Volcanic Plume CO2 Flux Measurements at Mount Etna by Mobile Differential Absorption Lidar

Volcanic eruptions are often preceded by precursory increases in the volcanic carbon dioxide (CO2) flux. Unfortunately, the traditional techniques used to measure volcanic CO2 require near-vent, in situ plume measurements that are potentially hazardous for operators and expose instruments to extreme conditions. To overcome these limitations, the project BRIDGE (BRIDging the gap between Gas Emissions and geophysical observations at active volcanoes) received funding from the European Research Council, with the objective to develop a new generation of volcanic gas sensing instruments, including a novel DIAL-Lidar (Differential Absorption Light Detection and Ranging) for remote (e.g., distal) CO2 observations. Here we report on the results of a field campaign carried out at Mt. Etna from 28 July 2016 to 1 August 2016, during which we used this novel DIAL-Lidar to retrieve spatially and temporally resolved profiles of excess CO2 concentrations inside the volcanic plume. By vertically scanning the volcanic plume at different elevation angles and distances, an excess CO2 concentration of tens of ppm (up to 30% above the atmospheric background of 400 ppm) was resolved from up to a 4 km distance from the plume itself. From this, the first remotely sensed volcanic CO2 flux estimation from Etna’s northeast crater was derived at ≈2850–3900 tons/day. This Lidar-based CO2 flux is in fair agreement with that (≈2750 tons/day) obtained using conventional techniques requiring the in situ measurement of volcanic gas composition.


Introduction
In the last two decades, there have been major advances in the instrumental monitoring of volcanic gas plume composition and fluxes [1]. These have included the first instrumental networks of scanning Differential Optical Absorption Spectrometers (DOAS) for volcanic SO 2 flux monitoring, the implementation of satellite-based volcanic gas observations, and the advent of sensor units for in situ gas monitoring [1][2][3][4][5][6]. Owing to this technical progress, volcanic gas plume composition and fluxes have increasingly been used to extract information on degassing mechanisms/processes [4], and to derive constraints on shallow volcano plumbing systems [5]. However, work still needs to be done to increase the number of volcanic gas species that can be detected in plumes, which remain few if compared to the countless number of chemicals quantified from fumarole direct sampling [1].
Studying volcanic gas plumes has additionally contributed to monitoring, and eventually allowed the prediction of volcano behavior [6]. In particular, it has been shown that, at open-vent persistently degassing volcanoes, volcanic eruptions are often preceded by anomalous increases of the volcanic CO 2 flux [7]. These initial observations have motivated attempts to systematically monitor the volcanic CO 2 flux, and to identify novel measurement strategies [8]. Until recently, however, attempts to remotely sense the volcanic CO 2 flux from distal locations have been limited in number [9,10], while the majority of the observations have involved in situ measurements in the proximity of hazardous active volcanic vents [3]. On Mt. Etna, for example, one of the largest volcanic CO 2 point sources on Earth [11], the volcanic CO 2 flux has systematically been measured since the mid-2000s by combining in situ measurement of the volcanic CO 2 /SO 2 ratio (with portable or permanent Multi-Component Gas Analyzer Systems, Multi-GAS; [12][13][14][15]) with remotely sensed SO 2 fluxes [16][17][18]. No successful report exists, at least to the best of our knowledge, of spectroscopy-based detection of Etna's volcanic CO 2 flux from a remote (distal) location.
Within the context of the ERC (European Research Council) starting the grant project BRIDGE (BRIDging the gap between Gas Emissions and geophysical observations at active volcanoes), we designed a new DIAL (Differential Absorption Lidar) [19], with the specific objective to remotely sense the volcanic CO 2 flux. Lidars have only recently been introduced in volcanic gas studies. A CO 2 laser-based lidar was used at Mt. Etna in 2008 [20] and at Stromboli Volcano in 2009 [21] to measure the volcanic plume water vapor flux. More recently, lidars were first been used to target volcanic CO 2 [9,10,[22][23][24]. Our lidar BILLI (BrIdge voLcanic LIdar) [22], for example, has recently been used to successfully retrieve three-dimensional tomographies of volcanic CO 2 in the plumes of Pozzuoli, Solfatara in 2014 [9] and Stromboli volcano in 2015 [23,24]. As such, although gas-sensing lidars remain far less exploited in volcanology than those targeting volcanic ash/particles [25,26], this novel application field may expand rapidly in the near future.
