Ground displacement assessment on Pico Volcano, Azores, by multitemporal InSAR data

Abstract. Interpreting the signal deriving from interferometric synthetic aperture radar (InSAR) analyses in volcanic islands, characterized by strong regional deformations and recurrent seismicity, is a complex and challenging issue. In these zones, the secondary effects connected to the SAR acquisition system cannot be neglected, and it is important to consider that delay phenomena of the electromagnetic waves, due to the propagation in the tropospheric layer and loss of SAR coherence because of dense vegetation, could affect the interferometric phase. This work focuses on Pico, the second largest and the youngest island of the Azores Archipelago (North Atlantic Ocean). This island consists of a central volcano and a fissure zone. These systems are inactive but recurrent microseismicity occurs in a rock volume hosting a partially crystallized magma storage system, which fed the recent activity of the central volcano. In the same area affected by microseismicity, the main volcanic edifice shows flank instability. All these elements support the hypothesis of possible reactivation of the shallow magmatic system. Aiming to check potential active ground displacements and to define their source, we collected two datasets of C-band Sentinel-1 SAR data, both in descending and ascending acquisition geometry, from January, 2017, to December, 2020. The application of the small baseline subset method of differential InSAR allowed drawing the mean ground velocity maps over the island and the displacement time series, useful to understand the deformation evolution. InSAR data only evidence areas affected by small-scale subsidence at the cinder cones of the fissure zone and along the southeastern slope of Pico volcano, where local debris flows activate during rainy periods.


Introduction
With a few exceptions, oceanic volcanoes are infrequently active and, consequently, few economic resources are dedicated to arrange large monitoring networks.In addition, the onset of a seismo-volcanic crisis is by far more rapid than the deployment and setting of a local network of sensors.Satellite remote sensing techniques are nowadays low-cost useful tools for rapidly monitoring volcanic areas, both for unrest detection 1 and during syn-eruptive phases to monitor lava emplacement and/or gas and ash emissions. 2,3rucial information for seismo-volcanic monitoring activities can be provided by integrating data interpretation from synthetic aperture radar (SAR) sensing from space, provided that data have been suitably filtered and cleaned from disturbing signals, such as atmospheric delay and SAR coherence loss, that very often mask out or hamper the reliable estimations of deformation processes.
At the oceanic island of Pico, central North Atlantic Ocean, morphological evidence revealed through field work and recurrent microseismicity suggest a possible instability of the central volcano that might be related to the emplacement of a magma body at depth.Aiming to define the spatiotemporal variability of this possible instability and its source, a multitemporal InSAR analysis of the island has been made covering the time interval from January 2017 to December 2020.This work also wants to possibly extend the previous analysis run by other authors 4 focused only on a deformation region along the southeastern flank of the Pico volcanic ridge.
The SAR images collected by the Sentinel-1 satellite mission of the ESA -European Space Agency have been exploited in this work.The stack of SAR images, on both ascending and descending orbits, has been processed with the small baseline subset (SBAS) method. 5,6lthough this method is robust and reliable, especially in areas where the loss of InSAR phase coherence is present, 7 the atmospheric delay strongly affects the interpretation of ground motion. 8Therefore, customized settings and processing approaches are needed for mitigating these undesired effects.
2 Geological Background

