Volcanic rift zones, characterized by repeated dike emplacements, are expected to delimit the upper portion of unstable flanks at basaltic edifices. We use nearly two decades of InSAR observations excluding wintertime acquisitions, to analyze the relationships between rift zones, dike emplacement and flank instability at Etna. The results highlight a general eastward shift of the volcano summit, including the northeast and south rifts. This steady-state eastward movement (1–2 cm/yr) is interrupted or even reversed during transient dike injections. Detailed analysis of the northeast rift shows that only during phases of dike injection, as in 2002, does the rift transiently becomes the upper border of the unstable flank. The flank's steady-state eastward movement is inferred to result from the interplay between magmatic activity, asymmetric topographic unbuttressing, and east-dipping detachment geometry at its base. This study documents the first evidence of steady-state volcano rift instability interrupted by transient dike injection at basaltic edifices.
 Dike emplacement along a rift zone on a volcano is a common trigger for flank instability [e.g., Cayol et al., 2000]. Therefore, it is reasonable to expect volcanic rift zones, characterized by repeated dike emplacements, to delimit the upper portions of unstable flanks at basaltic edifices, as has been proposed for Kilauea, Tenerife and Mount Etna [e.g., Cervelli et al., 2002; Walter et al., 2005; Ruch et al., 2010]. However, the relationship between flank instability and rift zone behavior, activity, and location on longer, inter-diking timescales is poorly understood. We use nearly two decades of InSAR observations at Etna to better understand this relationship.
 Mount Etna has three main rift zones, to the west (W), south (S), and to the northeast (NE). The latter rift consists of NE-striking eruptive fissures fed by shallow dikes propagating laterally from the central conduit (Figure 1) [Kieffer, 1985]. It is kinematically connected to the activity of the transtensive left-lateral Pernicana Fault System (PFS) (Figure 1) [Groppelli and Tibaldi, 1998]. The PFS bounds the northern margin of the volcano's unstable eastern flank, also affected by lateral spreading process [Borgia et al., 1992]. Motion of this flank is inhomogeneous and partitioned into sectors with distinct kinematics [Solaro et al., 2010; Bonforte et al., 2011]. The last two decades of activity of Etna have been characterized by a) inflation between 1994–2000, b) major flank eruptions in 2001 and 2002–2003 along the S and NE rifts, accompanied by dike intrusion and accelerated flank movement, and c) minor eruptions with limited deformation between 2004 and 2010) [e.g., Solaro et al., 2010].
 Following dike intrusions during the 2002–2003 eruption, the NE rift delimited the upper portion of the eastward moving flank [e.g., Acocella et al., 2003; Branca et al., 2003]. However, the longer-term kinematic behavior of this rift between periods of intrusive activity is poorly known. To characterize the relationship between rift kinematics and flank instability at Etna, we used long-term (1992–2010) InSAR time series of the volcano, giving unprecedented ground deformation data coverage of the volcano's summit. Results show, unexpectedly, that the NE and S rifts move eastward and are transiently affected by westward motion only during intrusive events.
 As snow coverage prevents coherence of the backscattered radar signal, long-term InSAR time series on the summit of seasonally snow-capped volcanoes, such as Mount Etna, are affected by decorrelation, which hinders detection of any summit deformation. To increase the summit coherence, we removed all the SAR images possibly affected by snow coverage. We then processed ERS satellite data between 1992 and 2010 and ENVISAT satellite data between 2003 and 2010, from both ascending and descending orbits, and computed 204 ascending and 194 descending interferograms. These were inverted by applying the Small Baseline Subset (SBAS) technique [e.g.,Pepe et al., 2005] to produce combined ERS-ENVISAT velocity maps and deformation time series. Ascending and descending maps and time series were used to separate vertical and East-West components. Our approach provided unprecedented spatial coverage of elevated areas. A spatial comparison to a similar dataset used bySolaro et al. , but limited to 1994–2008, shows an increase of the summit coverage of ∼14 km2.
 Horizontal (E-W component) ground deformation velocity maps clearly identify the boundary of the unstable portions of Etna (Figure 1). Following periods of linear and non-linear ground deformation behaviors previously observed in the deformation time series [e.g.,Solaro et al., 2010], we classify the data into three periods: P1 (1992–2000) with linear behavior (Figure 1a); P2 (2001–2003, including the 2001 and 2002–2003 eruptions), with non-linear behavior (Figure 1b); and P3 (2004–2010), with linear behavior (Figure 1c).
