Interaction between magma intrusion and flank dynamics at Mt. Etna in 2008, imaged by integrated dense GPS and DInSAR data

Transcription

1 Article Volume 14, Number 8 6 August 2013 doi: ISSN: Interaction between magma intrusion and flank dynamics at Mt. Etna in 2008, imaged by integrated dense GPS and DInSAR data Alessandro Bonforte, Francesco Guglielmino, and Giuseppe Puglisi Istituto Nazionale di Geofisica e Vulcanologia, Osservatorio Etneo, Piazza Roma 2 IT Catania, Italy (alessandro.bonforte@ct.ingv.it) [1] Global positioning system (GPS) and differential interferometric synthetic aperture radar (DInSAR) data, collected from July 2007 to July 2008 on Mt. Etna, are analyzed to define the dynamics preceding and accompanying the onset of the eruption on 13 May Short- and long-term comparisons have been made on both GPS and radar data, covering similar time windows. Thanks to the availability of three GPS surveys the year before the eruption onset, an increase in the seaward movement of the NE flank of the volcano has been detected in the few months before the dike intrusion. The GPS ground deformation pattern also shows a slight inflation centered on the western side of the volcano in the preeruptive long-term comparison (from July 2007 to May 2008). The GPS has been integrated with DInSAR data by the SISTEM approach, to take advantage of the different methodologies and provide high spatial sampling of the 3-D ground displacement pattern. We inverted the SISTEM results to model the pressure source causing the observed preeruptive inflation. The subsequent emplacement of the eruptive dike was imaged by two GPS surveys carried out on a dense network over the uppermost part of the volcano on 6 and 13 May, i.e., a few days before and a few hours after the beginning of the eruption. We inverted this comparison to define the position, geometry, and kinematics of the dike. The dike intrusion was also imaged by DInSAR data with temporal baselines of 2 3 months, which confirm strong displacements localized on the summit area, rapidly decreasing toward the middle flanks of the volcano, as detected by very short-term GPS data; furthermore, the comparison between DInSAR and GPS data highlighted the presence of a depressurizing source localized beneath the upper southwestern area, acting just after the dike intrusion. Finally, the long-period (1 year) GPS and DInSAR data were integrated by SISTEM to finely depict the 3-D ground deformation pattern with the highest spatial resolution. The long-period data allowed the complex kinematics of the volcano to be finely imaged and highlighting the interaction between flank dynamics and magma injection. Components: 9,072 words, 10 figures. Keywords: flank dynamics; eruption; volcanotectonics; GPS; DInSAR; data integration. Index Terms: 8485 Remote sensing of volcanoes: Volcanology; 8419 Volcano monitoring: Volcanology; 4302 Geological: Natural Hazards; 4316 Physical modeling: Natural Hazards; 4337 Remote sensing and disasters: Natural Hazards; 7280 Volcano seismology: Seismology; 8178 Tectonics and magmatism: Tectonophysics; 3294 Instruments and techniques: Mathematical Geophysics; 0545 Modeling: Computational Geophysics; 1952 Modeling: Informatics; 4255 Numerical modeling: Oceanography: General. Received 22 February 2013; Revised 29 May 2013; Accepted 30 May 2013; Published 6 August Bonforte, A., F. Guglielmino, and G. Puglisi (2013), Interaction between magma intrusion and flank dynamics at Mt. Etna in 2008, imaged by integrated dense GPS and DInSAR data, Geochem. Geophys. Geosyst., 14, , doi: American Geophysical Union. All Rights Reserved. 2818