Tectonic Setting
Pico volcano is the main geomorphological feature of a volcanic ridge, which also includes the edifices of the nearby Faial island, and it was generated from the interplay between magmatism from a fissure zone along a ∼N120°spreading system and N60°and N150°local tectonic systems. 9The island is composed of three overlapping volcanic systems: Topo volcano, Planalto da Achada fissure zone, and Pico volcano (Fig. 1).
Topo is the remnant of an extinct volcano, which is exposed only in the southern part of the island, eroded and marked by a large semicircular collapse scar (CS).
The Planalto da Achada fissure zone defines a 30-km long N120°-trending ridge of cinder cones.About 170 cinder cones, some of which have a considerable size, formed a steep ridge.Lava flows overlap the Topo edifice and are intercalated with basalts from the Pico volcano to the west.The oldest lava flow unit is 230 ka old 10 and notwithstanding the last magmatic event that occurred from 1562 to 1564; this system is volcanically and seismically inactive.
Pico volcano is the youngest of the Azores 11 (K/Ar age of 53 AE 5 ka), although its base has never been dated.It erupted at least 22 times in the last 1500 years, 12 with the last events in 1718 and in 1720 (Fig. 1).Many recent cinder cones and hornitos punctuate the basal shield volcano.A steep summit cone (up to 40 deg) rises to 2351 m above sea level (a.s.l.) without lateral cones starting from 1500 m a.s.l.The summit is truncated by a 550-m-wide subcircular pit-crater, partially occupied by a 110-m-high hornito and intersected by an N60°-trending eruptive fissure.
Recent eruptions (≤ 10 ka) occurred along three main tectonic directions 9 (Fig. 1): (a) the regional transtensive trend that cross-cuts the entire island with an N120°direction, evidenced by the alignment of cinder cones of Planalto da Achada fissure zone, on the eastern side of the island and the sparse cinder cones on the western flank of Pico volcano.Past GPS campaigns revealed a spreading velocity in the order of 3.5 mm/year for this system; 13 however, no seismicity or geochemical anomalies have been recorded so far.(b) The local transtensive trend with N60°d irection represented by a dextral fault that links the cinder cones of the São Mateus area, on the south coast of the island, to a series of dykes on the north-eastern coast.This fault caused the offset of the N120°regional trend.No seismicity has been recorded so far in this tectonic system.(c) The Lomba de Fogo-São João (LFSJ) normal fault with N150°direction that intersects the summit crater.Scarce instrumental seismicity (ML ≤ 2.8) characterizes the southern segment of this fault.Another normal fault parallel to the LFSJ is the Santo António fault, on the northeast flank of the volcano.Cinder cones along this fault are partially buried by younger lavas from other volcanic vents, suggesting that this fault is not associated with recent magmatism.The area between these two N150°faults is devoid of cinder cones, which suggests that the strain produced by the tectonic stress field is reduced here.

Ground Deformation Features
Some field works carried out during the last decade, corroborated by observations from the analyses of aerial photographs, revealed that a subcircular collapse structure with a circumference of ∼1.7 km truncates the top of the volcano.This structure is a collapse crater, further geomorphologically reworked by lateral slope failure.The crater wall is made up only of fractured lavas and is still preserved from SE to NW. Angular lava blocks of different sizes form a continuous deposit of loose debris, inside the crater, at the foot of the walls suggesting recent instability.Pāhoehoe lava, emitted from a hornito located to the east of the center of the crater, has flooded the crater.A large sector collapse has removed in recent times the northern sector of the crater wall, exposing a 40-m-thick section of a lava sequence [Figs.