 P1 is characterized by the inflation of the edifice, with the east flank moving eastward and the west flank moving west [Lundgren and Rosen, 2003; Solaro et al., 2010]. However, our data show that a diffuse boundary between the east- and westward moving portions lay ∼5 km to the west of the NE and S rifts, and formed an overall arcuate shape in map view (Figure 1a). This boundary implies that the entire summit area, including the NE and S rifts, moved eastward at a rate of 1–2 cm/yr. Interestingly, the northern part of the arcuate boundary lay northward of the PFS (Figure 1a), which is expected to delimit the northern margin of Etna's unstable flank [e.g., Neri et al., 2004]. In fact, this outer unstable part had an eastward velocity of ∼1 cm/yr, nearly half of that observed along PFS and FF (Fiumefreddo Fault, Figure 1a), immediately to the south. This implies that the PFS and FF did not mark the northern boundary of the unstable flank, at least during P1.
 During P2, following the emplacement of dikes along the NE and S rifts (2001–2003), there was an overall westward movement of the western upper flank (blue areas, Figure 1b). The portion of the summit that moved west was delimited by the northern and southern tips of the dikes emplaced along the NE and S rifts, respectively (solid red lines, Figure 1b). The NE rift moved up to 40 cm westward over this transient period, while the area to the south of PFS propagated eastward. Notably, no eastward motion was observed north of the PFS, implying that the fault system bounded the unstable flank. Any continuation in P2 of the eastward motion of the NE rift observed in P1 was masked by the larger displacements induced by diking.
 Movements during P3 (2003–2010) were generally similar to those in P1. However, the eastward motion of the NE rift doubled (up to 4 cm/yr, Figure 1c), whereas there was a minor decrease in the eastward motion along the S rift. In contrast to P1 but similar to P2, the PFS confined the northern portion of eastward motion of the unstable flank once more.
 We further investigated the relationship between flank instability and the volcanic rift zone along the NE rift (Figure 2). P1 shows major uplift and eastward motion of the entire NE rift (Figures 2a and 2b). The boundary between the portion moving eastward and that moving westward lies approximately 2 km to the west of the NE rift (Figure 2c). P2 is dominated by the 2002 dike intrusion, which shifted the entire rift zone westward (Figures 2d and 2e); thus, the horizontal component predominated (Figure 2f). P3 shows a consistent eastward movement throughout the NE rift (Figures 2g and 2h), whose motion was associated with subsidence of the entire edifice (Figure 2i). Compared to P1, the boundary between the portion moving eastward and that moving westward in P3 shifted >3 km to the west of the NE rift (immediately to the west of the profile shown in Figure 2i).
 Finally, we examined the deformation time series of the central part of the NE rift (marked by a red star in Figures 2j and 2k). The vertical component of this point clearly shows the 1994–2000 inflation period and post-2002 subsidence (Figure 2j). The E-W component shows a constant eastward shift, during both P1 and P3, whereas the 2002 dike induced a significant and sudden westward shift (Figure 2k). While in P3 the eastward velocity was higher than in P1, the transient deformation (∼0.5 m) of the 2002 dike emplacement offset nearly two decades of continuous eastward motion.
4. Discussion and Conclusions
 At the volcano scale, our results highlight a general arcuate boundary that delimited the eastward moving portion of the volcano from 1992 to 2010. This boundary is diffuse and may traduce a buried set of normal faults, or, may be alternatively, related to a non-brittle flexure zone associated to major faults activity, as the PFS. This portion included the entire summit area, including the entire NE and S rifts. Both rifts experienced a continuous eastward motion, even though there was a marked acceleration along the NE rift and a minor deceleration along the S rift in P3 compared to P1. This behavior strongly implies that the rifts did not coincide with the upper boundary of the unstable flank during this time.
 Specifically, during P1 the arcuate boundary of the unstable eastern flank lay beyond the major faults expected to delimit its motion. The eastward motion extended north of the PFS and FF. This situation mirrors what is observed to the south, where the eastward motion extended south of the ATF (Aci Trezza Fault, Figure 1a) and Ognina-Belpasso Faults, which are inferred to confine the eastward motion to the south [Froger et al., 2001; Solaro et al., 2010; Bonforte et al., 2011]. During the inflation of the volcano, then, as shown by the E-W velocity field, both the northern and southern boundaries of the unstable flank shared a similar and symmetric deformation profile. From the outside to the inside, this deformation is characterized by (a) diffuse deformation, (b) minor fault systems (FF to the north and Ognina-Belpasso to the south), and (c) major fault systems (PFS to the north and ATF to the south). This progression is a consistent feature along the sides of the eastern flank of Etna during volcano inflation.