2 1. Introduction [2] This work analyses the ground deformation preceding and accompanying the onset of the Mt. Etna eruption by integrating GPS data provided by the dense GPS network (Figure 1) that is measured periodically (at least once a year), and the ENVISAT SAR data acquired between 2007 and The exceptional instrumental quantity and quality of data, coupled with the frequent recent Etna eruptions, have enabled modeling different eruptive mechanisms, leading to a better understanding of the complex kinematics of the volcano. [3] Between 2007 and early 2008, the Mt. Etna volcano showed a recharging phase that climaxed with a new effusive eruption on May 2008 and lasted about 14 months. The recharge phase was accompanied by several lava-fountaining episodes from SE crater during 2007 [Andronico et al., 2008] and the last episode occurred on 10 May 2008, only 3 days before the onset of the 2008 eruption. The dike intrusion, occurring on 13 May, was accompanied by a strong seismic release and marked ground deformation; furthermore, violent lava fountaining took place along the eruptive fracture during the first hours of the eruption [Bonaccorso et al., 2011a; Aloisi et al., 2009]. [4] On the same day, lava fountains erupted from a N140 E fissure ( m above sea level (a.s.l.)), and the eruptive fissure propagated downslope southeastward, curving toward N120E and reaching a minimum elevation of 2620 m a.s.l., where lava flows erupted and rapidly expanded in the Valle del Bove. The seismicity affected the upper portion of the NE Rift, where a dry, extensive N-S fracture field began to form (see Figure 1). This eruption represents a particular event in the recent eruptive history of Mt. Etna, characterized by an intrusion mechanism that was never observed since monitoring networks exist on the volcano; in fact, the dike departed from the shallow central feeding system, without following the main conduit but also without propagating radially as usual for lateral eruptions on Etna [Bonaccorso et al., 2011a]. [5] We analyze the GPS and DInSAR data applying, when possible, the SISTEM approach to provide unprecedented detailed 3-D ground deformation maps. the SISTEM (Simultaneous and Integrated Strain Tensor Estimation from geodetic and satellite deformation Measurements) approach [Guglielmino et al., 2011a] has been recently developed to integrate GPS and SAR ground deformation data and make the most of the different methodologies to provide high spatial sampling of the 3-D ground displacement pattern. 2. Recent Eruptive History of Mt. Etna [6] Mount Etna is the largest active volcano in Europe, it is a Quaternary polygenetic volcano located on the east coast of Sicily, rising about 3300 m a.s.l.. The volcano stands within the deforming zone at the plate boundaries, with the Apenninic-Maghrebian chain (AMC in Figure 1) to the north and the Hyblean Foreland to the south (HF in Figure 1). The volcano has grown on the continental lithosphere, at the junction of two major structural lineaments, at the regional scale, which separates the continental lithosphere of the eastern Sicily from the oceanic lithosphere of the Ionian basin: the NE-SW striking Messina fault system (MF in Figure 1) and the NNW-SSE-oriented Malta Escarpment (ME in Figure 1) to the Hyblean Foreland [Monaco et al., 1997]. [7] During the past 400 years, the volcano has produced over 60 flank eruptions [Branca and Del Carlo, 2004], whereas the summit craters show persistent degassing and recurrent explosive activity at the vents on the crater floor. [8] The eastern flank of the volcanic edifice shows a fairly continuous seaward motion (flank dynamics), due to the interrelationship between gravity instability and magma intrusion [e.g., Borgia et al., 1992; Bonforte and Puglisi, 2006; Puglisietal., 2008; Acocella and Neri, 2009; Bonforte et al., 2011]. The margins of this unstable area comprise the Pernicana Fault System (PFS; Figure 1) to the north and the Ragalna Fault System (RFS; Figure 1) to the Southwest [Neri et al., 2007]. [9] In the decades, following the flank eruption, Mt Etna underwent persistent degassing, evolving to an almost continuous summit activity (after 1995) that increased in intensity until June During this period, the volcano was affected by a significant overall continuous but not constant inflation and a marked 2819

3 Figure 1. (a) Structural sketch map of Mt. Etna with main fault systems (PFS: Pernicana fault system; RNF: Ripe della Naca faults; TFS: Timpe fault system; SVF: S. Venerina fault; STF: S. Tecla fault; ACF: Acicatena fault; TF: Trecastagni fault; TMF: Tremestieri-Mascalucia fault; BOF: Belpasso-Ognina fault; RFS: Ragalna fault system; VdB: Valle del Bove); (b) the red circle indicates the GPS benchmarks; the dotted box indicates the zoom of the summit part, involved in the eruption, and reporting the lava flow (in pink), the eruptive fissure (black line), and the dry fractures (blue lines) with their propagation direction (dashed blue arrow). NEC: North-East Crater; VOR: Voragine Crater; BN: Bocca Nuova Crater; SEC: South-East Crater Coordinates are in kilometers, in WGS84 Ellipsoid UTM zone 33 Projection. acceleration of flank movements affecting the eastern to southern slopes of the volcano [Bonforte et al., 2008; Bonaccorso et al., 2011a]. This increasing activity and flank dynamics led to a flank eruption in July 2001 and to another one in October 2002, which followed the same path on the southern flank but also affected the NE rift [Behncke and Neri, 2003; Andronico et al., 2005; Bonforte et al., 2007a, 2009]. After these flank eruptions, the volcano showed an inflation/ deflation behavior accompanying two less powerful flank eruptions in and 2006 and decreasing dynamics of the eastern flank, following the exceptional sliding velocities recorded after the eruption [Bonaccorso et al., 2006, 2011; Bonforte et al., 2008]. [10] Unlike the and 2006 eruptions, which were fed by dikes radially propagating from the central conduit without significant seismicity and driven by the dynamics of the eastern flank of the volcano [Bonaccorso et al., 2006; Neri and Acocella, 2006; Neri et al., 2006], the 2008 intrusion was accompanied by clear geophysical and volcanological signals. 3. Ground Deformation Preceding the Eruption 3.1. GPS Data [11] Before the eruption, three GPS surveys were carried out in June 2007, February 2008, and April-May The first GPS survey, taken as the reference for this study, was carried out in June 2007 on the entire Mt. Etna GPS network. The whole GPS network (Figure 1) consisted of 20 permanent stations and 83 benchmarks (mostly self-centering), installed over the entire volcano and around it. The 63 GPS benchmarks are surveyed in static mode by carrying out sessions lasting from at least 4 24 h. Eleven of these benchmarks are arranged in two small areas (about 1km 2 ), across the Pernicana fault on its middle and lower sectors. Semikinematic occupations are used to survey 20 benchmarks aligned along a N-S profile reaching the summit area and along an E-W one on the middle southern flank of the volcano [Bonforte et al., 2004, 2007b; Puglisi and Bonforte, 2004] 2820

