**1. Introduction**

Active volcanoes are monitored by a variety of increasingly sophisticated volcanic gas sensing techniques [1–3], which are contributing to improved understanding and monitoring of volcanic activity. In particular, trends in volcanic gas composition and fluxes help detect subtle changes in the rates of magma ascent and degassing within shallow volcano plumbing systems, and allow to better confinepre-eruptive and syn-eruptive processes (gas in magmas being the main drivers of volcanic processes) [4]. Nevertheless, gas monitoring still lags behind more established seismic and geodetic techniques [5]. This is because the low temporal resolution of gas observations has traditionally hampered real-time analysis of fast-occurring volcanic processes, such as shallow intrusion of magma prior to eruption, and rapid gas ascent and release during explosive eruptions.

Volcanic SO2 emissions provide key information on the rates of magma ascent in shallow (<3 km) magma plumbing systems, and are extensively monitored worldwide using instrumental networks of scanning ultraviolet spectrometers using the Differential Optical Adsorption Spectroscopy (DOAS) technique [6]. The advantage of this method is that acquisition and processing are relatively easy to automate [7], which yield exceptionally continuous records of volcanic SO2 fluxes at relatively high

temporal resolution (tens of minutes) [8–10]. Biases in the technique include limited spatial resolution (making it impossible to distinguish contemporaneous degassing from different vents), temporal resolution inadequate to resolve individual explosive events, and errors related to poor knowledge of plume speed [11].

The recent advent of UV cameras [12] has paved the way to volcanic SO2 flux observations of much improved temporal and spatial resolution (see References [13–15] for recent reviews), which contributes to more effective integration between gas and geophysical datasets [16–29].

Even though space-based techniques are becoming increasingly performant in detecting volcanic SO2 fluxes, ground-based SO2 flux observations are still important and fundamental in monitoring basaltic volcanoes. While satellites become invaluable during paroxysmal explosive eruptions, when measurements from ground are complicated by volcanic ash within the plume. UV-camera measurements are more effective in monitoring more sluggish quiescent emissions in the low troposphere, and perhaps more useful to capture the early phases of unrest with escalating degassing activity [30].

The first examples of fully automated, permanent UV camera systems [14,28,31,32] are particularly promising, since they are opening the way to routinely monitoring volcanic SO2 flux at a high rate continuously (daily hours only). For example, Reference [32] has demonstrated the ability of UV cameras to capture a precursory phase of heightened SO2 flux in the weeks prior to the 2014 effusive eruption at the Stromboli volcano (Italy).

A current limitation of permanent UV camera systems is that, while data acquisition is fully autonomous, data processing is still time-consuming and operator-managed, e.g., data streamed by these systems are archived, and post-processed with ad-hoc codes [33]. To fully exploit the volcano monitoring potentials of UV cameras, automation of UV camera acquisition and processing routines is now timely and important.

The objectives of this study are: (i) to describe a new automatic routine for nearly real-time processing and visualization of UV camera data, (ii) to demonstrate the ability of the automatic routine to capturing temporal changes in SO2 flux regime at Mt. Etna, and (iii) to contribute to better constraining Etna's degassing and eruptive behavior in 2016. To this aim, we reported on automatically processed data streamed by a permanent UV camera deployed on Etna. In order to fully characterize the Etna's 2016 behavior, we integrated our SO2 flux results with independent geophysical parameters, traditionally used at Mt. Etna to constrain volcanic activity state and evolution [34–42]. More specifically, we used Moderate Resolution Imaging Spectroradiometer (MODIS) satellite-based thermal data obtained from the MIROVA (Middle InfraRed Observation of Volcanic Activity) system [43], ground-based thermal data streamed by monitoring cameras [44] of the Osservatorio Etneo (INGV-OE), and seismic tremor data [34,36,42].