Here, we report on the first successful use of BILLI at Mt. Etna. We show that, in our July-August 2016 Etna experiment, the lidar successfully resolved a volcanic CO 2 signal of a few tens of ppm (in excess to the background air) from more than 4 km of distance, and with good spatial (5 m) and temporal (10 s) resolution. These results are used to derive the first "remote" assessment of Etna's volcanic CO 2 flux. Our observations open new perspectives for routine volcanic CO 2 flux monitoring via lidars.

Field Set-Up on Mt. Etna
Observations on Mt. Etna were conducted from 28 July to 1 August 2016, including an initial phase of instrumental setup (28)(29). Successful CO 2 flux detections were obtained on 31 July, when optimal viewing conditions persisted over the day. The DIAL was mounted in a trailer loaded on a truck, parked at the INGV (Istituto Nazionale di Geofisica e Vulcanologia) observatory "Pizzi Deneri" (Figure 1). The observatory is located at 2823 m a.s.l., northeast of the summit crater of Mt. Etna (3329 m a.s.l.), and at about 3 km from the main degassing vents (Figure 1).
The lidar was used to scan the volcanic plume vertically, keeping a constant azimuth angle (230 • ) and varying the elevation angle from 7 • to 14 • (Figure 1). A full 7 • to 14 • vertical scan was completed in~15 min, and one atmospheric profile every 10 s was recorded throughout. With this instrumental set-up, the volcanic gas plume of the Etna's northeast crater was investigated (Figure 1), plumes from other craters being either too dilute (southeast crater) or only partially visible (central craters).
At our measurement conditions, two rock surfaces, corresponding to the eastern, outer flanks of the central crater, were intercepted by the laser beam at distances of 1.6 and 2.1 km, and at elevation angles from 7 • to 9 • . These rock surfaces retro-reflected the laser beam, yielding strong return signals (see below, Figure 2). The volcanic plume, e.g., high in-plume excess CO 2 concentrations, was detected in between the two above rock surfaces, and in the 2.2-4.2 km distance range.

DIAL
The main components of a lidar are the transmitter (laser) and the receiver (telescope). A lidar is merely an optical radar [19]: a laser pulse is transmitted to the atmosphere, and some of its photons are backscattered to the telescope by air molecules and aerosols (droplets, particles etc.). The optical power corresponding to this photon flux is transformed into an electronic signal by photodetector and preamplifier, and converted in digital signal by an ADC (analog-to-digital converter).
The chemico-physical properties of the atmosphere along the laser beam, at distance R (range) from the lidar, can be inferred from analysis of the detected signal as a function of t, the time interval between emission and detection. R and t are linked by the relation R = ct/2, where c is the speed of light. The returned signal to the lidars' telescope, as a function of R (or t), then yields an atmospheric profile ( Figure 2). In other words, an atmospheric profile is a range-resolved characterization of the lidar returned signal, which allows studying the air/plume optical properties along the light trajectory.
Air attenuates the laser pulse due to molecules and aerosol scattering and to the specific absorption of gases: if the laser wavelength coincides with absorption lines of a target gas, the attenuation will be stronger. A DIAL takes advantage of this effect: unlike a usual lidar, two wavelengths, ON and OFF, are transmitted, with only the former being absorbed by the target gas ( Figure 3).

DIAL
The main components of a lidar are the transmitter (laser) and the receiver (telescope). A lidar is merely an optical radar [19]: a laser pulse is transmitted to the atmosphere, and some of its photons are backscattered to the telescope by air molecules and aerosols (droplets, particles etc.). The optical power corresponding to this photon flux is transformed into an electronic signal by photodetector and preamplifier, and converted in digital signal by an ADC (analog-to-digital converter).