4 Figure 2. Results of the comparisons between GPS surveys carried out before the eruption onset. Arrows indicate the horizontal displacements; vertical displacements are reported as color map, according to the scale bar. Dashed blue lines indicate the position of RNF and SVF/STF faults (see Figure 1). [12] During the first months of 2008, before the eruption onset, two surveys on part of this network were carried out. The first survey of 2008 was carried out on the northeastern part of the network in February-March 2008 (hereafter the February 2008 survey), with the aim of monitoring the Pernicana fault. During this survey, the 11 stations of the two small networks across the Pernicana fault were surveyed and also a further five stations were installed in semipermanent mode on the northeastern flank of the volcano, carrying out 24 h measurement sessions for a week. [13] The second survey of 2008 was done on the southern part of the network, from the end of April to the first days of May 2008 (hereafter the May 2008 survey); six semipermanent GPS stations were installed on the southern flank of the volcano following a seismic swarm in the same area. In addition, during this survey, the northeastern part of the network was surveyed again by measuring both networks across the Pernicana fault and installing a further 12 semipermanent stations on the northeastern flank of the volcano and along the NE Rift. Data from each survey (both static and semikinematic) are processed using International GNSS Service (IGS) final precise ephemerides and antenna calibration tables, following the usual procedure adopted for Mt. Etna [Bonforte et al., 2008]. Processing results are then adjusted and referred to the same reference frame to obtain a consistent set of coordinates of all measured stations for each survey with associated errors (3 4 mm for the horizontal components and 6 7 mm for the vertical one). [14] By comparing the results of the June 2007 complete survey and that of February 2008 on the NE flank, the ground deformation pattern affecting the volcano during these 8 months can be investigated with greater detail on the entire upper and middle NE flank (Figure 2a). Station displacements indicate a widespread uplift, increasing toward the central-northern part of the volcano. Horizontal motions are almost negligible at all stations lying on the southern, western, and northern flanks of the volcano. On the contrary, a vigorous ESE-ward motion affects the entire northeastern sector, extending toward the lowermost eastern side. The strongest horizontal displacements have been observed close to the Pernicana fault, where stations moved more than 30 mm eastward in 8 months (equivalent to about 50 mm/yr). These displacements testify to a significant acceleration of the slip rate of the fault with respect to the usual 28 mm/yr [Tibaldi and Groppelli, 2002; Neri et al., 2004; Palano et al., 2006; Bonforte et al., 2007]. Displacements of the order of 20 mm affect the other stations on the NE sector of the volcano and on the lower eastern flank. Vertical motions show an interesting pattern: uplift affects GPS stations lying on upper NE part, whereas subsidence (about mm) has been measured on the lower NE and E sectors. This subsiding area is well confined to the North by the Pernicana and Ripe della Naca faults, whereas to the South, it seems to be close to the S. Tecla-S. Venerina faults. [15] By comparing the results of February and May 2008 surveys (before the eruption onset), a more significant inflation of the volcano is visible (Figure 2b), with a radial horizontal motion of the stations of the order of 10 mm, always accompanied by the ESE-ward motion of the eastern flank. Here, horizontal motions reach 20 mm at stations close to the 2821

5 Pernicana fault; these displacements become more significant when considering the shorter time window, just over 3 months, confirming the higher dynamics affecting this fault and showing a further increase (about 80 mm/yr). Also an inversion of the vertical ground deformation pattern, with respect to the previous period, can be noted. Subsidence (about mm) affects the upper part of the NE sector, whereas a smaller uplift affects stations under 1000 m of altitude; conversely, a vertical stability characterizes the lowermost eastern side of the volcano during this period. [16] The comparison between the results of June 2007 and May 2008 surveys allows a more complete ground deformation pattern to be imaged since more benchmarks were measured in May than in February, on both the southern and northeastern sides of the volcano (Figure 2c). This comparison makes the general inflation of the volcano more evident, thanks to the longer period. In particular, a radial pattern of the horizontal displacements from a center located on the upper-western side of the volcano is clearly visible. The ESE-ward motion affects the entire eastern side of the volcano, involving also GPS stations located on the southern part of the network. On this flank, the strongest displacements (up to mm over 11 months) are measured on the northern part, along the Pernicana fault, whereas a rotation of the displacement vectors is visible on the southern part, from SE-ward to ESE-ward moving from higher to lower altitude SAR Data [17] To investigate the deformation during the year before the onset of the eruption, we also performed a Differential Interferometry Synthetic Aperture Radar (DInSAR) analysis of ascending and descending ENVISAT-ASAR images. The procedure to generate the interferometric products was the so-called two-pass interferometry [Massonnet and Feigl, 1998]. [18] We analyzed the ascending 20 June March 2008 (Figure 3a) and the descending 27 June May 2008 (Figure 3b) pairs to have a coherent time window to the June 2007 May 2008 GPS survey (Figure 3c). We processed these data using the Jet Propulsion Laboratory (JPL)/Caltech Repeat Orbit Interferometry Package (ROI_PAC, version 3.0.1) [Rosen et al., 2004]. [19] Both the ENVISAT images show the general seaward movement of the eastern flank of Mt. Etna, confirming the displacements recorded by GPS. In addition, on the ascending data, an intense line of sight (LOS) displacement of about 1.5 fringes affecting the central summit area is clearly visible. The descending data confirm the uplift of the central western side of the volcano, even if it is less intense due to the look geometry and the greater noise SISTEM Data Integration [20] To derive high-accuracy three-dimensional surface motion maps, we applied the SISTEM method to integrate GPS and InSAR data. The standard formulation of the SISTEM approach as recently proposed by Guglielmino et al. [2011b] was improved to use both ascending and descending interferograms and better constrain the 3-D components of the displacements. The SISTEM results (Figure 3d) depicted the ground deformation occurring before the eruption onset well and show a slow inflation affecting the summit central area of the volcano (with an uplift of about 2 cm), coupled with an eastward movement of the eastern flank (about 5 cm). Furthermore, the SISTEM results also highlighted the subsiding area affecting the eastern flank, previously detected by GPS, with a maximum subsidence of about 3 cm. [21] The detailed 3-D ground deformation map from SISTEM, allowed clearly distinguishing the areas of the volcano affected by inflation and sliding dynamics. To avoid mixing the effects of two different dynamics, only the central and western portion of the whole data set was inverted, namely where the inflation effects are more visible Data Inversion and Modeling [22] To model the inflation acting during the preeruptive period (see Figure 4, INV1), we inverted the very detailed 3-D displacements coming from the SISTEM integration, using a Davis [1986] point-pressurized source, of ellipsoidal shape, arbitrarily oriented in the halfspace. For the minimization process, we used software based on the genetic algorithm (GA) optimization approach as proposed by Nunnari et al. [2005]. [23] The search grid parameters and results of the GA search are summarized in Figure 4. The solution converged to a final fitness value of 78% and the end result is in agreement with the effect produced by an inflating source, vertically elongated and located west of the summit craters at a depth of 3.5 km below sea level, with the semimajor axis a five times longer than the b and c axes. The three Euler angles (indicating the ellipsoid orientation) show that the ellipsoid dips about 87 toward SSW. 2822

