Geophysical Research Letters

Source model of deformation at Lazufre volcanic center, central Andes, constrained by InSAR time series

Authors


Corresponding author: J. Pearse, Alberta Geological Survey, 4999 98 Ave. NW, Edmonton, AB T6B 2X3, Canada. (jillpearse@gmail.com)

Abstract

[1] Recent interferometric synthetic aperture radar (InSAR) observations of the Lazufre region, central Andes, show large-scale uplift at a rate of about 3 cm/yr, beginning between 1998 and 2002. The initiation of this activity during the ERS satellite mission and its proximity to active volcanoes have made Lazufre the focus of several studies aimed at understanding its source geometry, its relation to nearby volcanoes, and whether or not the source area is expanding. There now exists a longer time series from multiple ERS/ENVISAT satellite tracks that allows a more comprehensive examination of the source geometry and its temporal and spatial variation, if any. We processed 15 years of InSAR data from three separate tracks; modeling the different look geometries provided greater resolution of source depth and geometry. We conclude that the source is a shallowly dipping sill at a depth of about 8 km, with no evidence of lateral expansion.

1 Introduction

[2] The Lazufre region is located on the border of Chile and Argentina (see Figure 1), within the Central Volcanic Zone (CVZ) of the Andes. The CVZ is an active volcanic arc extending from Peru through southwestern Bolivia, North Chile, and northwestern Argentina [De Silva, 1989; De Silva and Francis, 1991]. It includes a number of collapsed calderas—remnants of ancient “supervolcano” eruptions—which are now thought to be inactive. Observations of the CVZ using interferometric synthetic aperture radar (InSAR), however, show evidence of recent unrest, including four broad (tens of kilometers) areas of surface uplift consistent with inflation of large crustal magma bodies [Pritchard and Simons, 2002]. Lazufre is located in the southern part of the Altiplano-Puna plateau, which is an area of high elevation caused by subduction of the Nazca plate beneath South America. Crustal thickening in the Altiplano-Puna [De Silva, 1989] has provided large volumes of melt to fuel catastrophic caldera-forming eruptions over the past 20 million years, and several collapsed calderas (remnants of these massive eruptions) are visible throughout the region. Lazufre itself is not associated with a particular caldera; however, the broad uplift pattern (approximately 30 km wide) suggests a magma body comparable in size to other known supervolcanoes such as Yellowstone and Long Valley [Ruch et al., 2008]. The area contains several distinct volcanoes, including the Lastarria volcano, which shows fumarolic activity [Naranjo and Francis, 1987; De Silva and Francis, 1991].

Figure 1.

Topographic map of the Lazufre volcanic region.

[3] There have been several recent studies aimed at characterizing the deformation at Lazufre; however, ours includes three independent satellite tracks and extends the InSAR time series into June 2010. Pritchard and Simons [2002, 2004] were the first to detect the deformation at Lazufre, which they named for its proximity to two potentially active volcanoes, Lastarria and Cordon del Azufre. They note that the deformation appears to have begun in about 1998, but it was difficult to pinpoint the onset precisely or determine how rapidly it occurred due to the limited temporal coverage between 1995 and 2000. Ruch et al. [2009] extended the InSAR data set and inverted InSAR data from the ERS-1/2 and ENVISAT satellites spanning 1995–2008 (from one descending orbit) to estimate a source depth of 12–15 km for a tensile dislocation in an elastic half-space. Their time series shows an uplift rate that is fairly steady, possibly with a slight increase in rate over time. Ruch et al. [2008] inverted subsets of their stack of interferograms spanning 2003–2006 and concluded that the magmatic source is actually expanding laterally at a rate of about 4 km/yr.

[4] Previous studies of Lazufre use data from a single satellite track, so that only motion along that satellite line of sight (LOS) is observed. Data from a single look direction may be fit by a variety of source types, and there is no way to assess the source depth and geometry independently. A small sill source, for example, is difficult to distinguish from a larger sill at a shallower depth if only one LOS is used [Dieterich and Decker, 1975; Fialko et al., 2001a]. We have acquired and processed a long time series of ERS-1/2 and ENVISAT data (spanning 1995–2010) to investigate in more detail the spatial and time-dependent aspects of the Lazufre deformation. We acquired data on three separate orbital tracks (one descending and two ascending) in order to better resolve the ambiguity between source geometry and depth. We use a Markov chain Monte Carlo method constrained by the InSAR data to infer the size, shape, and depth of the inflating source, assuming an elastic half-space. The length of the time series allows us to divide the time into distinct segments and model them individually to see whether the uplift rate is varying with time and also whether the source has been expanding laterally.

2 Data

[5] We processed InSAR data from the European Space Agency's ERS and ENVISAT satellites, including one descending track from both the ERS and ENVISAT satellites and two ascending tracks from ENVISAT. SAR interferograms were computed using the JPL/Caltech ROI_PAC software (baseline plots are provided in the Supporting Information). Mean velocity maps obtained from performing a time series inversion on all interferograms from each ENVISAT track are shown in Figure 2. The ENVISAT data show LOS ground velocities toward the satellite (uplift) occurring at a rate of approximately 2.5 cm/yr.