 Our analysis shows that during the last two decades the entire NE rift, including its west side, has experienced an almost steady-state eastward motion. The exception is the 2002 diking event, which transiently shifted the NE rift westward, reversing the long-term eastward motion. We attribute the increase in eastward velocity of the NE rift from P1 to P3 to an unbuttressing effect due to the major slip of the northeast flank after 2002 [Solaro et al., 2010]. Alternatively, the inflation of the edifice during P1 may alter the E-W displacement estimation; with a resulting minor E-W displacement compared to P3, characterized by a general edifice deflation.
 The steady-state movement of the entire NE rift is thus only interrupted by transient dike injections. Long-term non-magmatic factors (e.g. flank instability, lateral spreading) play a role in the evolution of the magmatically active portion of the NE rift and highlights an unexpected behavior for active volcanic rift zones, where magmatic processes should dominate.
 Despite the similar slope value, the western slope of the NE rift has a shorter length than the eastern slope (Figure 3). This asymmetric topography provides differential unbuttressing conditions at the NE rift base, promoting slip towards the east [Froger et al., 2001; Solaro et al., 2010]. At depth, this process is probably enhanced by the gravitational load of the NE rift, which consists of a cluster of shallow laterally propagated frozen dikes located above less dense sediments [Siniscalchi et al., 2012] (Figure 3). The intersection between dikes and sediments is characterized by a strong mechanical contrast that, together with the eastward motion, may encourage the formation of a diffuse shear zone in the long-term. To explain the long-term rift kinematics, we propose a model in which the rift zone is bounded at depth by two E-dipping detachments (Figure 3). A first, shallow-rooted detachment probably located ∼2.5 km below the NE rift, and connects through the Piano Provenzana fault zone to a second, deeper detachment located ∼6 km below the NE rift (Figure 3) [Ruch et al., 2010; Siniscalchi et al., 2012], both depths are in agreement with seismic activity in the PFS area [Alparone et al., 2012]. Both detachments may form a structural gradient of ∼3.5 km (H, Figure 3) to the east that may further reduce the NE rift's stability. Therefore, the steady-state movement of the NE rift may result from the interplay between magmatic activity and asymmetric edifice topography, the gravitational load of the intrusive complex, and eastward dipping detachments.
 This study documents the deformation of the two most active volcanic rifts at Etna on a decadal scale, and provides, at least for the better-understood NE rift, a plausible explanation for this deformation. The resulting structural model geometry may deserve future advanced numerical or analogue models. To our knowledge, this is the first documentation of steady-state movement of volcanic rifts interrupted by transient dike injections, at least within a single volcano. A similar process has been proposed to occur along the SW rift of Kilauea (Hawaii), whose movement may have been the result of piggybacking on the south flank of the larger Mauna Loa volcano as it grew [Lipman et al., 2006]. More broadly, this study shows that volcanic rift zones may not be the upper boundaries of unstable volcano flanks, as other non-magmatic processes may control the evolution of volcanic rift zones. We therefore propose a general applicability of our model to other basaltic edifices (e.g., Tenerife, Kilauea, Mauna Loa, and Piton de la Fournaise) that are characterized by non-magmatic processes such as flank instability, asymmetric topography, affected by shallow, laterally propagated dike intrusions.
 We thank P. Lundgren and A. Gudmundsson for their insightful comments; we are grateful to Paul Byrne for his extensive edits and comments on an earlier version of this paper. This project was partially funded by INGV and the Italian DPC (DPC-INGV project V4 “Flank”). ERS and ENVISAT SAR data were provided by ESA through the Cat-1 project no. 4532 and the GEO Supersite initiative. The DEM was obtained from the SRTM archive. ERS-1/2 orbits are courtesy of the TU-Delft, The Netherlands. SAR data processing has been done at IREA-CNR, partially carried out under contract “Volcanic Risk System (SRV)” funded by the Italian Space Agency (ASI).
 The Editor thanks Paul Lundgren and Agust Gudmundsson for their assistance in evaluating this paper.