The chemico-physical properties of the atmosphere along the laser beam, at distance R (range) from the lidar, can be inferred from analysis of the detected signal as a function of t, the time interval between emission and detection. R and t are linked by the relation R = ct/2, where c is the speed of light. The returned signal to the lidars' telescope, as a function of R (or t), then yields an atmospheric profile ( Figure 2). In other words, an atmospheric profile is a range-resolved characterization of the lidar returned signal, which allows studying the air/plume optical properties along the light trajectory.
Air attenuates the laser pulse due to molecules and aerosol scattering and to the specific absorption of gases: if the laser wavelength coincides with absorption lines of a target gas, the attenuation will be stronger. A DIAL takes advantage of this effect: unlike a usual lidar, two wavelengths, ON and OFF, are transmitted, with only the former being absorbed by the target gas ( Figure 3). If the absorption line is narrow, and ON and OFF wavelengths are close enough, the target gas concentration along the lidar optical path can be derived from the ratio between the OFF and ON signals. In this application, we selected the following wavelengths ( Figure 3): ON, 2009.537 nm; OFF, 2008.484 nm. This selection was motivated by: (i) the CO2 absorption is relatively low, thus allowing the system to probe far ranges (beyond 4 km). If a stronger line had been used, the ON laser pulse would have extinguished before; (ii) the beam energy (depending mainly on the dye efficiency curve) and the detector responsivity are near their maximum; and (iii) the H2O absorption is very low ( Figure 3). Moreover, ON and OFF have been chosen so that the differential absorption of H2O is approximately zero (within the uncertainty of the spectroscopic data), thus minimizing the interference of water vapor to the carbon dioxide measurement.
In our case, the transmitter and the receiver are coaxial, and the lidar field-of-view can be aimed in the whole atmosphere thanks to a system made of two large elliptical mirrors [22]. This configuration allows the experimenters to scan the plume in both horizontal and vertical planes, thus If the absorption line is narrow, and ON and OFF wavelengths are close enough, the target gas concentration along the lidar optical path can be derived from the ratio between the OFF and ON signals. In this application, we selected the following wavelengths ( Figure 3): ON, 2009.537 nm; OFF, 2008.484 nm. This selection was motivated by: (i) the CO 2 absorption is relatively low, thus allowing the system to probe far ranges (beyond 4 km). If a stronger line had been used, the ON laser pulse would have extinguished before; (ii) the beam energy (depending mainly on the dye efficiency curve) and the detector responsivity are near their maximum; and (iii) the H 2 O absorption is very low (Figure 3). Moreover, ON and OFF have been chosen so that the differential absorption of H 2 O is approximately zero (within the uncertainty of the spectroscopic data), thus minimizing the interference of water vapor to the carbon dioxide measurement.
In our case, the transmitter and the receiver are coaxial, and the lidar field-of-view can be aimed in the whole atmosphere thanks to a system made of two large elliptical mirrors [22]. This configuration allows the experimenters to scan the plume in both horizontal and vertical planes, thus measuring CO 2 concentrations both outside and inside the volcanic plume [9]. From this, by scanning the volcanic gas plume from different angles and viewing directions, the CO 2 distribution in a cross-section of the volcanic plume can be retrieved. This, combined with independent knowledge of plume speed and altitude, allows the CO 2 flux to be retrieved.
The reader is referred to previous work [9,[22][23][24] for details on instrumental setup and data processing. The systematic error associated with the derived CO 2 concentrations is dominated by imprecision in wavelength setting [22]. This leads to inaccuracy in the differential absorption cross section, and thus in gas concentration. Thanks to a photo-acoustic cell filled with pure CO 2 at atmospheric pressure and temperature, the ON and OFF wavelengths were set before each atmospheric scan. The residual imprecision [23] of ±0.02 cm −1 (half laser linewidth: half width at half maximum of the energy transmitted by the laser system (J) vs the wavenumber (cm −1 )) implies a systematic error on CO 2 concentrations of 5.5% [24]. The statistical error of CO 2 measurement has been calculated by usual error propagation techniques from the standard deviation of the lidar signal. At 2.5 km, a mean range, it is about 2%, while it can exceed 5% at 4.2 km. Table 1 compares the instrumental set-up during the Mt. Etna field campaign, with those used at Solfatara [9,22], and Stromboli [23,24].  During its atmospheric propagation, the laser beam intensity approximately decreased:

Results
• exponentially, due to atmospheric extinction, according to the Lambert-Beer law, and; • as 1/R 2 , because the solid angle subtended by the receiver is A/R 2 , where A is the telescope effective area.