6 Figure 3. (a) Ascending 20 June 2007 to 26 March 2008, and (b) descending 27 June 2007 to 7 May 2008 phase interferograms; (c) GPS displacement vectors from June 2007 to 7 May (d) Results of the SIS- TEM integration: the arrows indicate the horizontal displacement vectors, whereas the vertical displacements are presented by a color map. The dashed-blue rectangle defines the inverted area for modeling the ellipsoidal pressure source indicated by the red star in Figure 3c.Figure 4. Summary of data inversions, resulting model parameters and search ranges for the three periods analyzed: Before (INV1), during (INV2), and after (INV3) the dike intrusion. Coordinates are in WGS84, UTM 33N projection. Depth is referred to the sea level. 4. Ground Deformation Across the Eruption Onset 4.1. Snapshot of Eruption Onset Imaged by GPS Data [24] By comparing the results of the GPS measurements carried out on 6 May with those made on 13 May (a few hours after the eruption onset), the ground deformation due to the dike injection can be detected and analyzed (Figure 5). Displacements of the order of tens of centimeters have been measured on the upper part of Mt. Etna. More than 65 cm northeastward displacement was measured at EPDN, whereas southwestward motions of about 36 cm and 29 cm were measured at EPLU and ECPN stations, respectively. Ground deformation disappears below an altitude of m a.s.l. on the northern, western, and southern flanks of the volcano. Only on the northeastern 2823

7 Figure 4. Summary of the data inversions, resulting model parameters and search ranges for the 3 periods analyzed: before (INV1), during (INV2) and after (INV3) the dike intrusion. Coordinates are in WGS84, UTM 33N projection. Depth are referred to the sea level. sector of the volcano does the ground deformation extend downward, clearly driven by the NE Rift- Pernicana fault system. GPS benchmarks above the NE Rift show northeastward motions that rapidly decrease from 30 cm to 4 cm from higher to lower stations. These displacements compared to the stability of EDAM station located about 2 km northwestward, define a mainly transcurrent kinematics of the Rift (Figure 5). These kinematics are induced by the NE-ward displacement produced by the dike injection. This pattern of ground deformation immediately demonstrated that no intrusion was propagating along the NE rift, unlike what was observed during the 2002 dike intrusion, when the extension across the NE rift facilitated the dike propagation [Bonforte et al., 2007]. Below the rift, in the NE sector, horizontal displacements progressively decrease to 1 2 cm, whereas the entire sector shows a significant uplift reaching a maximum of 8 cm just east of the Ripe della Naca fault (EMAG benchmark). Starting from the lower part of the NE rift, 3-D displacements show an uplift that increases toward NE, reaching the maximum value close to the Pernicana fault. This deformation abruptly disappears a few kilometers down slope. This particular feature may be due to the higher mobility of the NE flank, decoupled from the rest of the volcano; the flank is pushed by the intrusion against the Pernicana fault and this could cause, in the very short term, a kind of local and temporary thrusting, as already observed in the past [Puglisi et al., 2001]. [25] To interpret the deformation pattern measured, we performed an analytical inversion (see Figure 4, INV2) of the GPS data under the assumption of a homogeneous, isotropic, and elastic half-space by using Okada s [1992] model, following the previously described procedure. A dislocation source in the Okada model is defined by 10 parameters following the Aki and Richards [1980] definition. The search grid parameters and results of the GA search and the results of the inversion are shown in Table 2. The solution converged to a final fitness value of 78%. [26] The model gives a good fit to the data, especially for the horizontal motion, with an average misfit of 0.8 cm for east and north components and 2824