Figure 2.

Average line-of-sight velocity maps from each ENVISAT track: (a) track 282 (descending orbit), produced from a stack of 37 interferograms spanning August 2005 to June 2010; (b) track 318 (ascending): 27 interferograms spanning November 2005 to March 2010; and (c) track 404 (ascending): 34 interferograms spanning March 2003 to May 2010. Red indicates motion toward the satellite. The black box in Figure 2a outlines the location of the source model shown in Figures 5 and 6.

[6] To understand the time dependence of the deformation, we performed a least squares time series analysis of each ENVISAT data set, which gives displacement values at time points corresponding to each SAR acquisition [Usai and Achilli, 1999; Lundgren et al., 2001]. We use a variation on the small baseline subset approach [Berardino et al., 2002]. Results are plotted in Figure 3. Each ENVISAT time series shows an apparently constant rate of inflation; when the ERS acquisitions along track 282 are included in the track 282 data set, it appears that no discernible deformation occurred within the first few years of the first ERS acquisition in 1995. The sparseness of acquisitions between 1995 and 2000 makes it difficult to pinpoint the onset of uplift, but it appears that Lazufre began its current cycle of inflation sometime between 1998 and 2003 (see Figure 3). Unfortunately, since acquisitions ceased in 2010, there is no way to ascertain whether the deformation is persisting at a constant rate or whether it is speeding up (as suggested by the Ruch et al. [2009] time series) or slowing down. Time series from the ascending tracks (318 and 404) suggest that the rate may be slowing, but this cannot be confirmed using the available data.

Figure 3.

Time series of LOS deformation for all data sets. Blue crosses: all acquisitions from both ERS (1995–2001) and ENVISAT (2003–2010) on track 282; red circles: ENVISAT track 318; green stars: ENVISAT track 404. Line of sight displacement at time points corresponding to individual SAR acquisitions is plotted in centimeters. Vertical dashed lines indicate time segment divisions.

3 Modeling and Analysis

[7] We used the Markov chain Monte Carlo (MCMC) method as formulated by Fukuda and Johnson [2010] to invert the data for the deformation source, assuming an elastic medium, and a grid of 18 × 20 tensile patches on a plane. We solved for location (easting and northing) depth, strike and dip of the source plane, and tensile opening of each patch in a joint inversion of the average displacement maps for all three ENVISAT tracks. Results of this MCMC inversion are seen in Figures 4 and 5. Average displacement data from each of the three ENVISAT tracks are plotted in Figure 4, along with simulated displacement maps for the modeled source (projected onto each of the three tracks' line of sight), and residuals. The synthetic interferograms from the inferred source show a good fit with the data from all three tracks. Within the assumptions of an elastic half-space and a planar distributed tensile dislocation as the source type, the depth and source geometry appear well constrained. Histograms from the Monte Carlo simulations are included in the Supporting Information. The best fitting source is shown in Figure 5. The source plane is centered at a depth of approximately 8 km, with a strike of about 10.5°, and dipping down to the southeast at about 10°. The source appears to be an elongated sill with dimensions of approximately 20 km × 30 km and a maximum tensile opening rate of about 5 cm/yr. This depth is a little shallower than the 10 km estimate of Ruch and Walter [2010] and Anderssohn et al. [2009] and the 12–14 km estimate of Ruch et al. [2010], both obtained from inverting a single track of data (t282) for a single uniform tensile opening. We were better able to resolve the ambiguity between source geometry and depth through the projection of the horizontal and vertical deformation components into the three tracks' different radar LOS directions. Because we used a distributed source and the dip is shallow, the strike of the overall distributed plane is likely not as well constrained and may explain differences between our strike (around 10°) and that of Ruch and Walter [2010] (~28°), which appears better aligned with stress indicators.

Figure 4.

Results of Monte Carlo slip inversions, including the observed mean velocity map, the synthetic mean velocity map produced by the modeled source, and the residual (observed minus synthetic for ENVISAT tracks: (a) 282, (b) 318, and (c) 404. The same source model was used for all the synthetic displacement data.

Figure 5.

Source model obtained from Monte Carlo slip inversions: (a) map view and (b) cross-section view from south. The source geometry is an oblong sill, dipping down to the east at about 10°, with maximum displacement occurring at a depth of approximately 8 km. The maximum tensile opening is about 5 cm/yr.