Seismicity
Seismic data were recorded from the permanent monitoring network operated by the Centro de Informação e Vigilância Sismovulcânica dos Açores on Pico Island.The seismic network Fig. 1 DEM of the island of Pico showing the three volcanic edifices, historic lava flows, and the tectonic features. 9The thick dashed line marks the N120°trend associated with the regional spreading of the Pico-Faial system.This system is dislocated by the N60°transtensive system marked by the thin dashed line.The thin line passing through the crater marks the LFSJ fault, along which the 1718 eruption occurred.The other thin line marks the S. Antonio fault, which follows the same trend as the previous fault.The inset shows the location of the islands in the Atlantic Ocean and the main tectonic structures (TF, Terceira rift; MAR, mid-Atlantic ridge; EAFZ, East Azores fracture zone; GF, Gloria fault).
comprises four short-period stations: three deployed around the Pico volcano and one on the eastern end of the island.
Since the replacement of some seismic stations occurred at the beginning of 2017 and aiming to consider only high-quality data, we selected 150 events following these criteria: earthquakes were recorded from at least three stations and with root mean square of travel time residuals (RMS) of ≤ 0.3 s (using the standard seismic velocity model for the Azores region 14,15 ).These high-quality events were relocated using the HYP program, a modified version of HYPOCENTER, 16 integrated into the SEISAN 10.3 analysis software 17 and a five-layer one-dimensional P-wave velocity model adapted for Pico based on the model for offshore Faial. 18he analysis of these events showed recurrent microseismicity (ML ≤ 2.8) along the southern side of the LFSJ [Fig.3(a)], centered at a depth of 5.6 and 6.7 km, where the N150°a nd N60°systems intersect each other and define the small reservoir that fed the activity of the summit Pico volcano crater during the last 5 ka. 9    Both descending and ascending acquisition geometries have been elaborated using the SARscape software (sarmap SA, 5.6.2version), covering a whole time span that ranges from January, 2017, to December, 2020.The details of such a dataset are reported in Table 1.The temporal sampling of the SAR images is 12 days, i.e., we used only the images taken by the Sentinel-1A platform, the first of the twin spacecraft launched between 2014 and 2015.
To estimate the mean ground velocity (MGV) maps and the displacement time series (DTS) over the island, we have applied the SBAS method. 5,6SBAS is a multitemporal SAR interferometric technique that exploits many interferograms calculated by pairing SAR images following specific spatiotemporal constraints, aiming at preserving SAR phase coherence and point measurement density.
The geographical location of Pico Island and its land cover had a critical impact on the first multitemporal InSAR processing.In such areas, the effects connected to the SAR acquisition system cannot be neglected, and it is important to consider the interferometric phase delay phenomena, because of the propagation in the tropospheric layer of the electromagnetic signals, loss of coherence because of dense vegetation, 21 and the issue related to phase inconsistency that can induce biases in the final velocity maps. 22Therefore, we have addressed the problem of the optimization of the abovementioned spatiotemporal constraints for mitigating these undesired effects.
We found out that the best results are obtained considering the minimum and maximum time interval between acquisitions (temporal baseline) equal to 36 and 180 days, excluding the short temporal baseline interferograms according to Ref. 22, to decrease the overall phase error.For our elaborations, it was considered a maximum orbital separation (spatial baseline) not exceeding 115 m for the ascending track and 105 m for the descending one.These settings allowed us to obtain 792 and 817 interferograms for ascending [Fig.4(a)] and descending [Fig.4

(b)] orbits,
Table 1 Main features of the Sentinel-1 dataset used for the interferometric processing: acquisition geometry, orbital track number of the satellite, number of images used during each processing, number of interferograms evaluated as reliable for the results production, time interval considered in each processing, and angle of incidence at the scene center.respectively, that have been used as input in the single value decomposition (SVD) algorithm that produced the first estimation of MGV maps.
A multilooking operation equal to 7 and 2 for the range and azimuth direction was applied to obtain a final resolution on the ground of 30 m.The ALOS Word 3D 30-m digital elevation model (DEM) was used to remove the phase topographic contribution, and the Delaunay MCF unwrapping method 23 was used in the processing.Note that the unwrapping step has been done after the phase-noise filtering.The latter is an extension of the Goldstein method 24 that takes the coherence-dependent alpha parameter to perform selective filtering; therefore, the incoherent areas are filtered more than the coherent ones.This approach permits to significantly improve the fringes visibility, minimizing the decorrelation effects and the signal loss. 25oints below a coherence threshold of 0.2 have not been unwrapped.
A final very important step of the SBAS processing is the temporal and spatial filtering of the interferometric stacks.This step is needed to estimate and remove atmospheric-related SAR phase and correct the MGV from phase-to-altitude correlated signals. 8,26These latter patterns were visible in most of our ascending [Fig.5(a)] and descending interferograms; therefore, to retrieve correct measurements, it was necessary to apply a specific filter implemented in the SARscape software 27 (atmosphere height correlation filter) that allowed us to estimate and remove the height-related component of the atmosphere from each interferogram.The filter assumes a power law correlation between altitude and tropospheric delay.An example of a filtered interferogram is shown in Fig. 5(b).

Rainfalls
Rainfall data have also been analyzed to check possible correlation with the observed deformation patterns. 28Data from four weather stations are available on the island. 29The first station is in São Caetano, 3.2 km SW of the summit cone, at 746 m a.s.l.A second station is in Cabeço do Teicho, 4.6 km NE of the summit cone, at the side of the transversal road cutting through the island, at 847 m a.s.l.A third station is in Cabecinho, 7.5 km west of the eastern tip of the island, along the fissure zone at 564 m a.s.l.The last station is in Canada das Bandeiras, at the junction with the transversal road heading from Madalena, 8.2 km away, 628 m. a.s.l.We have used only the pluviometric data of São Caetano due to the proximity of some InSAR points showing ground motion; indeed, the output of InSAR processing does not present measurement points close to the other stations.