8 On the other hand, predicted vertical motion fits the observed one around the dike position well, confirming the particular and local deformation caused by the mobility of the NE flank of the volcano, controlled by the Pernicana fault. [27] The results of the inversion define a shallow dike whose southern tip is located beneath the eruptive fracture, extending NNW-ward for about 3 km. The dike shows an opening of almost 5 m and is westward dipping (72 ), in line with the seismicity pattern shown in Bonaccorso et al. [2011a], extending downward toward the central conduit of the volcano. Figure 5. GPS displacements referring to the 6?13 May 2008 surveys comparison. Red vectors in the map and dots in the plot indicate the observed displacements, blue vectors and dots indicate the expected ones from the Okada model (black dashed line in the map). Coordinates are in kilometers, in WGS84 Ellipsoid UTM zone 33 Projection. 2 cm for the vertical one. The reduced 2 is equal to 1 considering the a posteriori standard deviation of 1.2 and 1.3 cm, respectively, for east and north displacements and 3.9 cm for the vertical component. As also visible from Figure 5, despite a good fit for the horizontal displacements, there are high residuals for the vertical ground motion, mostly concentrated on the upper NE flank of the volcano Short-Term Deformations Across the Eruption Onset Imaged by DInSAR Data [28] To further investigate the ground deformation due to the dike intrusion, we also considered two ENVISAT interferometric pairs, acquired in ascending and descending view geometries. DIn- SAR data provide a more detailed ground deformation sampling with respect to GPS, even if in 1- D (along LOS) and over a rather longer time span. [29] A DInSAR pair of ascending ENVISAT images spanning 3 months (Bperp 279 m), obtained by combining the 26 March and 4 June 2008 images (Figure 6), confirms the general stability of the edifice at lower altitude, as shown by GPS, and the local deformation induced by the dike intrusion extending from the uppermost part of the volcano down to about 1700 m a.s.l.. The information is available only on the eastern side of the volcano and up to about 2500 m of altitude, because the top of the mountain was covered by the snow in March. However, the shape of the deforming area on the eastern sector of the volcano (moving away to the sensor for this looking geometry) confirms what was observed by GPS, with a greater extension of the deformation field toward NE and driven by the local fault system; indeed fringes show a clear lobe toward NE and are abruptly cut by the NE rift and upper Pernicana fault. [30] A DInSAR pair of descending ENVISAT images spanning 2 months (Bperp 154 m), obtained by combining the 7 May and 16 July 2008 images not biased by snow coverage (Figure6b), has been processed. On the other hand, as for all descending views on Etna, it is missing in information on the uppermost eastern side due to the strong geometrical distortion. This pair very clearly shows the ground deformation pattern on 2825

9 Figure 6. (a) Ascending 26 March 2008 to 4 June 2008, and (b) descending 7 May and 16 July 2008 phase interferograms; the same (c) ascending and (d) descending interferograms are reported as unwrapped displacement maps. the upper western side of the volcano. At least five fringes (i.e., 14 cm) of LOS displacement (moving away from the sensor for this looking geometry) affect the upper part of the volcano. [31] The descending DInSAR data confirm the rapid decay of the ground deformation; indeed, it disappears below 1700 m a.s.l., as already shown by the GPS and by ascending DInSAR data. On the eastern side, almost all information is lost except a lobe of about one fringe that is still visible on the NE sector, interrupted by the Pernicana fault and NE Rift. [32] We produced a simulated descending interferogram using the dike parameters from our previous GPS-only inversion (INV2) and found that the GPS dike parameters were insufficient to describe all of the deformation observed in the interferogram. In the first row of Figure 7a, the observed and simulated interferograms are reported together with the residuals between them. It is clear that the residuals show more than two fringes around the summit crater area, revealing that the displacement caused by the intrusion of the eruptive dike was not strong enough to produce as many fringes as observed on the western side of the volcano. Considering that this descending pair spans from a few days before the intrusion to a few months afterward and that the residual fringes testify to an additional ground motion away from the sensor (meaning westward and/or subsidence in descending geometry), we inverted this interferogram (see Figure 4, INV3) by adding a depressurizing source acting during and after the dike intrusion, as suggested by Aloisi et al. [2011]. The results of this second inversion are shown in the second row of the Figure 7a; 2826

10 Figure 7. Observed and simulated interferograms and residuals for (a) descending and (b) ascending data described in Figure 7. The top rows of each box report the results of the simulation considering only the dike (black line) modeled by GPS data, shown in Figure 5. The bottom rows of each box report the results of the simulation considering additional pressure sources (see text for details). 2827

11 Figure 8. Long-period (1 year) GPS displacements measured on the entire network. Dashed blue lines indicate the position of RNF and SVF/STF fault systems (see Figure 1). in this case, the residuals remain within 60.5 interferometric fringe over the studied area. The final result agrees with the effect produced by the dike intrusion superimposed to a deflating source (red star in Figure 7a), vertically elongated and located beneath the summit craters at a depth of 1 km below sea level (SX ¼ 500,210 m, SY ¼ 4,177,200 m, in UTM reference frame), with the semimajor axis, a more than 10 times longer than the b and c axes. The three Euler angles (indicating the ellipsoid orientation) show that the ellipsoid dips about 76 toward SSE. [33] We adopted the same procedure to also compare the ground deformation detected by ascending DInSAR data with the effect produced by the intrusion. In the first row of Figure 7b, we report the observed interferogram with the simulated one considering the effects of only the dike and the resulting residuals; even if we have no information on the upper part of the volcano, it is evident from the residuals that the dike produces too many fringes on the western flank. As a matter of fact, in this geometry, the motion of the ground is toward the sensor which means westward and/or uplift. Considering that the ascending pair spans from a few months before the dike intrusion to a month after, we had to produce a simulated interferogram by considering the effects of all the three sources previously detected: (i) the pressurizing source active before the eruption (INV1), (ii) the dike intrusion and finally (INV2), (iii) the depressurizing source active after the eruption (INV3). We highlight here that the pressurizing source strength has been opportunely scaled, under the assumption of a linear evolution of the ground deformation pattern over time, and the inflating source stops when the deflation one begins to be active. The resulting interferogram is shown in the second row of Figure 7b; it highlights that the subsidence produced by the depressurizing source acts, for the ascending geometry, in an opposite way with respect to the westward displacement produced by the dike, resulting in less fringes that better fit the 2828