[8] To assess whether the source is expanding laterally with time, we divided each time series into three time segments of approximately 3 years each (2002–2005, 2005–2008, 2008–2010.5) with one SAR scene shared between consecutive segments. The time segment divisions are marked with vertical lines on the time series plots (Figure 3). Note that the last time segment spans only 2.5 years because no data were acquired beyond June 2010. We computed average velocity maps for each time segment separately. For each time interval, we solved for the distributed opening with a Laplacian smoothed solution using a singular value decomposition least squares inversion. In these cases we used the overall geometry of the tensile source that was solved for using the MCMC method for the entire time series mean velocities (Figure 5). This assumes that the extended sill-shaped source is not moving vertically within the crust. The slip models produced by inverting time segments 1, 2, and 3 are shown in Figures 6a, 6b, and 6c. The observed and synthetic displacement maps and residuals are included in the Supporting Information.

Figure 6.

Slip models produced from Monte Carlo inversion of separate time segments: (a) 2002–2005, (b) 2005–2008, and (c) 2008–2010.5. Red color shows opening rate of each patch in meters per year.

[9] The separate time segment inversions result in sources with similar size, shape, and deformation rate. There is some variation between the different time segments, with the segment 2 source appearing larger than the others. This is probably due to an uneven distribution of points in the time series: segment 2 contains fewer acquisitions from track 282 than the others, and that section of the time history appears the noisiest (Figure 3). Also, acquisitions along the ascending tracks 318 and 404 only began in 2005, so inversion of the first time segment included only acquisitions from track 282, which may explain why the model source appears less smooth than the others. Given the variation from segment 1 to 2 and the lack of any resolvable change in shape from segment 2 to 3, we conclude that there is no apparent change in the shape or size of the deformation source during the 2002–2010 time period.

4 Discussion and Conclusions

[10] By processing InSAR data from ERS-1/2 and ENVISAT from 1995 to 2010 from three separate look directions, we were able to confirm that the onset of the current active inflation episode at Lazufre occurred between 1998 and 2002, after which it has continued at a fairly constant rate of about 3 cm/yr. Inverse models using a Monte Carlo method show that the source is a shallowly dipping oblong sill-shaped magma body centered at a depth of about 8 km, opening at a fairly constant rate of about 5 cm/yr. Furthermore, we were able to show using inverse modeling of three separate time segments that there is no evidence of lateral expansion of the magma body during the time of observation.

[11] Our observations and model results are consistent with a large steadily inflating sill-like magma body intruding into elastic crust. Unfortunately, we have no observations at Lazufre prior to 1995 and there is no known evidence of previous caldera-forming eruptions at that site, so it is difficult to characterize the long-term behavior of the Lazufre magma body. Lazufre is unique in that respect: all other known regions exhibiting inflation on this scale, such as Yellowstone [Chang et al., 2007], Long Valley [Hill et al., 1985; Langbein et al., 1993; Newman et al., 2006], Campi Flegrei [Di Vito et al., 1999], as well as the other large inflationary sources in the Central Volcanic Zone [Pritchard and Simons, 2002], are associated with calderas and evidence of long histories of eruptive activity.

[12] The apparent steadiness of the uplift is also unusual, given that other neovolcanic areas such as Yellowstone and Long Valley, for example, show episodic inflation and deflation on scales of months to years. Socorro (New Mexico) [Fialko and Simons, 2001; Fialko et al., 2001b] and Uturuncu (Bolivia) [Pritchard and Simons, 2004] do show steady uplift, but at significantly slower rates than at Lazufre (2 mm/yr for Socorro and 1 cm/yr for Uturuncu), and the inflationary sources are deeper (~20 km). The steadiness of uplift in those cases has been attributed to well-established longer-timescale processes (thermal inertia and viscoelastic relaxation in the case of Socorro [Pearse and Fialko, 2010] and a diapir emerging buoyantly from a large midcrustal sill in the case of Uturuncu [Fialko and Pearse, 2012]). Instead, the inflationary source at Lazufre appears to be new, began abruptly, and maintained a steady rate of inflation.

[13] The apparent constancy of the deformation rate may be due to the fact that our 15 years of measurements provide only a small snapshot of the inflationary period of the volcanic cycle. Also, since we observe no significant changes in the geometry of the magma body since the onset of inflation, either the sill-like intrusion formed rapidly—enough that its lateral expansion could not be detected using the available SAR scenes acquired around the time of onset of the deformation—or this represents the top of a finite volume magma body that has undergone new injection of magma or other fluids into an existing magma system, since geodetic data are rather insensitive to the deeper shape of an extended pressure source [Yun et al., 2006]. This suggests the possibility of an existing magma chamber at Lazufre that episodically fills on timescales that are much longer than our period of observation.

Acknowledgments

[14] ERS and Envisat SAR data were provided courtesy of the European Space Agency through M. Pritchard. The manuscript benefited from thoughtful reviews by M. Shirzaei and M. Poland. The research described in this paper was supported under contract with the National Aeronautics and Space Administration at the Jet Propulsion Laboratory, California Institute of Technology.

[15] Andrew V. Newman thanks Manoochehr Shirzaei and Michael Poland for their assistance in evaluating this paper.

Ancillary