Results
Figure 6 shows the final mean ground displacement maps after the application of the SBAS method to the ascending [Fig.6(a)] and descending [Fig.6(b)] Sentinel-1A dataset.Ground displacement rate is in millimeters per year (mm/year) and negative line of sight (LOS) velocities show ground movements away from the satellite sensor, whereas positive LOS velocities imply ground movements approaching the satellite sensor.Note that these maps provide relative velocities, i.e., they are referred to a point in the map that is assumed stable.This reference point is indicated with the black triangle symbol on the maps, and it is located in Madalena village.Both the ascending and descending displacement maps show the highest negative LOS velocities beside the cinder cones of the fissure zone central area and along the southeastern slope of the Pico volcano.
It is well known that SAR can provide deformation measurements only along its LOS.Actually, it is monodimensional information.This limitation can be overcome by exploiting the two orbit datasets.Indeed, the two MGV and DTS can be combined and, therefore, retrieve an estimation of the nearly vertical motion and nearly east-west motion.Unfortunately, any movements in the north-south direction are not detectable because of the intrinsic weak sensitivity to this direction of the SAR polar-orbiting systems, such as Sentinel-1.Following the formulation provided in Appendix A of Ref. 30, which considers the displacements taken over the same point from descending and ascending orbits and the look angles of both acquisition modes, we produced the maps reported in Fig. 7.
Figure 7(a) shows the vertical displacement maps, whereas Fig. 7(b) shows the horizontal (east-west) displacement map.Maps of the vertical displacement rate show negative values if the ground motion is downward, whereas positive values if surface uplift occurs.The maps of the displacement rate in the horizontal direction display negative values if the motion is oriented to the west, whereas positive values if the movement is oriented toward the east.We can note that the east-west map does not present significant movements, whereas some ground movements are visible in the vertical direction.Inside the MGV maps, the blue diamond refers to the weather station in São Caetano.
The thin black box within the MGV maps shows the position corresponding to the close-up of the summit area of the Pico volcano presented in the insets of Figs.7(a) and 7(b).The inset contains the ground deformation features revealed from some field works carried out during the last decade and described in Sec.2.2.Within the enlargements of the Pico summit cone, our results show the absence of noteworthy vertical and horizontal motions near the rim of the old central crater (OCC), at the eastern TF and CS.
The blue and magenta boxes in the vertical MGV map [Fig.7(a)] represent the two areas where patterns of significant vertical ground motions were detected, i.e., the cinder cones of the fissure zone (magenta box) and the area along the southeastern flank of Pico volcano (blue box), in agreement with the slope of these two zones.The ground velocity ranges around −7 mm∕year in the SE slope of Pico volcano and between −5 and −15 mm∕year in the central fissure zone.
To estimate the interferometric results errors, for both vertical [Fig.7(a)] and horizontal [Fig.7(b)] displacement maps, the associated precision using the formula reported in Ref. 31 was also calculated.The output, derived from parameters, such as interferometric coherence and wavelength, is shown in Fig. 8.