12 Figure 9. Long-period (1 year) ENVISAT (a) ascending and (b) descending phase interferograms. observed deformation, as visible from the final residuals. 5. Longer Period (1 Year) Ground Deformation 5.1. June 2007-June 2008 GPS Data [34] Long-term GPS comparison allows a highly detailed map of the 3-D ground deformation field to be drawn, thanks to the more than 100 GPS stations processed (Figure 8) during the two surveys. The largest deformations in this data set result from the dike intrusion; this is evident when compared with the short-term data. Indeed, very strong displacements have been measured at all stations on the summit area and upper NE rift, belonging to the N-S kinematic profile. With this comparison, we exploit many GPS benchmarks with respect to the previous ones, for a more detailed and extended sampling of the ground deformation field all over the volcano. More information is available near the summit, from benchmarks all around the summit craters that were not measured in the previous months due to the snow coverage. Northeastward displacement of over 1 m affected NS08 benchmark, very close to the eastern border of the fracture field propagating toward the northern part of the volcano. On the other side of this fracture field, a maximum southwestward displacement of 38 cm was measured at NS10 benchmark, confirming the asymmetric ground deformation field, as previously detected by the short-term comparisons. Smaller displacement values at ECPN and EPLU stations (with respect to the short-period comparison) are due to the slight contraction measured after the beginning of the eruption and revealed by previous DInSAR data inversion. At EPDN station, this effect is masked by the eastward motion of this station during the months preceding the eruption. However, at the first order, the difference between this ground deformation pattern and the one resulting from the short term (the 6 13 May comparison) is proportionally very small, and we can attribute the deformation near the summit of the volcano mainly to dike intrusion. From the long-term deformation, it is even more evident how the strong displacements of the summit rapidly decrease downward and completely disappear on the northern, western and southern flanks at about 1700 m a.s.l.. Only on the eastern side of the volcano are the ground displacements significant. Here, displacements increase from South to North, passing from about 2 cm on the lower southeastern sector up to about 7 cm on the southern side of the Pernicana fault that entirely encloses the deformation, because the stations on its northern side show no significant displacement. Also the NE rift represents an abrupt closure of the ground deformation, evidenced by the very small displacement recorded at TMPR station, lying a few hundred meters NW. On the 2829

13 Figure 10. East, north, and up displacement components resulting from the SISTEM integration of longperiod data shown in Figures 8 and 10. Thick black line indicates the position of the eruptive fissure; blue solid lines indicate the dry fractures opened on the northern flank; dashed blue lines indicate the position of RNF and SVF/STF fault systems (see Figure 1). Coordinates are in meters, in WGS84 Ellipsoid UTM zone 33 projection. southern side, ground deformation is progressively enclosed by various structures. For example, the activity of the S. Tecla-S. Venerina fault system is evident from the different vertical motion of the stations lying on its northern side (stable) with respect to those on its southern side (subsiding) July 2007-June 2008 DInSAR Data [35] To investigate the long period covered by the two complete GPS surveys, we analyzed the corresponding ascending and descending ENVISAT- ASAR images. We processed the ENVISAT ascending pair spanning from 25 July 2007 to 4 June 2008 (Figure 9a) and the 1 August 2007, to 17 July 2008 descending interferogram (Figure 9b). The ground deformation pattern resulting from these interferograms is similar to the previously discussed March 2008-June 2008 short-term pair. Like the GPS, the interferograms confirm that the overall ground deformation pattern in the 1 year comparison is mostly dominated by the displacement induced by the dike intrusion episode and that the strongest ground deformation is confined to areas over 1700 m a.s.l.. The interferograms show a better overall coherence with respect to the previous short term ones, especially on the upper southern and northern flanks of the volcano, providing us more complete and detailed images of the complex kinematics affecting the volcano. In these 1 year pairs, indeed, there is no snow coverage and the loss of coherence is limited only to those areas having unfavorable slope geometries. [36] The detailed analysis of ascending interferogram showed, on the eastern side of the volcano, that the fringes extend locally at lower altitudes, forming two main lobes: the first one consists of two fringes extending toward SE, affecting the inner part of the Valle del Bove and extending downward to about 1000 m of altitude, more clearly visible than in the short term (see Figures 6a and 9a); the second one is stronger and affects the northeastern side of the volcano, from the NE rift to the upper part of the Pernicana fault that abruptly interrupts the fringe pattern. On the upper NE flank of the volcano, the gradient of the ground deformation is very strong, producing at least three fringes inside the rift itself. Ground deformation totally disappears north of the NE-rift, confirming the decoupling role played by this structure, as observed by GPS. Due to the unfavorable morphology for the ascending view, there is very little information in this interferogram on the upper western flank of the volcano, but at least two very close fringes appear on the northwestern part, suggesting a strong deformation gradient affecting the decorrelated area. [37] The descending view allows the deformation affecting the western side of the volcano to be better imaged than the ascending one. This interferogram shows that on the western flank the ground deformation pattern is more uniform, with a roughly semicircular shape of the fringes, without significant disturbances. This feature again confirms the more elastic behavior of this side of the volcano also in the long term [Bonforte et al., 2008]. On the eastern side, a stronger and more extended deformation is visible with respect to the short-term pair (see Figures 6b and 9b), due to the continuous accumulation of displacement on this side of the volcano. The decoupling role of the Pernicana fault is confirmed; it cuts all fringes affecting the eastern flank also at lower altitude, preventing their propagation toward the north. 2830