For these reasons, it is a common practice in lidar science to express results using a RCS, this being the logarithm of the product of the signal times the square of the range. To improve the SNR (signal-to-noise ratio), the RCS was obtained by averaging 50 laser shots for each lidar return, and a 13-point Savitzky-Golay filter was applied [28].
During a vertical scan, the measured range-resolved RCS profiles varied as the laser elevation was sequentially increased. Below 7.25 • elevation, the laser beam hit a first rock surface at about a 1.6 km distance. Laser beam retro-reflection at this rock surface produced, in the lidar return signal, a strong, narrow RCS peak at R = 1.6 km. At 7.25 • elevation (Figure 2a), only part of the beam was intercepted by the R = 1.6 km rock surface, while the remaining part impinged on the rock surface at R = 2.1 km, producing a second narrow RCS peak. For geometrical reasons, an elevation increase corresponded to an increase in the range at which the rock surfaces were encountered, e.g., the second rock surface was encountered at R = 2.1 at 7.25 • elevation, shown in Figure 2a, and at R = 2.3 km at 8 • elevation, shown in Figure 2b. No rock surface was hit by the laser beam at elevations >9 • , e.g., note the absence of narrow RCS peaks in Figure 2c,d.
Back-scattering of the laser beam by the volcanic plume produced wide and jagged RCS peaks, therefore very distinct from the narrow and defined peaks produced by beam retro-reflection at rock surfaces (compare the two peak shapes in Figure 2b).
The volcanic plume was detected at range distances in between the two rock surfaces up to a 9 • elevation (e.g., Figure 2b), or beyond them at a 9 • to 14 • elevation (Figure 2c,d). A broad, irregular RCS peak in the lidar returns, corresponding to the volcanic plume, was resolved up to a maximum measurement range of 4.2 km (Figure 2d).
We used the procedure detailed in References [9,24] to convert the RCS profiles into range-resolved profiles of in-plume excess CO 2 concentrations (see Figure 2b',c',d',d"). This procedure involves calculating the excess CO 2 concentration corresponding to each i-th ADC channel of the lidar profile from: where ∆R and RCS i are, respectively, the range interval and range corrected signal corresponding to each ADC channel; R 1 and R 2 are the range distances of the two above rock surfaces; and ∆C is the average excess CO 2 concentration in the air/plume parcel between them (this is obtained from the intensity contrast of lidar returns produced by the two rock surfaces). The term "excess" implies that the reported CO 2 concentrations are after subtraction of the ambient atmospheric background, and therefore correspond to the "volcanic" CO 2 levels in the plume. The ambient atmospheric CO 2 background was obtained from the processing of lidar returns in the 0-1.6 km range distance, where no plume signal was detected (see Reference [9] for details of calculations).
At an 8 • elevation, shown in Figure 2b', the volcanic plume was evidenced by a band of excess CO 2 concentrations of ≤125 ppm. These excess CO 2 concentrations agree well with those derived by in situ in-plume measurements with conventional techniques (e.g., the Multi-GAS), from which in-plume CO 2 concentrations of tens to hundreds of ppm above ambient air are typically obtained [12]. The plume appears to be about 300 m thick; this relatively narrow plume's cross-section was probably justified by the fact that, at such an 8 • elevation, the laser beam intercepted the volcanic plume at below the summit crater's rim altitude. Due to its close proximity to the crater slopes, the volcanic plume was, at least partly, protected from the local wind field, a fact that reduced its atmospheric dispersion. The volcanic plume was still relatively narrow at a~9 • elevation (Figure 2c'), where the laser beam pointed just above the summit crater's rim. At even higher elevations, the volcanic plume was wider and scattered by the wind, and the returned RCS often presented multiple peaks (Figure 2d,d').