Discussion
For the two areas where patterns of noteworthy vertical ground motions were recorded, i.e., the area along the southeastern flank of Pico volcano [blue box in Fig. 7(a)] and the cinder cones of the fissure zone [magenta box in Fig. 7(a)], we have analyzed the DTS to follow the ground movement evolution.As an example, we have selected one point in each area: point A in the southeastern flank of the Pico volcano [Fig.9(a)] and point B in the fissure zone [Fig.9(b)].
The time series plots show a dominant linear trend of downward deformation [Figs.9(c) and 9(d)].Small long-wavelength oscillations of a few mm seem to be present in the point A movements [Fig.9(c)].These oscillations have suggested the comparison with the rainfall data coming from the close São Caetano weather station, which are superimposed on the plot (light blue line).We have found that after peaks of rainfalls (black dashed lines) the ground increases its downward velocity, as shown in the period from November, 2017, to June, 2018, and from March to June, 2019, just after almost no motion between the rainfall minima and the rainfall maxima.Indeed, "flat" trends are visible in the months preceding the rainy periods (i.e., on the left side of the vertical dashed lines).Moreover, the absence of displacements recorded by the SAR from the end of June, 2018, to early February, 2019, is consistent with the period of almost no rainfall that lasted from April to October, 2018.Figures 9(e  Overall, in the two areas investigated in more detail, a prevalent downward movement is noted and the absence of significant displacements in the EW direction in both cases is visible in Fig. 10. The multitemporal InSAR analysis performed with Sentinel-1A data from January, 2017, to December, 2020, revealed that the island of Pico is overall stable.The approach applied to remove systematic biases and to filter out the strong atmospheric component that afflicted many of the interferograms generated through SBAS processing seems to have provided reliable results, as confirmed by the high InSAR measurement precision (Fig. 8).The method used in this work agrees with the recently published strategy adopted for the study of ground deformations on another oceanic volcanic island, 32 where the same solution to the phase bias and atmospheric removal issues was found.
MGV maps highlighted that the ground deformation features, i.e., crater depression, TFs, and CSs, identified during field campaigns performed in the last decade, do not show any ongoing displacement [insets in Figs.7(a) and 7(b)].The only two areas within the island where movements have been recorded by SAR sensors, as visible both within the displacements maps measured along the satellite LOS and inside the maps representing the average vertical displacement rate, are along the southeastern flank of the main volcanic edifice (Pico volcano) and along the cinder cones in the central part of the fissure zone (Fig. 9).
Vertical movements recorded along the SE steep flank of the Pico volcano, characterized by an average rate of 7 mm/year, are linked to slow-moving grain flow, i.e., the sliding of coarse pyroclastic materials.The dynamics of the movement also seem to be controlled by the accumulation of water during the wettest months, as accelerations of the downward displacements are noted following the saturation of the volcanic materials (e.g., Ref. 33).
Maximum vertical displacement rates (about 15 mm/year) were recorded at the recent cinder cones along the fissure zone.This area is free from vegetation and the pyroclastic material of the cones is made of fine grain size materials that slide gravitationally downward, consistent with the morphology of these volcanic structures [Fig.9(f)].Therefore, the evolution trend of the recorded movements inside the Pico island suggests that both the difference in soil moisture content and the grain size between different volcanic soil types affect the deformation rate.
Seismic data were analyzed during the same time interval considered for the SAR elaborations.We believe that, given the low magnitude and deep hypocenters of the earthquakes that occurred from 2017 to 2020, the ground deformation measured by SAR at the two previously cited areas is not driven by seismicity but rather by gravity.Moreover, the depletion of the water table cannot be the cause of the observed movements since the main groundwater reservoir at Pico is placed at sea level, and only a few of the sparse perched bodies above are characterized by very reduced outflow (0.001 m 3 ∕s). 34nfortunately, because of the lack of InSAR coherence over the main body of the suspected landslide observed by other authors, 4 we did not study the ground deformation along the southeastern flank of the Pico volcanic ridge.To obtain reliable data and draw solid conclusions about that area through SAR results and in the absence of other external data, extensive coverage of the slump area is needed.
Future development of this work could be the updating of the multitemporal SAR results obtained with Sentinel-1 data acquired in the C-band and the processing of new data acquired with short revisit times and with longer wavelengths (e.g., L-band) to maintain InSAR coherence even in the most densely vegetated areas.
We might hypothesize that no shallow magmatic intrusions occurred during the analyzed period or that any possible magma ascent, associated with a reactivation of the shallow magmatic system of the Pico volcano, might produce deformations that are not measurable at the surface by the satellite geodetic techniques used in this work. 35Conclusions SAR remote sensing from satellites is a powerful tool for the analysis of ground deformation in volcanic environments.We have applied for the first time the SBAS multitemporal InSAR technique to study ground motion on the entire island of Pico, aiming at investigating possible active ground displacements and at a better characterization of its source.Our results did not show any relevant ongoing deformation in the timeframe of our observations.Main conclusive remarks about the observed elements of concern are: • Seismicity during the 2017 to 2020 period is too weak (ML ≤ 2.8) and too deep (∼5 to 6 km) to cause any displacement at the surface.Even the fault crossing the magma reservoir, the source of seismicity, does not show crustal deformation.• Summit crater depression is stable.It is not widen and deepen, and the TFs on the southern slope of the volcano are not moving.CSs show stability in the period of observation.• Intense rainfall periods cause small mass movements in unconsolidated sediments at steep cinder cones and at the intermediate/bottom of CSs.
To conclude, oceanic islands are challenging study areas because atmospheric and phase coherence disturbances affect the InSAR signal.Such interferences can be strong enough to generate false interferometric fringes and yield misleading results.Adequate filtering and processing strategy are fundamental to obtain reliable results.
2(a) and 2(b)].Wellpronounced radial CSs characterize the whole eastern half of the summit cone volume.These scars are free of vegetation, suggesting that they are recent.Traction fractures (TFs) (up to 1.8 m wide and several meters deep) are present 750 m south of the summit hornito, at the boundary of the former crater terrace, between two narrow CSs [Figs.2(b) and 2(c)].Many of these fractures follow the N60°direction and define an area of slope instability prone to collapse.