14 5.3. SISTEM Data Integration [38] We applied the SISTEM integration method also to the long-period GPS and DInSAR data (both ascending and descending views) to derive 3-D surface motion maps. These long-period data sets are the most complete data sets available for the studied period (more than 100 GPS stations and good coherence of the DInSAR data over almost all the investigated area). The high spatial resolution of the SISTEM results (Figure 10) finely depict the 3-D ground deformation pattern well, highlighting how the deformation related to the dike intrusion is confined to the upper part of the volcano. In addition, the high detail of the displacement pattern constrains the position of the dike at surface well, in perfect agreement with the position of the eruptive and dry fractures detected by geological surveys [Bonaccorso et al., 2011a]. [39] The asymmetric pattern of the ground deformation is clearly visible on the SISTEM results, because the pattern of the displacement is well shaped for all the three components of ground motion; the displacement maps highlight the stronger horizontal deformation with respect to the vertical one, that is, together with the local extent of the displacement pattern, gives an immediate confirmation of the shallow depth of the dike. [40] The E-W components show the highest displacement values and perfectly trace the location of the eruptive and dry fractures that are visible in the field, following also their curvature at the southernmost part, where the eruptive vents formed. On the northern flank, the E-W displacement field confirms the existence of an area affected by a marked E-W extension, where the dry fractures opened. While the westward displacements are fairly uniformly distributed over the upper western flank, the eastward ones depict a more complex shape; indeed, the eastward displacements extend to the north forming a lobe over the entire upper NE rift (symmetric to the northern lobe of the westward displacements) but they also extend eastward, involving the entire upper northeastern flank down to the Ripe della Naca faults that abruptly reduce the eastward motion. Both westward and eastward displacements are separated and shaped, on the southern flank of the volcano, by the trace of the fracture field formed in 1989, as mapped in Barreca et al. [2013]. This displacement pattern suggests that an E-W stretching affected this feature; in particular, the westward displacements follow the western side of this structure, down to an altitude of about 2000 m. [41] A similar pattern was imaged for the N-S displacement component; the curvature at the southern tip of the eruptive fractures is more evident here, as well as the northward extension of the dike beyond the summit area. From the distribution of the N-S motion, the left-lateral kinematics of the dike are evident. Also this map shows a clear lobe of the northward displacements affecting the entire upper part of the NE rift. This time it is not symmetric to the southward ones as for the E-W components. This asymmetric pattern of the N-S components is probably due to the northward propagation of the dike, in agreement with the opening of the dry fractures in the days following the dike emplacement. Again, the Ripe della Naca faults show an active role in shaping the displacement pattern, this time marking an abrupt increase of the southward displacement affecting the lower eastern flank of the volcano; this displacement seems to be confined, on the SE flank, by the S.Tecla-S. Venerina fault system. The southward displacements on the southern flank of the volcano are clearly shaped by the 1989 fracture system, revealing a temporary left-lateral kinematics of this discontinuity, as was already observed during the intrusion of the 2001 eccentric dyke [Gambino, 2004; Bonforte et al., 2008]. This pattern of the ground deformation seems to confirm the role of the 1989 fracture system in conditioning and distributing the ground deformation on the upper southern flank of the volcano, separating two blocks moving with different kinematics as suggested in Bonforte et al. [2004, 2008]. [42] Finally, the vertical components evidence a local uplift of the uppermost north part of the volcano over the intrusion. A subsidence affects the upper northeastern flank, just east of the NE rift, evidencing a westward tilt of the block comprised between the NE rift and the Ripe della Naca faults, bordered northward by the Pernicana fault as also detected in 2010 by Guglielmino et al. [2011]. Another subsiding area has been detected just east of the Valle del Bove, between the Ripe della Naca and upper part of S. Tecla-S. Venerina faults. Other local subsidence affects the lowermost northeastern and southeastern flanks of the volcano. 6. Discussion and Conclusion [43] Integration of the GPS and DInSAR data presented in this paper enable detailing in time and 2831