As explained before (Section 2), a sequence of atmospheric profiles was acquired as the lidar vertically scanned the horizon, from a 7 • to 14 • (max) elevation. All CO 2 profiles (e.g., Figure 2), taken at different elevations during a single lidar rotation sequence, were combined and integrated to obtain a CO 2  The results are illustrated in the form of contour maps of excess CO 2 concentrations, plotted as a function of the range and elevation. The colored spots correspond to areas of high excess CO 2 (the natural background is dark blue), and therefore illustrate the spatial distribution and temporal evolution of the volcanic plume (the yellow lines delimit the positions of the laser beam reflections off the rock surfaces). In all the maps we obtained (see examples in Figure 4), the structure of the plume was well resolved. The plume was tracked as a cluster of high CO 2 concentration spots, trending from about a 9 • elevation and R = 2.4 km (the vent rim) to a 13 • elevation and R = 2.5-2.9 km. As such, our CO 2 concentration maps were consistent with a gently lofting volcanic plume (Figure 4), with vertical and horizontal movements driven by thermal buoyancy and by the local wind field pattern. The maps indicate the plume was being dispersed away from the lidar during our observations, since the range of volcanic plume detection increased with the elevation in all the maps. The maps of Figure 4 set the basis for the calculation of the volcanic CO2 flux. In analogy with previous work [9], we obtained the volcanic CO2 flux by integrating the background-corrected (excess) CO2 concentrations over a plume cross-section (from the maps of Figure 4), which allowed us to derive the plume CO2 molecular density. This was then multiplied by the plume transport speed to obtain the CO2 flux (ΦCO2, in Kg•s −1 ), as: where vp is the plume transport speed (in m/s); NmolCO2−total is the total-plume CO2 molecular density (expressed in molecules•m −1 ); and PMCO2 and NA are, respectively, the CO2 molecular weight and Avogadro's constant. The term NmolCO2−total was obtained by integrating the effective average excess CO2 concentrations (C , [ppm]) over the entire plume cross-section, according to: where Nh is the atmospheric number density (molecules•m −3 ) at the crater's summit height, and Ai represents the i-th effective plume area. The plume transport speed was inferred at 9.7 ± 0.8 m/s from the processing of plume images taken on the same day by the permanent UV camera system (UV4) in use at the Pizzi Deneri observatory since 2014; see Reference [29] for details on the instrument. The UV camera images were processed using an optical flow sub-routine using the Lukas/Kanade algorithm [30,31], integrated in the Vulcamera software [32] (same methodology as described in [29]).
Our derived CO2 fluxes are illustrated in Figure 5. The CO2 flux varied from 1235 to 8050 tons/day during the measurement interval, and averaged at 2850 ± 1800 tons/day. The maps of Figure 4 set the basis for the calculation of the volcanic CO 2 flux. In analogy with previous work [9], we obtained the volcanic CO 2 flux by integrating the background-corrected (excess) CO 2 concentrations over a plume cross-section (from the maps of Figure 4), which allowed us to derive the plume CO 2 molecular density. This was then multiplied by the plume transport speed to obtain the CO 2 flux (ΦCO 2 , in Kg·s −1 ), as: where v p is the plume transport speed (in m/s); NmolCO 2−total is the total-plume CO 2 molecular density (expressed in molecules·m −1 ); and PM CO2 and N A are, respectively, the CO 2 molecular weight and Avogadro's constant. The term N molCO2−total was obtained by integrating the effective average excess CO 2 concentrations (C exc,i [ppm]) over the entire plume cross-section, according to: where N h is the atmospheric number density (molecules·m −3 ) at the crater's summit height, and A i represents the i-th effective plume area. The plume transport speed was inferred at 9.7 ± 0.8 m/s from the processing of plume images taken on the same day by the permanent UV camera system (UV4) in use at the Pizzi Deneri observatory since 2014; see Reference [29] for details on the instrument. The UV camera images were processed using an optical flow sub-routine using the Lukas/Kanade algorithm [30,31], integrated in the Vulcamera software [32] (same methodology as described in [29]).
Our derived CO 2 fluxes are illustrated in Figure 5. The CO 2 flux varied from 1235 to 8050 tons/day during the measurement interval, and averaged at 2850 ± 1800 tons/day.