Figure 3 (
b) shows the frequency histogram of the seismic events recorded in the time interval previously described.

Fig. 3
Fig. 3 (a) Distribution of recent seismicity below the island of Pico and (b) histogram with the frequency of the seismic events.

Fig. 2
Fig. 2 Aerial views of the summit of Pico volcano.(a) Its summit cone comprises two collapse craters, the last one is partially filled by the lavas from a nested hornito.Sector collapses affect the two craters for at least the last 2500 years, generating rock-and-debris flows.These scars are particularly pronounced in the northern flank, where they are from 200 to 250 m wide and extend up to 1.2 km along the upper cone.(b) These collapses exposed a recent lava pile filling the crater.(c) Hundreds of meters-long traction fissures populate the southwestern flank of the older crater, at 2050 m a.s.l..These fissures are several meters deep and quite fresh and mark limited volumes of pāhoehoe lava, prone to collapse.

Fig. 4
Fig.4The so-called time-position plot showing the network of computed interferograms for (a) ascending and (b) descending SAR data.Time-position plot provides the normal distance from a reference image (called super master) chosen automatically during the network links generation (y axis) and the input acquisition dates (x axis).The plot highlights the dense network of connections and redundant interferograms that reflects a more robust solution to the SVD algorithm.

Fig. 5
Fig. 5 Ascending wrapped interferogram presented in slant range coordinates, i.e., flipped northsouth, computed with Sentinel-1 data acquired on September 27, 2020, and November 2, 2020, (a) without the atmosphere height correlation filter and (b) after the application of the filter.

Fig. 7
Fig. 7 (a) Vertical and (b) east-west displacement maps were drawn using Sentinel-1A data.The blue diamond in the vertical motion map refers to the weather station in São Caetano, and the two blue and magenta boxes highlight the locations where some ground motions have been detected.The black box, centered on the summit of the Pico volcano, indicates the area that is shown in detail in the inset that contains the ground deformation features described inside Sec.2.2.Abbreviations inside the inset are: OCC, old central crater; TF, traction fractures; CS, collapse scars; NCW, northern crater wall; SH, summit hornito.
) and 9(f) allow appreciating the present morphology of the two areas investigated through DTS analysis.The satellite images of the Pico volcano main cone [Fig.9(e)] and of the fissure zone [Fig.9(f)] were acquired in December, 2020, and are freely available in the Google Earth software.Both images show pyroclastic

Fig. 8
Fig.8(a) Vertical displacement precision map and (b) horizontal displacement precision map.These output products, which are derived from parameters, such as coherence and wavelength, provide an estimate of the interferometric results error.The higher the precision value, the lower the measurement precision.

Fig. 9
Fig. 9 Tridimensional maps of the areas affected by deformation: (a) the SE flank of Pico volcano, (b) the cinder cone area along the fissure zone.Below each 3D map, the graph showing the time series of nearly vertical displacements in (c) points A and (d) B is reported.(e), (f) Satellite images of the two areas investigated through DTS.Images were captured in December, 2020, and are freely available in the Google Earth software.

Fig. 10
Fig. 10 Tridimensional maps of the areas affected by vertical deformation, which show no significant displacements in the nearly EW direction along (a) the SE flank of Pico volcano and (b) at the cinder cone area along the fissure zone.Blue and magenta boxes are referred to Fig. 7(a).