15 Figure 11. Sketch view of the ground deformation sources active before (INV1) during (INV2) and just after (INV3) the onset of the eruption. The orange ellipses indicate the location of the source of the tremor as reported in Bonaccorso et al. [2011a]. The large light blue arrows indicate the kinematics of the eastern flank during the investigated period. The small dark blue arrows indicate the propagation of the dry fractures on the upper northern flank. space the ground deformation field evolution preceding, accompanying and following the May 2008 dike intrusion (Figure 11). [44] In the preeruptive period, the SISTEM integration (June 2007-May 2008 time spanning) finely depicted the dynamic of the volcano, dominated by an eastward sliding of the northeastern flank with a high rate that increased in the first month of 2008 before the eruption. This dynamic was accompanied by a slight but visible inflation on the northern and western flanks. Inversion of the SISTEM results (INV1) allowed detecting a vertically elongated ellipsoidal source pressurizing beneath the upper western flank of the volcano at about 3.5 km below sea level and dipping 87 toward SSW. The position is in agreement with the usual path of the uprising magma from depth, beneath the western side as previously imaged by Bonforte et al. [2008] and reference therein and with the deep source detected by tilt data in Bonaccorso et al. [2011b] and interpreted as the storage of new gas-rich magma, periodically ascending to a shallower level and triggering lava fountains. [45] The dike emplacement has been finely imaged by a very short interval (6 13 May 2008) GPS comparison, over a dense network covering the summit area of the volcano, above 2000 m a.s.l. The inversion of this data set gave a highly constrained modeling of the dike (INV2), located exactly beneath the eruptive fracture, extending NNW-ward for about 3 km, dipping westward toward the central conduit. This model concurs with previous ones proposed by Aloisi et al. [2009] on the basis of CGPS data and with the tremor source location reported in Bonaccorso et al. [2011a]. This geometry confirms the dike injection started from the shallow feeding system of the volcano, the same that fed all the previous lava fountain episodes [Bonaccorso et al., 2011b]. The intrusion was probably deviated from the usual conduit and 2832

16 driven eastward by the local stretching induced by the eastward sliding of the northeastern flank measured during the previous months. Differently from 2002, when the magma intruded into the NE rift, in 2008 magma did not propagate radially from the main conduit but it rose upward, deflecting toward the closer surface represented by the steep western wall of the Valle del Bove. Petrological analyses confirm that the magma erupted during the first month of eruption came from the same body feeding the lava fountains; only on 17 June did the magma begin to show a more primitive composition, indicating that the eruption started to be fed by a new magma coming from a deeper source [Bonaccorso et al., 2011b]. This hypothesis is in good agreement with the deflation measured by InSAR data encompassing the first 2 months of eruption; indeed, InSAR data inversion (INV3) detected a depressurizing source at about 1.5 km below the sea level, beneath the tremor source. As happened for the silent eruption [Bonaccorso et al., 2006], the dike allowed the outflow of magma from a shallow storage inside the volcanic edifice. After the depletion of this shallow storage, magma began to be drained through the entire feeding system directly from the deeper storage levels. Unlike the silent eruption, which was induced by an exceptional sliding rate of the eastern flank coupled with a strong inflation of the whole volcano edifice, the seismicity recorded during the 2008 intrusion and ground deformation modeling confirm the forceful role of the injecting magma; furthermore, by contrast to the eruption, intense lava fountaining and intense strombolian activity characterized the eruption onset at the eruptive fissure [Bonaccorso et al., 2011a]. [46] These features make the eruption an intermediate case between a lateral eruption and the silent one. Indeed, by considering both the dike geometry and the source of the volcanic tremor [Bonaccorso et al., 2011a], it is possible to infer that, in this case, the high-rate spreading of the eastern flank only deviated the path of a gasrich magma batch eastward, rising forcefully from the same shallow reservoir feeding the lava fountaining. Magma intrusion also propagated northward, forming a well-defined west-dipping dike, but it was not able to reach the surface. In a feedback process, the forceful dike intrusion further pushed the NE flank of the volcano, which was affected by a consequent significant eastward displacement, well shaped and enclosed by the NE rift and Pernicana fault. [47] The unprecedented high-resolution 3-D displacement maps provided by the long-period SIS- TEM integration enabled depicting the dynamics of the volcano in response to the onset of the 2008 eruption with precision; in particular, the kinematics of the dike intrusion and its northward propagation has been shown, as well as the complex dynamics of the eastern flank. The eastern flank shows some interesting features revealing the active role of some important faults: (i) the Pernicana fault drives and encloses the ground deformation pattern on the north eastern side of the volcano confirming its usual role [Bonforte et al., 2007; Guglielmino et al., 2011b]; (ii) the Ripe della Naca faults decouple this side of the volcano into an upper and lower block with a westward tilt of the upper block according to Ruch et al. [2010], and with a southward displacement of the lower one; (iii) the NW-SE S. Tecla-S. Venerina fault system encloses the ground deformation pattern on the southeastern side with a main right-lateral kinematics (as previously detected by Bonforte et al. [2011] during a noneruptive period). [48] The detailed ground deformation pattern derived from DInSAR images and GPS surveys on a very dense network integrated by the SISTEM method revealed the complex kinematics of the volcano preceding, encompassing and following the May 2008 eruption onset, highlighting the interaction between flank dynamics and magma injection even in the case of a shallow dike injection. Acknowledgments [49] The authors would like to thank the GPS Permanent network staff of INGV-OE for data availability and the GPS surveys group for their fundamental work in carrying out periodic measurements on the geodetic networks; we are grateful to Stephen Conway for his precious contribution in improving the English language of the manuscript. This work was partially funded by the Task D7 Enhancement of the remote sensing laboratory of the project Extension and enhancement of the volcanic and seismic monitoring systems of Sicily, funded by the Sicilian regional government. The ENVISAT data were provided in the frame of the ESA CAT project. References Acocella, V., and M. Neri (2009), Dike propagation in volcanic edifices: Overview and possible developments, Tectonophysics, 471, 67 77, doi: /j.tecto Aki, K., and P. G. Richard (1980), Quantitative Seismology, Freeman, New York. 2833