lava fountaining, with often large ash ejection in the atmosphere affecting nearby cities and local air traffic.

As the activity of Etna is mostly effusive, sometimes accompanied by small-to-moderate explosions, it often yields to small VEI (1-2).

During 2008-2009 and early 2010, no significant explosive activity was reported. Since October 2010, more paroxysmal eruptive periods, characterized by strong Strombolian activity, lava fountaining, and often dense dark ash emissions were reported. However, such information is not as precise as continuous near-field observations. The Smithsonian database is more useful for explosive volcanoes that rarely erupt or at least not continuously, and where the amount of observations is limited.

The main objective of this study is to assess the potential of the European infrasound network to monitor Etna by analyzing nearand far-field recordings from 2008 until now. In order to calibrate the existing network and evaluate the performance of the future ARISE (Atmospheric dynamics Research Infra Structure in Europe (http://arise-project.eu/) network, frequency-dependent attenuation relations are integrated into a network performance modeling technique. We first present the infrasound network and describe array processing methods. Then, the capability of the IS48 IMS station to detect Etna is analyzed by considering a detailed description of both the background noise at the receiver and the dynamics of structure of the stratosphere. Finally, we evaluate the performance of the existing IMS network and quantify its improvement by adding experimental stations.

The IMS network is unique by its global and homogeneous coverage. Significant advances in array designs and processing methods as well as the development of highly sensitive sensors and efficient wind-noise filtering systems allow now detecting low-amplitude coherent signals from remote volcanoes with an unprecedented precision [10-29]. In particular, at a distance of about 550 km, Etna is permanently monitored by the IS48 IMS station (35.80˚N, 9.32˚E) located in Tunisia. In case of major eruption, signals can be detected by other IMS stations like IS26 in Germany and IS43 in Russia, at a distance of 1240 and 2680 km, respectively. In addition to the existing operating IMS network, we consider in this study the four experimental arrays OHP, AMT, CEA and Flers (Figure 1). Within the course of the ARISE project, other arrays, like the ones operated by the Royal Netherlands Meteorological Institute (KNMI, The Netherlands), the Federal Institute for Geosciences and Natural Resources (BGR, Germany) and the Atomic Weapons Establishment (AWE Blacknest, UK), will provide additional far-field recordings.

In the near-field, one permanent small aperture (~250 m) four-element array, ETN, operated by the University of Firenze (UNIFI), routinely records the Etna activity since 2007. Each array element is equipped with a differential pressure transducer with a sensitivity of 20 mV/Pa [30]. The wide frequency response (0.01 - 100 Hz) and the 200 Pa peak-to-peak pressure range allow a full-recovery of the signals of interest. ETN is deployed on the southern flank of Etna volcano, at an elevation of about 2000 m a.s.l. and at a distance of approximately 5 km from the summit craters. This site allows a clear azimuthal discrimination of infrasound radiated from most of the Etna summit craters, and thus represents an essential contribution to accurately monitor its degassing and volcanic activity. Data are processed in real-time using a cross-correlation based method. Since September 2007, almost 2.4 ×10⁶ detections (about one detection per minute) related to both degassing and explosive activity were measured.

IS48 well detects Etna (550 km, 65.4˚) and Stromboli (618 km, 55.9˚), the nearest active volcanoes. Itconsists of seven separate MB2000-type microbarometers connected to a central recording facility with inter-sensor-spacing

ranging from 150 m to 1.6 km. The microbarometers operate from DC up to 27 Hz with a flat frequency response from 0.02 to 4 Hz and an electronic self-noise level of 2 mPa RMS (<18 dB below the minimum acoustic noise at 1Hz). Infrasound data are routinely processed with the Progressive Multi-Channel Correlation method (PMCC) [31]. The processing is performed consecutively using an adaptive window length and log-spaced frequency bands allowing the full frequency band of interest (0.02 - 4 Hz) to be processed in one single run [32].

Figure 2 presents the results of the PMCC automatic processing results at IS48 filtered in the 0.1 - 4 Hz band from 2006 to 2012. Several sources of infrasonic waves are identified:

§  From 0.1 to 0.3 Hz, microbaroms produced by large interacting open-ocean swell systems [33-35] nearly continuously detected in winter from North Atlantic Ocean with a back-azimuth between 310˚ and 320˚, and in summer from the Mediterranean Sea with a back-azimuth about 90˚.

§  Above 0.5 Hz, in summer, persistent detections associated with eruptions of Mt. Etna and Stromboli with back-azimuth between 50˚ and 80˚. Paroxysmal events but also small-to-moderate explosions are almost continuously recorded between May and September.

§  Above 1 Hz, detections possibly related to industrial activity (oil and gas fields, refineries in Libya and Algeria) with back-azimuth between 120˚ and 210˚.

Monitoring infrasound at IS48over several years reveals a clear seasonal transition in the bearings along with the stratospheric general circulation between summer and winter for both microbarom and volcano signals. Furthermore, these seasonal variations reverse whether sources are locate east or west of the array.This oscillation clearly captured in climatological wind models [4-36] controls to the first order the direction from where signals are expected to be detected.

Figure 3 presents one example of a major eruptive episode of Etna on April 1^st, 2012 which was well detected by several stations up to station IS43 in Russia. The measured celerity (horizontal propagation range divided by travel time) near 300 m/s at all stations suggests a propagation through the stratospheric waveguide [3,4]. To locate the source, the three nearest stations (IS48, IS26 and CEA) are considered by applying a simple cross bearing method. The source is located at about 75 km to north-east of the true location (38.27˚N, 15.56˚E).