Discussion
As long-term volcanic gas records have increased in number and quality over the last few decades [33], full empirical evidence has emerged for precursory increases of the volcanic CO2 flux emissions prior to eruption of mafic to intermediate volcanoes [7]. However, remote direct measurements of the volcanic CO2 flux, which are intrinsically safer for operators and more prone to provide continuous, long-term observations, have remained impossible until recently [9,10].
Our results here support the ability of the DIAL-Lidar BILLI to profile atmospheric CO2 concentrations over large optical paths (Figures 2 and 4), and to remotely sense the CO2 flux from distal (up to 4 km distant) sources ( Figure 5). This instrument thus promises a real step ahead in the remote observation of volcanic gas emissions. Improved CO2 flux measurements are not only vital for better gas-based volcano monitoring, but are also needed to better constrain the global volcanic CO2 budget, which is still inaccurately known [8].
The volcanic CO2 flux from Mt. Etna has been assessed in the past by either in-plume airborne CO2 profiling [11], or by indirect methods involving in situ measurements of plume CO2/SO2 ratios, via either the Multi-GAS [3,[12][13][14][15] or Fourier Transform InfraRed Spectrometers (FTIR; [16]). To the best of our knowledge, our results are the first to report a direct, remote quantification of Etna's CO2 flux.
Our lidar results show that, in the circa 5-h-long observational widow, the CO2 flux from Etna's northeast crater varied from 1235 to 8050 tons/day ( Figure 5). The CO2 flux was somewhat higher, typically >4000 tons/day and up to 8050 tons/day, after 3 p.m. local time, relative to the 12-3 p.m. period (<4000 tons/day) ( Figure 5). No change in activity was yet observed at the northeast crater, which continued to exhibit quiescent degassing over the entire measurement interval. We therefore consider the observed variation as part of the normal fluctuations in degassing activity that occur at Etna, likely in response to temporal variations in the magma/gas transport rate in the volcano's feeding conduits [15][16][17][18]. By taking the arithmetic mean of the individual CO2 flux measurements in Figure 5, we would obtain a time-averaged CO2 flux of 2850 ± 1800 tons/day for 31 July 2016. In view of the non-stationary CO2 emission behavior captured by our high-temporal resolution measurement ( Figure 5), we also perform an independent exercise in which we calculated the total CO2 output from the northeast crater by integrating (in the time domain) the available CO2 flux measurements, each representative of 13-18 min of observation (the mean duration of scans was 15 min). From this, we obtained that ≈796 tons of CO2 were cumulatively released during 5 h of observations, implying a time-averaged CO2 flux of 3900 tons/day. This is about 30% higher than, but within one standard deviation of, the CO2 flux obtained above from a simple arithmetic mean approach (2850 ± 1800 tons/day).

Discussion
As long-term volcanic gas records have increased in number and quality over the last few decades [33], full empirical evidence has emerged for precursory increases of the volcanic CO 2 flux emissions prior to eruption of mafic to intermediate volcanoes [7]. However, remote direct measurements of the volcanic CO 2 flux, which are intrinsically safer for operators and more prone to provide continuous, long-term observations, have remained impossible until recently [9,10].
Our results here support the ability of the DIAL-Lidar BILLI to profile atmospheric CO 2 concentrations over large optical paths (Figures 2 and 4), and to remotely sense the CO 2 flux from distal (up to 4 km distant) sources ( Figure 5). This instrument thus promises a real step ahead in the remote observation of volcanic gas emissions. Improved CO 2 flux measurements are not only vital for better gas-based volcano monitoring, but are also needed to better constrain the global volcanic CO 2 budget, which is still inaccurately known [8].
The volcanic CO 2 flux from Mt. Etna has been assessed in the past by either in-plume airborne CO 2 profiling [11], or by indirect methods involving in situ measurements of plume CO 2 /SO 2 ratios, via either the Multi-GAS [3,[12][13][14][15] or Fourier Transform InfraRed Spectrometers (FTIR; [16]). To the best of our knowledge, our results are the first to report a direct, remote quantification of Etna's CO 2 flux.
Our lidar results show that, in the circa 5-h-long observational widow, the CO 2 flux from Etna's northeast crater varied from 1235 to 8050 tons/day ( Figure 5). The CO 2 flux was somewhat higher, typically >4000 tons/day and up to 8050 tons/day, after 3 p.m. local time, relative to the 12-3 p.m. period (<4000 tons/day) ( Figure 5). No change in activity was yet observed at the northeast crater, which continued to exhibit quiescent degassing over the entire measurement interval. We therefore consider the observed variation as part of the normal fluctuations in degassing activity that occur at Etna, likely in response to temporal variations in the magma/gas transport rate in the volcano's feeding conduits [15][16][17][18]. By taking the arithmetic mean of the individual CO 2 flux measurements in Figure 5, we would obtain a time-averaged CO 2 flux of 2850 ± 1800 tons/day for 31 July 2016. In view of the non-stationary CO 2 emission behavior captured by our high-temporal resolution measurement ( Figure 5), we also perform an independent exercise in which we calculated the total CO 2 output from the northeast crater by integrating (in the time domain) the available CO 2 flux measurements, each representative of 13-18 min of observation (the mean duration of scans was 15 min). From this, we obtained that ≈796 tons of CO 2 were cumulatively released during 5 h of observations, implying a time-averaged CO 2 flux of 3900 tons/day. This is about 30% higher than, but within one standard deviation of, the CO 2 flux obtained above from a simple arithmetic mean approach (2850 ± 1800 tons/day).
In the attempt to add confidence to our results, we compared our lidar-based CO 2 flux with independent estimates based upon a more conventional technique that involves a combination of SO 2 fluxes and plume CO 2 /SO 2 ratios ( Figure 6). Our permanent UV camera system (UV4) at Pizzi Deneri indicated, for the morning of the same 31 July, a time-averaged SO 2 flux of 645 ± 125 tons/day. This is the mean (±1 standard deviation) of 4 h of observations at a 0.5 Hz rate ( Figure 6; same methodology as in [29]). Our inferred northeast crater's SO 2 flux (645 ± 125 tons/day) corresponded to about 30% of the total volcano's SO 2 emissions (≈2200 tons/day). These latter emissions were inferred using the same UV camera system, and were thus primarily determined by the central craters (not targeted by our DIAL-Lidar). The northeast crater's volcanic plume was in situ measured by a portable Multi-GAS instrument (the same as in [12,13]) two days later. These in situ observations yielded a (molar) CO 2 /SO 2 ratio of ≈6, demonstrating the usual [12,16] CO 2 -poor composition of the northeast crater (the simultaneously observed CO 2 /SO 2 ratio of the central crater's plume was ≈16). We consider our Multi-GAS-derived composition on 2 August as still representative of the northeast crater's emissions on 31 July, since volcanic activity at that crater did not exhibit any substantial change in between the two days. By combining the two sets of data together, we converted the SO 2 flux time-series into a 4-h-long CO 2 flux time-series ( Figure 6), from which an averaged (arithmetic mean) UV-Camera + MultiGAS CO 2 flux of ≈2750 tons/day was obtained. This is close to our lidar-based estimates above (≈2850-3900 tons/day). We caution that the two independent CO 2 flux time-series (from lidar and UV-Camera + MultiGAS) are not temporally overlapping, since the UV camera system ran only in the morning, when sunlight conditions were optimal [29], while our successful CO 2 flux measurement with the lidar started a few hours later in the afternoon. In addition, the UV-Camera + MultiGAS used a constant CO 2 /SO 2 ratio (of six) throughout the entire UV camera temporal window, while it is valid only as a first approximation. However, the close CO 2 flux values inferred from lidar and UV-Camera + MultiGAS provide mutual validation for the two independent techniques.
Geosciences 2017, 7, 9 10 of 13 In the attempt to add confidence to our results, we compared our lidar-based CO2 flux with independent estimates based upon a more conventional technique that involves a combination of SO2 fluxes and plume CO2/SO2 ratios ( Figure 6). Our permanent UV camera system (UV4) at Pizzi Deneri indicated, for the morning of the same 31 July, a time-averaged SO2 flux of 645 ± 125 tons/day. This is the mean (±1 standard deviation) of 4 h of observations at a 0.5 Hz rate ( Figure 6; same methodology as in [29]). Our inferred northeast crater's SO2 flux (645 ± 125 tons/day) corresponded to about 30% of the total volcano's SO2 emissions (≈2200 tons/day). These latter emissions were inferred using the same UV camera system, and were thus primarily determined by the central craters (not targeted by our DIAL-Lidar). The northeast crater's volcanic plume was in situ measured by a portable Multi-GAS instrument (the same as in [12,13]) two days later. These in situ observations yielded a (molar) CO2/SO2 ratio of ≈6, demonstrating the usual [12,16] CO2-poor composition of the northeast crater (the simultaneously observed CO2/SO2 ratio of the central crater's plume was ≈16). We consider our Multi-GAS-derived composition on 2 August as still representative of the northeast crater's emissions on 31 July, since volcanic activity at that crater did not exhibit any substantial change in between the two days. By combining the two sets of data together, we converted the SO2 flux timeseries into a 4-h-long CO2 flux time-series ( Figure 6), from which an averaged (arithmetic mean) UV-Camera + MultiGAS CO2 flux of ≈2750 tons/day was obtained. This is close to our lidar-based estimates above (≈2850-3900 tons/day). We caution that the two independent CO2 flux time-series (from lidar and UV-Camera + MultiGAS) are not temporally overlapping, since the UV camera system ran only in the morning, when sunlight conditions were optimal [29], while our successful CO2 flux measurement with the lidar started a few hours later in the afternoon. In addition, the UV-Camera + MultiGAS used a constant CO2/SO2 ratio (of six) throughout the entire UV camera temporal window, while it is valid only as a first approximation. However, the close CO2 flux values inferred from lidar and UV-Camera + MultiGAS provide mutual validation for the two independent techniques. Figure 6. Time-series of CO2 flux emissions from the northeast crater (in red) obtained from the UV-Camera + Multi-Gas technique. These were calculated by converting the SO2 flux time-series (in black) obtained by the UV4 permanent UV camera system on 31 July 2016 (from 8 a.m. to 12 p.m., local civil time) using a CO2/SO2 ratio (molar) of six. The plume speed time-series calculated from the UV camera on the same 31 July is also shown (blue trend).  Figure 6. Time-series of CO 2 flux emissions from the northeast crater (in red) obtained from the UV-Camera + Multi-Gas technique. These were calculated by converting the SO 2 flux time-series (in black) obtained by the UV4 permanent UV camera system on 31 July 2016 (from 8 a.m. to 12 p.m., local civil time) using a CO 2 /SO 2 ratio (molar) of six. The plume speed time-series calculated from the UV camera on the same 31 July is also shown (blue trend).

Conclusions
We have shown for the first time that the volcanic CO 2 flux can be detected with lidar from up to a 4 km distance. During our Mt. Etna field campaign, our DIAL-Lidar BILLI vertically scanned the volcanic plume while profiling CO 2 concentrations every 10 s, with a spatial resolution of 5 m. With this configuration, we successfully detected an excess volcanic CO 2 signal of a few tens of ppm, with relatively low systematic and statistical errors (5.5% and 2%, respectively). By integrating the results of the atmospheric profile taken at different heading angles, and covering a full scan of the plume, the volcanic CO 2 flux was derived (after integration, and in combination with the plume transport speed) at ≈2850-3900 tons/day. This lidar-based flux is close to that independently obtained by in situ observations of the volcanic plume (≈2750 tons/day), which combined Multi-GAS in situ sensing of the plume composition and remotely sensed (UV camera) SO 2 fluxes.
Clearly, additional field tests are required to validate our novel technique even further. Still, our results suggest BILLI is a major advance in ground-based observations of volcanic plumes. The instrument allows the remote measurements of volcanic CO 2 (and particles, if desired) from distal (safe) areas, and with unprecedented temporal resolution and high spatial coverage. Further development is now required to make this technology an operational tool for routine volcanic gas observations. Efforts are currently being undertaken to reduce the weight and power requirements (the current prototype is~1100 kg and requires 6.5 kW), and to implement more user-friendly operational routines and software. These implementations are required to widen the application range of the lidar, and to allow its use in remote/harsh volcanic environments.